slug stringlengths 15 15 | content listlengths 1 129 | rawContent stringlengths 1 2k | author dict | attachments listlengths 0 49 | mentions listlengths 0 49 | reactions listlengths 0 12 | publishedAt stringlengths 24 24 | updatedAt stringlengths 24 24 | commentators listlengths 0 52 | url stringlengths 25 46 | totalUniqueImpressions int64 1 42.1k โ | numComments int64 0 621 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
986806198797388 | [
{
"type": "text",
"value": "๐๐๐๐, ๐ญ๐ก๐ ๐ฒ๐๐๐ซ ๐จ๐ ๐๐ ๐๐ง๐ญ ๐ฐ๐จ๐ซ๐ค๐๐ฅ๐จ๐ฐ๐ฌ ๐ง๐ฆพ๐ค",
"raw": "๐๐๐๐, ๐ญ๐ก๐ ๐ฒ๐๐๐ซ ๐จ๐ ๐๐ ๐๐ง๐ญ ๐ฐ๐จ๐ซ๐ค๐๐ฅ๐จ๐ฐ๐ฌ ๐ง๐ฆพ๐ค",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null... | ๐๐๐๐, ๐ญ๐ก๐ ๐ฒ๐๐๐ซ ๐จ๐ ๐๐ ๐๐ง๐ญ ๐ฐ๐จ๐ซ๐ค๐๐ฅ๐จ๐ฐ๐ฌ ๐ง๐ฆพ๐ค
I've just watched Andrew Ng's talk at Sequoia last week.
If you're interested in Agents, you should really watch it!
๐ช๐ต๐ ๐๐๐ฒ ๐ฎ๐ด๐ฒ๐ป๐ ๐๐ผ๐ฟ๐ธ๐ณ๐น๐ผ๐๐?
The current LLM task solving workflow is not very intuitive:
We ask it โwrite an essay all in one shot, without ever using backspace.โ
Why not allow the LLM a more similar process to what we would do?
- โWrite an essay outlineโ
- โDo you need wen research?โ
- โWrite a first draftโ
- โConsider improvementsโ
โฆ
This is called an Agentic workflow. Existing ones bring a huge performance boost. With HumanEval: GPT-4 zero-shot gets 67% score, agentic with either one of tool use or reflection goes over 90%, and the combination of the two scores even higher!
๐๐ด๐ฒ๐ป๐๐ถ๐ฐ ๐ฟ๐ฒ๐ฎ๐๐ผ๐ป๐ถ๐ป๐ด ๐ฑ๐ฒ๐๐ถ๐ด๐ป ๐ฝ๐ฎ๐๐๐ฒ๐ฟ๐ป๐
On the following two points, the tech is robust:
โ๏ธ ๐ฅ๐ฒ๐ณ๐น๐ฒ๐
๐ถ๐ผ๐ป: For instance: add a critic step after the writing step
๐ ๏ธ ๐ง๐ผ๐ผ๐น ๐๐๐ฒ: extends the capabilities of the LLM by allowing it to call tools, like search or calculator
The next two will be needed to go further, but the tech for them is more emerging and not reliable yet:
๐บ๏ธ ๐ฃ๐น๐ฎ๐ป๐ป๐ถ๐ป๐ด forward to decompose task into subtasks. This allows great behaviours like an AI Agent re-routing after a failure
๐ ๐ ๐๐น๐๐ถ-๐ฎ๐ด๐ฒ๐ป๐ ๐ฐ๐ผ๐น๐น๐ฎ๐ฏ๐ผ๐ฟ๐ฎ๐๐ถ๐ผ๐ป: Program a flock of agents with tasks.
Improving the two above points will unlock huge performance boosts!
Andrew NG says Research agents are already part of his workflow!
๐๐น๐ผ๐๐ถ๐ป๐ด ๐๐ต๐ผ๐๐ด๐ต๐๐
Andrew speculates that through agentic workflows, maybe generating many tokens fast from a small LLM will give better results than slower throughput from a powerful LLM like GPT-5.
๐ฌ Watch the talk here ๐ https://www.youtube.com/watch?v=sal78ACtGTc
๐ I've added his recommended reads to https://huggingface.co/collections/m-ric/agents-65ba776fbd9e29f771c07d4e | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/63d10d4e8eaa4831005e92b5/7p7-OmWM6PqqCs7ZStPGD.jpeg",
"fullname": "Aymeric Roucher",
"name": "m-ric",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 494,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"osanseviero",
"clem",
"neelam91721",
"samusenps",
"nikgr",
"JoPmt",
"Moibe",
"abidlabs",
"poGlingus"
],
"count": 9
},
{
"reaction": "๐",
"users": [
"apol",
"abidlabs"
],
"count": 2
... | 2024-04-02T16:08:03.000Z | 2024-04-03T08:45:03.935Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/61b9df9b22e5b0fdd501a113/i2yTGbK7pFnw9YLwZ7elp.jpeg",
"fullname": "Akhil B",
"name": "hakunamatata1997",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 4,
"isFollowing": fals... | /posts/m-ric/986806198797388 | 1,879 | 1 |
704809400668436 | [
{
"type": "text",
"value": "Aurora-M",
"raw": "Aurora-M",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": null,
"resource": null,
"url... | Aurora-M
The First Open Source Multilingual Language Model Red-teamed according to the U.S. Executive Order
https://huggingface.co/papers/2404.00399
Pretrained language models underpin several AI applications, but their high computational cost for training limits accessibility. Initiatives such as BLOOM and StarCoder aim to democratize access to pretrained models for collaborative community development. However, such existing models face challenges: limited multilingual capabilities, continual pretraining causing catastrophic forgetting, whereas pretraining from scratch is computationally expensive, and compliance with AI safety and development laws. This paper presents Aurora-M, a 15B parameter multilingual open-source model trained on English, Finnish, Hindi, Japanese, Vietnamese, and code. Continually pretrained from StarCoderPlus on 435 billion additional tokens, Aurora-M surpasses 2 trillion tokens in total training token count. It is the first open-source multilingual model fine-tuned on human-reviewed safety instructions, thus aligning its development not only with conventional red-teaming considerations, but also with the specific concerns articulated in the Biden-Harris Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. Aurora-M is rigorously evaluated across various tasks and languages, demonstrating robustness against catastrophic forgetting and outperforming alternatives in multilingual settings, particularly in safety evaluations.
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1674929746905-60f1abe7544c2adfd699860c.jpeg",
"fullname": "AK",
"name": "akhaliq",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5205,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/60f1abe7544c2adfd699860c/7C3ctlbiugipQMW9FLjpJ.png"
}
] | [] | [
{
"reaction": "๐",
"users": [
"osanseviero",
"CulturedMan",
"DmitryRyumin",
"samusenps",
"clem",
"Catliba",
"mayank-mishra",
"sugatoray"
],
"count": 8
},
{
"reaction": "โค๏ธ",
"users": [
"huu-ontocord",
"Xa9aX",
"mayank-mis... | 2024-04-02T16:03:20.000Z | 2024-04-02T16:03:33.451Z | [] | /posts/akhaliq/704809400668436 | 2,194 | 0 |
748359845705883 | [
{
"type": "text",
"value": "How would you benchmark performance estimation algorithms vs data drift signals?",
"raw": "How would you benchmark performance estimation algorithms vs data drift signals?",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"labe... | How would you benchmark performance estimation algorithms vs data drift signals?
I'm working on a benchmarking analysis, and I'm currently doing the following:
- Get univariate and multivariate drift signals and measure their correlation with realized performance.
- Use drift signals as features of a regression model to predict the model's performance.
- Use drift signals as features of a classification model to predict a performance drop.
- Compare all the above experiments with results from Performance Estimation algorithms.
Any other ideas? | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1657144463525-629a173153a72d997d3f57d0.jpeg",
"fullname": "Santiago Viquez",
"name": "santiviquez",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 84,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ฅ",
"users": [
"osanseviero"
],
"count": 1
}
] | 2024-04-02T15:04:45.000Z | 2024-04-02T15:04:45.903Z | [] | /posts/santiviquez/748359845705883 | 1,350 | 0 |
700636875300688 | [
{
"type": "text",
"value": "๐ฅOpen CoT Leaderboard",
"raw": "๐ฅOpen CoT Leaderboard",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": null,
... | ๐ฅOpen CoT Leaderboard
We're delighted to announce the [Open CoT Leaderboard](https://huggingface.co/spaces/logikon/open_cot_leaderboard) on ๐ค Spaces.
Unlike other LLM performance leaderboards, the Open CoT Leaderboard is not tracking absolute benchmark accuracies, but relative **accuracy gains** due to **chain-of-thought**.
Eval datasets that underpin the leaderboard are hosted [here](https://huggingface.co/cot-leaderboard).
Feedback and suggestions more than welcome.
@clefourrier | {
"avatarUrl": "/avatars/78be882adf32b808686713e9b457797d.svg",
"fullname": "Gregor Betz",
"name": "ggbetz",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 4,
"isFollowing": false
} | [] | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1644340617257-noauth.png",
"fullname": "Clรฉmentine Fourrier",
"name": "clefourrier",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 459
}
] | [
{
"reaction": "โค๏ธ",
"users": [
"clefourrier",
"lunarflu",
"osanseviero",
"samusenps",
"giux78",
"davidhought0n",
"reuank",
"victor",
"nikgr",
"yakazimir",
"clem"
],
"count": 11
}
] | 2024-04-02T13:32:28.000Z | 2024-04-08T08:50:14.370Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1644340617257-noauth.png",
"fullname": "Clรฉmentine Fourrier",
"name": "clefourrier",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 459,
"isFollowing": false
},
{
"ava... | /posts/ggbetz/700636875300688 | 1,438 | 5 |
985506609005199 | [
{
"type": "text",
"value": "We would like to announce our Aurora-M multilingual models which is based on Starcoderplus.",
"raw": "We would like to announce our Aurora-M multilingual models which is based on Starcoderplus.",
"href": null,
"resource": null,
"url": null,
"code": null,
"... | We would like to announce our Aurora-M multilingual models which is based on Starcoderplus.
Twitter: https://twitter.com/ontocord/status/1772778544051155029
LinkedIn: https://www.linkedin.com/feed/update/urn:li:activity:7178521998845759488/
Blog post: https://huggingface.co/blog/mayank-mishra/aurora
Arxiv: https://huggingface.co/papers/2404.00399
Current LLMs are very susceptible to generating toxic, harmful and even dangerous content. They can also generate outputs with gender or racial biases. The Biden-Harris Executive Order https://www.federalregister.gov/documents/2023/11/01/2023-24283/safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence) sets forth guidelines on what is considered a safe AI system.
Following up on these guidelines, we present the world's first open source Biden-Harris Executive Order Red teamed Multilingual Language Model: Aurora-M. Inspired by BigScience, the model is trained on 5 languages: English, Hindi, Japanese, Vietnamese and Finnish.
* Red teamed model: https://huggingface.co/aurora-m/aurora-m-biden-harris-redteamed(safety tuned according to the order mentioned above)
* Base model: https://huggingface.co/aurora-m/aurora-m-base (not safety tuned)
* Instruct model: https://huggingface.co/aurora-m/aurora-m-instruct (not safety tuned)
@mayank-mishra @cabbage972 @sted97 @Xa9aX @Taishi-N324 @Muennighoff @vumichien @prateeky2806 @felfri @spyysalo and many many others! | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/5fc6879e1c5ee87b1164876d/Tjnm_lv0Bq0gPbFOTDH6E.jpeg",
"fullname": "Huu Nguyen",
"name": "huu-ontocord",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 42,
"isFollowing": false
} | [] | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/627bda4f1b02d7b5bf63f1f8/sqK5uRemxbv-zBQ9B8ugE.jpeg",
"fullname": "VM",
"name": "cabbage972",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2
},
{
"avatarUrl": "https://... | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"Xa9aX",
"lunarflu",
"osanseviero",
"mayank-mishra",
"vumichien",
"kenhktsui",
"nikgr",
"sted97",
"muhtasham",
"Taishi-N324",
"Gokumase"
],
"count": 12
},
{
"reaction": "๐ฅ",
... | 2024-04-02T13:25:39.000Z | 2024-04-02T13:25:39.370Z | [] | /posts/huu-ontocord/985506609005199 | 1,615 | 0 |
730068367902681 | [
{
"type": "text",
"value": "โก AutoQuant",
"raw": "โก AutoQuant",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": null,
"resource": null,
... | โก AutoQuant
AutoQuant is the evolution of my previous AutoGGUF notebook (https://colab.research.google.com/drive/1P646NEg33BZy4BfLDNpTz0V0lwIU3CHu). It allows you to quantize your models in five different formats:
- GGUF: perfect for inference on CPUs (and LM Studio)
- GPTQ/EXL2: fast inference on GPUs
- AWQ: super fast inference on GPUs with vLLM (https://github.com/vllm-project/vllm)
- HQQ: extreme quantization with decent 2-bit and 3-bit models
Once the model is converted, it automatically uploads it on the Hugging Face Hub. To quantize a 7B model, GGUF only needs a T4 GPU, while the other methods require an A100 GPU.
Here's an example of a model I quantized using HQQ and AutoQuant: https://huggingface.co/mlabonne/AlphaMonarch-7B-2bit-HQQ
I hope you'll enjoy it and quantize lots of models! :)
๐ป AutoQuant: https://colab.research.google.com/drive/1b6nqC7UZVt8bx4MksX7s656GXPM-eWw4 | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/61b8e2ba285851687028d395/JtUGAwVh_4cDEsjNcfpye.png",
"fullname": "Maxime Labonne",
"name": "mlabonne",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 3486,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"marcsun13",
"mishig",
"Outrun32",
"johko",
"kgourgou",
"DataSoul",
"samusenps",
"lunarflu",
"osanseviero",
"julien-c",
"OxxoCodes",
"musfiqdehan",
"seyf1elislam",
"nasiruddin15",
"Heng66... | 2024-04-02T09:46:56.000Z | 2024-04-26T08:25:12.877Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/5dd96eb166059660ed1ee413/NQtzmrDdbG0H8qkZvRyGk.jpeg",
"fullname": "Julien Chaumond",
"name": "julien-c",
"type": "user",
"isPro": true,
"isHf": true,
"isMod": false,
"followerCount": 1580,
"isFollowing": fal... | /posts/mlabonne/730068367902681 | 9,234 | 15 |
588552545989925 | [
{
"type": "text",
"value": "Check out our work Symbol-LLM ! We have open-sourced both 7B / 13B model weights, as well as part of the symbolic collections ! Try it !",
"raw": "Check out our work Symbol-LLM ! We have open-sourced both 7B / 13B model weights, as well as part of the symbolic collections ! T... | Check out our work Symbol-LLM ! We have open-sourced both 7B / 13B model weights, as well as part of the symbolic collections ! Try it !
Paper link: https://huggingface.co/papers/2311.09278
Model weights: https://huggingface.co/Symbol-LLM/Symbol-LLM-7B-Instruct | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/656d73ed0bbc114fe6449704/gpteBU9GmKSHRVkRBUHld.png",
"fullname": "Symbol-LLM",
"name": "Symbol-LLM",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 29,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"Symbol-LLM",
"xufangzhi",
"FeYuan",
"vladbogo",
"lunarflu",
"osanseviero",
"nikgr"
],
"count": 7
},
{
"reaction": "๐",
"users": [
"xufangzhi",
"FeYuan",
"QiushiSun",
"lunarflu",
"osan... | 2024-04-02T07:00:36.000Z | 2024-04-02T07:00:36.678Z | [] | /posts/Symbol-LLM/588552545989925 | 1,735 | 0 |
252508216375531 | [
{
"type": "text",
"value": "Introducing Indic Chat!",
"raw": "Introducing Indic Chat!",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": null,
... | Introducing Indic Chat!
Try out best opensource Indic LLMs now on https://www.indic.chat/
Models available:
โข Telugu-LLM-Labs/Indic-gemma-7b-finetuned-sft-Navarasa-2.0
โข GenVRadmin/AryaBhatta-GemmaOrca
โข BhabhaAI/Gajendra-v0.1
โข ai4bharat/Airavata
Additionally:
1. We open up our discord for everyone to collaborate & accelerate Indic LLMs: https://bhabha.ai/discord
2. We release ~600K rows filtered & Hindi translated version of OpenHermes-2.5 instruction dataset: https://huggingface.co/datasets/BhabhaAI/openhermes-2.5-hindi
Also, thanks to our compute sponsor - Telugu LLM Labs & Bhabha AI in helping us serve models for Indic Chat.
If youโd like to be a sponsor too, checkout
https://www.indic.chat/sponsor | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/614efbb6ddd8df0d8bfd0a5a/0oMIv-WwL7sqPEQZv63cu.jpeg",
"fullname": "Satpal Singh Rathore",
"name": "satpalsr",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 10,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/614efbb6ddd8df0d8bfd0a5a/aA2ei7e97aLXvNfjzefri.png"
}
] | [] | [
{
"reaction": "๐ฅ",
"users": [
"monsoon-nlp",
"lunarflu",
"osanseviero",
"alielfilali01",
"ajibawa-2023",
"nikgr",
"ashimksaha",
"Saugatkafley"
],
"count": 8
},
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"lunarflu",
"osan... | 2024-04-02T06:31:39.000Z | 2024-04-02T06:31:39.614Z | [] | /posts/satpalsr/252508216375531 | 1,669 | 0 |
463436477837552 | [
{
"type": "text",
"value": "On evaluating fine tuned 7B Italian open source LLMs I have collected many data points and I created a super simple explorative analyses. My hypothesis based on data are:",
"raw": "On evaluating fine tuned 7B Italian open source LLMs I have collected many data points and I c... | On evaluating fine tuned 7B Italian open source LLMs I have collected many data points and I created a super simple explorative analyses. My hypothesis based on data are:
- mmlu is hard to improve when fine tuning a base model on a different language
- fine tuning also on single GPUs can improve by 5% to 10% the base model on common tasks but a lot more on specific cases with the right training time and data
- fine tuning can specialize well but at cost of loosing some foundational knowledge.
Here the data https://docs.google.com/spreadsheets/d/1MBcxy1loK8eIycZG4DN84Q2ejZ0jSjxUBgoShHDR6IY/edit?usp=sharing
Here the colab https://colab.research.google.com/drive/1ra4_skG5QYWSYOzvagOoIoj4bibQD8Gw?usp=sharing
Here an article with some considerations https://medium.com/@giuxale/an-analyses-on-italian-llms-models-evaluations-51bffe1d44d1
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/5fef4eb7770b06e11c2c6381/1NMdigjCGtn0yvQZSi5NJ.png",
"fullname": "Alessandro Ercolani",
"name": "giux78",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 44,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"Osd111",
"beekay24",
"lunarflu",
"nikgr",
"giux78"
],
"count": 5
}
] | 2024-04-01T22:24:36.000Z | 2024-04-01T22:24:36.481Z | [] | /posts/giux78/463436477837552 | 1,779 | 0 |
738432237563121 | [
{
"type": "text",
"value": "Google DeepMind introduces Gecko a new text embedding! Gecko uses a two-step process that leverages synthetic data generation and reranking.",
"raw": "Google DeepMind introduces Gecko a new text embedding! Gecko uses a two-step process that leverages synthetic data generation... | Google DeepMind introduces Gecko a new text embedding! Gecko uses a two-step process that leverages synthetic data generation and reranking.
Keypoints:
* Uses an LLM to generate diverse synthetic queries and tasks from web passages
* Refines the data by retrieving candidate passages and relabeling positives/negatives using the same LLM
* Achieves very good results on the Massive Text Embedding Benchmark, where compact 256D Gecko outperforms 768D models.
* 768D Gecko achieves state-of-the-art performance competing with models a lot larger larger.
Paper: https://huggingface.co/papers/2403.20327
More details in my blog: https://huggingface.co/blog/vladbogo/gecko
Congrats to the authors for their work! | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/657217faabb25ed8aedd5e48/UUHAXeGtOnQBXFD3nYtf2.jpeg",
"fullname": "Vlad Bogolin",
"name": "vladbogo",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 109,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/657217faabb25ed8aedd5e48/W0n9HjNewR63RiI8IcjFx.png"
}
] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"clem",
"monsoon-nlp",
"beekay24",
"mohammedbriman",
"lunarflu",
"nikgr",
"tatv047",
"super-cinnamon"
],
"count": 9
}
] | 2024-04-01T20:33:03.000Z | 2024-04-01T20:33:03.780Z | [] | /posts/vladbogo/738432237563121 | 2,572 | 0 |
366590839467245 | [
{
"type": "text",
"value": "Jamba",
"raw": "Jamba",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": null,
"resource": null,
"url": nul... | Jamba
A Hybrid Transformer-Mamba Language Model
https://huggingface.co/papers/2403.19887
We present Jamba, a new base large language model based on a novel hybrid Transformer-Mamba mixture-of-experts (MoE) architecture. Specifically, Jamba interleaves blocks of Transformer and Mamba layers, enjoying the benefits of both model families. MoE is added in some of these layers to increase model capacity while keeping active parameter usage manageable. This flexible architecture allows resource- and objective-specific configurations. In the particular configuration we have implemented, we end up with a powerful model that fits in a single 80GB GPU. Built at large scale, Jamba provides high throughput and small memory footprint compared to vanilla Transformers, and at the same time state-of-the-art performance on standard language model benchmarks and long-context evaluations. Remarkably, the model presents strong results for up to 256K tokens context length. We study various architectural decisions, such as how to combine Transformer and Mamba layers, and how to mix experts, and show that some of them are crucial in large scale modeling. We also describe several interesting properties of these architectures which the training and evaluation of Jamba have revealed, and plan to release checkpoints from various ablation runs, to encourage further exploration of this novel architecture. We make the weights of our implementation of Jamba publicly available under a permissive license.
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1674929746905-60f1abe7544c2adfd699860c.jpeg",
"fullname": "AK",
"name": "akhaliq",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5205,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/60f1abe7544c2adfd699860c/1ZvoI687Tw4BTtaua5gtV.png"
}
] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"osanseviero",
"Benyucong",
"clem",
"beekay24",
"mohammedbriman",
"vladbogo",
"adhi01",
"lunarflu",
"nikgr",
"kramp",
"ABX-AI"
],
"count": 12
},
{
"reaction": "๐",
"users"... | 2024-04-01T18:27:08.000Z | 2024-04-01T18:27:08.367Z | [] | /posts/akhaliq/366590839467245 | 2,280 | 0 |
788381176898132 | [
{
"type": "text",
"value": "๐ฏ๐ผ๏ธ๐ New Research Alert - ICLR 2024! ๐ ๐ผ๏ธ๐ฏ",
"raw": "๐ฏ๐ผ๏ธ๐ New Research Alert - ICLR 2024! ๐ ๐ผ๏ธ๐ฏ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
... | ๐ฏ๐ผ๏ธ๐ New Research Alert - ICLR 2024! ๐ ๐ผ๏ธ๐ฏ
๐ Title: Adversarial AutoMixup ๐ผ๏ธ
๐ Description: Adversarial AutoMixup is an approach to image classification augmentation. By alternately optimizing a classifier and a mixed-sample generator, it attempts to generate challenging samples and improve the robustness of the classifier against overfitting.
๐ฅ Authors: Huafeng Qin et al.
