Hugging Face's logo

MedIT Solutions
company
Verified
https://meditsolutions.pl
Activity Feed Request to join this org

AI & ML interests

None defined yet.

Recent Activity

mkurman  updated a model 23 days ago
meditsolutions/NeuroBLAST-1.9B-Instruct-Early-Preview
mkurman  updated a collection 23 days ago
NeuroBLAST
mkurman  published a model 23 days ago
meditsolutions/NeuroBLAST-1.9B-Instruct-Early-Preview
View all activity

mkurman 
posted an update 22 days ago
view post
Post
258
🚀 Big news! NeuroBLAST, the outstanding new architecture, has officially arrived on HF! After three intense months of training my 1.9 billion SLM on my trusty RTX 3090 Ti, I’m happy to announce the results. While it’s not perfect just yet, I’ve dedicated countless hours to optimizing costs while crafting clever layer connections that mimic the brain's centers. Plus, I’ve introduced a new memory-like layer that’s sure to turn heads! I can’t wait to dive deep into this journey in my upcoming blog post. Stay tuned for the full scoop! 🔥

meditsolutions/NeuroBLAST-1.9B-Instruct-Early-Preview
mkurman 
updated a model 23 days ago
mkurman 
updated a collection 23 days ago
mkurman 
published a model 23 days ago
mkurman 
updated a model 29 days ago
mkurman 
posted an update 5 months ago
mkurman 
updated a collection 5 months ago
mkurman 
in meditsolutions/medit-one-140M-9B-tokens-checkpoint 5 months ago

Question on meaning of parameter of this model

2
#2 opened 5 months ago by
JLouisBiz
mkurman 
posted an update 5 months ago
view post
Post
937
Just released NVAMP Loss!

✔️ modification of the cross-entropy loss function designed specifically for training LLMs.
✔️ twist on the standard cross-entropy loss by emphasizing the importance of outlier prediction errors and dynamically normalizing token-level variance.
✔️ more stable and efficient training, leading to models that generalize better.

Check it out, give it a spin, and let me know what you think!

Licensed under the Apache 2.0 license and ready to use. Happy training! 🔥🤖

https://github.com/mkurman/nvamp-loss
mkurman 
in meditsolutions/medit-one-140M-9B-tokens-checkpoint 5 months ago

Can't install

2
#1 opened 5 months ago by
JLouisBiz
mkurman 
posted an update 5 months ago
mkurman 
published a model 5 months ago
mkurman 
posted an update 5 months ago
mkurman 
posted an update 5 months ago
mkurman 
updated a model 5 months ago
mkurman 
published a model 5 months ago
mkurman 
posted an update 6 months ago
view post
Post
3711
Introducing a new architecture, MedIT One – a single-token transformer with LSTM-like recurrence.

It is extremely fast in training and inference, but we lack funding for large-scale training. Enjoy 🍓

https://github.com/MedITSolutionsKurman/medit-one
mkurman 
posted an update 6 months ago
view post
Post
2048
I've been working on something cool: a GRPO with an LLM evaluator that can also perform SFT on the feedback data - if you want. Check it out 😊

Any 🌟are more than welcome 🤗

https://github.com/mkurman/grpo-llm-evaluator
mkurman 
posted an update 6 months ago
view post
Post
1595
Blurred-Thoughts Supervised-Finetuning 🙈

After hours of working with GitHub Copilot to organize the code, I'm keen to announce the release of Blurred Thoughts Supervised-Finetuning (BT-SFT), a new method for fine-tuning LLMs to produce more diverse and creative responses.

BT-SFT introduces:
✅ Smart tokenization method randomly masks tokens within <think> ... </think> tags, promoting the model to generate diverse responses that align better with its probability distribution instead of memorizing the thought process from distilled data.
✅ Reward function that ensures responses are well-structured.

Explore and contribute to the project available in my GitHub repository:
https://github.com/mkurman/blurred-thoughts-SFT

Keep me updated on your experiments with BT-SFT! 🐐