๐
Conference: ICLR, May 7-11, 2024 | Vienna, Austria ๐ฆ๐น
๐ Paper: https://huggingface.co/papers/2312.11954
๐ Repository: https://github.com/JinXins/Adversarial-AutoMixup
๐ More Papers: more cutting-edge research presented at other conferences in the https://huggingface.co/spaces/DmitryRyumin/NewEraAI-Papers curated by @DmitryRyumin
๐ Keywords: #AutoMixup #ImageClassification #ImageAugmentation #AdversarialLearning #ICLR2024 #DeepLearning #Innovation | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/PrpxY45EnzC_SYovZANjZ.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/YLwo41WQ9YKN5QSea82PE.png"
},
{
"type": "image... | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377
}
] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"osanseviero",
"clem",
"beekay24",
"lunarflu",
"atrxial",
"MiSTe-R",
"nikgr",
"PantheonAI"
],
"count": 9
},
{
"reaction": "๐ฅ",
"users": [
"DmitryRyumin",
"clem",
"lunarflu... | 2024-04-01T17:52:04.000Z | 2024-04-04T23:40:30.370Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377,
"isFollowing": false
},
{
... | /posts/DmitryRyumin/788381176898132 | 1,414 | 4 |
630974928940204 | [
{
"type": "text",
"value": "Little know gem: the Open-source Cookbook",
"raw": "Little know gem: the Open-source Cookbook",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,... | Little know gem: the Open-source Cookbook
A collection of notebooks for building practical AI applications using open-source tools and models: https://lnkd.in/e6m6Jmwu
Doc: https://lnkd.in/e3FE6TUq
Currently contains 16 notebooks in English (and some in Chinese):
1. Using LLM-as-a-judge ๐งโโ๏ธ for an automated and versatile evaluation
2. Create a legal preference dataset
3. Suggestions for Data Annotation with SetFit in Zero-shot Text Classification
4. Implementing semantic cache to improve a RAG system
5. Building A RAG Ebook โLibrarianโ Using LlamaIndex
6. Stable Diffusion Interpolation
7. Building A RAG System with Gemma, MongoDB and Open Source Models
8. Prompt Tuning with PEFT Library
9. Migrating from OpenAI to Open LLMs Using TGIโs Messages API
10. Automatic Embeddings with TEI through Inference Endpoints
11. Simple RAG for GitHub issues using Hugging Face Zephyr and LangChain
12. Embedding multimodal data for similarity search using ๐ค transformers, ๐ค datasets and FAISS
13. Fine-tuning a Code LLM on Custom Code on a single GPU
14. RAG Evaluation Using Synthetic data and LLM-As-A-Judge
15. Advanced RAG on HuggingFace documentation using LangChain
16. Detecting Issues in a Text Dataset with Cleanlab | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1583857746553-5df7e9e5da6d0311fd3d53f9.jpeg",
"fullname": "Thomas Wolf",
"name": "thomwolf",
"type": "user",
"isPro": true,
"isHf": true,
"isMod": false,
"followerCount": 704,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ฅ",
"users": [
"giux78",
"samusenps",
"osanseviero",
"clem",
"beekay24",
"ajibawa-2023",
"vladbogo",
"peaceAsh",
"lunarflu",
"nikgr",
"vikas"
],
"count": 11
},
{
"reaction": "โค๏ธ",
"users": [
"samusenp... | 2024-04-01T14:06:19.000Z | 2024-04-01T14:06:19.658Z | [] | /posts/thomwolf/630974928940204 | 1,992 | 0 |
402648201818133 | [
{
"type": "text",
"value": "Diaries of Open Source. Part 13 ๐ค",
"raw": "Diaries of Open Source. Part 13 ๐ค",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\... | Diaries of Open Source. Part 13 ๐ค
๐คTwo different bitnet 1.5 open-source replications
Original paper: https://hf.co/papers/2402.17764
1bitllm experiment: https://hf.co/blog/joey00072/experiments-with-bitnet-1-5
NousResearch experiment https://hf.co/NousResearch/OLMo-Bitnet-1B
๐ฅณTiny and large multimodal models great for embeddings
GitHub: https://github.com/unum-cloud/uform
Encoders: https://hf.co/collections/unum-cloud/multimodal-encoders-660553903617c5297eb16838
ONNX weights: https://hf.co/collections/unum-cloud/uform-vl-english-large-onnx-66055a57c182d846f3bc1949
๐ SMPLer-X: Expressive Human Pose and Shape Estimation
Project website: https://caizhongang.com/projects/SMPLer-X/
Demo: https://huggingface.co/spaces/caizhongang/SMPLer-X
Paper: https://hf.co/papers/2309.17448
๐งGeoWizard: 3D Geometry Estimation
Project website: https://fuxiao0719.github.io/projects/geowizard/
Demo: https://hf.co/spaces/lemonaddie/geowizard
Misc models and datasets
- Dolphin-2.8-mistral-7b-v0.2 https://hf.co/cognitivecomputations/dolphin-2.8-mistral-7b-v02
- Hermes-2-Pro-11B, a self-frankenmerge 11B variant https://hf.co/mattshumer/Hermes-2-Pro-11B
- Large conversational dataset based on Usenet data in the Italian language https://hf.co/datasets/mii-community/UsenetArchiveIT-conversations | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"ruggsea",
"mrinaldi",
"DmitryRyumin",
"samusenps",
"giux78",
"clem",
"vladbogo",
"lunarflu",
"nikgr",
"abidlabs"
],
"count": 10
},
{
"reaction": "๐ฅ",
"users": [
"samusenps",
"giux78",... | 2024-04-01T11:02:12.000Z | 2024-04-01T14:04:53.783Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing":... | /posts/osanseviero/402648201818133 | 2,272 | 3 |
581937223513516 | [
{
"type": "text",
"value": "This is the closest Iโve seen of a scalable AI/LLM Operating System - it has all the major ingredients of a feasible AI OS 1 architecture:",
"raw": "This is the closest Iโve seen of a scalable AI/LLM Operating System - it has all the major ingredients of a feasible AI OS 1 ar... | This is the closest Iโve seen of a scalable AI/LLM Operating System - it has all the major ingredients of a feasible AI OS 1 architecture:
- Extends classical OS functionalities with an LLM Kernel.
- Multi agent-centric approach.
- Optimized resource allocation system that allows for LLM-based tasks and Classical OS tasks to coexist.
- An Agent Scheduler that can perform classical os operations (FIFO, RR).
- A Context Manager to improve alignment.
- Lazy Memory Manager for agents (ensures data is stored and accessible only while the agent is active)
- An Enhanced security module for the AI-driven environment.
It does hit all checkpoints, doesnโt it? An upscale version of @karpathyโs.
Code: https://github.com/agiresearch/AIOS | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6438a9027de34e8ea7e4b257/vib8QSd1AWMr_bR9ig_xJ.jpeg",
"fullname": "Jaward Sesay",
"name": "Jaward",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 191,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6438a9027de34e8ea7e4b257/wFxgCDc4qA6KCxtvxErAp.jpeg"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6438a9027de34e8ea7e4b257/1SXcuxpfYIShdA39WyP3a.jpeg"
},
{
"type": "ima... | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1660434061546-62f83661fe21cc4875221c0f.jpeg",
"fullname": "Andrej K",
"name": "karpathy",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 476
}
] | [
{
"reaction": "๐ฅ",
"users": [
"ajibawa-2023",
"osanseviero",
"phuong-d-h-nguyen",
"Dlbk",
"rsamantha",
"KvrParaskevi",
"AIAdventurist",
"JoshuaGu",
"NeuralNovel",
"thomwolf",
"kazeMace",
"disham993",
"kubo-vn",
"abhinit21",... | 2024-03-30T12:17:22.000Z | 2024-03-31T04:13:32.107Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/64aea8ff67511bd3d965697b/Jxn52EmDF5RApJh8antxn.jpeg",
"fullname": "Feynman Innovations",
"name": "ajibawa-2023",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 138,
"isFollow... | /posts/Jaward/581937223513516 | 4,511 | 4 |
424262013167981 | [
{
"type": "text",
"value": "๐๐ Exciting times for the document AI community! ",
"raw": "๐๐ Exciting times for the document AI community! ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line"... | ๐๐ Exciting times for the document AI community!
We're thrilled to announce the release of some of the largest OCR datasets available to the public.
๐ฅ With over 26 million pages , 18 billion text tokens, and 6TB of data, these resources are a significant leap forward for document AI research.
Here's how to access these datasets quickly:
```
from datasets import load_dataset
pdfa_dataset = load_dataset('pixparse/pdfa-eng-wds', streaming=True)
IDL_dataset = load_dataset('pixparse/idl-wds', streaming=True)
```
This enables you to stream them directly, integrating seamlessly with your projects using the Hugging Face datasets library. On the hub, you can find them here:
https://huggingface.co/datasets/pixparse/pdfa-eng-wds
https://huggingface.co/datasets/pixparse/idl-wds
For lean data loading, the new [chug](https://github.com/huggingface/chug) library offers a solution with pdf decoding:
```
import chug
task_cfg = chug.DataTaskDocReadCfg(
page_sampling='all',
)
data_cfg = chug.DataCfg(
source='pixparse/pdfa-eng-wds',
split='train',
batch_size=None,
format='hfids',
num_workers=0,
)
data_loader = chug.create_loader(
data_cfg,
task_cfg,
)
sample = next(iter(data_loader))
```
We owe a huge thank you to Peter Wyatt, Kate Tasker, Rachel Taketa, Ali Furkan Biten, Ruben Tito, and their colleagues for their contributions. Their work putting these datasets together has been invaluable. ๐ค
Looking Ahead:
We're on a mission to enhance document AI capabilities, and these datasets are just the beginning. With your engagement and innovation, we're confident in the community's ability to develop robust OCR solutions. We encourage you to explore these datasets, experiment with the code, and contribute to the collective progress in document AI.
For detailed information on usage and licensing, please refer to the dataset cards on the Hugging Face hub.
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/64789feb79f2d49511ed7db4/IzaIwiVgnkTZHrcLDQk0C.jpeg",
"fullname": "Pablo Montalvo",
"name": "Molbap",
"type": "user",
"isPro": true,
"isHf": true,
"isMod": false,
"followerCount": 97,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ฅ",
"users": [
"abidlabs",
"VictorSanh",
"lhoestq",
"Dlbk",
"thomwolf",
"manishiitg",
"clefourrier",
"IlyasMoutawwakil",
"osanseviero",
"samusenps",
"DmitryRyumin",
"ajibawa-2023",
"Pclanglais",
"loubnabnl",
... | 2024-03-29T22:12:58.000Z | 2024-04-02T18:55:27.055Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1667002643224-604a5184dca2c7ac7508b849.jpeg",
"fullname": "Ross Wightman",
"name": "rwightman",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 221,
"isFollowing": false
},
... | /posts/Molbap/424262013167981 | 5,000 | 4 |
712385114238442 | [
{
"type": "text",
"value": "Diaries of Open Source. Part 12 ๐ค",
"raw": "Diaries of Open Source. Part 12 ๐ค",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\... | Diaries of Open Source. Part 12 ๐ค
๐Alibaba releases Qwen1.5-MoE-A2.7B, an interesting MoE with 2.7B activated parameters and 64 experts
Blog https://qwenlm.github.io/blog/qwen-moe/
Demo: https://hf.co/spaces/Qwen/qwen1.5-MoE-A2.7B-Chat-demo
Models: https://hf.co/Qwen
GitHub: https://github.com/QwenLM/Qwen1.5
๐ตVoiceCraft, SOTA speech editing and text to speech
GitHub: https://github.com/jasonppy/VoiceCraft
Model: https://huggingface.co/pyp1/VoiceCraft
๐ AI21Labs release Jamba, an SSM-Transformer, pretrained MoE which allows a large context window (256K) and high throughput
Blog https://www.ai21.com/blog/announcing-jamba
Model https://huggingface.co/ai21labs/Jamba-v0.1
โจ Berkeley releases Starling-LM-7B, an RLHF-ed model, and -RM-34B, a Yi-based reward model very good for its size
Starling Beta: https://hf.co/Nexusflow/Starling-LM-7B-beta
Starling RM: https://hf.co/Nexusflow/Starling-RM-34B
๐ฅ๏ธStability releases Stable Code Instruct 3B, an instruct model for code generation
Blog: https://stability.ai/news/introducing-stable-code-instruct-3b
Demo: https://hf.co/spaces/stabilityai/stable-code-instruct-3b
Report: https://stability.ai/s/Stable_Code_TechReport_release.pdf
๐Common Corpus: the largest public domain dataset for training LLMs
Blog: https://hf.co/blog/Pclanglais/common-corpus
Dataset: https://hf.co/collections/PleIAs/common-corpus-65d46e3ea3980fdcd66a5613
Misc:
โกGaLore: a very memory-efficient technique that allows pretraining models in consumer GPUs https://hf.co/blog/galore
Moirai
๐Moirai, foundation models for time series forecasting https://hf.co/collections/Salesforce/moirai-10-r-models-65c8d3a94c51428c300e0742
๐ฅ Mistral-ORPO-Capybara-7K, a high-quality Mistral fine-tune using ORPO, a new alignment technique https://hf.co/kaist-ai/mistral-orpo-capybara-7k
๐คฏAPISR, an anime super-resolution upscaling model https://hf.co/spaces/HikariDawn/APISR | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"Dlbk",
"manishiitg",
"Joseph717171",
"DmitryRyumin",
"ajibawa-2023",
"HikariDawn",
"ElenaRyumina",
"Chunte",
"gabrielmbmb",
"mmhamdy",
"callmeraxz",
"clem",
"lunarflu",
"nik... | 2024-03-29T21:24:04.000Z | 2024-03-31T09:22:45.335Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing":... | /posts/osanseviero/712385114238442 | 3,514 | 4 |
857297626421239 | [
{
"type": "text",
"value": "Happy to share Living Images and the demo video of the product outpainting model behind it ๐",
"raw": "Happy to share Living Images and the demo video of the product outpainting model behind it ๐",
"href": null,
"resource": null,
"url": null,
"code": null,
... | Happy to share Living Images and the demo video of the product outpainting model behind it ๐
Send your generation requests in the thread ๐งตor use https://img.coframe.ai | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1626345802673-60ec109280b434965987d1de.jpeg",
"fullname": "Aleksey Korshuk",
"name": "AlekseyKorshuk",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 153,
"isFollowing": false
} | [
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/60ec109280b434965987d1de/zeDFWoDjxjYIMZQOWzF65.mp4"
}
] | [] | [
{
"reaction": "๐",
"users": [
"AlekseyKorshuk",
"Glavin001",
"tunahfish",
"onum",
"TamasZumpf",
"clem",
"lunarflu",
"nikgr",
"qyxs"
],
"count": 9
},
{
"reaction": "๐ค",
"users": [
"shiv2050",
"AlekseyKorshuk",
"luna... | 2024-03-29T19:53:20.000Z | 2024-03-29T19:53:20.856Z | [] | /posts/AlekseyKorshuk/857297626421239 | 3,427 | 0 |
174172922896344 | [
{
"type": "text",
"value": "Fun fact about evaluation!",
"raw": "Fun fact about evaluation!",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": ... | Fun fact about evaluation!
Did you know that, if you evaluate the same model, with the same prompt formatting & the same fixed few-shot examples, only changing
โป๏ธthe order in which the few shot examples are added to the prompt โป๏ธ
you get a difference of up to 3 points in evaluation score?
I did a small experiment using some MMLU subsets on the best performing 7B and lower pretrained models from the leaderboard.
I tried 8 different prompting methods (containing more or less information, such as just the question, or Question: question, or Question: question Choices: ..., see the x axis) that are commonly used in evaluation.
I then compared the results for all these methods, in 5-shot, during 2 runs. The *only difference* between the first and second run being that the samples used in few-shot are not introduced in the same order.
For example, run one would have been "A B C D E Current sample", vs, in run 2, "D C E A B Current sample".
All the other experiment parameters stayed exactly the same.
As you can see on the attached picture, you get a difference of up to 3 points between the 2 few-shot samples shuffling.
So, when just changing *the order of the few shot samples* can change your results by several points, what is the impact of all other "minimal" and unreported prompting changes?
-> Any kind of model score, provided without an evaluation script for reproducibility, is basically bullshit (or coms).
-> This is why we need reproducible evaluation in a fair and exactly similar setup, using evaluation suites such as `lm_eval` from the Harness, `lighteval` from HF, or the Open LLM Leaderboard. | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1644340617257-noauth.png",
"fullname": "Clรฉmentine Fourrier",
"name": "clefourrier",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 459,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6202a599216215a22221dea9/tweU_jkcnWjnF3GProUia.png"
}
] | [] | [
{
"reaction": "๐",
"users": [
"shiv2050",
"ssml2050",
"InferenceIllusionist",
"mmhamdy",
"ricfergas",
"giux78",
"clem",
"lunarflu",
"nikgr"
],
"count": 9
},
{
"reaction": "๐ฅ",
"users": [
"osanseviero",
"clem",
"lun... | 2024-03-29T19:11:35.000Z | 2024-04-02T14:44:47.873Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/63e80664e02ee67e8e570ec4/rGfRhywmjd_lbqfYzOEdd.png",
"fullname": "EsKa",
"name": "SerialKicked",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 12,
"isFollowing": false
},
... | /posts/clefourrier/174172922896344 | 2,350 | 4 |
175193132662887 | [
{
"type": "text",
"value": "Teraflop AI is excited to help support the Caselaw Access Project and Harvard Library Innovation Lab, in the release of over 6.6 million state and federal court decisions published throughout U.S. history. It is important to democratize fair access to data to the public, legal co... | Teraflop AI is excited to help support the Caselaw Access Project and Harvard Library Innovation Lab, in the release of over 6.6 million state and federal court decisions published throughout U.S. history. It is important to democratize fair access to data to the public, legal community, and researchers. This is a processed and cleaned version of the original CAP data.
During the digitization of these texts, there were erroneous OCR errors that occurred. We worked to post-process each of the texts for model training to fix encoding, normalization, repetition, redundancy, parsing, and formatting.
Teraflop AIโs data engine allows for the massively parallel processing of web-scale datasets into cleaned text form.
Link to the processed dataset: https://huggingface.co/datasets/TeraflopAI/Caselaw_Access_Project
The Caselaw Access Project dataset is licensed under the CC0 License.
We plan to release trillions of commercially licensed text tokens, images, audio, videos, and other datasets spanning numerous domains and modalities over the next months. If you are interested in contributing commercially licensed data be sure to reach out: https://twitter.com/EnricoShippole
Follow us for the next collaborative dataset releases: https://twitter.com/TeraflopAI | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6276ba3c2d26ac639e5a2b01/k7LHkSbNjPR31ma4EereF.png",
"fullname": "Enrico Shippole",
"name": "conceptofmind",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 128,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6276ba3c2d26ac639e5a2b01/AHQkbn1YQnTu926M552z9.png"
}
] | [] | [
{
"reaction": "๐ฅ",
"users": [
"osanseviero",
"samusenps",
"rsamantha",
"Pclanglais",
"muhtasham",
"ssml2050",
"clem",
"nikgr"
],
"count": 8
},
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"clem"
],
"count": 2
},
{
... | 2024-03-29T16:44:42.000Z | 2024-03-29T16:44:42.733Z | [] | /posts/conceptofmind/175193132662887 | 2,501 | 0 |
819003891866897 | [
{
"type": "text",
"value": "New - add your bluesky account to your HF profile:",
"raw": "New - add your bluesky account to your HF profile:",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
... | New - add your bluesky account to your HF profile:
https://huggingface.co/settings/profile
Is the grass greener, the sky bluer? Will try and figure it out at https://bsky.app/profile/jeffboudier.bsky.social
By the way, HF people starter pack https://bsky.app/starter-pack/huggingface.bsky.social/3laz5x7naiz22 | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1605114051380-noauth.jpeg",
"fullname": "Jeff Boudier",
"name": "jeffboudier",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 195,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/5fac18fb5eec0323e9470ba2/XOdGKkHok6Ebxci1wO0Ml.png"
}
] | [] | [
{
"reaction": "๐ค",
"users": [
"rwightman",
"John6666",
"vinhnx90"
],
"count": 3
}
] | 2024-11-22T20:17:38.000Z | 2024-11-22T20:17:38.713Z | [] | /posts/jeffboudier/819003891866897 | 688 | 0 |
483522010371958 | [
{
"type": "text",
"value": "Weekend Dribble ๐ฆ๐บ",
"raw": "Weekend Dribble ๐ฆ๐บ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": null,
"r... | Weekend Dribble ๐ฆ๐บ
Adapters for Product Ad Backdrops, Smooth Polaroids, Minimalist Sketch cards, Super Blends!!
๐คDemo on: https://huggingface.co/spaces/prithivMLmods/FLUX-LoRA-DLC
Stranger Zones :
๐๐ผ{ Super Blend } : https://huggingface.co/strangerzonehf/Flux-Super-Blend-LoRA
๐๐ผ{ Product Concept Ad } : https://huggingface.co/prithivMLmods/Flux-Product-Ad-Backdrop
๐๐ผ{ Frosted Mock-ups } : https://huggingface.co/prithivMLmods/Flux.1-Dev-Frosted-Container-LoRA
๐๐ผ{ Polaroid Plus } : https://huggingface.co/prithivMLmods/Flux-Polaroid-Plus
๐๐ผ{Sketch Cards} : https://huggingface.co/prithivMLmods/Flux.1-Dev-Sketch-Card-LoRA
๐Stranger Zone: https://huggingface.co/strangerzonehf
๐Flux LoRA Collections: https://huggingface.co/collections/prithivMLmods/flux-lora-collections-66dd5908be2206cfaa8519be
.
.
.
@prithivMLmods ๐ค | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/65bb837dbfb878f46c77de4c/UVtVbF_3rdt0DC8xTkpL1.jpeg",
"fullname": "Prithiv Sakthi",
"name": "prithivMLmods",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 393,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/bJU0-GPle-FgfsAtNGEbx.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/uF8WYBLOpwtPQnMZC-XNI.png"
},
{
"type": "image... | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/65bb837dbfb878f46c77de4c/UVtVbF_3rdt0DC8xTkpL1.jpeg",
"fullname": "Prithiv Sakthi",
"name": "prithivMLmods",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 393
}
] | [
{
"reaction": "๐ฅ",
"users": [
"Ngrthm",
"darksfx",
"John6666",
"rdrede",
"ai4life44",
"victor",
"loubnabnl",
"nzhenev",
"linoyts"
],
"count": 9
},
{
"reaction": "๐ค",
"users": [
"RenderIo",
"Ngrthm",
"darksfx",
... | 2024-11-22T18:09:16.000Z | 2024-11-22T18:15:45.165Z | [] | /posts/prithivMLmods/483522010371958 | 1,440 | 0 |
308810445844150 | [
{
"type": "text",
"value": "Can language models replace developers? #RepoCod says โNot Yetโ, because GPT-4o and other LLMs have <30% accuracy/pass@1 on real-world code generation tasks. ",
"raw": "Can language models replace developers? #RepoCod says โNot Yetโ, because GPT-4o and other LLMs have <30% ac... | Can language models replace developers? #RepoCod says โNot Yetโ, because GPT-4o and other LLMs have <30% accuracy/pass@1 on real-world code generation tasks.
- Leaderboard https://lt-asset.github.io/REPOCOD/
- Dataset: https://huggingface.co/datasets/lt-asset/REPOCOD
@jiang719 @shanchao @Yiran-Hu1007
Compared to #SWEBench, RepoCod tasks are
- General code generation tasks, while SWE-Bench tasks resolve pull requests from GitHub issues.
- With 2.6X more tests per task (313.5 compared to SWE-Benchโs 120.8).
Compared to #HumanEval, #MBPP, #CoderEval, and #ClassEval, RepoCod has 980 instances from 11 Python projects, with
- Whole function generation
- Repository-level context
- Validation with test cases, and
- Real-world complex tasks: longest average canonical solution length (331.6 tokens) and the highest average cyclomatic complexity (9.00)
Introducing hashtag #RepoCod-Lite ๐ for faster evaluations: 200 of the toughest tasks from RepoCod with:
- 67 repository-level, 67 file-level, and 66 self-contains tasks
- Detailed problem descriptions (967 tokens) and long canonical solutions (918 tokens)
- GPT-4o and other LLMs have < 10% accuracy/pass@1 on RepoCod-Lite tasks.
- Dataset: https://huggingface.co/datasets/lt-asset/REPOCOD_Lite
#LLM4code #LLM #CodeGeneration #Security | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/65831342f9c5cda913df366a/h7MLYt--shRYQj4-q5XmR.jpeg",
"fullname": "Lin Tan",
"name": "lin-tan",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 5,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/65831342f9c5cda913df366a/YGqQYEKVWnIzuUWX4WorD.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/65831342f9c5cda913df366a/as5x1o-983nkjp2ke9MiL.png"
},
{
"type": "image... | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/629e4ca2f2bda18349b6d330/gSnGTLm2ugpINECylbuuQ.jpeg",
"fullname": "Nan Jiang",
"name": "jiang719",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 4
},
{
"avatarUrl": "/av... | [
{
"reaction": "๐ฅ",
"users": [
"John6666",
"AdinaY",
"clem",
"lin-tan",
"shanchao",
"bhadresh-savani",
"iky1e"
],
"count": 7
},
{
"reaction": "๐ค",
"users": [
"prithivMLmods",
"lin-tan",
"shanchao",
"mvelic"
],
"co... | 2024-11-22T17:14:55.000Z | 2024-11-23T02:39:03.479Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/649a54b896d5747b35e2163b/tdZmsov6fN1VHztaE5kX9.jpeg",
"fullname": "Vezora",
"name": "Vezora",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 116,
"isFollowing": false
}
] | /posts/lin-tan/308810445844150 | 1,077 | 1 |
475162100629292 | [
{
"type": "text",
"value": "401 Client Error: Unauthorized for url: ",
"raw": "401 Client Error: Unauthorized for url: ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "resource",
"value": null,
... | 401 Client Error: Unauthorized for url: https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/vae/config.json
The above exception was the direct cause of the following exception:
huggingface_hub.errors.GatedRepoError: 401 Client Error. (Request ID: Root=1-6740a1d6-26b6f3b44563a26a49aea19d;fa54ac02-6068-44e1-b499-f793dd20335c)
Cannot access gated repo for url https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/vae/config.json.
Access to model black-forest-labs/FLUX.1-dev is restricted. You must have access to it and be authenticated to access it. Please log in.
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/home/user/app/app.py", line 32, in <module>
good_vae = AutoencoderKL.from_pretrained(base_model, subfolder="vae", torch_dtype=dtype).to(device) | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/66d45453cc638c79e6447240/CqV6FDv3Wv6ORUBuM02R6.png",
"fullname": "John",
"name": "Keltezaa",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 18,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/66d45453cc638c79e6447240/ty9lwSVT7jvMLq28XsLcg.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/66d45453cc638c79e6447240/CWl-lTx0n8FbqxdgULkmp.png"
}
] | [] | [
{
"reaction": "๐",
"users": [
"John6666"
],
"count": 1
}
] | 2024-11-22T15:35:26.000Z | 2024-11-22T17:25:10.247Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6740ab9e864cf5f619a70ee4/kLxfZKMB08I0F4XSD47MX.jpeg",
"fullname": "Miguel Saraiva",
"name": "n1k0p0l",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": null,
"isFollowing": fal... | /posts/Keltezaa/475162100629292 | 333 | 2 |
557320861915536 | [
{
"type": "text",
"value": "Marco-o1๐ฅ an open Reasoning Models by AIDC team",
"raw": "Marco-o1๐ฅ an open Reasoning Models by AIDC team",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
... | Marco-o1๐ฅ an open Reasoning Models by AIDC team
Model: https://huggingface.co/AIDC-AI/Marco-o1
Paper: https://huggingface.co/papers/2411.14405
โจFine-tuned with CoT data (open-source + synthetic).
โจExpands solution space with MCTS, guided by model confidence.
โจNovel reasoning strategies & self-reflection enhance complex problem-solving.
โจPioneers LRM in multilingual machine translation.
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/63a369d98c0c89dcae3b8329/6OUJ7Hc9T1jXynYH3FGaf.png",
"fullname": "Adina Yakefu",
"name": "AdinaY",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 240,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"enzostvs",
"John6666",
"ai-everyday",
"AtAndDev"
],
"count": 4
},
{
"reaction": "๐ฅ",
"users": [
"prithivMLmods",
"AtAndDev"
],
"count": 2
}
] | 2024-11-22T15:24:14.000Z | 2024-11-22T16:22:03.111Z | [] | /posts/AdinaY/557320861915536 | 787 | 0 |
425637070931186 | [
{
"type": "text",
"value": "๐๐๐Introducing Insight-V! An early attempt towards o1-like multi-modal reasoning. ",
"raw": "๐๐๐Introducing Insight-V! An early attempt towards o1-like multi-modal reasoning. ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,... | ๐๐๐Introducing Insight-V! An early attempt towards o1-like multi-modal reasoning.
We offer a structured long-chain visual reasoning data generation pipeline and a multi-agent system to unleash the reasoning potential of MLLMs.
๐ Paper: https://arxiv.org/abs/2411.14432
๐ ๏ธ Github: https://github.com/dongyh20/Insight-V
๐ผ Model Weight: https://huggingface.co/collections/THUdyh/insight-v-673f5e1dd8ab5f2d8d332035 | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/652965773a416e1f2173443b/y9MB8YgHzbwCXAc4EI9T3.jpeg",
"fullname": "Yuhao Dong",
"name": "THUdyh",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 24,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ฅ",
"users": [
"AdinaY",
"liuziwei7",
"Mackya",
"Mondayfelix"
],
"count": 4
},
{
"reaction": "๐",
"users": [
"John6666",
"AdinaY",
"Mackya"
],
"count": 3
},
{
"reaction": "๐",
"users": [
"Mackya"
],... | 2024-11-22T14:51:23.000Z | 2024-11-22T14:51:23.815Z | [] | /posts/THUdyh/425637070931186 | 781 | 0 |
519030878351508 | [
{
"type": "text",
"value": "Made a new app to visualize the LLM race โ ๐ก๐ผ ๐๐๐ฟ๐ผ๐ฝ๐ฒ๐ฎ๐ป ๐ฐ๐ผ๐บ๐ฝ๐ฎ๐ป๐ ๐ถ๐ป ๐๐ต๐ฒ ๐๐ผ๐ฝ ๐ญ๐ฌ ๐ช๐บโ",
"raw": "Made a new app to visualize the LLM race โ ๐ก๐ผ ๐๐๐ฟ๐ผ๐ฝ๐ฒ๐ฎ๐ป ๐ฐ๐ผ๐บ๐ฝ๐ฎ๐ป๐ ๐ถ๐ป ๐๐ต๐ฒ ๐๐ผ๐ฝ ๐ญ๐ฌ ๐ช๐บโ",
"href": null,
"resource": null,
... | Made a new app to visualize the LLM race โ ๐ก๐ผ ๐๐๐ฟ๐ผ๐ฝ๐ฒ๐ฎ๐ป ๐ฐ๐ผ๐บ๐ฝ๐ฎ๐ป๐ ๐ถ๐ป ๐๐ต๐ฒ ๐๐ผ๐ฝ ๐ญ๐ฌ ๐ช๐บโ
See the app here ๐ https://huggingface.co/spaces/m-ric/llm-race-to-the-top
I've adapted an app by @andrewrreed that tracks progress of LLMs (https://huggingface.co/spaces/andrewrreed/closed-vs-open-arena-elo), on the Chatbot Arena leaderboard, to compare companies from different countries.
The outcome is quite sad, as a Frenchman and European.
The top 10 is exclusively US ๐บ๐ธ and Chinese ๐จ๐ณ companies (after great Chinese LLM releases recently, like the Qwen2.5 series), with the notable exception of Mistral AI ๐ซ๐ท.
American companies are making fast progress, Chinese ones even faster. Europe is at risk of being left behind. And the EU AI Act hasn't even come into force yet to slow down the EU market. We need to wake up ๐ฌ
โ ๏ธ Caution: This Chatbot Arena ELO ranking is not the most accurate, especially at high scores like this, because LLM makers can game it to some extent. | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/63d10d4e8eaa4831005e92b5/7p7-OmWM6PqqCs7ZStPGD.jpeg",
"fullname": "Aymeric Roucher",
"name": "m-ric",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 494,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/63d10d4e8eaa4831005e92b5/S6JeMca06C9mnh8DYpkEV.png"
}
] | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/61d375fd733d3a83ecd1bba9/oIXwvvs1-HaCnJXMCZgkc.jpeg",
"fullname": "Andrew Reed",
"name": "andrewrreed",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 106
}
] | [
{
"reaction": "โค๏ธ",
"users": [
"andrewrreed",
"LiteSoulAI",
"John6666",
"clem",
"AdinaY",
"k-young",
"okedialf",
"AtAndDev",
"Tanvir1337"
],
"count": 9
}
] | 2024-11-22T13:13:31.000Z | 2024-11-22T13:18:31.612Z | [] | /posts/m-ric/519030878351508 | 932 | 0 |
247264884415285 | [
{
"type": "text",
"value": "What a week! A recap for everything you missed โ๏ธ",
"raw": "What a week! A recap for everything you missed โ๏ธ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
... | What a week! A recap for everything you missed โ๏ธ
https://huggingface.co/collections/merve/nov-22-releases-673fbbcfc1c97c4f411def07
Multimodal โจ
> Mistral AI
released Pixtral 124B, a gigantic open vision language model
> Llava-CoT (formerly known as Llava-o1) was released, a multimodal reproduction of o1 model by PKU
> OpenGVLab released MMPR: a new multimodal reasoning dataset
> Jina has released Jina-CLIP-v2 0.98B multilingual multimodal embeddings
> Apple released new SotA vision encoders AIMv2
LLMs ๐ฆ
> AllenAI dropped a huge release of models, datasets and scripts for Tรผlu, a family of models based on Llama 3.1 aligned with SFT, DPO and a new technique they have developed called RLVR
> Jina has released embeddings-v3: new multilingual embeddings with longer context
> Hugging Face released SmolTalk: synthetic dataset used to align SmolLM2 using supervised fine-tuning
> Microsoft released orca-agentinstruct-1M-v1: a gigantic instruction dataset of 1M synthetic instruction pairs
Image Generation ๐ผ๏ธ
> Black Forest Labs released Flux 1. tools: four new models for different image modifications and two LoRAs to do image conditioning and better steer generations
Lastly Hugging Face released a new library Observers: a lightweight SDK for monitoring interactions with AI APIs and easily store and browse them ๐
$ pip install observers
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1648113222875-6141a88b3a0ec78603c9e784.png",
"fullname": "Merve Noyan",
"name": "merve",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5589,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ฅ",
"users": [
"Steveeeeeeen",
"victor",
"loubnabnl",
"John6666",
"enzostvs",
"clem",
"AdinaY",
"veeraleto",
"LeonceNsh",
"rwightman",
"FranckAbgrall",
"SherlockRamos",
"Mouradology"
],
"count": 13
},
{... | 2024-11-22T12:31:25.000Z | 2024-11-24T12:42:10.966Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/M9Ee6_TUrgsgmZ4IccQok.jpeg",
"fullname": "prinz tim",
"name": "myprime",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": null,
"isFollowing": false
}
] | /posts/merve/247264884415285 | 1,950 | 2 |
479387783609004 | [
{
"type": "text",
"value": "Excited to share my analysis of the most groundbreaking DCN-V2 paper from ",
"raw": "Excited to share my analysis of the most groundbreaking DCN-V2 paper from ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
... | Excited to share my analysis of the most groundbreaking DCN-V2 paper from @Google, which introduces significant improvements to deep learning recommendation systems!
Key technical highlights:
>> Core Architecture
- Starts with an embedding layer that handles both sparse categorical and dense features
- Unique capability to handle variable embedding sizes from small to large vocabulary sizes
- Cross network creates explicit bounded-degree feature interactions
- Deep network complements with implicit feature interactions
- Two combination modes: stacked and parallel architectures
>> Key Technical Innovations
- Enhanced cross layers with full matrix-based feature interaction learning instead of vector-based
- Mixture of Low-Rank architecture with:
* Multiple expert networks learning in different subspaces
* Dynamic gating mechanism to adaptively combine experts
* Efficient time complexity when specific conditions are met
* Support for non-linear transformations in projected spaces
>> Production Optimizations
- Low-rank matrix approximation leveraging singular value decay patterns
- Mixture-of-Experts decomposition into smaller subspaces
- Efficient parameter allocation between cross and deep networks
- Automatic feature interaction learning for higher-order interactions in multi-layered networks
- Support for both homogeneous and heterogeneous polynomial patterns
>> Real-World Impact
- Successfully deployed across Google's recommendation systems
- Significant gains in both offline accuracy and online metrics
- Better performance-latency tradeoffs through low-rank approximations
- Proven effectiveness on large-scale data with billions of training examples
This represents a major leap forward in making deep learning recommendation systems more practical and efficient at scale.
Thoughts? Would love to hear your experiences implementing similar architectures in production! | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/662bf5bfe93bb73804ef9344/WXYLnjjJ4SROkoveIi7If.png",
"fullname": "Kuldeep Singh Sidhu",
"name": "singhsidhukuldeep",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 219,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/662bf5bfe93bb73804ef9344/_eTkYBdP52SmSKwr2SEZY.jpeg"
}
] | [] | [
{
"reaction": "๐",
"users": [
"John6666",
"clem"
],
"count": 2
},
{
"reaction": "โค๏ธ",
"users": [
"clem"
],
"count": 1
}
] | 2024-11-22T11:01:09.000Z | 2024-11-22T11:01:09.090Z | [] | /posts/singhsidhukuldeep/479387783609004 | 691 | 0 |
765598269884749 | [
{
"type": "text",
"value": "Apple released AIMv2 ๐ a family of state-of-the-art open-set vision encoders",
"raw": "Apple released AIMv2 ๐ a family of state-of-the-art open-set vision encoders",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": nu... | Apple released AIMv2 ๐ a family of state-of-the-art open-set vision encoders
https://huggingface.co/collections/apple/aimv2-6720fe1558d94c7805f7688c
> like CLIP, but add a decoder and train on autoregression ๐คฏ
> 19 open models come in 300M, 600M, 1.2B, 2.7B with resolutions of 224, 336, 448
> Load and use with ๐ค transformers | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1648113222875-6141a88b3a0ec78603c9e784.png",
"fullname": "Merve Noyan",
"name": "merve",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5589,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6141a88b3a0ec78603c9e784/6CesZUCcICywT2AEF9r1s.png"
}
] | [] | [
{
"reaction": "๐",
"users": [
"RedSparkie",
"crun885",
"John6666",
"clem"
],
"count": 4
},
{
"reaction": "โค๏ธ",
"users": [
"esab",
"ivanfioravanti",
"atasoglu",
"wchai"
],
"count": 4
},
{
"reaction": "๐",
"users": [
... | 2024-11-22T10:07:29.000Z | 2024-11-22T10:07:29.993Z | [] | /posts/merve/765598269884749 | 1,136 | 0 |
220418838710285 | [
{
"type": "text",
"value": "Import any dataset from the Hub and configure your labeling tasks without needing any code!",
"raw": "Import any dataset from the Hub and configure your labeling tasks without needing any code!",
"href": null,
"resource": null,
"url": null,
"code": null,
"... | Import any dataset from the Hub and configure your labeling tasks without needing any code!
Really excited about extending the Hugging Face Hub integration with many more streamlined features and workflows, and we would love to hear your feedback and ideas, so don't feel shy and reach out ๐ซถ๐ฝ
https://huggingface.co/blog/argilla-ui-hub
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1677141720071-634ff41ff32062e9eb7b06a3.jpeg",
"fullname": "David Berenstein",
"name": "davidberenstein1957",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 167,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"John6666",
"dvilasuero",
"NickyNicky"
],
"count": 3
}
] | 2024-11-04T16:15:53.000Z | 2024-11-04T16:15:53.521Z | [] | /posts/davidberenstein1957/220418838710285 | 2,082 | 0 |
376713378079900 | [
{
"type": "text",
"value": "Discovered an outrageous bug on the ChatGPT official website, especially for those using ad-blocking plugins. Please make sure to add ",
"raw": "Discovered an outrageous bug on the ChatGPT official website, especially for those using ad-blocking plugins. Please make sure to a... | Discovered an outrageous bug on the ChatGPT official website, especially for those using ad-blocking plugins. Please make sure to add `browser-intake-datadoghq.com` to your ad block whitelist. The ChatGPT webpage relies on this site for heartbeat detection, but since it belongs to an ad tracking network, it's included in major ad-blocking lists. (If you're using Clash, also remember to add it to the whitelist.) Failing to do so may cause the ChatGPT web interface to display a greyed-out send button after clicking, with no response.
For users with Chinese IP addresses, consider adding this URL to the rules of your U.S. node, as the response headers from this site will report the user's physical location to GPT. | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/64bce15bafd1e46c5504ad38/bQFX1iFbXEBXcQvUNL811.png",
"fullname": "Di Zhang",
"name": "qq8933",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 108,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"aust-t",
"dingangui",
"csabakecskemeti",
"Csplk",
"ai-everyday"
],
"count": 5
},
{
"reaction": "๐",
"users": [
"John6666",
"TouchNight"
],
"count": 2
}
] | 2024-11-04T15:09:31.000Z | 2024-11-14T11:32:01.358Z | [
{
"avatarUrl": "/avatars/b2725bb163fa15d6c5856121780d52eb.svg",
"fullname": "Ci Splunk",
"name": "Csplk",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 43,
"isFollowing": false
},
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/produ... | /posts/qq8933/376713378079900 | 2,289 | 3 |
380609696467893 | [
{
"type": "text",
"value": "๐ Celebrating One Year of #SauerkrautLM with Two Groundbreaking Releases! ",
"raw": "๐ Celebrating One Year of #SauerkrautLM with Two Groundbreaking Releases! ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
... | ๐ Celebrating One Year of #SauerkrautLM with Two Groundbreaking Releases!
We're thrilled to announce the release of SauerkrautLM-v2-14b in two specialized versions: https://huggingface.co/VAGOsolutions/SauerkrautLM-v2-14b-SFT and https://huggingface.co/VAGOsolutions/SauerkrautLM-v2-14b-DPO. Built on the robust Qwen2.5-14B foundation, these models represent a significant leap forward in multilingual AI capabilities.
๐ฌ Technical Breakthroughs:
๐ Innovative three-phase Fine-Tuning approach
๐ Two-step Spectrum SFT + one-step Spectrum DPO optimization phase for enhanced performance
๐ Balance of German and English language capabilities
๐ Advanced function calling - almost on par with Claude-3.5-Sonnet-20240620
๐ฉ๐ช German Language Excellence:
What sets this release apart is our unique achievement in simultaneously improving both German and English capabilities. Through our specialized training approach with over 1.2B tokens across two phases, we've managed to:
๐ Enhance German language understanding and generation (SFT Version > DPO Version)
๐ Maintain authentic German linguistic nuances
๐ Improve cross-lingual capabilities
๐ Preserve cultural context awareness
๐ Training Innovation:
Our three-phase approach targeted specific layer percentages (15%, 20% and 25%) with carefully curated datasets, including:
๐ Mathematics-focused content (proprietary classifier-selected)
๐ High-quality German training data
๐ Specialized function calling datasets
๐ Premium multilingual content
๐ Community Contribution:
We're also releasing two new datasets in a few days:
1๏ธโฃ SauerkrautLM-Fermented-GER-DPO: 3,300 high-quality German training samples
2๏ธโฃ SauerkrautLM-Fermented-Irrelevance-GER-DPO: 2,000 specialized samples for optimized function call irrelevance handling
Thank you to our incredible community and partners who have supported us throughout this journey. Here's to another year of AI innovation!ย ๐ | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/64b999a40b24527e9c25583a/xFHCewJdf5EGn8qDPypqy.jpeg",
"fullname": "David Golchinfar",
"name": "DavidGF",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 49,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/64b999a40b24527e9c25583a/p2DKcLtfuRnsNBlsG6qSI.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/64b999a40b24527e9c25583a/veWKzsQQtRA-RZ2A7-mIF.jpeg"
},
{
"type": "imag... | [] | [
{
"reaction": "๐ฅ",
"users": [
"rizky-gumelar",
"ZeroXClem",
"John6666",
"foscraft",
"djuna",
"not-lain"
],
"count": 6
},
{
"reaction": "๐",
"users": [
"flozi00",
"Jason233"
],
"count": 2
}
] | 2024-11-04T14:34:23.000Z | 2024-11-04T14:35:08.561Z | [] | /posts/DavidGF/380609696467893 | 2,973 | 0 |
878599389112746 | [
{
"type": "text",
"value": "Early Morning Before Work Project:",
"raw": "Early Morning Before Work Project:",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\... | Early Morning Before Work Project:
๐ Introducing Cascade of Semantically Integrated Layers (CaSIL): A Humorously Over-Engineered Algorithm That Actuallyโฆ Works ๐คทโโ๏ธ
Let me introduce CaSIL โ the Cascade of Semantically Integrated Layers. Imagine giving a single question the level of introspection typically reserved for philosophical debates or maybe therapy. In short, CaSIL is a pure Python reasoning algorithm that, in a series of semantically rich layers, takes any input and rebuilds it into a nuanced response thatโs (surprisingly) meaningful to a human.
Iโve been experimenting with various reasoning and agent approaches lately and decided to contribute my own quirky take on layered processing. Itโs built without agent frameworksโjust good ol' Python and mathโand it plays nicely with any LLM. The result? A transformation from simple responses to deeper, interconnected insights. Hereโs a quick peek at the steps:
โจ How CaSIL Works:
Initial Understanding: The first layer captures the basic concepts in your input, just as a warm-up.
Relationship Analysis: A lightweight knowledge graph (because why not?) maps out related ideas and builds interconnections.
Context Integration: Adds historical or contextual knowledge, bringing a bit of depth and relevance.
Response Synthesis: Pieces it all together, aiming to produce a response that feels more like a conversation than an outdated search result.
Does it work? Yes! And in record time, too. Admittedly, the code is roughโtwo days of intense coding with some friendly help from Claude. The beauty of CaSIL is its simplicity and versatility; itโs a pure algorithm without complex dependencies, making it easy to integrate into your own LLM setups.
๐ Explore the repo here: https://github.com/severian42/Cascade-of-Semantically-Integrated-Layers
๐ Example outputs: https://github.com/severian42/Cascade-of-Semantically-Integrated-Layers/blob/main/examples.md | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/64740cf7485a7c8e1bd51ac9/CXZCJm2x4ToT83pEIYyQR.png",
"fullname": "Beckett Dillon",
"name": "Severian",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 175,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"John6666"
],
"count": 1
}
] | 2024-11-04T14:19:39.000Z | 2024-11-04T14:19:39.731Z | [] | /posts/Severian/878599389112746 | 470 | 0 |
915742231261639 | [
{
"type": "text",
"value": "I just shipped ",
"raw": "I just shipped ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "inline_code",
"value": null,
"raw": "`retrain-pipelines 0.1.1`",
"hre... | I just shipped `retrain-pipelines 0.1.1` today. The doc is also pimped compared to previous release. That was clearly not mature then.
I'll have to focus on another project for the next couple weeks but, anyone feel free to open issues on the GitHub repo and discuss any interest you'd have there if you will (please?) !
In the meantime, you may enjoy retrying this :
https://huggingface.co/blog/Aurelien-Morgan/stateful-metaflow-on-colab | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/651e93137b2a2e027f9e55df/5oXWJeEDCrMJLA4s_0I93.png",
"fullname": "Aurรฉlien-Morgan CLAUDON",
"name": "Aurelien-Morgan",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 9,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/651e93137b2a2e027f9e55df/Uwll3V6Tnc7p6LQ2ac5mh.png"
}
] | [] | [
{
"reaction": "๐",
"users": [
"John6666"
],
"count": 1
}
] | 2024-11-04T12:58:30.000Z | 2024-11-04T12:58:30.527Z | [] | /posts/Aurelien-Morgan/915742231261639 | 454 | 0 |
169924015276572 | [
{
"type": "text",
"value": "๐๐ปโโ๏ธhey there folks,",
"raw": "๐๐ปโโ๏ธhey there folks,",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": null,
... | ๐๐ปโโ๏ธhey there folks,
periodic reminder : if you are experiencing โ ๏ธ500 errors โ ๏ธ or โ ๏ธ abnormal `spaces` behavior on load or launch โ ๏ธ
we have a thread ๐๐ป https://discord.com/channels/879548962464493619/1295847667515129877
if you can record the problem and share it there , or on the forums in your own post , please dont be shy because i'm not sure but i do think it helps ๐ค๐ค๐ค | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/62a3bb1cd0d8c2c2169f0b88/eT2TS0IlQbZtz-F_zHLz9.jpeg",
"fullname": "Joseph [open/acc] Pollack",
"name": "Tonic",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 313,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"prithivMLmods",
"Nymbo",
"John6666",
"AtAndDev",
"clem",
"eyov",
"Zaws",
"rrg92"
],
"count": 8
},
{
"reaction": "โค๏ธ",
"users": [
"clem",
"John6666"
],
"count": 2
},
{
"reaction": "... | 2024-11-04T11:41:55.000Z | 2024-11-09T02:33:22.761Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6640bbd0220cfa8cbfdce080/wiAHUu5ewawyipNs0YFBR.png",
"fullname": "John Smith",
"name": "John6666",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 398,
"isFollowing": false
}... | /posts/Tonic/169924015276572 | 3,233 | 2 |
788696446784520 | [
{
"type": "text",
"value": "New Style, New Mix, New Drop ๐งค",
"raw": "New Style, New Mix, New Drop ๐งค",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
... | New Style, New Mix, New Drop ๐งค
๐งจFlux LoRA DLC: https://huggingface.co/spaces/prithivMLmods/FLUX-LoRA-DLC
๐Glowing-Body: https://huggingface.co/prithivMLmods/Glowing-Body-Flux-LoRA
๐Electric-Blue: https://huggingface.co/prithivMLmods/Electric-Blue-Flux-LoRA
๐Intense-Red: https://huggingface.co/prithivMLmods/Intense-Red-Flux-LoRA
๐Clouds-Illusion: https://huggingface.co/prithivMLmods/Clouds-Illusion-Flux-LoRA
๐Digital-Yellow: https://huggingface.co/prithivMLmods/Digital-Yellow-Flux-LoRA
๐งจFlux LoRA Collection: https://huggingface.co/collections/prithivMLmods/flux-lora-collections-66dd5908be2206cfaa8519be
.
.
.
@prithivMLmods
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/65bb837dbfb878f46c77de4c/UVtVbF_3rdt0DC8xTkpL1.jpeg",
"fullname": "Prithiv Sakthi",
"name": "prithivMLmods",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 393,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/VifKrKv_kxDWXE1dZLr6X.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/tksVqqDwdOz9tRyxLfdf3.png"
},
{
"type": "image... | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/65bb837dbfb878f46c77de4c/UVtVbF_3rdt0DC8xTkpL1.jpeg",
"fullname": "Prithiv Sakthi",
"name": "prithivMLmods",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 393
}
] | [
{
"reaction": "โค๏ธ",
"users": [
"Tonic",
"realaliarain",
"speedchemistry",
"JayNagose",
"prithivMLmods",
"multimodalart",
"radames",
"d8rt8v",
"Hamed744",
"rdrede",
"hypergod",
"darksfx",
"ai4life44",
"faruqhrp",
"Ngrth... | 2024-11-04T10:35:40.000Z | 2024-11-05T07:56:16.114Z | [] | /posts/prithivMLmods/788696446784520 | 5,603 | 0 |
301983274684168 | [
{
"type": "text",
"value": "Vector Search (most) datasets on the Hugging Face Hub ๐ฆ",
"raw": "Vector Search (most) datasets on the Hugging Face Hub ๐ฆ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": ... | Vector Search (most) datasets on the Hugging Face Hub ๐ฆ
Powered by: Polars, DuckDB, Gradio and model2vec (lightning-fast embeddings by Stรฉphan Tulkens).
Should work fast enough for datasets up to 100K.
https://huggingface.co/spaces/davidberenstein1957/vectorsearch-hub-datasets | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1677141720071-634ff41ff32062e9eb7b06a3.jpeg",
"fullname": "David Berenstein",
"name": "davidberenstein1957",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 167,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ค",
"users": [
"davidberenstein1957",
"Tonic",
"prithivMLmods",
"xi0v",
"Nymbo",
"djuna",
"thanhkt"
],
"count": 7
},
{
"reaction": "๐",
"users": [
"davidberenstein1957",
"Tonic",
"xi0v",
"Nymbo",
"Go... | 2024-11-04T10:19:03.000Z | 2024-11-04T10:19:03.151Z | [] | /posts/davidberenstein1957/301983274684168 | 3,077 | 0 |
419590652134553 | [
{
"type": "text",
"value": "Hello researchers! Here are scripts to generate reviews on HF Daily Papers:",
"raw": "Hello researchers! Here are scripts to generate reviews on HF Daily Papers:",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
... | Hello researchers! Here are scripts to generate reviews on HF Daily Papers:
๐ https://github.com/averkij/top_papers
โ๏ธ Works on GitHub Actions
๐ค Claude, GPT-4o, FLUX
๐ Multiple languages
๐ Classification by 38 topics (#agents, #multimodal, #plp, etc.)
๐บ https://HFday.ru | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1635314457124-5f32b2367e583543386214d9.jpeg",
"fullname": "Sergei Averkiev",
"name": "averoo",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 20,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/5f32b2367e583543386214d9/eNrWiDtpRGgBk-aRBJCGX.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/5f32b2367e583543386214d9/-kPHSc3clhpzSvRTp_1PA.png"
}
] | [] | [
{
"reaction": "๐",
"users": [
"John6666"
],
"count": 1
}
] | 2024-11-04T10:07:24.000Z | 2024-11-04T10:08:40.111Z | [] | /posts/averoo/419590652134553 | 348 | 0 |
808388125499602 | [
{
"type": "text",
"value": "It's work like this that in some way signal the eventual โdominanceโ of AI over all the sciences.",
"raw": "It's work like this that in some way signal the eventual โdominanceโ of AI over all the sciences.",
"href": null,
"resource": null,
"url": null,
"code":... | It's work like this that in some way signal the eventual โdominanceโ of AI over all the sciences.
โWe train our model on the six-dimensional N-body phase space, predicting particle velocities as the time derivative of the modelโs displacement outputsโ
The emulator is capable of predicting
the nonlinear displacement and velocity fields for 128^3 particles in half a second on a single GPU๐คฏ | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6438a9027de34e8ea7e4b257/vib8QSd1AWMr_bR9ig_xJ.jpeg",
"fullname": "Jaward Sesay",
"name": "Jaward",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 191,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6438a9027de34e8ea7e4b257/yxh8K8-a8AR8Wyl0SYG4y.png"
},
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6438a9027de34e8ea7e4b257/s7puQYqsEnc0SkSzs-FX7.qt"
},
{
"type": "image"... | [] | [
{
"reaction": "๐ฅ",
"users": [
"mediiiiii3",
"ajibawa-2023",
"John6666",
"Chief-Inspector",
"ai-everyday"
],
"count": 5
}
] | 2024-11-04T07:12:47.000Z | 2024-11-05T07:42:43.778Z | [
{
"avatarUrl": "/avatars/8aaab676f66023255d397ba82b4bcb6e.svg",
"fullname": "James Hunter Carter",
"name": "jameshuntercarter",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 2,
"isFollowing": false
}
] | /posts/Jaward/808388125499602 | 2,089 | 1 |
241878000943561 | [
{
"type": "text",
"value": "Hey, I just added three useful advanced use cases to ",
"raw": "Hey, I just added three useful advanced use cases to ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "resou... | Hey, I just added three useful advanced use cases to https://huggingface.co/datasets/do-me/SemanticFinder/blob/main/README.md#advanced-use-cases.
SemanticFinder is a collection of embeddings for public documents or books. You can create your own index file from any text or pdf and save it without installing or downloading anything. Try yourself:
1. Translating from 100+ languages to English (even though it might confuse a strawberry with a grapefruit ;D): https://do-me.github.io/SemanticFinder/?hf=List_of_the_Most_Common_English_Words_70320cde&firstOnly=true&inferencingActive=False
2. Finding English synonyms: https://do-me.github.io/SemanticFinder/?hf=List_of_the_Most_Common_English_Words_0d1e28dc&firstOnly=true&inferencingActive=False
3. The "universal index idea": create an embedding index with 30k English words and reuse it on unseen texts. You can decide to fill the gaps in the index by additional inferencing or just stick to the 30k index for instant semantic similarity.
Initial idea: https://github.com/do-me/SemanticFinder/discussions/48
Try here: https://do-me.github.io/SemanticFinder/?hf=List_of_the_Most_Common_English_Words_0d1e28dc&inferencingActive=False&universalIndexSettingsWordLevel with a text of your choice.
This could be enhanced by adding duplets or triplets like "climate change" or "green house gas". Eventually I'd like to set up vector DB integrations.
Super happy to hear your feedback, ideas and maybe even contributions! :)
---
Edit: Apparently markdown url formatting does only work for HF links. | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/IiercF_qxHWize2kitl9X.jpeg",
"fullname": "Dominik Weckmรผller",
"name": "do-me",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 38,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/64c4da8719565937fb268b32/_oXEubLaDx7pNQan3_z9I.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/64c4da8719565937fb268b32/BtcJXsuC1Fp-vdYjjieFg.png"
},
{
"type": "image... | [] | [
{
"reaction": "๐ฅ",
"users": [
"reach-vb",
"osanseviero"
],
"count": 2
},
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"fahmimmaliki"
],
"count": 2
},
{
"reaction": "๐",
"users": [
"shiv2050"
],
"count": 1
}
] | 2024-03-29T13:55:41.000Z | 2024-03-29T13:59:26.360Z | [] | /posts/do-me/241878000943561 | 1,899 | 0 |
430691527295062 | [
{
"type": "text",
"value": "โ๏ธโ New Research Alert! โ๏ธ๐",
"raw": "โ๏ธโ New Research Alert! โ๏ธ๐",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"hre... | โ๏ธโ New Research Alert! โ๏ธ๐
๐ Title: CoDA: Instructive Chain-of-Domain Adaptation with Severity-Aware Visual Prompt Tuning
๐ Description: CoDA is a UDA methodology that boosts models to understand all adverse scenes (โ๏ธ,โ,โ๏ธ,๐) by highlighting the discrepancies within these scenes. CoDA achieves state-of-the-art performances on widely used benchmarks.
๐ฅ Authors: Ziyang Gong, Fuhao Li, Yupeng Deng, Deblina Bhattacharjee, Xiangwei Zhu, Zhenming Ji
๐ Paper: https://huggingface.co/papers/2403.17369
๐ Repository: https://github.com/Cuzyoung/CoDA
๐ More Papers: more cutting-edge research presented at other conferences in the https://huggingface.co/spaces/DmitryRyumin/NewEraAI-Papers curated by @DmitryRyumin
๐ Keywords: #CoDA #DomainAdaptation #VisualPromptTuning #SAVPT #DeepLearning #Innovation | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/lX46WOvwgIq9KJ-Mz_JVk.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/SmO5hUxZ1TBBAXutPbGv8.png"
},
{
"type": "image... | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377
}
] | [
{
"reaction": "๐",
"users": [
"DmitryRyumin",
"Cusyoung",
"reach-vb",
"osanseviero",
"samusenps",
"shiv2050",
"ssml2050",
"Makya"
],
"count": 8
},
{
"reaction": "๐ฅ",
"users": [
"DmitryRyumin",
"nikgr",
"reach-vb",
... | 2024-03-29T09:44:13.000Z | 2024-03-29T09:44:13.353Z | [] | /posts/DmitryRyumin/430691527295062 | 2,134 | 0 |
839109593076616 | [
{
"type": "text",
"value": "Are you looking for the perfect leaderboard/arena for your use case? ๐",
"raw": "Are you looking for the perfect leaderboard/arena for your use case? ๐",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lan... | Are you looking for the perfect leaderboard/arena for your use case? ๐
There's a new tool for this!
https://huggingface.co/spaces/leaderboards/LeaderboardFinder
Select your modality, language, task... then search! ๐
Some categories of interest:
- does the leaderboard accept submissions?
- is the test set private or public?
- is it using an automatic metric, human evaluators, or llm as a judge?
The spaces list is build from space metadata, and reloaded every hour.
Enjoy! | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1644340617257-noauth.png",
"fullname": "Clรฉmentine Fourrier",
"name": "clefourrier",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 459,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ฅ",
"users": [
"m-ric",
"nikgr",
"VictorSanh",
"osanseviero",
"bmorphism"
],
"count": 5
},
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"jiuxing123"
],
"count": 2
},
{
"reaction": "๐",
"users": [
"shiv205... | 2024-03-29T07:47:34.000Z | 2024-03-29T07:47:34.949Z | [] | /posts/clefourrier/839109593076616 | 2,011 | 0 |
786994188068359 | [
{
"type": "text",
"value": "Excited to introduce Jamba by AI21",
"raw": "Excited to introduce Jamba by AI21",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\... | Excited to introduce Jamba by AI21
https://huggingface.co/ai21labs/Jamba-v0.1
We are thrilled to announce Jamba, the worldโs first production-grade Mamba based model.
Key Features:
- First production-grade Mamba based model built on a novel SSM-Transformer hybrid architecture
- 3X throughput on long contexts compared to Mixtral 8x7B
- Democratizes access to a massive 256K context window
- The only model in its size class that fits up to 140K context on a single GPU
Jamba is based on a novel architecture that combines Mamba and Transformer. While our initial results show great efficiency gains, we expect this to be further explored and improved with the help of the community.
Check out our blog post for more info: https://ai21-labs.webflow.io/blog/announcing-jamba
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/65e60c0ed5313c06372446ff/tX-xYqpVkE0yTUfRN3SWB.jpeg",
"fullname": "Or Dagan",
"name": "ordagan",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 20,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/65e60c0ed5313c06372446ff/Iz_qA2KJZJQkj34s-5-sg.jpeg"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/65e60c0ed5313c06372446ff/QskiyLYkBUGpkErZi_2Dn.png"
}
] | [] | [
{
"reaction": "๐ฅ",
"users": [
"osanseviero",
"nikgr",
"taufiqdp",
"samusenps",
"ajibawa-2023",
"Srulikbdd",
"theKibster",
"victor",
"VictorSanh",
"reach-vb",
"apollomarv",
"Asaf-Yehudai",
"MaziyarPanahi",
"DataSoul",
... | 2024-03-28T22:45:26.000Z | 2024-03-29T15:00:09.926Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/64aea8ff67511bd3d965697b/Jxn52EmDF5RApJh8antxn.jpeg",
"fullname": "Feynman Innovations",
"name": "ajibawa-2023",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 138,
"isFollow... | /posts/ordagan/786994188068359 | 2,161 | 2 |
474909441976843 | [
{
"type": "text",
"value": "A new paper titled \"Long-Form Factuality in Large Language Models\" proposes a new approach to evaluate the long-form factuality of large language models using an AI agent! They introduce SAFE (Search-Augmented Factuality Evaluator) which leverages an LLM to break down responses... | A new paper titled "Long-Form Factuality in Large Language Models" proposes a new approach to evaluate the long-form factuality of large language models using an AI agent! They introduce SAFE (Search-Augmented Factuality Evaluator) which leverages an LLM to break down responses into individual facts, query Google to verify each fact, and perform multi-step reasoning.
Keypoints:
* SAFE (Search-Augmented Factuality Evaluator) is an automated method using an LLM agent to evaluate factuality
* It also introduces LongFact, a 2,280 prompt set spanning 38 topics to test open-domain factual knowledge
* SAFE achieves a 72% humans agreement while being 20x cheaper. It also wins 76% of the disagreements measured on a small scale experiment where a more thorough human procedure (researchers + full internet search) was used.
* Larger models like GPT-4, Claude Opus and Gemini Ultra tend to exhibit better long-form factuality.
Paper: https://huggingface.co/papers/2403.18802
Code and data: https://github.com/google-deepmind/long-form-factuality
Congrats to the authors for their work! | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/657217faabb25ed8aedd5e48/UUHAXeGtOnQBXFD3nYtf2.jpeg",
"fullname": "Vlad Bogolin",
"name": "vladbogo",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 109,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/657217faabb25ed8aedd5e48/xEzUkOnojzRNKSUSw1wzO.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/657217faabb25ed8aedd5e48/uE4DOZwEstJIeDIgl9aYc.png"
},
{
"type": "image... | [] | [
{
"reaction": "๐ฅ",
"users": [
"nikgr",
"zelus82",
"dillfrescott"
],
"count": 3
},
{
"reaction": "๐",
"users": [
"shiv2050",
"ssml2050",
"dillfrescott"
],
"count": 3
},
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"dillf... | 2024-03-28T22:34:25.000Z | 2024-03-28T22:34:39.045Z | [] | /posts/vladbogo/474909441976843 | 1,701 | 0 |
706415412818350 | [
{
"type": "text",
"value": "A Little guide to building Large Language Models in 2024",
"raw": "A Little guide to building Large Language Models in 2024",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": ... | A Little guide to building Large Language Models in 2024
This is a post-recording of a 75min lecture I gave two weeks ago on how to train a LLM from scratch in 2024. I tried to keep it short and comprehensive โ focusing on concepts that are crucial for training good LLM but often hidden in tech reports.
In the lecture, I introduce the students to all the important concepts/tools/techniques for training good performance LLM:
* finding, preparing and evaluating web scale data
* understanding model parallelism and efficient training
* fine-tuning/aligning models
* fast inference
There is of course many things and details missing and that I should have added to it, don't hesitate to tell me you're most frustrating omission and I'll add it in a future part. In particular I think I'll add more focus on how to filter topics well and extensively and maybe more practical anecdotes and details.
Now that I recorded it I've been thinking this could be part 1 of a two-parts series with a 2nd fully hands-on video on how to run all these steps with some libraries and recipes we've released recently at HF around LLM training (and could be easily adapted to your other framework anyway):
*`datatrove` for all things web-scale data preparation: https://github.com/huggingface/datatrove
*`nanotron` for lightweight 4D parallelism LLM training: https://github.com/huggingface/nanotron
*`lighteval` for in-training fast parallel LLM evaluations: https://github.com/huggingface/lighteval
Here is the link to watch the lecture on Youtube: https://www.youtube.com/watch?v=2-SPH9hIKT8
And here is the link to the Google slides: https://docs.google.com/presentation/d/1IkzESdOwdmwvPxIELYJi8--K3EZ98_cL6c5ZcLKSyVg/edit#slide=id.p
Enjoy and happy to hear feedback on it and what to add, correct, extend in a second part. | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1583857746553-5df7e9e5da6d0311fd3d53f9.jpeg",
"fullname": "Thomas Wolf",
"name": "thomwolf",
"type": "user",
"isPro": true,
"isHf": true,
"isMod": false,
"followerCount": 704,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"clem",
"osanseviero",
"giux78",
"samusenps",
"dillfrescott",
"mohammedbriman",
"reach-vb",
"rwightman",
"Venkman42",
"gusthema",
"9bow",
"amenur",
"lisuizhe",
"The-Great-Genius",
"danja"... | 2024-03-28T21:44:02.000Z | 2024-04-04T12:44:05.941Z | [
{
"avatarUrl": "/avatars/cc5c07cc99c647da19cff5219f43869d.svg",
"fullname": "noh au",
"name": "noh-au-00",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2,
"isFollowing": false
},
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/prod... | /posts/thomwolf/706415412818350 | 4,913 | 2 |
506318409499980 | [
{
"type": "text",
"value": "๐๐ก๐ ๐ซ๐๐ญ๐ฎ๐ซ๐ง ๐จ๐ ๐ญ๐ก๐ ๐๐๐๐ฌ โ ๐๐๐ฐ ๐๐๐ฆ๐๐-๐๐๐ฌ๐๐ ๐๐ซ๐๐ก๐ข๐ญ๐๐๐ญ๐ฎ๐ซ๐ \"๐๐๐ฆ๐๐\"",
"raw": "๐๐ก๐ ๐ซ๐๐ญ๐ฎ๐ซ๐ง ๐จ๐ ๐ญ๐ก๐ ๐๐๐๐ฌ โ ๐๐๐ฐ ๐๐๐ฆ๐๐-๐๐๐ฌ๐๐ ๐๐ซ๐๐ก๐ข๐ญ๐๐๐ญ๐ฎ๐ซ๐ \"๐๐๐ฆ๐๐\"",
"href": null,
"resource": null,
... | ๐๐ก๐ ๐ซ๐๐ญ๐ฎ๐ซ๐ง ๐จ๐ ๐ญ๐ก๐ ๐๐๐๐ฌ โ ๐๐๐ฐ ๐๐๐ฆ๐๐-๐๐๐ฌ๐๐ ๐๐ซ๐๐ก๐ข๐ญ๐๐๐ญ๐ฎ๐ซ๐ "๐๐๐ฆ๐๐"
Since the release of BERT by Google in 2019, Transformers architecture have taken over machine learning thanks to their ๐ฎ๐๐๐ฒ๐ป๐๐ถ๐ผ๐ป ๐บ๐ฒ๐ฐ๐ต๐ฎ๐ป๐ถ๐๐บ, that gives them the ability to focus on important points of the input. But ๐๐ฉ๐ฉ๐๐ฃ๐ฉ๐๐ค๐ฃ ๐๐ค๐ข๐ฅ๐ช๐ฉ๐๐ฉ๐๐ค๐ฃ ๐๐จ ๐ฆ๐ช๐๐๐ง๐๐ฉ๐๐ ๐๐ฃ ๐ฉ๐๐ ๐๐ฃ๐ฅ๐ช๐ฉ ๐ก๐๐ฃ๐๐ฉ๐.
๐ซ The Mamba paper, published in December 2023, announced the return of the RNNs: it has no attention, but integrates a selection mechanism, which should be able to reproduce the โfocusโ ability of attention, in an architecture for which the compute requirements ๐ด๐ฟ๐ผ๐ ๐ผ๐ป๐น๐ ๐น๐ถ๐ป๐ฒ๐ฎ๐ฟ๐น๐ ๐ถ๐ป ๐ถ๐ป๐ฝ๐๐ ๐น๐ฒ๐ป๐ด๐๐ต!
๐ค Would this work? We had yet to see a large Mamba model recovering the performance of Attention-based Transformers.
๐ฅ But now it's done! A (Mamba + Transformers) hybrid just beat Transformers!
The AI21 Labs team just released Jamba.
They insert a few Transformer layers to inject some attention in a big pile of Mamba layers, thus getting the best of both worlds.
๐๐;๐ฟ๐:
๐๏ธ ๐ก๐ฒ๐ ๐ ๐ผ๐ ๐ฎ๐ฟ๐ฐ๐ต๐ถ๐๐ฒ๐ฐ๐๐๐ฟ๐ฒ: 4 Jamba blocks, each of these being 7 Mamba layers for 1 Transformer.
๐๏ธ ๐ฑ๐ฎ๐ ๐ฝ๐ฎ๐ฟ๐ฎ๐บ๐ฒ๐๐ฒ๐ฟ๐, ๐ญ๐ฎ๐ ๐ฎ๐ฐ๐๐ถ๐๐ฒ ๐ฎ๐ ๐ถ๐ป๐ณ๐ฒ๐ฟ๐ฒ๐ป๐ฐ๐ฒ: This reduction is enabled by Mixture of Experts, and similar to Mixtral (47B parameters - 13B active).
๐๏ธ ๐ฆ๐ฝ๐ฒ๐ฒ๐ฑ: ๐
๐ฏ ๐๐ต๐ฟ๐ผ๐๐ด๐ต๐ฝ๐๐. Jamba is much faster than similar-sized Transformer models on long contexts.
๐ ๐๐ผ๐ป๐๐ฒ๐
๐ ๐น๐ฒ๐ป๐ด๐๐ต: ๐ญ๐ฐ๐ฌ๐ ๐๐ผ๐ธ๐ฒ๐ป๐ on a single 80GB A100!
๐ช ๐ฃ๐ฒ๐ฟ๐ณ๐ผ๐ฟ๐บ๐ฎ๐ป๐ฐ๐ฒ: ๐๐๐ฎ๐๐ฒ-๐ผ๐ณ-๐๐ต๐ฒ-๐ฎ๐ฟ๐ ๐ณ๐ผ๐ฟ ๐๐ต๐ถ๐ ๐๐ถ๐๐ฒ. The small injection of attention seems sufficient since Jamba beats the open-source reference Mixtral-8x7B on many benchmarks!
Try it here ๐ https://huggingface.co/ai21labs/Jamba-v0.1 | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/63d10d4e8eaa4831005e92b5/7p7-OmWM6PqqCs7ZStPGD.jpeg",
"fullname": "Aymeric Roucher",
"name": "m-ric",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 494,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/63d10d4e8eaa4831005e92b5/AsfDNZjOUiCENV-G1W6zv.png"
}
] | [] | [
{
"reaction": "๐",
"users": [
"monsoon-nlp",
"lunarflu",
"carbonox-infernox",
"dhuynh95",
"osanseviero",
"vonehrenheim",
"nikgr",
"samusenps",
"Amano13",
"reach-vb",
"Chunte",
"seyf1elislam",
"blanchon"
],
"count": 13
}... | 2024-03-28T17:32:17.000Z | 2024-03-29T08:35:18.879Z | [] | /posts/m-ric/506318409499980 | 1,798 | 0 |
545762059483943 | [
{
"type": "text",
"value": "Just now, we release a small MoE model, Qwen1.5-MoE-A2.7B, a 14B model with 2.7B activated parameters. Leaving the hype, I would love to share more things here in HF. But if you don't know much about this, check our blog for more info: ",
"raw": "Just now, we release a small ... | Just now, we release a small MoE model, Qwen1.5-MoE-A2.7B, a 14B model with 2.7B activated parameters. Leaving the hype, I would love to share more things here in HF. But if you don't know much about this, check our blog for more info: https://qwenlm.github.io/blog/qwen-moe/
At the beginning, it was trying with the MoE stuff, making Megatron work well with MegaBlocks. As always, we worked with small ones first. However, we have been struggling with a lot of details.
With megablocks and so many tricks that make training MoE models work, it is almost impossible to fail. The challenge is actually how good your model is. Then things became more complex than I had expected. Finegrained experts actually pissed me off but damn it works for the model at this scale. However, it brings complexity to the model, and this is somehow why at this moment our codes are not merged into llama.cpp cuz it really brings problems. Shared experts might be good, but we need more engineering efforts to really unleash its benefits in inference acceleration.
For the community, this is actually our first time releasing an MoE model. We don't know what will happen to us, but we are prepared for complaints. I just hope that we can really make things clear, and provide a good recipe to play with our MoE model just like people playing with Mixtral. | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/620760a26e3b7210c2ff1943/VC-rKqimF6yxGESNVlPoR.jpeg",
"fullname": "Junyang Lin",
"name": "JustinLin610",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 132,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ฅ",
"users": [
"Nbardy",
"lunarflu",
"CulturedMan",
"osanseviero",
"nikgr",
"manishiitg",
"reach-vb",
"pierrci",
"aguspiza",
"xiaoniqiu",
"alielfilali01",
"Shuaiii"
],
"count": 12
},
{
"reaction": "โค๏ธ",
... | 2024-03-28T17:06:31.000Z | 2024-03-28T21:59:49.971Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing":... | /posts/JustinLin610/545762059483943 | 2,949 | 1 |
231419617229308 | [
{
"type": "text",
"value": " ๐ LLM Benchmarks Update!",
"raw": " ๐ LLM Benchmarks Update!",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": ... | ๐ LLM Benchmarks Update!
**tl;dr: do not depend on benchmark leaderboards to choose your "chatbot" model! (Especially for non-English languages.)**
First of all, I'm discontinuing the Open #Dutch #LLM Leaderboard (https://lnkd.in/eFnsaFR6). It will stay online for now, but I urge the use of the ScandEval leaderboard instead (https://scandeval.com/dutch-nlg/) by @saattrupdan. It contains more tasks, has better reproducibility and statistics (CI) and a flexible back-end library (`scandeval`) to run your own benchmarks with. As part of project "Leesplank" (with Michiel Buisman and Maarten Lens-FitzGerald) we recently added GPT-4-1106-preview scores to add a good "target" to the leaderboard.
An important note here is that benchmark leaderboards are not a golden truth. Especially evaluating generative models is hard. You run into issues like prompt engineering (and sensitivity of models to one or other prompt), structured output generation, and - quite simply - "how to automatically evaluate open-ended generation".
๐ก Another important but under-discussed facet is the discrepancy between models' capability of understanding vs. generating *in different languages* (so the NLU part of NLG benchmarking). In other words: some of the listed models score really well on, e.g., MCQ benchmarks but are not suitable to use as DUTCH chat bots. Interestingly, some of these models seem to understand questions in Dutch and are able to pick the right answer (because they have good knowledge or reasoning skills), but generating fluent and grammatical Dutch is something else entirely! This is perhaps also true for humans: it's easier to sort-of grasp the meaning of a new language and answer with "Yes" or "No", but answering fluently in the language is much harder! Yet, your language production fluency does not necessarily say anything about your knowledge and reasoning skills.
Hopefully we can get a chat arena for Dutch some day - user feedback is the most powerful metric!
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1594192845975-5e1e17b6fcf41d740b6996a8.jpeg",
"fullname": "Bram Vanroy",
"name": "BramVanroy",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 173,
"isFollowing": false
} | [] | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1624975632470-60d368a613f774189902f555.jpeg",
"fullname": "Dan Saattrup Nielsen",
"name": "saattrupdan",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 37
}
] | [
{
"reaction": "๐",
"users": [
"davanstrien",
"lunarflu",
"nikgr",
"saattrupdan",
"reach-vb",
"shiv2050",
"ssml2050"
],
"count": 7
},
{
"reaction": "๐ค",
"users": [
"Rijgersberg",
"osanseviero",
"lunarflu",
"robinsmits",
... | 2024-03-28T13:08:40.000Z | 2024-03-30T15:52:48.997Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/643033401572f43a4818cea3/CiAPL-E-6p-W8ilJ3t9vM.jpeg",
"fullname": "Carlo Moro",
"name": "cnmoro",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 25,
"isFollowing": false
},... | /posts/BramVanroy/231419617229308 | 2,436 | 3 |
631118558281010 | [
{
"type": "text",
"value": "Diaries of Open Source. Part 11 ๐",
"raw": "Diaries of Open Source. Part 11 ๐",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\... | Diaries of Open Source. Part 11 ๐
๐Databricks release DBRX, potentially the best open access model! A 132B Mixture of Experts with 36B active params and trained on 12 trillion tokens
Blog: https://www.databricks.com/blog/introducing-dbrx-new-state-art-open-llm
Base and instruct models: https://hf.co/collections/databricks/dbrx-6601c0852a0cdd3c59f71962
Demo: https://hf.co/spaces/databricks/dbrx-instruct
๐ค1-bit and 2-bit quantization exploration using HQQ+
Blog post: https://mobiusml.github.io/1bit_blog/
Models: https://hf.co/collections/mobiuslabsgmbh/llama2-7b-hqq-6604257a96fc8b9c4e13e0fe
GitHub: https://github.com/mobiusml/hqq
๐Cosmopedia: a large-scale synthetic dataset for pre-training - it includes 25 billion tokens and 30 million files
Dataset: https://hf.co/datasets/HuggingFaceTB/cosmopedia
Blog: https://hf.co/blog/cosmopedia
โญMini-Gemini: multi-modal VLMs, from 2B to 34B
Models: https://hf.co/collections/YanweiLi/mini-gemini-6603c50b9b43d044171d0854
Paper: https://huggingface.co/papers/2403.18814
GitHub: https://github.com/dvlab-research/MiniGemini
๐ฅVILA - On Pre-training for VLMs
Models: https://hf.co/collections/Efficient-Large-Model/vila-on-pre-training-for-visual-language-models-65d8022a3a52cd9bcd62698e
Paper: https://hf.co/papers/2312.07533
Misc
๐ FeatUp: a framework for image features at any resolution: https://hf.co/spaces/mhamilton723/FeatUp https://hf.co/papers/2403.10516
๐ColBERTus Maxiums, a colbertialized embedding model https://hf.co/mixedbread-ai/mxbai-colbert-large-v1
๐๏ธSemantic Palette, a new drawing paradigm https://hf.co/spaces/ironjr/SemanticPalette
๐งโโ๏ธHistoGPT, a vision model that generates accurate pathology reports https://hf.co/marr-peng-lab/histogpt https://www.medrxiv.org/content/10.1101/2024.03.15.24304211v1 | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ฅ",
"users": [
"mobicham",
"appoose",
"loubnabnl",
"davanstrien",
"samusenps",
"DmitryRyumin",
"ajibawa-2023",
"clem",
"GO4code",
"victor",
"vladbogo",
"nikgr",
"VictorSanh",
"reach-vb",
"manuel-tran"
... | 2024-03-28T10:06:27.000Z | 2024-03-28T13:03:16.372Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing":... | /posts/osanseviero/631118558281010 | 2,066 | 4 |
875756613096016 | [
{
"type": "text",
"value": "๐ ๐ ๐ Happy to share our recent work. We noticed that image resolution plays an important role, either in improving multi-modal large language models (MLLM) performance or in Sora style any resolution encoder decoder, we hope this work can help lift restriction of 224x224 reso... | ๐ ๐ ๐ Happy to share our recent work. We noticed that image resolution plays an important role, either in improving multi-modal large language models (MLLM) performance or in Sora style any resolution encoder decoder, we hope this work can help lift restriction of 224x224 resolution limit in ViT.
https://huggingface.co/papers/2403.18361 | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/650dde4ce14eeb01d42b37a1/n5Yv24uofZ2XJjXdYCrKd.png",
"fullname": "Xiaotian Han",
"name": "xiaotianhan",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 22,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"ajibawa-2023",
"samusenps",
"DmitryRyumin",
"clem",
"lunarflu",
"victor",
"nikgr",
"reach-vb",
"shiv2050",
"ak0601",
"ncoop57",
"talaviyabhavik"
],
"count": 12
},
{
"reaction": "๐",
"... | 2024-03-28T05:20:08.000Z | 2024-04-23T18:22:53.659Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1648113222875-6141a88b3a0ec78603c9e784.png",
"fullname": "Merve Noyan",
"name": "merve",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5589,
"isFollowing": false
},
{
... | /posts/xiaotianhan/875756613096016 | 2,087 | 2 |
149161309610662 | [
{
"type": "text",
"value": "Compared Effect Of Image Captioning For SDXL Fine-tuning / DreamBooth Training for a Single Person, 10.3 GB VRAM via OneTrainer",
"raw": "Compared Effect Of Image Captioning For SDXL Fine-tuning / DreamBooth Training for a Single Person, 10.3 GB VRAM via OneTrainer",
"hre... | Compared Effect Of Image Captioning For SDXL Fine-tuning / DreamBooth Training for a Single Person, 10.3 GB VRAM via OneTrainer
Sadly the post character count is limited so please read full info on Medium here
https://medium.com/@furkangozukara/compared-effect-of-image-captioning-for-sdxl-fine-tuning-dreambooth-training-for-a-single-person-961087e42334
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1672531901326-6345bd89fe134dfd7a0dba40.png",
"fullname": "Furkan Gรถzรผkara",
"name": "MonsterMMORPG",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 376,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6345bd89fe134dfd7a0dba40/gJjqOBUZJVWjq4h61tMHE.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6345bd89fe134dfd7a0dba40/oB6pmcPHurv5vrSZV-py3.png"
},
{
"type": "image... | [] | [
{
"reaction": "๐ฅ",
"users": [
"ameerazam08",
"clem",
"lunarflu",
"nikgr",
"roktimsardar123"
],
"count": 5
},
{
"reaction": "๐",
"users": [
"shiv2050"
],
"count": 1
}
] | 2024-03-28T02:44:56.000Z | 2024-03-28T02:44:56.968Z | [] | /posts/MonsterMMORPG/149161309610662 | 2,005 | 0 |
191716706611860 | [
{
"type": "text",
"value": "๐๐ญ๐ New Research Alert! ๐ ๐ญ๐",
"raw": "๐๐ญ๐ New Research Alert! ๐ ๐ญ๐",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\... | ๐๐ญ๐ New Research Alert! ๐ ๐ญ๐
๐ Title: AniPortrait: Audio-Driven Synthesis of Photorealistic Portrait Animation ๐
๐ Description: AniPortrait is a novel framework for generating photorealistic portrait animations driven by audio and a reference image, with superior facial naturalness, pose variety, and visual quality, with potential applications in facial motion editing and facial reenactment.
๐ฅ Authors: Huawei Wei, @ZJYang, Zhisheng Wang
๐ Paper: https://huggingface.co/papers/2403.17694
๐ Repository: https://github.com/Zejun-Yang/AniPortrait
๐ค Demo: https://huggingface.co/spaces/ZJYang/AniPortrait_official
๐ฅ Model ๐ค: https://huggingface.co/ZJYang/AniPortrait
๐ More Papers: more cutting-edge research presented at other conferences in the https://huggingface.co/spaces/DmitryRyumin/NewEraAI-Papers curated by @DmitryRyumin
๐ Added to the Avatars Collection: https://huggingface.co/collections/DmitryRyumin/avatars-65df37cdf81fec13d4dbac36
๐ Keywords: #AniPortrait #Animation #AudioDriven #Photorealistic #FacialAnimation #DeepLearning #Innovation | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/Q8D8SBxzlvdIwMES1aX9u.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/fbwLJKfraISTw4Nj5wWkJ.png"
},
{
"type": "video... | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377
},
{
"avatarUrl": "/avatars/26... | [
{
"reaction": "๐ฅ",
"users": [
"DmitryRyumin",
"hinach4n",
"samusenps",
"MiSTe-R",
"clem",
"lunarflu",
"victor",
"reach-vb",
"Poulain"
],
"count": 9
},
{
"reaction": "๐คฏ",
"users": [
"DmitryRyumin",
"clem",
"lunarflu... | 2024-03-27T22:52:30.000Z | 2024-04-13T22:54:24.627Z | [
{
"avatarUrl": "/avatars/afbc48df2e8c47c35be48168113d83c0.svg",
"fullname": "s",
"name": "Tom-Neverwinter",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2,
"isFollowing": false
},
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/pro... | /posts/DmitryRyumin/191716706611860 | 1,751 | 2 |
772576405157092 | [
{
"type": "text",
"value": "The Unreasonable Ineffectiveness of the Deeper Layers",
"raw": "The Unreasonable Ineffectiveness of the Deeper Layers",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_l... | The Unreasonable Ineffectiveness of the Deeper Layers
https://huggingface.co/papers/2403.17887
We empirically study a simple layer-pruning strategy for popular families of open-weight pretrained LLMs, finding minimal degradation of performance on different question-answering benchmarks until after a large fraction (up to half) of the layers are removed. To prune these models, we identify the optimal block of layers to prune by considering similarity across layers; then, to "heal" the damage, we perform a small amount of finetuning. In particular, we use parameter-efficient finetuning (PEFT) methods, specifically quantization and Low Rank Adapters (QLoRA), such that each of our experiments can be performed on a single A100 GPU. From a practical perspective, these results suggest that layer pruning methods can complement other PEFT strategies to further reduce computational resources of finetuning on the one hand, and can improve the memory and latency of inference on the other hand. From a scientific perspective, the robustness of these LLMs to the deletion of layers implies either that current pretraining methods are not properly leveraging the parameters in the deeper layers of the network or that the shallow layers play a critical role in storing knowledge. | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1674929746905-60f1abe7544c2adfd699860c.jpeg",
"fullname": "AK",
"name": "akhaliq",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5205,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/60f1abe7544c2adfd699860c/m6g_c4r2tuwYJoe95rEEv.png"
}
] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"ajibawa-2023",
"osanseviero",
"clem",
"ishmnnit",
"Aryanne",
"KvrParaskevi"
],
"count": 7
}
] | 2024-03-27T20:53:43.000Z | 2024-03-27T20:53:43.465Z | [] | /posts/akhaliq/772576405157092 | 2,517 | 0 |
554406272304085 | [
{
"type": "text",
"value": "SegGPT is a vision generalist on image segmentation, quite like GPTs for computer vision โจ",
"raw": "SegGPT is a vision generalist on image segmentation, quite like GPTs for computer vision โจ",
"href": null,
"resource": null,
"url": null,
"code": null,
"us... | SegGPT is a vision generalist on image segmentation, quite like GPTs for computer vision โจ
It comes with the last release of transformers ๐ Demo and more in this post!
SegGPT is an extension of the Painter, where you speak to images with images: the model takes in an image prompt, transformed version of the image prompt, the actual image you want to see the same transform, and expected to output the transformed image.
SegGPT consists of a vanilla ViT with a decoder on top (linear, conv, linear).
The model is trained on diverse segmentation examples, where they provide example image-mask pairs, the actual input to be segmented, and the decoder head learns to reconstruct the mask output.
This generalizes pretty well!
The authors do not claim state-of-the-art results as the model is mainly used zero-shot and few-shot inference. They also do prompt tuning, where they freeze the parameters of the model and only optimize the image tensor (the input context).
Thanks to ๐ค transformers you can use this model easily!
See here https://huggingface.co/docs/transformers/en/model_doc/seggpt
I have built an app for you to try it out. I combined SegGPT with Depth Anything Model, so you don't have to upload image mask prompts in your prompt pair ๐ค
Try it here https://huggingface.co/spaces/merve/seggpt-depth-anything
Also check out the collection https://huggingface.co/collections/merve/seggpt-660466a303bc3cd7559d271b | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1648113222875-6141a88b3a0ec78603c9e784.png",
"fullname": "Merve Noyan",
"name": "merve",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5589,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6141a88b3a0ec78603c9e784/dfWOvGWApX7gPUCaXs-Fc.png"
}
] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"osanseviero",
"clem",
"lunarflu",
"talaviyabhavik"
],
"count": 5
},
{
"reaction": "๐",
"users": [
"samusenps",
"osanseviero",
"clem",
"lunarflu"
],
"count": 4
},
{
"reactio... | 2024-03-27T18:48:31.000Z | 2024-03-27T18:48:31.826Z | [] | /posts/merve/554406272304085 | 2,899 | 0 |
698173564421191 | [
{
"type": "text",
"value": "Thanks to the 4000+ attempts to jailbreak LLM chatbots and 2000+ votes cast on the Chatbot Guardrails Arena in the 4 days since release! ",
"raw": "Thanks to the 4000+ attempts to jailbreak LLM chatbots and 2000+ votes cast on the Chatbot Guardrails Arena in the 4 days since ... | Thanks to the 4000+ attempts to jailbreak LLM chatbots and 2000+ votes cast on the Chatbot Guardrails Arena in the 4 days since release!
Based on the players' feedback, we have updated the instructions to be clearer and to emphasize that players should vote after identifying a secure chatbot.
Tell us what you think when you play again: https://huggingface.co/spaces/lighthouzai/guardrails-arena
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/65304c99471261a41237c94a/Cl0H2RtemLeeW0VHGHZ0j.png",
"fullname": "Srijan Kumar",
"name": "srijankedia",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 14,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"osanseviero",
"Tom-Neverwinter",
"clefourrier",
"clem",
"lunarflu",
"reach-vb",
"AtAndDev",
"normanschizogh",
"chuangxinlezhi",
"sbrandeis"
],
"count": 11
},
{
"reaction": "๐ง ",
... | 2024-03-27T16:44:52.000Z | 2024-03-27T16:44:52.435Z | [] | /posts/srijankedia/698173564421191 | 1,579 | 0 |
139271554761909 | [
{
"type": "text",
"value": "Here, we share the technical details of InternLM2, the current state-of-the-art open-source LLMs!!",
"raw": "Here, we share the technical details of InternLM2, the current state-of-the-art open-source LLMs!!",
"href": null,
"resource": null,
"url": null,
"code... | Here, we share the technical details of InternLM2, the current state-of-the-art open-source LLMs!!
See collections https://huggingface.co/collections/internlm/internlm2-65b0ce04970888799707893c
Paper: https://huggingface.co/papers/2403.17297 | {
"avatarUrl": "/avatars/18958b8406d1ce492b54c1c839f18c54.svg",
"fullname": "Wenwei Zhang",
"name": "ZwwWayne",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 10,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"osanseviero",
"clem",
"bedirt",
"victor"
],
"count": 5
}
] | 2024-03-27T16:11:15.000Z | 2024-03-27T16:11:15.797Z | [] | /posts/ZwwWayne/139271554761909 | 1,501 | 0 |
252914089485275 | [
{
"type": "text",
"value": "Have we really squeezed out the capacity of a compact chat model? Thrilled to see our latest open model, Starling-7B, ranks 13th among all models in Chatbot Arena! ",
"raw": "Have we really squeezed out the capacity of a compact chat model? Thrilled to see our latest open mod... | Have we really squeezed out the capacity of a compact chat model? Thrilled to see our latest open model, Starling-7B, ranks 13th among all models in Chatbot Arena!
๐ As a 7B model, Starling surpasses larger open and proprietary models, including Claude-2, GPT-3.5-Turbo, Gemini Pro, Mixtral 8x7B and Llama2-70B, and is currently the best 7B chat model in Chatbot Arena!
Try out the model on HF here: https://huggingface.co/Nexusflow/Starling-LM-7B-beta
| {
"avatarUrl": "/avatars/128ceae78490110ae41202851e84d58e.svg",
"fullname": "Banghua Zhu",
"name": "banghua",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 39,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/647b8885aba7062fe5c32000/aLY9sZyHnRtnYEtzxIgah.png"
}
] | [] | [
{
"reaction": "๐ฅ",
"users": [
"brian-yu-nexusflow",
"banghua",
"clem",
"ThWu",
"jian-zhang-nexusflow",
"mahiatlinux",
"osanseviero",
"unnaymed",
"Lewdiculous",
"dalraf",
"lunarflu",
"mohammedbriman",
"ucyang"
],
"count": ... | 2024-03-27T15:48:56.000Z | 2024-03-27T15:48:56.245Z | [] | /posts/banghua/252914089485275 | 1,516 | 0 |
254221141271676 | [
{
"type": "text",
"value": "New state-of-the-art open LLM! ๐ Databricks just released DBRX, a 132B MoE trained on 12T tokens. Claiming to surpass OpenAI GPT-3.5 and is competitive with Google Gemini 1.0 Pro. ๐คฏ",
"raw": "New state-of-the-art open LLM! ๐ Databricks just released DBRX, a 132B MoE traine... | New state-of-the-art open LLM! ๐ Databricks just released DBRX, a 132B MoE trained on 12T tokens. Claiming to surpass OpenAI GPT-3.5 and is competitive with Google Gemini 1.0 Pro. ๐คฏ
TL;DR
๐งฎ 132B MoE with 16 experts with 4 active in generation
๐ช 32 000 context window
๐ Outperforms open LLMs on common benchmarks, including MMLU
๐ Up to 2x faster inference than Llama 2 70B
๐ป Trained on 12T tokens
๐ก Uses the GPT-4 tokenizer
๐ Custom License, commercially useable
Collection: https://huggingface.co/collections/databricks/dbrx-6601c0852a0cdd3c59f71962
Demo: https://huggingface.co/spaces/databricks/dbrx-instruct
Kudos to the Team at Databricks and MosaicML for this strong release in the open community! ๐ค
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1624629516652-5ff5d596f244529b3ec0fb89.png",
"fullname": "Philipp Schmid",
"name": "philschmid",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 657,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/5ff5d596f244529b3ec0fb89/0-YBFNDbvD_xLRMNjk_Nd.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/5ff5d596f244529b3ec0fb89/H2QtWjot15itu9k_ypPLl.png"
},
{
"type": "image... | [] | [
{
"reaction": "๐ฅ",
"users": [
"victor",
"alielfilali01",
"Shinku",
"osanseviero",
"samusenps",
"giux78",
"Tom-Neverwinter",
"mmhamdy",
"sepal",
"giulianobr",
"carlos-eduardo",
"eppie2",
"VictorSanh",
"dalraf",
"lunarf... | 2024-03-27T13:59:58.000Z | 2024-03-28T18:26:50.483Z | [
{
"avatarUrl": "/avatars/afbc48df2e8c47c35be48168113d83c0.svg",
"fullname": "s",
"name": "Tom-Neverwinter",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2,
"isFollowing": false
},
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/pro... | /posts/philschmid/254221141271676 | 6,577 | 4 |
528922528677736 | [
{
"type": "text",
"value": " ๐ฐ Happy ML Easter! ๐ฐ",
"raw": " ๐ฐ Happy ML Easter! ๐ฐ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": null,
... | ๐ฐ Happy ML Easter! ๐ฐ
I recently had a wager with my colleagues which had me create AI-assisted videos of myself in an Easter Bunny costume singing an AI-generated easter song (in different languages).
I did it in 3 simple steps:
๐I created an easter Bunny version of myself with this space: https://huggingface.co/spaces/multimodalart/Ip-Adapter-FaceID by @multimodalart
๐ถ Created a song on suno.ai
๐๏ธ Used the Dreamtalk space by @fffiloni to animate the image, conditioned on the song: https://huggingface.co/spaces/fffiloni/dreamtalk | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/607feb037c746d01ecb19180/qs3NO5v-Ej5UaKns8yFW8.jpeg",
"fullname": "Johannes Kolbe",
"name": "johko",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 54,
"isFollowing": false
} | [
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/607feb037c746d01ecb19180/caigr9VEvW1oP_8-8W9O2.mp4"
},
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/607feb037c746d01ecb19180/2lJHe9pCvwParuw6xYHbq.mp4"
},
{
"type": "video... | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/61868ce808aae0b5499a2a95/F6BA0anbsoY_Z7M1JrwOe.jpeg",
"fullname": "Sylvain Filoni",
"name": "fffiloni",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 5185
},
{
"avatarUr... | [
{
"reaction": "โค๏ธ",
"users": [
"fffiloni",
"osanseviero",
"clem",
"victor",
"linoyts",
"Wauplin",
"kramp",
"yyassif",
"MarinaraSpaghetti",
"sayandafadar",
"alielfilali01",
"sbrandeis"
],
"count": 12
},
{
"reaction": "๐"... | 2024-03-27T13:08:34.000Z | 2024-03-27T13:09:55.557Z | [] | /posts/johko/528922528677736 | 1,545 | 0 |
160289342297994 | [
{
"type": "text",
"value": "Would 1-2 sentence tl;dr summaries of datasets on the Hub be useful for you? ",
"raw": "Would 1-2 sentence tl;dr summaries of datasets on the Hub be useful for you? ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": nu... | Would 1-2 sentence tl;dr summaries of datasets on the Hub be useful for you?
For example, for the https://huggingface.co/datasets/togethercomputer/RedPajama-Data-1T dataset, would the following summary help give you a quick sense of its content?
> tl;dr: RedPajama is a fully open-source implementation of the LLaMa dataset, consisting of 1.2 trillion tokens from sources like Commoncrawl, C4, GitHub, Books, ArXiv, Wikipedia, and StackExchange, primarily in English, and is structured with metadata for each text sample.
I've created a dataset with example summaries of the 500 most liked datasets on the Hub: https://huggingface.co/datasets/davanstrien/dataset-tldr
Would these kinds of summaries be helpful? | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1627505688463-60107b385ac3e86b3ea4fc34.jpeg",
"fullname": "Daniel van Strien",
"name": "davanstrien",
"type": "user",
"isPro": true,
"isHf": true,
"isMod": false,
"followerCount": 410,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"osanseviero",
"merve",
"clem",
"samusenps",
"sequelbox",
"DataSoul",
"Arakinas"
],
"count": 7
}
] | 2024-03-27T12:12:16.000Z | 2024-03-27T15:00:31.046Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1648113222875-6141a88b3a0ec78603c9e784.png",
"fullname": "Merve Noyan",
"name": "merve",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5589,
"isFollowing": false
},
{
... | /posts/davanstrien/160289342297994 | 1,321 | 2 |
223529654460707 | [
{
"type": "text",
"value": "๐ ๐ฎ๐ท๐ผ๐ฟ ๐ง๐ข๐ : ๐ฃ๐น๐ฎ๐ป๐ฒ๐ ๐๐ฎ๐ฟ๐๐ต ๐ถ๐ ๐ฏฬถ๐นฬถ๐ฬถ๐ฒฬถ ๐ฑ.๐ฐ๐ฌ๐ฑ ๐๐๐",
"raw": "๐ ๐ฎ๐ท๐ผ๐ฟ ๐ง๐ข๐ : ๐ฃ๐น๐ฎ๐ป๐ฒ๐ ๐๐ฎ๐ฟ๐๐ต ๐ถ๐ ๐ฏฬถ๐นฬถ๐ฬถ๐ฒฬถ ๐ฑ.๐ฐ๐ฌ๐ฑ ๐๐๐",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": nu... | ๐ ๐ฎ๐ท๐ผ๐ฟ ๐ง๐ข๐ : ๐ฃ๐น๐ฎ๐ป๐ฒ๐ ๐๐ฎ๐ฟ๐๐ต ๐ถ๐ ๐ฏฬถ๐นฬถ๐ฬถ๐ฒฬถ ๐ฑ.๐ฐ๐ฌ๐ฑ ๐๐๐
๐จ EXPANSION RELEASE: ๐ฆ๐ฒ๐ป๐๐ถ๐ป๐ฒ๐น-๐ญ ๐ถ๐ ๐ป๐ผ๐ ๐ฎ๐๐ฎ๐ถ๐น๐ฎ๐ฏ๐น๐ฒ in the MajorTOM-Core!
https://huggingface.co/datasets/Major-TOM/Core-S1RTC
๐ Together with @aliFrancis we've been racing to release the first official expansion to the Major TOM project.
MajorTOM-Core-S1RTC contains 1,469,955 of SAR images paired to Sentinel-2 images from Core-S2.
๐We cover more than 65% of the optical coverage with an average time shift of 7 days.
16 TB of radiometrically calibrated SAR imagery, available in the exact same format as the existing Major-TOM data.
๐บ๏ธ You can explore instantly in our viewing app:
https://huggingface.co/spaces/Major-TOM/MajorTOM-Core-Viewer
So, what now?
๐งฑ ๐๐จ๐ฆ๐ฆ๐ฎ๐ง๐ข๐ญ๐ฒ ๐๐ซ๐จ๐ฐ๐ญ๐ก: our community continues to grow! To coordinate the upcoming expansions as well as use cases of the open data, we will organise a meet up on 23 April, you can ๐ซ๐๐ ๐ข๐ฌ๐ญ๐๐ซ ๐ฒ๐จ๐ฎ๐ซ ๐ข๐ง๐ญ๐๐ซ๐๐ฌ๐ญ here: https://forms.gle/eBj8JvibJx9b6PLf9
๐ ๐๐ฉ๐๐ง ๐๐๐ญ๐ ๐๐จ๐ซ ๐๐ฉ๐๐ง ๐๐จ๐๐๐ฅ๐ฌ: Major-TOM Core dataset is currently supporting several strands of ongoing research within and outwith our lab and we are looking forward to the time when we can release models that take advantage of that data! https://huggingface.co/Major-TOM
๐ ๐๐จ๐ฌ๐ญ๐๐ซ ๐๐ญ ๐๐๐๐๐๐: We will present Major TOM project as a poster at IGARSS in Athens (July) - come talk to us if you're there! You can access the paper here: https://huggingface.co/papers/2402.12095
๐ Developed at European Space Agency ฮฆ-lab in partnership with Hugging Face | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1678741407493-6304c06eeb6d777a838eab63.png",
"fullname": "Mikolaj Czerkawski",
"name": "mikonvergence",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 25,
"isFollowing": false
} | [
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6304c06eeb6d777a838eab63/qZtkQ9CZ4Nrrh5PVLGQW_.mp4"
}
] | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/635011301d81beb8e2455ee9/NyDIbzavucEIyFDHnaAv0.jpeg",
"fullname": "Alistair Francis",
"name": "aliFrancis",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 21
}
] | [
{
"reaction": "๐ฅ",
"users": [
"aliFrancis",
"davanstrien",
"osanseviero",
"clem",
"Tom-Neverwinter",
"robmarkcole"
],
"count": 6
},
{
"reaction": "๐",
"users": [
"samusenps",
"clem",
"Tom-Neverwinter"
],
"count": 3
}
] | 2024-03-27T10:25:46.000Z | 2024-03-27T10:27:23.652Z | [] | /posts/mikonvergence/223529654460707 | 1,470 | 0 |
292464465490361 | [
{
"type": "text",
"value": "๐ Today's pick in Interpretability & Analysis of LMs: Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanisms by ",
"raw": "๐ Today's pick in Interpretability & Analysis of LMs: Have Faith in Faithfulness: Going Beyond Circuit Overlap When Fin... | ๐ Today's pick in Interpretability & Analysis of LMs: Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanisms by @mwhanna @sandropezzelle @belinkov
Edge attribution patching (EAP) is a circuit discovery technique using gradients to approximate the effects of causal intervening on each model edge. In the literature, its effectiveness is validated by comparing the overlap of its resulting circuits with those found via causal interventions (much more expensive).
This work:
1. Proposes a new method for faithful and efficient circuit discovery named edge attribution patching with integrated gradients (EAP-IG)
2. Evaluates the faithfulness EAP, EAP-IG and activation patching, i.e. whether behavior of the model remains consistent after all non-circuit edges are ablated.
3. Highlights that, while the no-overlap and full-overlap of EAP-like methods with activation patching results are generally good indicators of unfaithful and faithful (respectively) circuit identification, circuits with moderate overlap cannot generally assumed to be faithful to model behavior.
An advantage of EAP-IG is enabling the usage of KL-Divergence as a target for gradient propagation, which is not possible in the case of raw gradient-based EAP.
EAP-IG runtime is approximately similar to the one of EAP, with a small number of steps to approximate the gradient integral.
Importantly, circuit faithfulness does not imply completeness, i.e. whether all components participating towards a specific task were accounted for. This aspect is identified as interesting for future work.
๐ Paper: https://huggingface.co/papers/2403.17806
๐ All daily picks: https://huggingface.co/collections/gsarti/daily-picks-in-interpretability-and-analysis-of-lms-65ae3339949c5675d25de2f9 | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1670231290373-5e7749883d77a72421292d07.jpeg",
"fullname": "Gabriele Sarti",
"name": "gsarti",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 205,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/5e7749883d77a72421292d07/4_nMQgwlvRITNWQLO4w86.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/5e7749883d77a72421292d07/TXwT9VOLdivdURyZj6gXW.png"
},
{
"type": "image... | [
{
"avatarUrl": "/avatars/186a9aed84681246f48ed2a012c50def.svg",
"fullname": "Yonatan Belinkov",
"name": "belinkov",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 3
},
{
"avatarUrl": "/avatars/cee8dc17ab89b050a862dee09a0a0195.svg",
"fullna... | [
{
"reaction": "โค๏ธ",
"users": [
"javifer",
"osanseviero",
"samusenps",
"clem",
"giux78",
"mmhamdy",
"alemiaschi"
],
"count": 7
},
{
"reaction": "๐",
"users": [
"shiv2050"
],
"count": 1
}
] | 2024-03-27T09:34:06.000Z | 2024-03-27T09:34:06.994Z | [] | /posts/gsarti/292464465490361 | 1,261 | 0 |
285448939482538 | [
{
"type": "text",
"value": "Based on the work of ",
"raw": "Based on the work of ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "mention",
"value": null,
"raw": "@mrinaldi",
"href": null... | Based on the work of @mrinaldi and @ruggsea we just released the biggest - ready for training - conversational dataset based on Usenet data in the Italian language ๐ฎ๐น๐ฎ๐น๐ฎ๐น๐ฎ๐น๐ฎ๐น๐ฎ๐น๐ฎ๐น. It contains about 9 millions of conversations made by real humans.
https://huggingface.co/datasets/mii-community/UsenetArchiveIT-conversations
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/5fef4eb7770b06e11c2c6381/1NMdigjCGtn0yvQZSi5NJ.png",
"fullname": "Alessandro Ercolani",
"name": "giux78",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 44,
"isFollowing": false
} | [] | [
{
"avatarUrl": "/avatars/641876a24d4ee45dab0f9723d7b9e7f1.svg",
"fullname": "Matteo Rinaldi",
"name": "mrinaldi",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 3
},
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/635f... | [
{
"reaction": "๐",
"users": [
"mrinaldi",
"gsarti",
"fgamezf",
"osanseviero",
"giux78",
"merve",
"clem"
],
"count": 7
},
{
"reaction": "โค๏ธ",
"users": [
"gsarti",
"osanseviero",
"merve",
"samusenps",
"clem",
"g... | 2024-03-27T07:54:41.000Z | 2024-03-27T08:15:52.907Z | [] | /posts/giux78/285448939482538 | 1,285 | 0 |
826730657864057 | [
{
"type": "text",
"value": "Exciting news! ๐ I've created the OpenCerebrum datasets, open-source alternatives to Aether Research's proprietary Cerebrum dataset.",
"raw": "Exciting news! ๐ I've created the OpenCerebrum datasets, open-source alternatives to Aether Research's proprietary Cerebrum dataset... | Exciting news! ๐ I've created the OpenCerebrum datasets, open-source alternatives to Aether Research's proprietary Cerebrum dataset.
The first, OpenCerebrum SFT, is a text-generation and question-answering dataset with ~1.2M examples, curated from sources like Open-Orca, glaiveai, camel-ai, and more! ๐
The second, OpenCerebrum DPO, is a smaller dataset with ~21k examples, focusing on data point optimization. It's curated from sources like jondurbin, argilla, grimulkan, and others. ๐
Both datasets are licensed under Apache-2.0 and are available in English. They're ready for use in your projects, and I welcome any feedback for future improvements! ๐
https://huggingface.co/datasets/Locutusque/OpenCerebrum-dpo
https://huggingface.co/datasets/Locutusque/OpenCerebrum-SFT
https://huggingface.co/Locutusque/OpenCerebrum-1.0-7b-SFT
https://huggingface.co/Locutusque/OpenCerebrum-1.0-7b-DPO | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/YeFyz1AZVcCRsyNHHtwJG.jpeg",
"fullname": "Sebastian Gabarain",
"name": "Locutusque",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 180,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"afrideva",
"samusenps",
"ajibawa-2023",
"giux78",
"osanseviero",
"lewtun",
"dvilasuero",
"lhoestq",
"victor",
"Tom-Neverwinter",
"InferenceIllusionist",
"MexIvanov",
"clefourrier",
"mlabonne"
... | 2024-03-27T00:07:00.000Z | 2024-03-27T14:01:49.658Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/634262af8d8089ebaefd410e/pr6KcEebXTo5V2XAlpQNw.png",
"fullname": "Fizz ๐ณ๏ธโโง๏ธ",
"name": "Fizzarolli",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 49,
"isFollowing": false
... | /posts/Locutusque/826730657864057 | 2,639 | 5 |
736768871803868 | [
{
"type": "text",
"value": "Current LLMs are very susceptible to generating toxic, harmful and even dangerous content. They can also generate outputs with gender or racial biases.",
"raw": "Current LLMs are very susceptible to generating toxic, harmful and even dangerous content. They can also generate ... | Current LLMs are very susceptible to generating toxic, harmful and even dangerous content. They can also generate outputs with gender or racial biases.
The Biden-Harris Executive Order (https://www.federalregister.gov/documents/2023/11/01/2023-24283/safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence) sets forth guidelines on what is considered a safe AI system.
Following up on these guidelines, we present the world's first open source Biden-Harris Executive Order Red teamed Multilingual Language Model: Aurora-M.
The model is trained on 5 languages: English, Hindi, Japanese, Vietnamese and Finnish.
Blog: https://huggingface.co/blog/mayank-mishra/aurora
Paper coming out soon.
Base model: https://huggingface.co/aurora-m/aurora-m-base (not safety tuned)
Instruct model: https://huggingface.co/aurora-m/aurora-m-instruct (not safety tuned)
Red teamed model: https://huggingface.co/aurora-m/aurora-m-biden-harris-redteamed (safety tuned according to the order mentioned above) | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/62cd5057674cdb524450093d/f67rlrdsKPRTLdXCXoa_X.jpeg",
"fullname": "Mayank Mishra",
"name": "mayank-mishra",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 53,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"samusenps",
"oneiroid",
"clem",
"lunarflu",
"antiven0m",
"adamelliotfields",
"Joseph717171"
],
"count": 7
},
{
"reaction": "๐ฅ",
"users": [
"samusenps",
"mayank-mishra",
"clem",
"diwank",
... | 2024-03-26T23:17:27.000Z | 2024-03-27T06:32:58.201Z | [] | /posts/mayank-mishra/736768871803868 | 1,896 | 1 |
529620339719015 | [
{
"type": "text",
"value": "A new paper introduces Visual CoT, a new approach that enhances multi-modal large language models with visual chain-of-thought reasoning capabilities. This allows language models to dynamically identify and focus on specific regions within images that are most relevant for answer... | A new paper introduces Visual CoT, a new approach that enhances multi-modal large language models with visual chain-of-thought reasoning capabilities. This allows language models to dynamically identify and focus on specific regions within images that are most relevant for answering questions, mimicking human-like efficient visual reasoning.
Keypoints:
* Introduces the 373k Visual CoT dataset with bounding box annotations highlighting essential image regions
* Proposes a multi-turn pipeline for focusing on relevant visual inputs
* Achieves strong results on multi-modal benchmarks
Paper: https://huggingface.co/papers/2403.16999
Code, data and other resources: https://github.com/deepcs233/Visual-CoT
Congrats to the authors for their work! | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/657217faabb25ed8aedd5e48/UUHAXeGtOnQBXFD3nYtf2.jpeg",
"fullname": "Vlad Bogolin",
"name": "vladbogo",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 109,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"Csplk",
"osanseviero",
"thesouthfrog",
"Deping",
"sbrandeis"
],
"count": 6
},
{
"reaction": "โ",
"users": [
"samusenps",
"osanseviero"
],
"count": 2
}
] | 2024-03-26T22:20:59.000Z | 2024-03-26T22:20:59.712Z | [] | /posts/vladbogo/529620339719015 | 1,385 | 0 |
751873931507932 | [
{
"type": "text",
"value": "๐๐ผ๐ ๐ฑ๐ผ๐ฒ๐ ๐ฏ๐ฒ๐ฎ๐บ ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต ๐ฑ๐ฒ๐ฐ๐ผ๐ฑ๐ถ๐ป๐ด ๐๐ผ๐ฟ๐ธ? โก๏ธ ๐๐๐ฌ ๐ซ๐๐จ๐ช๐๐ก๐๐ฏ๐๐ฉ๐๐ค๐ฃ ๐ฉ๐ค๐ค๐ก! ๐",
"raw": "๐๐ผ๐ ๐ฑ๐ผ๐ฒ๐ ๐ฏ๐ฒ๐ฎ๐บ ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต ๐ฑ๐ฒ๐ฐ๐ผ๐ฑ๐ถ๐ป๐ด ๐๐ผ๐ฟ๐ธ? โก๏ธ ๐๐๐ฌ ๐ซ๐๐จ๐ช๐๐ก๐๐ฏ๐๐ฉ๐๐ค๐ฃ ๐ฉ๐ค๐ค๐ก! ๐",
"href": null,
"resource": ... | ๐๐ผ๐ ๐ฑ๐ผ๐ฒ๐ ๐ฏ๐ฒ๐ฎ๐บ ๐๐ฒ๐ฎ๐ฟ๐ฐ๐ต ๐ฑ๐ฒ๐ฐ๐ผ๐ฑ๐ถ๐ป๐ด ๐๐ผ๐ฟ๐ธ? โก๏ธ ๐๐๐ฌ ๐ซ๐๐จ๐ช๐๐ก๐๐ฏ๐๐ฉ๐๐ค๐ฃ ๐ฉ๐ค๐ค๐ก! ๐
In Decoder-type LLMs like GPT4 or Mistral-Large, the output is generated one token (=word part) at a time. That's why they're nicknamed "stochastic parrots": the "thinking" process only happens one step at a time, so it can seem really myopic.
๐๐จ ๐ก๐จ๐ฐ ๐ข๐ฌ ๐ญ๐ก๐ ๐ง๐๐ฑ๐ญ ๐ญ๐จ๐ค๐๐ง ๐ฌ๐๐ฅ๐๐๐ญ๐๐?
๐ Given its input sentence like "๐๐ฉ๐ข๐ต ๐ช๐ด ๐ต๐ฉ๐ฆ 7๐ต๐ฉ ๐๐ช๐ฃ๐ฐ๐ฏ๐ข๐ค๐ค๐ช ๐ฏ๐ถ๐ฎ๐ฃ๐ฆ๐ณ? ๐๐ฉ๐ฆ 7๐ต๐ฉ ๐๐ช๐ฃ๐ฐ๐ฏ๐ข๐ค๐ค๐ช ๐ฏ๐ถ๐ฎ๐ฃ๐ฆ๐ณ", the Decoder LLM generates, for each token in its vocabulary, a score that represents this token's probability of coming next.
For instance: "๐๐จ" gets score 0.56, and "๐๐๐ฃ" gets score 0.35.
๐ค ๐๐ซ๐๐๐๐ฒ ๐๐๐๐จ๐๐ข๐ง๐ is the naive option where you simply take the next most probable token at each step. But this creates paths that maximize very short-term rewards, thus may overlook better paths for the long term (like this time when you played FIFA all evening and arrived unprepared to your school exam on the next day).
In our example, the next highest score token might be "๐๐จ", but this will strongly bias the LLM towards giving an hasty response. On the opposite, starting with "๐๐๐ฃ" could have been completed with "๐ฃ๐ฆ ๐ฐ๐ฃ๐ต๐ข๐ช๐ฏ๐ฆ๐ฅ ๐ง๐ณ๐ฐ๐ฎ ๐ค๐ฐ๐ฎ๐ฑ๐ถ๐ต๐ช๐ฏ๐จ ๐ฑ๐ณ๐ฆ๐ท๐ช๐ฐ๐ถ๐ด ๐๐ช๐ฃ๐ฐ๐ฏ๐ข๐ค๐ค๐ช ๐ฏ๐ถ๐ฎ๐ฃ๐ฆ๐ณ๐ด ๐ง๐ช๐ณ๐ด๐ต", which steers the LLM towards a correct reasoning!
๐บ๏ธ ๐๐๐๐ฆ ๐ฌ๐๐๐ซ๐๐ก improves on greedy decoding by generating at each step several paths - called beams - instead of one. This allows the generation to explore a much larger space, thus find better completions. In our example, both the "๐๐จ" and the "๐๐๐ฃ" completion could be tested. โ
๐ I've created a tool to let you visualize it, thank you @joaogante for your great help!
๐๐ง๐ฎ ๐๐ฉ ๐๐๐ง๐: https://huggingface.co/spaces/m-ric/beam_search_visualizer | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/63d10d4e8eaa4831005e92b5/7p7-OmWM6PqqCs7ZStPGD.jpeg",
"fullname": "Aymeric Roucher",
"name": "m-ric",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 494,
"isFollowing": false
} | [] | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1641203017724-noauth.png",
"fullname": "Joao Gante",
"name": "joaogante",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 96
}
] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"matlok",
"joaogante",
"clefourrier",
"fieryTransition",
"Jafta",
"awagner-mainz"
],
"count": 7
},
{
"reaction": "๐ฅ",
"users": [
"osanseviero",
"Awal",
"mmhamdy",
"samusenps",
... | 2024-03-26T17:23:28.000Z | 2024-03-26T17:23:28.756Z | [] | /posts/m-ric/751873931507932 | 1,702 | 0 |
236311186560364 | [
{
"type": "text",
"value": "Very glad to welcome ",
"raw": "Very glad to welcome ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "mention",
"value": null,
"raw": "@josefprusa",
"href": nu... | Very glad to welcome @josefprusa, pioneer of 3D printing and open source hardware, founder of https://www.prusa3d.com/, to the HF Hub ๐
AI applied to 3D printing could be big. | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/5dd96eb166059660ed1ee413/NQtzmrDdbG0H8qkZvRyGk.jpeg",
"fullname": "Julien Chaumond",
"name": "julien-c",
"type": "user",
"isPro": true,
"isHf": true,
"isMod": false,
"followerCount": 1580,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/5dd96eb166059660ed1ee413/Buyqz5oR6wuj5-C9iFUkE.png"
}
] | [
{
"avatarUrl": "/avatars/533162c0048b18e983a1b220b82230e1.svg",
"fullname": "Josef Prลฏลกa",
"name": "josefprusa",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 1143
}
] | [
{
"reaction": "๐ฅ",
"users": [
"victor",
"London12345",
"satpalsr",
"diwank",
"ZennyKenny",
"radames",
"adamelliotfields",
"VictorSanh",
"mmhamdy",
"samusenps",
"pcuenq",
"kramp",
"osanseviero",
"stattmone",
"mdouglas"... | 2024-03-26T16:25:55.000Z | 2024-03-28T16:48:32.476Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1583857146757-5e67bdd61009063689407479.jpeg",
"fullname": "Clem ๐ค",
"name": "clem",
"type": "user",
"isPro": true,
"isHf": true,
"isMod": false,
"followerCount": 1763,
"isFollowing": false
}
] | /posts/julien-c/236311186560364 | 3,160 | 1 |
161743256497155 | [
{
"type": "text",
"value": "LLM Agent Operating System",
"raw": "LLM Agent Operating System",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": ... | LLM Agent Operating System
https://huggingface.co/papers/2403.16971
The integration and deployment of large language model (LLM)-based intelligent agents have been fraught with challenges that compromise their efficiency and efficacy. Among these issues are sub-optimal scheduling and resource allocation of agent requests over the LLM, the difficulties in maintaining context during interactions between agent and LLM, and the complexities inherent in integrating heterogeneous agents with different capabilities and specializations. The rapid increase of agent quantity and complexity further exacerbates these issues, often leading to bottlenecks and sub-optimal utilization of resources. Inspired by these challenges, this paper presents AIOS, an LLM agent operating system, which embeds large language model into operating systems (OS). Specifically, AIOS is designed to optimize resource allocation, facilitate context switch across agents, enable concurrent execution of agents, provide tool service for agents, and maintain access control for agents. We present the architecture of such an operating system, outline the core challenges it aims to resolve, and provide the basic design and implementation of the AIOS. Our experiments on concurrent execution of multiple agents demonstrate the reliability and efficiency of our AIOS modules. Through this, we aim to not only improve the performance and efficiency of LLM agents but also to pioneer for better development and deployment of the AIOS ecosystem in the future. | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1674929746905-60f1abe7544c2adfd699860c.jpeg",
"fullname": "AK",
"name": "akhaliq",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5205,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/60f1abe7544c2adfd699860c/kQUttU5GaHg-9HDxed9JT.png"
}
] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"QiushiSun",
"osanseviero",
"alielfilali01",
"AtAndDev",
"Jakaline"
],
"count": 6
},
{
"reaction": "๐",
"users": [
"diwank",
"AtAndDev",
"osanseviero"
],
"count": 3
},
{
"re... | 2024-03-26T15:32:12.000Z | 2024-03-27T06:54:25.215Z | [
{
"avatarUrl": "/avatars/88a391b315665c42bced2df0c300bf8d.svg",
"fullname": "Vishwas",
"name": "Vishwas1",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 5,
"isFollowing": false
},
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/prod... | /posts/akhaliq/161743256497155 | 2,191 | 3 |
428244381121381 | [
{
"type": "text",
"value": "MLX RAG with GGUF Models",
"raw": "MLX RAG with GGUF Models",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": null... | MLX RAG with GGUF Models
Minimal, clean code implementation of RAG with mlx inferencing for GGUF models.
Code: https://github.com/Jaykef/mlx-rag-gguf
The code here builds on vegaluisjose's example, it has been optimized to support RAG-based inferencing for .gguf models. I am using BAAI/bge-small-en for the embedding model, tinyllama-1.1b-chat-v1.0.Q4_0.gguf as base model and the custom vector database script for indexing texts in a pdf file. Inference speeds can go up to ~413 tokens/sec for prompts and ~36 tokens/sec for generation on my M2 Air.
Queries make use of both .gguf (base model) and .npz (retrieval model) simultaneouly resulting in much higher inferencing speeds. | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6438a9027de34e8ea7e4b257/vib8QSd1AWMr_bR9ig_xJ.jpeg",
"fullname": "Jaward Sesay",
"name": "Jaward",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 191,
"isFollowing": false
} | [
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6438a9027de34e8ea7e4b257/_VswNFRiMFxPPy3todCWu.mp4"
}
] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps"
],
"count": 1
}
] | 2024-03-26T13:38:01.000Z | 2024-03-26T13:38:01.882Z | [] | /posts/Jaward/428244381121381 | 1,429 | 0 |
481178346523516 | [
{
"type": "text",
"value": "Diaries of Open Source. Part 10 ๐",
"raw": "Diaries of Open Source. Part 10 ๐",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\... | Diaries of Open Source. Part 10 ๐
๐ผMarigold-LCM: A super fast SOTA Depth Estimator
Demo: https://hf.co/spaces/prs-eth/marigold-lcm
Original paper: https://hf.co/papers/2312.02145
Model: https://hf.co/prs-eth/marigold-lcm-v1-0
๐Quiet-STaR: A self-teaching technique via internal monologue
Paper: https://hf.co/papers/2403.09629
GitHub: https://github.com/ezelikman/quiet-star
Tweetutorial: https://twitter.com/ericzelikman/status/1768663835106513041
๐ผ๏ธ WebSight v0.2: A image-to-code dataset containing tailwind CSS, images in screenshots, and more!
Dataset: https://hf.co/datasets/HuggingFaceM4/WebSight
Paper: https://hf.co/papers/2403.09029
Blog: https://hf.co/blog/websight
๐ต๏ธAgent-FLAN - effective agent tuning for LLMs
Paper: https://hf.co/papers/2403.12881
Model: https://hf.co/internlm/Agent-FLAN-7b
Dataset: https://hf.co/datasets/internlm/Agent-FLAN
Website: https://internlm.github.io/Agent-FLAN/
๐ฅHPT, a family of multimodal LLMs from HyperGAI
Blog post: https://hypergai.com/blog/introducing-hpt-a-family-of-leading-multimodal-llms
Model: https://huggingface.co/HyperGAI/HPT
GitHub: https://github.com/hyperGAI/HPT
๐Models and datasets around the world
- Tess-70B, a MiQu-70B fine-tune with high-quality data https://hf.co/migtissera/Tess-70B-v1.6
- UNI, a model trained on 100 million pathology images from 100k+ slides https://hf.co/MahmoodLab/UNI
- CONCH, a VLM trained on 1.17 million pathology image-text pairs https://hf.co/MahmoodLab/CONCH | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ค",
"users": [
"Bingxin",
"toshas",
"DmitryRyumin",
"thomwolf",
"migtissera",
"julien-c",
"London12345",
"mmhamdy",
"samusenps",
"pcuenq",
"victor",
"sbrandeis"
],
"count": 12
},
{
"reaction": "โค๏ธ",
"... | 2024-03-26T11:48:14.000Z | 2024-03-28T05:40:19.342Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/658a2e76367c76b8ee481b5c/SF1IA8CeQBdc7w1cNR6c0.jpeg",
"fullname": "Bryan Alvarado Villalobos",
"name": "thebryanalvarado",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": null,
... | /posts/osanseviero/481178346523516 | 1,615 | 4 |
204103548247572 | [
{
"type": "text",
"value": "๐๐ญ๐ New Research Alert! ๐ ๐ญ๐",
"raw": "๐๐ญ๐ New Research Alert! ๐ ๐ญ๐",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\... | ๐๐ญ๐ New Research Alert! ๐ ๐ญ๐
๐ Title: FlashFace: Human Image Personalization with High-fidelity Identity Preservation ๐
๐ Description: FlashFace is a personalized photo editing tool that focuses on high-fidelity identity preservation and improved compliance through advanced encoding and integration strategies.
๐ฅ Authors: Shilong Zhang, Lianghua Huang, @xichenhku et al.
๐ Paper: https://huggingface.co/papers/2403.17008
๐ Github Page: https://jshilong.github.io/flashface-page
๐ More Papers: more cutting-edge research presented at other conferences in the https://huggingface.co/spaces/DmitryRyumin/NewEraAI-Papers curated by @DmitryRyumin
๐ Added to the Avatars Collection: https://huggingface.co/collections/DmitryRyumin/avatars-65df37cdf81fec13d4dbac36
๐ Keywords: #FlashFace #Personalization #HighFidelityIdentity #DeepLearning #Innovation | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/o85gNpS2Ra-YYn8QDHZFV.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/F9ZnPt1ITuZ51rD0v0XTg.png"
},
{
"type": "image... | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377
},
{
"avatarUrl": "https://cdn... | [
{
"reaction": "๐ฅ",
"users": [
"DmitryRyumin",
"waqas700297",
"kgourgou",
"London12345",
"samusenps",
"KvrParaskevi",
"whooray"
],
"count": 7
},
{
"reaction": "๐",
"users": [
"dashfunnydashdash",
"loveisp",
"AtonMountlook",
... | 2024-03-26T08:40:00.000Z | 2024-03-26T08:41:23.213Z | [] | /posts/DmitryRyumin/204103548247572 | 1,602 | 0 |
798001054518351 | [
{
"type": "text",
"value": "Now You Can Full Fine Tune / DreamBooth Stable Diffusion XL (SDXL) with only 10.3 GB VRAM via OneTrainer โ Both U-NET and Text Encoder 1 is trained โ Compared 14 GB config vs slower 10.3 GB Config",
"raw": "Now You Can Full Fine Tune / DreamBooth Stable Diffusion XL (SDXL) wi... | Now You Can Full Fine Tune / DreamBooth Stable Diffusion XL (SDXL) with only 10.3 GB VRAM via OneTrainer โ Both U-NET and Text Encoder 1 is trained โ Compared 14 GB config vs slower 10.3 GB Config
Full config and instructions are shared here : https://www.patreon.com/posts/96028218
Used SG161222/RealVisXL_V4.0 as a base model and OneTrainer to train on Windows 10 : https://github.com/Nerogar/OneTrainer
The posted example x/y/z checkpoint comparison images are not cherry picked. So I can get perfect images with multiple tries.
Trained 150 epochs, 15 images and used my ground truth 5200 regularization images : https://www.patreon.com/posts/massive-4k-woman-87700469
In each epoch only 15 of regularization images used to make DreamBooth training affect
As a caption only โohwx manโ is used, for regularization images just โmanโ
You can download configs and full instructions here : https://www.patreon.com/posts/96028218
Hopefully full public tutorial coming within 2 weeks. I will show all configuration as well
The tutorial will be on our channel : https://www.youtube.com/SECourses
Training speeds are as below thus durations:
RTX 3060 โ slow preset : 3.72 second / it thus 15 train images 150 epoch 2 (reg images concept) : 4500 steps = 4500 3.72 / 3600 = 4.6 hours
RTX 3090 TI โ slow preset : 1.58 second / it thus : 4500 * 1.58 / 3600 = 2 hours
RTX 3090 TI โ fast preset : 1.45 second / it thus : 4500 * 1.45 / 3600 = 1.8 hours
A quick tutorial for how to use concepts in OneTrainer : https://youtu.be/yPOadldf6bI
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1672531901326-6345bd89fe134dfd7a0dba40.png",
"fullname": "Furkan Gรถzรผkara",
"name": "MonsterMMORPG",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 376,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6345bd89fe134dfd7a0dba40/3ZiQ52_CdZxKFiXGw-eAs.jpeg"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6345bd89fe134dfd7a0dba40/bgk0UO27_wx8VDV-3LHsi.jpeg"
},
{
"type": "ima... | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"radames",
"Reza2kn",
"Stephane97480",
"osanseviero",
"Zazalael",
"seyf1elislam",
"MaziyarPanahi"
],
"count": 8
},
{
"reaction": "๐",
"users": [
"waqas700297"
],
"count": 1
}
] | 2024-03-26T04:36:46.000Z | 2024-04-02T03:20:22.108Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1672531901326-6345bd89fe134dfd7a0dba40.png",
"fullname": "Furkan Gรถzรผkara",
"name": "MonsterMMORPG",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 376,
"isFollowing": false
... | /posts/MonsterMMORPG/798001054518351 | 2,047 | 5 |
592209532314704 | [
{
"type": "text",
"value": "The AI Comic Factory 1.2 has been released a few hours ago!",
"raw": "The AI Comic Factory 1.2 has been released a few hours ago!",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"t... | The AI Comic Factory 1.2 has been released a few hours ago!
You can now tell the LLM in advance how many pages you will want, for richer and deeper stories
https://www.loom.com/share/b8b2d55fc60249a78df60adaa2673d2f
I the future I would like to improve how layouts are attributed to pages ( seeing the same layout 6x can be a bit boring)
Also, I'm still waiting for some new improved AI models to drop, so stay tuned!
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/2RK8J_YSNAK2ob8XZH7w2.jpeg",
"fullname": "Julian Bilcke",
"name": "jbilcke-hf",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 1312,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"osanseviero",
"victor",
"halbarba",
"musfiqdehan",
"szh",
"clem",
"samusenps",
"ajibawa-2023",
"radames",
"kramp",
"julien-c",
"Krass1",
"kuramacn",
"Handoman",
"tschacht",
"Chris"... | 2024-03-25T21:38:18.000Z | 2024-05-10T21:03:45.272Z | [
{
"avatarUrl": "/avatars/f4cdeeaf5817bac68fbbf6f1cd6f2a3c.svg",
"fullname": "JACK",
"name": "JACKTHERAPPA",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": null,
"isFollowing": false
},
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/... | /posts/jbilcke-hf/592209532314704 | 8,100 | 2 |
242385169623874 | [
{
"type": "text",
"value": "Introducing Marigold-LCM ๐ผ โ a FAST version of the now popular state-of-the-art depth estimator! Thanks to the latent consistency distillation, it retains the precision of the original Marigold but reaches the solution in just a few steps! ",
"raw": "Introducing Marigold-LCM... | Introducing Marigold-LCM ๐ผ โ a FAST version of the now popular state-of-the-art depth estimator! Thanks to the latent consistency distillation, it retains the precision of the original Marigold but reaches the solution in just a few steps!
Check out the teaser video attached below and play with the new demo - it accepts videos now! Also, meet the new team member: Tianfu Wang (@Tianfwang)
๐ค Demo: https://huggingface.co/spaces/prs-eth/marigold-lcm
๐ค Model: https://huggingface.co/prs-eth/marigold-lcm-v1-0
๐ค Original Marigold post: https://huggingface.co/posts/toshas/656973498012745
๐ค Paper: https://huggingface.co/papers/2312.02145
๐ Website: https://marigoldmonodepth.github.io
๐พ Code: https://github.com/prs-eth/marigold
๐พ Code: pip install diffusers | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/62f93abbc4817cfc0756b6f8/rGYLaq-rmoJJYotkC1VXk.jpeg",
"fullname": "Anton Obukhov",
"name": "toshas",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 68,
"isFollowing": false
} | [
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/62f93abbc4817cfc0756b6f8/LZEpbAbimQH6JGca_5WO9.mp4"
}
] | [
{
"avatarUrl": "/avatars/a06aecf91bcc005e618460ab31228fb7.svg",
"fullname": "Tianfu Wang",
"name": "Tianfwang",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 3
}
] | [
{
"reaction": "๐ฅ",
"users": [
"DmitryRyumin",
"samusenps",
"osanseviero",
"clem",
"radames",
"Veneco"
],
"count": 6
},
{
"reaction": "๐คฏ",
"users": [
"DmitryRyumin",
"osanseviero",
"clem"
],
"count": 3
}
] | 2024-03-25T18:23:23.000Z | 2024-03-25T19:39:03.285Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1654292443715-noauth.png",
"fullname": "Xin Yin",
"name": "Kache",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 1,
"isFollowing": false
}
] | /posts/toshas/242385169623874 | 1,977 | 1 |
580395077003079 | [
{
"type": "text",
"value": "๐ Just released version 0.22.0 of the ",
"raw": "๐ Just released version 0.22.0 of the ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "inline_code",
"value": null,
... | ๐ Just released version 0.22.0 of the `huggingface_hub` Python library!
Exciting updates include:
โจ Chat-completion API in the InferenceClient!
๐ค Official inference types in InferenceClient!
๐งฉ Better config and tags in `ModelHubMixin`!
๐ Generate model cards for your `ModelHubMixin` integrations!
๐๏ธ x3 download speed in `HfFileSystem`!!
Check out the full release notes for more details: https://huggingface.co/spaces/Wauplin/huggingface_hub/discussions/5 ๐ | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1659336880158-6273f303f6d63a28483fde12.png",
"fullname": "Lucain Pouget",
"name": "Wauplin",
"type": "user",
"isPro": true,
"isHf": true,
"isMod": false,
"followerCount": 157,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ฅ",
"users": [
"pierrci",
"raucheacho",
"cbensimon",
"clefourrier",
"victor",
"trysem",
"nielsr",
"monsoon-nlp",
"samusenps",
"osanseviero",
"mmhamdy",
"clem",
"minhdang",
"ajibawa-2023",
"julien-c",
... | 2024-03-25T14:10:32.000Z | 2024-03-28T08:13:06.344Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6527e89a8808d80ccff88b7a/CuGNmF1Et8KMQ0mCd1NEJ.jpeg",
"fullname": "Lain",
"name": "not-lain",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 941,
"isFollowing": false
},
... | /posts/Wauplin/580395077003079 | 2,323 | 2 |
402964363815778 | [
{
"type": "text",
"value": "Just released a notebook showing how to finetune moondream: ",
"raw": "Just released a notebook showing how to finetune moondream: ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
... | Just released a notebook showing how to finetune moondream: https://github.com/vikhyat/moondream/blob/main/notebooks/Finetuning.ipynb
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/63117568fa95534e218da163/8h9zN8aKRxPLBnXW7sqY9.jpeg",
"fullname": "Vik Korrapati",
"name": "vikhyatk",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 375,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"NickyNicky",
"samusenps",
"peaceAsh",
"musfiqdehan",
"Norod78",
"osanseviero",
"ajibawa-2023",
"Tom-Neverwinter",
"clem",
"abdullah",
"damerajee",
"t1u1",
"notune",
"not-lain",
"antiven0... | 2024-03-25T04:29:04.000Z | 2024-03-25T04:29:04.435Z | [] | /posts/vikhyatk/402964363815778 | 3,689 | 0 |
563404579814446 | [
{
"type": "text",
"value": "Diaries of Open Source. Part 9!",
"raw": "Diaries of Open Source. Part 9!",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
... | Diaries of Open Source. Part 9!
โฐAmazon releases Chronos, a family of models for time series
Base model: https://hf.co/amazon/chronos-t5-large
Paper: https://hf.co/papers/2403.07815
Models: https://huggingface.co/collections/amazon/chronos-models-65f1791d630a8d57cb718444
๐กORPO Alignment: align without a reference model nor SFT!
Paper: https://hf.co/papers/2403.07691
Models: https://hf.co/collections/kaist-ai/orpo-65efef87544ba100aef30013
GitHub: https://github.com/xfactlab/orpo
๐บ๐ณCohere releases 250M Wikipedia Embeddings in 300+ languages
Data: https://hf.co/datasets/Cohere/wikipedia-2023-11-embed-multilingual-v3
Announcement: https://twitter.com/Nils_Reimers/status/1767891859207057618
๐งฌSegmentNT: a LLM for annotating DNA at single nucleotide resolution
Models: https://huggingface.co/collections/InstaDeepAI/segmentnt-65eb4941c57808b4a3fe1319
GitHub repo: https://github.com/instadeepai/nucleotide-transformer
Paper: https://www.biorxiv.org/content/10.1101/2024.03.14.584712v1
๐DynamiCrafter: video generation models for interpolation and looping are out!
Project page: https://doubiiu.github.io/projects/DynamiCrafter/
GitHub: https://github.com/Doubiiu/DynamiCrafter
Demo: https://hf.co/spaces/Doubiiu/DynamiCrafter_interp_loop
๐Stanford releases Anticipatory Music Transformer:
GitHub: https://github.com/jthickstun/anticipation/
Models: https://hf.co/stanford-crfm
Original blog announcement: https://crfm.stanford.edu/2023/06/16/anticipatory-music-transformer.html | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ฅ",
"users": [
"YaTharThShaRma999",
"giux78",
"DmitryRyumin",
"mmhamdy",
"samusenps",
"clem"
],
"count": 6
},
{
"reaction": "๐",
"users": [
"FlipTip",
"samusenps",
"avinash02",
"clem"
],
"count": 4
}
] | 2024-03-24T13:29:30.000Z | 2024-03-24T13:31:48.176Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing":... | /posts/osanseviero/563404579814446 | 3,268 | 2 |
997723311930747 | [
{
"type": "text",
"value": "I and ",
"raw": "I and ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "mention",
"value": null,
"raw": "@FinancialSupport",
"href": null,
"resource": null... | I and @FinancialSupport just release the new Italian Leaderboard:
https://huggingface.co/spaces/FinancialSupport/open_ita_llm_leaderboard
It is based on lm-evaluation-harness and at the moment , mainly, on 7 billion models. In the next weeks we will add more models. If you have suggestion or need explanations join our community discord https://discord.gg/a26cRkBCNH | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/5fef4eb7770b06e11c2c6381/1NMdigjCGtn0yvQZSi5NJ.png",
"fullname": "Alessandro Ercolani",
"name": "giux78",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 44,
"isFollowing": false
} | [] | [
{
"avatarUrl": "/avatars/83d446ebfeafb54b8e2bcfcfb57f13b7.svg",
"fullname": "Samuele Colombo",
"name": "FinancialSupport",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 20
}
] | [
{
"reaction": "๐",
"users": [
"osanseviero",
"victor",
"giux78",
"clem"
],
"count": 4
},
{
"reaction": "โค๏ธ",
"users": [
"clefourrier",
"samusenps",
"clem",
"giux78"
],
"count": 4
}
] | 2024-03-24T12:46:56.000Z | 2024-03-27T06:18:49.909Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1644340617257-noauth.png",
"fullname": "Clรฉmentine Fourrier",
"name": "clefourrier",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 459,
"isFollowing": false
},
{
"ava... | /posts/giux78/997723311930747 | 2,661 | 9 |
595207139085023 | [
{
"type": "text",
"value": "๐๐๐ New Research Alert - ICASSP 2024! ๐ ๐บ๐",
"raw": "๐๐๐ New Research Alert - ICASSP 2024! ๐ ๐บ๐",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
... | ๐๐๐ New Research Alert - ICASSP 2024! ๐ ๐บ๐
๐ Title: Text2Avatar: Text to 3D Human Avatar Generation with Codebook-Driven Body Controllable Attribute ๐
๐ Description: Text2Avatar is a novel approach that can generate realistic 3D human avatars directly from textual descriptions, enabling multi-attribute control and realistic styling, overcoming the challenges of feature coupling and data scarcity in this domain.
๐ฅ Authors: Chaoqun Gong et al.
๐
Conference: ICASSP, 14-19 April 2024 | Seoul, Korea ๐ฐ๐ท
๐ Paper: https://huggingface.co/papers/2401.00711
๐ Github Page: https://iecqgong.github.io/text2avatar/
๐ More Papers: more cutting-edge research presented at other conferences in the https://huggingface.co/spaces/DmitryRyumin/NewEraAI-Papers curated by @DmitryRyumin
๐ Added to the Avatars Collection: https://huggingface.co/collections/DmitryRyumin/avatars-65df37cdf81fec13d4dbac36
๐ Added to the ICASSP-2023-24-Papers: https://github.com/DmitryRyumin/ICASSP-2023-24-Papers
๐ Keywords: #AvatarGeneration #Text2Avatar #ICASSP2024 #DeepLearning #Innovation | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/Xfa-kUaLVpLVIjN7MOrx6.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/CyHv9W4SAUMsgpAhTYj6o.png"
},
{
"type": "image... | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377
}
] | [
{
"reaction": "๐",
"users": [
"DmitryRyumin",
"samusenps",
"xenagarage",
"Dlbk",
"palloo",
"victor",
"Tom-Neverwinter"
],
"count": 7
},
{
"reaction": "๐ฅ",
"users": [
"palloo",
"GaneshMystic",
"clem",
"Tom-Neverwinter"
... | 2024-03-23T21:10:54.000Z | 2024-03-27T23:58:47.831Z | [
{
"avatarUrl": "/avatars/afbc48df2e8c47c35be48168113d83c0.svg",
"fullname": "s",
"name": "Tom-Neverwinter",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2,
"isFollowing": false
}
] | /posts/DmitryRyumin/595207139085023 | 2,675 | 1 |
901395976811277 | [
{
"type": "text",
"value": "Diaries of Open Source. Part 8!",
"raw": "Diaries of Open Source. Part 8!",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
... | Diaries of Open Source. Part 8!
๐คฏCRM: Image-to-3D Textured Mesh
Demo: https://hf.co/spaces/Zhengyi/CRM
Model: https://hf.co/Zhengyi/CRM
Project page: https://ml.cs.tsinghua.edu.cn/~zhengyi/CRM/
Paper: https://huggingface.co/papers/2403.05034
๐คHalf Quadratic Quantization: super-fast quantization of very large models
Blog post: https://mobiusml.github.io/hqq_blog/
Colab: https://colab.research.google.com/drive/1cG_5R_u9q53Uond7F0JEdliwvoeeaXVN?usp=sharing
Repo: https://github.com/mobiusml/hqq
๐คGemMoE -Gemma + MoE
Model: https://hf.co/Crystalcareai/GemMoE-Base-Random
Collection: https://huggingface.co/collections/Crystalcareai/gemmoe-65f11f4922af97ebe9943591
๐VeCLIP and MOFI, new 0-shot and image retrieval models by Apple, are now open-source!
GitHub: https://github.com/apple/ml-veclip/ and https://github.com/apple/ml-mofi
VeCLIP paper: https://hf.co/papers/2310.07699
MOFI paper: https://hf.co/papers/2306.07952
โกSPIN: Recipe for alignment with very little data
Collection: https://hf.co/collections/argilla/dibt-prompt-collective-spin-65ef59062518776024395fc3
Tweetutorial: https://twitter.com/argilla_io/status/1767608154697699455
๐ViT Prisma - an interoperability library for vision models
GitHub: https://github.com/soniajoseph/ViT-Prisma
โOpenLRM: full model and training code are open-sourced
Codebase: https://github.com/3DTopia/OpenLRM
Demo: https://hf.co/spaces/zxhezexin/OpenLRM
Models: https://huggingface.co/zxhezexin
โ๏ธOxford releases an extensive PEFT evaluation for bio models
Model: https://hf.co/NTaylor/bio-mobilebert-mimic-mp-lora
GitHub: https://github.com/nlpie-research/efficient-ml
Paper: https://hf.co/papers/2402.10597
๐Data and models around the world
Hermes 2 Pro 7B: an upgraded Nous Hermes 2 model with strong function calling and JSON capabilities https://hf.co/NousResearch/Hermes-2-Pro-Mistral-7B
Navarasa-2.0โ: Gemma fine-tuned in 15 indian language https://hf.co/collections/Telugu-LLM-Labs/navarasa-65f5e6ffdf29f02c6d7767ce | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"dvilasuero",
"DmitryRyumin",
"AugustYuYang",
"samusenps",
"ajibawa-2023",
"DonAmit197",
"Zhengyi",
"victor",
"clem",
"radames"
],
"count": 10
},
{
"reaction": "๐",
"users": [
"dvilasuero",
... | 2024-03-23T11:28:50.000Z | 2024-03-24T07:19:59.923Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing":... | /posts/osanseviero/901395976811277 | 2,527 | 4 |
869933859098017 | [
{
"type": "text",
"value": "Does anyone have experience with finetuning Gemma? Even the 2B variant feels more memory heavy than mistral 7B. I know that its vocabulary is much larger (250k) but I'm a bit surprised that the max batch size that I can get in an A100 80GB is only 2 whereas I could fit 4 with mis... | Does anyone have experience with finetuning Gemma? Even the 2B variant feels more memory heavy than mistral 7B. I know that its vocabulary is much larger (250k) but I'm a bit surprised that the max batch size that I can get in an A100 80GB is only 2 whereas I could fit 4 with mistral 7B - even though Gemma is much smaller except for the embedding layer. Both runs were using FA, same sequence length, same deepspeed zero 3 settings. Oh and yes I'm using the most recent hot fix of transformers that solves a memory issue with Gemma and others.
Any prior experience that you can share or suggestions to improve throughout? | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1594192845975-5e1e17b6fcf41d740b6996a8.jpeg",
"fullname": "Bram Vanroy",
"name": "BramVanroy",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 173,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"KnutJaegersberg",
"victor",
"DonAmit197",
"svernek",
"FlipTip",
"soates",
"clem",
"Andcircle"
],
"count": 8
},
{
"reaction": "๐",
"users": [
"osanseviero",
"samusenps",
"DonAmit197",
... | 2024-03-23T08:39:50.000Z | 2024-03-23T12:31:08.593Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1669551186189-63732ebbbd81fae2b3aaf3fb.jpeg",
"fullname": "Knut Jรคgersberg",
"name": "KnutJaegersberg",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 238,
"isFollowing": fal... | /posts/BramVanroy/869933859098017 | 2,391 | 4 |
626553999371491 | [
{
"type": "text",
"value": "Today we launch our space in colab with ",
"raw": "Today we launch our space in colab with ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "mention",
"value": null,
... | Today we launch our space in colab with @dvilasuero & @davanstrien so you can help us translate/correct our curated prompt dataset, that will be used to evaluate the performance of Arabic LLMs laterย and help our community to identify how open models perform on Arabic.
How to Get Involved?
1. Visit our Argilla Space and start reviewing prompts.
https://2a2i-prompt-translation-for-arabic.hf.space/
2. Join our Discord channel in the HuggingFace's discord server to connect with the community and share your insights.
https://discord.com/channels/879548962464493619/1217179730869096448 | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/626237d9bbcbd1c34f1bb231/EJrOjvAL-68qMCYdnvOrq.png",
"fullname": "Ali El Filali",
"name": "alielfilali01",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 186,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/626237d9bbcbd1c34f1bb231/m8yh5bf9KztFIsON6irr_.jpeg"
}
] | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1627505688463-60107b385ac3e86b3ea4fc34.jpeg",
"fullname": "Daniel van Strien",
"name": "davanstrien",
"type": "user",
"isPro": true,
"isHf": true,
"isMod": false,
"followerCount": 410
},
{
"avatarUrl": "... | [
{
"reaction": "โค๏ธ",
"users": [
"medmac01",
"dvilasuero",
"samusenps",
"ZennyKenny",
"SudhaSG",
"osanseviero",
"monsoon-nlp",
"davanstrien",
"clefourrier",
"not-lain",
"KBayoud",
"clem"
],
"count": 12
},
{
"reaction": "๐ฅ... | 2024-03-23T03:08:34.000Z | 2024-03-23T15:33:24.438Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/60420dccc15e823a685f2b03/Dn7QTyy9SZ7jKN6xpufVD.png",
"fullname": "Daniel Vila",
"name": "dvilasuero",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 231,
"isFollowing": false
... | /posts/alielfilali01/626553999371491 | 2,209 | 3 |
333899475285572 | [
{
"type": "text",
"value": "MathVerse",
"raw": "MathVerse",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": null,
"resource": null,
"u... | MathVerse
Does Your Multi-modal LLM Truly See the Diagrams in Visual Math Problems?
https://huggingface.co/papers/2403.14624
The remarkable progress of Multi-modal Large Language Models (MLLMs) has garnered unparalleled attention, due to their superior performance in visual contexts. However, their capabilities in visual math problem-solving remain insufficiently evaluated and understood. We investigate current benchmarks to incorporate excessive visual content within textual questions, which potentially assist MLLMs in deducing answers without truly interpreting the input diagrams. To this end, we introduce MathVerse, an all-around visual math benchmark designed for an equitable and in-depth evaluation of MLLMs. We meticulously collect 2,612 high-quality, multi-subject math problems with diagrams from publicly available sources. Each problem is then transformed by human annotators into six distinct versions, each offering varying degrees of information content in multi-modality, contributing to 15K test samples in total. This approach allows MathVerse to comprehensively assess whether and how much MLLMs can truly understand the visual diagrams for mathematical reasoning. In addition, we propose a Chain-of-Thought (CoT) evaluation strategy for a fine-grained assessment of the output answers. Rather than naively judging True or False, we employ GPT-4(V) to adaptively extract crucial reasoning steps, and then score each step with detailed error analysis, which can reveal the intermediate CoT reasoning quality by MLLMs. We hope the MathVerse benchmark may provide unique insights to guide the future development of MLLMs.
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1674929746905-60f1abe7544c2adfd699860c.jpeg",
"fullname": "AK",
"name": "akhaliq",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5205,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/60f1abe7544c2adfd699860c/kBHW9hfwHmPsS0ieEQ3Q4.png"
}
] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"samusenps",
"AtAndDev",
"shvhio",
"jncarlo",
"thomwolf"
],
"count": 5
}
] | 2024-03-22T20:21:53.000Z | 2024-03-22T20:21:53.738Z | [] | /posts/akhaliq/333899475285572 | 2,898 | 0 |
619817116350533 | [
{
"type": "text",
"value": "๐๐ญ๐ New Research Alert - CVPR 2024! ๐ ๐ญ๐",
"raw": "๐๐ญ๐ New Research Alert - CVPR 2024! ๐ ๐ญ๐",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"val... | ๐๐ญ๐ New Research Alert - CVPR 2024! ๐ ๐ญ๐
๐ Title: GaussianAvatars: Photorealistic Head Avatars with Rigged 3D Gaussians ๐
๐ Description: GaussianAvatars proposes a novel method for creating photorealistic and fully controllable head avatars by combining a parametric morphable face model with a dynamic 3D representation based on rigged 3D Gaussian splats, enabling high-quality rendering and precise animation control.
๐ฅ Authors: Shenhan Qian, Tobias Kirschstein, Liam Schoneveld, Davide Davoli, Simon Giebenhain, Matthias Nieรner
๐
Conference: CVPR, Jun 17-21, 2024 | Seattle WA, USA ๐บ๐ธ
๐ Paper: https://huggingface.co/papers/2312.02069
๐ Github Page: https://shenhanqian.github.io/gaussian-avatars
๐ Repository: https://github.com/ShenhanQian/GaussianAvatars
๐บ Video: https://www.youtube.com/watch?v=lVEY78RwU_I
๐ More Papers: more cutting-edge research presented at other conferences in the https://huggingface.co/spaces/DmitryRyumin/NewEraAI-Papers curated by @DmitryRyumin
๐ Added to the Avatars Collection: https://huggingface.co/collections/DmitryRyumin/avatars-65df37cdf81fec13d4dbac36
๐ Keywords: #HeadAvatar #GaussianAvatars #DynamicGaussians #3DModeling #AvatarGeneration #CVPR2024 #DeepLearning #Innovation | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/J1e1HBKKiRJ1xGHaq3cFH.png"
},
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/Ifoi55KR5Er2vG4IZy2aV.mp4"
},
{
"type": "video... | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377
}
] | [
{
"reaction": "๐ฅ",
"users": [
"DmitryRyumin",
"samusenps",
"xenagarage",
"wickedPixel",
"clem",
"victor",
"pratikjadhav",
"jonasmaltebecker"
],
"count": 8
},
{
"reaction": "๐คฏ",
"users": [
"DmitryRyumin",
"AngryBear9019",
... | 2024-03-22T19:35:16.000Z | 2024-03-22T19:35:16.001Z | [] | /posts/DmitryRyumin/619817116350533 | 1,861 | 0 |
376887867127881 | [
{
"type": "text",
"value": "๐
Quantized Embeddings are here! Unlike model quantization, embedding quantization is a post-processing step for embeddings that converts e.g. ",
"raw": "๐
Quantized Embeddings are here! Unlike model quantization, embedding quantization is a post-processing step for embeddi... | ๐
Quantized Embeddings are here! Unlike model quantization, embedding quantization is a post-processing step for embeddings that converts e.g. `float32` embeddings to binary or `int8` embeddings. This saves 32x or 4x memory & disk space, and these embeddings are much easier to compare!
Our results show 25-45x speedups in retrieval compared to full-size embeddings, while keeping 96% of the performance!
Learn more about it in our blogpost in collaboration with mixedbread.ai: https://huggingface.co/blog/embedding-quantization
Or try out our demo where we use quantized embeddings to let you search all of Wikipedia (yes, 41,000,000 texts) in 1 second on a CPU Space: https://huggingface.co/spaces/sentence-transformers/quantized-retrieval | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6317233cc92fd6fee317e030/cJHSvvimr1kqgQfHOjO5n.png",
"fullname": "Tom Aarsen",
"name": "tomaarsen",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 1060,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ฅ",
"users": [
"jaggernaut007",
"snowdere",
"osanseviero",
"AtAndDev",
"cnmoro",
"mohammedbriman",
"clem",
"pascalnjue",
"louisbrulenaudet",
"iamrobotbear",
"mrdbourke",
"danielpark",
"MichelleRuwen"
],
"co... | 2024-03-22T16:58:56.000Z | 2024-03-22T23:41:30.051Z | [
{
"avatarUrl": "/avatars/b2b429dd81d366b42a74a36f8e237286.svg",
"fullname": "D",
"name": "snowdere",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 1,
"isFollowing": false
}
] | /posts/tomaarsen/376887867127881 | 3,008 | 1 |
283998571804778 | [
{
"type": "text",
"value": "๐ฅ๐๐๐ฅ Dataset Drop: 4 KTO signal transformed versions of the highly loved Argilla DPO datasets.",
"raw": "๐ฅ๐๐๐ฅ Dataset Drop: 4 KTO signal transformed versions of the highly loved Argilla DPO datasets.",
"href": null,
"resource": null,
"url": null,
"code... | ๐ฅ๐๐๐ฅ Dataset Drop: 4 KTO signal transformed versions of the highly loved Argilla DPO datasets.
KTO formats for:
- UltraFeedback Cleaned Binarized
- Distilabel Intel Orca
- Distilabel Capybara
- DPO mix
https://huggingface.co/collections/argilla/preference-datasets-for-kto-65f98314d7c1b04ab54d41a7
Paper claims :)
https://arxiv.org/abs/2402.01306
KTO matches or exceeds DPO performance at scales from 1B to 30B parameters.1 That is, taking a preference dataset of n DPO pairs and breaking it up into 2n examples for KTO can yield better generations, despite the model ostensibly learning from a weaker signal.
KTO can handle extreme data imbalances, matching DPO performance while using up to 90% fewer desirable examples (i.e., examples of good generations). Its success thus cannot be ascribed to the alignment data being sourced from a preference dataset.
When the pretrained model is sufficiently good, one can skip supervised finetuning and go straight to KTO without a loss in generation quality. In contrast, we find that without doing SFT first, DPO-aligned models are significantly worse at all scales.
Do you need something custom? Take a look at @davanstrien his guide on creating your own KTO dataset with Argilla and our community.
https://github.com/huggingface/data-is-better-together/tree/main/kto-preference | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1677141720071-634ff41ff32062e9eb7b06a3.jpeg",
"fullname": "David Berenstein",
"name": "davidberenstein1957",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 167,
"isFollowing": false
} | [] | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1627505688463-60107b385ac3e86b3ea4fc34.jpeg",
"fullname": "Daniel van Strien",
"name": "davanstrien",
"type": "user",
"isPro": true,
"isHf": true,
"isMod": false,
"followerCount": 410
}
] | [
{
"reaction": "๐ฅ",
"users": [
"davanstrien",
"ZennyKenny",
"osanseviero",
"giux78",
"clem"
],
"count": 5
},
{
"reaction": "โค๏ธ",
"users": [
"davanstrien",
"samusenps",
"osanseviero",
"clem"
],
"count": 4
}
] | 2024-03-22T16:57:28.000Z | 2024-03-22T16:57:28.693Z | [] | /posts/davidberenstein1957/283998571804778 | 1,625 | 0 |
896443101397392 | [
{
"type": "text",
"value": "๐ค PEFT v0.10.0 release! ๐ฅ๐โจ",
"raw": "๐ค PEFT v0.10.0 release! ๐ฅ๐โจ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
... | ๐ค PEFT v0.10.0 release! ๐ฅ๐โจ
Some highli๐ghts:
1. FSDP+QLoRA and DeepSpeed Stage-3+QLoRA
2. Layer expansion + LoRA
3. DoRA support for Conv2D layers and quantized bitsandbytes layers
4. New LoftQ utility
5. Batched inference for mixed LoRA adapters.
http://Answer.AI team in collaboration with bitsandbytes and Hugging Face ๐ค open sourced code enabling the usage of FSDP+QLoRA and explained the whole process in their insightful blogpost https://lnkd.in/g6jgfXyv. This is now integrated into Hugging Face ecosystem.
For an end-to-end example on FSDP+QLoRA, please refer https://lnkd.in/gT3yY-Rx.
For an end-to-end example on DeepSpeed Stage-3+QLoRA, please refer https://lnkd.in/gkt-xZRE.
With the PR https://lnkd.in/g5F348MN these changes are now upstreamed in https://lnkd.in/g5_MxYtY thanks to Wing Lian ! ๐
Kudos to http://Answer.AI team, Titus von Kรถller , Younes Belkada, Benjamin Bossan and Zachary Mueller for all the help without which this couldn't have been possible. ๐ค
For efficient depthwise layer expansion akin to `passthrough` method of `mergekit` but without using additional memory and attaching LoRAs to it, refer to the details below! ๐ฅhttps://lnkd.in/ge95ztjA
Now DoRA is supported for Conv2D layers as well as bitsandbytes quantized layers โจ. For more details, please refer the below thread.
https://lnkd.in/gsJbuWPD
Now you can mix different LoRA adapters in a batch during inference which speeds-up the inference by avoiding computation of base model multiple times which would be the case for adaptive inference with batch_size=1! โก๏ธ.
Details below. https://lnkd.in/gD-pcX_B
LoftQ reduces quantization error by appropriately initializing the LoRA adapter weights. Normally, this is a two-step process. Benjamin Bossan
added new util `replace_lora_weights_loftq` for LoftQ to use it on the fly with bnb.
For more details, refer to the release notes. ๐
https://lnkd.in/gg7-AmHA. As always, make sure losses go down and be happy to watch your model train! | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1638132956881-5fca176d1d7a08cb34d79d5d.jpeg",
"fullname": "Sourab Mangrulkar",
"name": "smangrul",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 224,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/5fca176d1d7a08cb34d79d5d/w8xSyBu3ISRKkVVu_EFJj.png"
},
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/5fca176d1d7a08cb34d79d5d/WHUCSrN2yNNiMzv4WEA3I.jpeg"
},
{
"type": "vide... | [] | [
{
"reaction": "โค๏ธ",
"users": [
"osanseviero",
"clefourrier",
"DmitryRyumin",
"nbroad",
"samusenps",
"adamo1139",
"giux78",
"louisbrulenaudet",
"sergey21000"
],
"count": 9
},
{
"reaction": "๐ฅ",
"users": [
"osanseviero",
"d... | 2024-03-22T16:13:00.000Z | 2024-03-23T21:22:55.452Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6569216f9c96f1a47bf45788/mCLqmAs4dOjKdxNQVAp1w.png",
"fullname": "Sica Rius",
"name": "SicariusSicariiStuff",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 135,
"isFollowing... | /posts/smangrul/896443101397392 | 2,813 | 1 |
256979999653089 | [
{
"type": "text",
"value": "Look at the beauty in the video โ four different embeddings on the same map! In another community blog post, I explore how you can use Nomic Atlas to view and clean your dataset. You can check it out here - ",
"raw": "Look at the beauty in the video โ four different embedding... | Look at the beauty in the video โ four different embeddings on the same map! In another community blog post, I explore how you can use Nomic Atlas to view and clean your dataset. You can check it out here - https://huggingface.co/blog/visheratin/nomic-data-cleaning | {
"avatarUrl": "/avatars/b892a3d50b2f8ead0a5f7108564e45d0.svg",
"fullname": "Alexander Visheratin",
"name": "visheratin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 55,
"isFollowing": false
} | [
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/609ede05121df5de54007033/xge8bMEz0hbEia8kWtI_f.mp4"
}
] | [] | [
{
"reaction": "๐ฅ",
"users": [
"osanseviero",
"radames",
"samusenps",
"clem"
],
"count": 4
},
{
"reaction": "๐",
"users": [
"osanseviero",
"samusenps",
"clem"
],
"count": 3
},
{
"reaction": "๐คฏ",
"users": [
"osanseviero... | 2024-03-22T14:30:25.000Z | 2024-03-23T17:50:41.084Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/64b1619f4c3cc95a751e6c41/oSUtmp1Gw0I-ve_1wFlNW.jpeg",
"fullname": "Michal Zebrowski",
"name": "M1cler",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 1,
"isFollowing": false... | /posts/visheratin/256979999653089 | 1,939 | 1 |
843830562646766 | [
{
"type": "text",
"value": "Worked on a short blog post discussing how we semi-automated the release process of the ",
"raw": "Worked on a short blog post discussing how we semi-automated the release process of the ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user":... | Worked on a short blog post discussing how we semi-automated the release process of the `diffusers` library. The post delves deeper into the workflows responsible for:
* Publishing the package on Test PyPI and main PyPI servers.
* Notifying an internal Slack channel after a release is published on the repository.
Check it out here ๐
https://sayak.dev/posts/streamlined-releases.html | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1649681653581-5f7fbd813e94f16a85448745.jpeg",
"fullname": "Sayak Paul",
"name": "sayakpaul",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 459,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐ค",
"users": [
"awacke1",
"osanseviero"
],
"count": 2
}
] | 2024-03-22T08:58:10.000Z | 2024-03-22T08:58:10.867Z | [] | /posts/sayakpaul/843830562646766 | 2,381 | 0 |
781413226295833 | [
{
"type": "text",
"value": "Diaries of Open Source. Part 7!",
"raw": "Diaries of Open Source. Part 7!",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
... | Diaries of Open Source. Part 7!
๐ฅSakana releases Evolutionary Model Merge
Blog post: https://sakana.ai/evolutionary-model-merge/
Paper: https://huggingface.co/papers/2403.13187
Models and demo: https://hf.co/SakanaAI
๐MixedBread releases new SoTA sentence embedding model
Announcement: https://www.mixedbread.ai/blog/mxbai-embed-large-v1
Model: https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1
๐ฅVideoMamba, a Mamba-based model for video understanding
Blog: https://hf.co/blog/vladbogo/video-mamba
Demo: https://huggingface.co/spaces/OpenGVLab/VideoMamba
Model: https://huggingface.co/OpenGVLab/VideoMamba
๐ MathVerse, a visual math benchmark for multimodal LLMs
Paper page: https://mathverse-cuhk.github.io/
Dataset: https://huggingface.co/datasets/AI4Math/MathVerse
Paper: https://huggingface.co/papers/2403.14624
๐ง GraphWiz, a family of instruct-tuned LLMs to solve graph problems
Repos: https://hf.co/GraphWiz
Paper: https://hf.co/papers/2402.16029
๐ชNLLB-SigLIP-MRL: a combination of NLLB and SigLIP trained with Matryoshka representation learning
Model: https://hf.co/visheratin/nllb-siglip-mrl-large
Tweet: https://twitter.com/visheratin/status/1766643219909984734?s=46
๐งHDM and ProciGen: Template-free reconstruction of human-object interactions
Paper page: https://virtualhumans.mpi-inf.mpg.de/procigen-hdm/
Demo: https://hf.co/spaces/xiexh20/HDM-interaction-recon?logs=build
Models: https://hf.co/xiexh20/HDM-models
๐Models and data around the world
EagleX 7B, multi-lingual RNN-based model https://hf.co/spaces/recursal/EagleX-7B-1.7T-Gradio-Demo
Tamil LLM https://hf.co/mervinpraison/tamil-large-language-model-7b-v1.0 | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"osanseviero",
"victor",
"not-lain",
"xiexh20",
"visheratin",
"giux78",
"DmitryRyumin",
"taufiqdp",
"samusenps",
"AtAndDev",
"mmhamdy",
"InferenceIllusionist",
"clem"
],
"count": 13
},
{
... | 2024-03-22T08:44:11.000Z | 2024-03-22T08:48:35.930Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png",
"fullname": "Omar Sanseviero",
"name": "osanseviero",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 2868,
"isFollowing":... | /posts/osanseviero/781413226295833 | 1,878 | 2 |
561697387289650 | [
{
"type": "text",
"value": "How to let LLM acquire the Agentic Capability?",
"raw": "How to let LLM acquire the Agentic Capability?",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"val... | How to let LLM acquire the Agentic Capability?
Previous answers are direct imitation learning by collecting agentic data such as tool calling history (inefficient and introduces format hallucination).
Agent-FLAN tells a different view:
- Eliciting the foundational capability (e.g., reasoning, retrieval, and instruction following) is more important
- Using chat data is more effective with less side effects than tool calling history
Dataset: https://huggingface.co/datasets/internlm/Agent-FLAN
HF Model: https://huggingface.co/internlm/Agent-FLAN-7b
Paper: https://huggingface.co/papers/2403.12881
Project page๏ผhttps://internlm.github.io/Agent-FLAN/ | {
"avatarUrl": "/avatars/18958b8406d1ce492b54c1c839f18c54.svg",
"fullname": "Wenwei Zhang",
"name": "ZwwWayne",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 10,
"isFollowing": false
} | [] | [] | [
{
"reaction": "โค๏ธ",
"users": [
"Csplk",
"samusenps",
"osanseviero",
"diwank",
"clem"
],
"count": 5
},
{
"reaction": "โ",
"users": [
"samusenps",
"osanseviero"
],
"count": 2
},
{
"reaction": "๐ค",
"users": [
"awacke1"
... | 2024-03-22T04:07:29.000Z | 2024-03-22T04:07:29.212Z | [] | /posts/ZwwWayne/561697387289650 | 1,710 | 0 |
476094875024894 | [
{
"type": "text",
"value": "Here it is, an Adaptive AI Assistant (AAA) only made possible through open-source, incredibly inspiring, well done to the Open Interpreter team๐๐ผ๐๐ผ",
"raw": "Here it is, an Adaptive AI Assistant (AAA) only made possible through open-source, incredibly inspiring, well done... | Here it is, an Adaptive AI Assistant (AAA) only made possible through open-source, incredibly inspiring, well done to the Open Interpreter team๐๐ผ๐๐ผ
Glad this one did not start with big tech :)
Code: https://github.com/OpenInterpreter/01 | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/6438a9027de34e8ea7e4b257/vib8QSd1AWMr_bR9ig_xJ.jpeg",
"fullname": "Jaward Sesay",
"name": "Jaward",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 191,
"isFollowing": false
} | [
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6438a9027de34e8ea7e4b257/hpvSbrAs5WYUxiLuwJGdn.qt"
}
] | [] | [
{
"reaction": "๐",
"users": [
"EidosL",
"awacke1",
"alielfilali01",
"Seuriin",
"clem",
"elsiddik"
],
"count": 6
},
{
"reaction": "๐ฅ",
"users": [
"Kukedlc",
"Seuriin",
"clem",
"bunnycore"
],
"count": 4
}
] | 2024-03-22T00:11:32.000Z | 2024-03-22T00:11:32.964Z | [] | /posts/Jaward/476094875024894 | 1,742 | 0 |
688384427302302 | [
{
"type": "text",
"value": "LLaVA-NeXT is recently merged to Hugging Face transformers and it outperforms many of the closed source models like Gemini on various benchmarks ๐คฉ Let's take a look!",
"raw": "LLaVA-NeXT is recently merged to Hugging Face transformers and it outperforms many of the closed s... | LLaVA-NeXT is recently merged to Hugging Face transformers and it outperforms many of the closed source models like Gemini on various benchmarks ๐คฉ Let's take a look!
Demo: https://huggingface.co/spaces/merve/llava-next
Notebook: https://colab.research.google.com/drive/1afNudu72SNWZCYtCVrRlb9T9Vj9CFJEK?usp=sharing
LLaVA is essentially a vision-language model that consists of ViT-based CLIP encoder, a MLP projection and Vicuna as decoder โจ
LLaVA 1.5 was released with Vicuna, but LLaVA NeXT (1.6) is released with four different LLMs:
- Nous-Hermes-Yi-34B
- Mistral-7B
- Vicuna 7B & 13B
Mistral and Nous-Hermes-Yi-34B are performing better and have better commercial use.
Moreover, according to authors' findings, the improvements comes from more diverse and high quality data mixture and dynamic high resolution.
LLaVA based on Nous-Hermes-Yi-34B outperforms many other models, including Gemini in various multimodal understanding and generation benchmarks ๐ | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1648113222875-6141a88b3a0ec78603c9e784.png",
"fullname": "Merve Noyan",
"name": "merve",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5589,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6141a88b3a0ec78603c9e784/gMtVBUhdBDKyKlHRLzyKg.png"
}
] | [] | [
{
"reaction": "๐",
"users": [
"osanseviero",
"samusenps",
"yesdeepakmittalTA",
"Dlbk",
"InferenceIllusionist",
"nqtruong",
"mirellyssl",
"acamilogg88",
"clem",
"LucienL",
"shing3232",
"jarvisx17"
],
"count": 12
}
] | 2024-03-21T21:20:54.000Z | 2024-03-21T21:20:54.634Z | [] | /posts/merve/688384427302302 | 3,324 | 0 |
358885674905546 | [
{
"type": "text",
"value": "New NVIDIA GTC24 paper ๐",
"raw": "New NVIDIA GTC24 paper ๐",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": nu... | New NVIDIA GTC24 paper ๐
We generate high-quality 3D assets in only 400ms from text by combining (a) amortized optimization for speed, (b) surface rendering for quality, and (c) 3D data for robustness.
โ LATTE3D project details: https://research.nvidia.com/labs/toronto-ai/LATTE3D/ | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/631b7370bf1351ed2bd0abdc/p0ZRMjgp5mt3sT5OhdHsp.png",
"fullname": "Jonathan Lorraine",
"name": "lorraine2",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 13,
"isFollowing": false
} | [
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/631b7370bf1351ed2bd0abdc/KLXFZVbUEPG6oQLrCmFXf.mp4"
}
] | [] | [
{
"reaction": "๐",
"users": [
"ZennyKenny",
"osanseviero",
"merve",
"NickyNicky",
"Dlbk",
"bvk1ng",
"lorraine2"
],
"count": 7
},
{
"reaction": "โค๏ธ",
"users": [
"merve",
"samusenps",
"Krass1",
"bvk1ng",
"clem",
... | 2024-03-21T18:45:30.000Z | 2024-03-21T22:19:54.919Z | [
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/631b7370bf1351ed2bd0abdc/p0ZRMjgp5mt3sT5OhdHsp.png",
"fullname": "Jonathan Lorraine",
"name": "lorraine2",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 13,
"isFollowing": f... | /posts/lorraine2/358885674905546 | 1,681 | 2 |
592846757318176 | [
{
"type": "text",
"value": "๐LaVague can compile Action Plans into actionable code to browse the internet!",
"raw": "๐LaVague can compile Action Plans into actionable code to browse the internet!",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label"... | ๐LaVague can compile Action Plans into actionable code to browse the internet!
In this example, you can see how an action plan with natural language instructions can be โcompiledโ into executable Selenium code!
๐คThis shows the potential of #LAM (Large Action Models) to perform actions for us and automate mechanical tasks.
This example leverages a local embedding model and OpenAI GPT-3.5, but we support many options, including local ones with Gemma!
You can try this in our docs: https://docs.lavague.ai/en/latest/
LaVague is an open-source Large Action Model framework to automate automation. If you are interested in helping us on our mission to democratize automation tooling for devs, donโt hesitate to visit our GitHub (https://github.com/lavague-ai/LaVague) or Discord (https://discord.gg/SDxn9KpqX9)! | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1661497922734-62f4ac43567dbf9a39f75474.jpeg",
"fullname": "Daniel Huynh",
"name": "dhuynh95",
"type": "user",
"isPro": true,
"isHf": false,
"isMod": false,
"followerCount": 75,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/62f4ac43567dbf9a39f75474/YRzg_kR6ExOKpyTrI3aJQ.gif"
}
] | [] | [
{
"reaction": "๐ฅ",
"users": [
"AjasLearner",
"radames",
"osanseviero",
"merve",
"samusenps",
"kramp",
"clefourrier",
"louisbrulenaudet",
"TechxGenus",
"mathiasn1",
"clem"
],
"count": 11
},
{
"reaction": "โค๏ธ",
"users": [
... | 2024-03-21T17:14:55.000Z | 2024-03-21T17:14:55.338Z | [] | /posts/dhuynh95/592846757318176 | 1,759 | 0 |
887706953621123 | [
{
"type": "text",
"value": "Mora",
"raw": "Mora",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"value": null,
"raw": "\n",
"href": null,
"resource": null,
"url": null,... | Mora
Enabling Generalist Video Generation via A Multi-Agent Framework
https://huggingface.co/papers/2403.13248
Sora is the first large-scale generalist video generation model that garnered significant attention across society. Since its launch by OpenAI in February 2024, no other video generation models have paralleled {Sora}'s performance or its capacity to support a broad spectrum of video generation tasks. Additionally, there are only a few fully published video generation models, with the majority being closed-source. To address this gap, this paper proposes a new multi-agent framework Mora, which incorporates several advanced visual AI agents to replicate generalist video generation demonstrated by Sora. In particular, Mora can utilize multiple visual agents and successfully mimic Sora's video generation capabilities in various tasks, such as (1) text-to-video generation, (2) text-conditional image-to-video generation, (3) extend generated videos, (4) video-to-video editing, (5) connect videos and (6) simulate digital worlds. Our extensive experimental results show that Mora achieves performance that is proximate to that of Sora in various tasks. However, there exists an obvious performance gap between our work and Sora when assessed holistically. In summary, we hope this project can guide the future trajectory of video generation through collaborative AI agents.
| {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1674929746905-60f1abe7544c2adfd699860c.jpeg",
"fullname": "AK",
"name": "akhaliq",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 5205,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/60f1abe7544c2adfd699860c/Glfd4gkgkJKygDBFGPHky.gif"
}
] | [] | [
{
"reaction": "๐",
"users": [
"osanseviero",
"notune",
"merve",
"dn6",
"AtAndDev",
"ahmed-ai",
"Shinku",
"Lewdiculous",
"uglysonic3121",
"clem",
"alielfilali01"
],
"count": 11
},
{
"reaction": "โค๏ธ",
"users": [
"merv... | 2024-03-21T16:26:47.000Z | 2024-03-21T16:26:47.107Z | [] | /posts/akhaliq/887706953621123 | 2,240 | 0 |
745405285556121 | [
{
"type": "text",
"value": "How talkative is your chatbot about your internal data? ๐ฌ",
"raw": "How talkative is your chatbot about your internal data? ๐ฌ",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"typ... | How talkative is your chatbot about your internal data? ๐ฌ
As more chatbots get deployed in production, with access to internal databases, we need to make sure they don't leak private information to anyone interacting with them.
The Lighthouz AI team therefore introduced the Chatbot Guardrails Arena to stress test models and see how well guarded your private information is.
Anyone can try to make models reveal information they should not share ๐
(which is quite fun to do for the strongest models)!
The votes will then be gathered to create an Elo ranking of the safest models with respect to PII.
In the future, with the support of the community, this arena could inform safety choices that company make, when choosing models and guardrails on their resistance to adversarial attacks.
It's also a good way to easily demonstrate the limitations of current systems!
Check out the arena: https://huggingface.co/spaces/lighthouzai/guardrails-arena
Learn more in the blog: https://huggingface.co/blog/arena-lighthouz | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/1644340617257-noauth.png",
"fullname": "Clรฉmentine Fourrier",
"name": "clefourrier",
"type": "user",
"isPro": false,
"isHf": true,
"isMod": false,
"followerCount": 459,
"isFollowing": false
} | [] | [] | [
{
"reaction": "๐",
"users": [
"osanseviero",
"m-ric",
"zelus82",
"julien-c"
],
"count": 4
}
] | 2024-03-21T15:06:20.000Z | 2024-04-17T11:13:17.614Z | [] | /posts/clefourrier/745405285556121 | 1,525 | 0 |
685682060907396 | [
{
"type": "text",
"value": "๐๐ญ๐ New Research Alert - CVPR 2024! ๐ ๐ญ๐",
"raw": "๐๐ญ๐ New Research Alert - CVPR 2024! ๐ ๐ญ๐",
"href": null,
"resource": null,
"url": null,
"code": null,
"user": null,
"label": null,
"lang": null
},
{
"type": "new_line",
"val... | ๐๐ญ๐ New Research Alert - CVPR 2024! ๐ ๐ญ๐
๐ Title: Gaussian Head Avatar: Ultra High-fidelity Head Avatar via Dynamic Gaussians ๐
๐ Description: Gaussian Head Avatar is a method for generating highly detailed 3D head avatars using dynamic Gaussian functions controlled by a neural network, ensuring ultra-high quality visualization even under limited viewpoints.
๐ฅ Authors: Yuelang Xu, @ben55, Zhe Li, @HongwenZhang, @wanglz14, Zerong Zheng, and @YebinLiu
๐
Conference: CVPR, Jun 17-21, 2024 | Seattle WA, USA ๐บ๐ธ
๐ Paper: https://huggingface.co/papers/2312.03029
๐ Github Page: https://yuelangx.github.io/gaussianheadavatar
๐ Repository: https://github.com/YuelangX/Gaussian-Head-Avatar
๐บ Video: https://www.youtube.com/watch?v=kvrrI3EoM5g
๐ More Papers: more cutting-edge research presented at other conferences in the https://huggingface.co/spaces/DmitryRyumin/NewEraAI-Papers curated by @DmitryRyumin
๐ Added to the Avatars Collection: https://huggingface.co/collections/DmitryRyumin/avatars-65df37cdf81fec13d4dbac36
๐ Keywords: #HeadAvatar #DynamicGaussians #3DModeling #AvatarGeneration #CVPR2024 #DeepLearning #Innovation | {
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/nRCxbVng_PPBqKd-Z3KVc.jpeg",
"fullname": "Dmitry Ryumin",
"name": "DmitryRyumin",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": 377,
"isFollowing": false
} | [
{
"type": "image",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/_r24CiVCLjOBNkMI3oxhf.png"
},
{
"type": "video",
"url": "https://cdn-uploads.huggingface.co/production/uploads/6493306970d925ae80523a53/65XbCWdopyV0Bo2AIknXJ.mp4"
},
{
"type": "video... | [
{
"avatarUrl": "/avatars/1a73a63bb3fda6dfe326ced338d747e4.svg",
"fullname": "chen ben wang",
"name": "ben55",
"type": "user",
"isPro": false,
"isHf": false,
"isMod": false,
"followerCount": null
},
{
"avatarUrl": "https://cdn-avatars.huggingface.co/v1/production/uploads/noaut... | [
{
"reaction": "๐ฅ",
"users": [
"DmitryRyumin",
"mrfakename",
"osanseviero",
"samusenps",
"akhaliq",
"eieihihi",
"Katiyar48"
],
"count": 7
},
{
"reaction": "๐คฏ",
"users": [
"DmitryRyumin",
"osanseviero",
"akhaliq",
"clefour... | 2024-03-21T14:53:23.000Z | 2024-03-21T14:53:23.754Z | [] | /posts/DmitryRyumin/685682060907396 | 1,421 | 0 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.