Improve language tag
#5
by
lbourdois
- opened
README.md
CHANGED
|
@@ -1,115 +1,129 @@
|
|
| 1 |
-
---
|
| 2 |
-
library_name: transformers
|
| 3 |
-
license: apache-2.0
|
| 4 |
-
base_model:
|
| 5 |
-
- Qwen/Qwen2.5-32B-Instruct
|
| 6 |
-
tags:
|
| 7 |
-
- llama-factory
|
| 8 |
-
- full
|
| 9 |
-
- generated_from_trainer
|
| 10 |
-
|
| 11 |
-
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
-
|
| 73 |
-
-
|
| 74 |
-
-
|
| 75 |
-
|
| 76 |
-
-
|
| 77 |
-
-
|
| 78 |
-
|
| 79 |
-
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
-
|
| 88 |
-
-
|
| 89 |
-
-
|
| 90 |
-
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 115 |
- 🤖 [OpenThinker-32B-Unverified model](https://huggingface.co/open-thoughts/OpenThinker-32B-Unverified)
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
base_model:
|
| 5 |
+
- Qwen/Qwen2.5-32B-Instruct
|
| 6 |
+
tags:
|
| 7 |
+
- llama-factory
|
| 8 |
+
- full
|
| 9 |
+
- generated_from_trainer
|
| 10 |
+
datasets:
|
| 11 |
+
- open-thoughts/open-thoughts-114k
|
| 12 |
+
language:
|
| 13 |
+
- zho
|
| 14 |
+
- eng
|
| 15 |
+
- fra
|
| 16 |
+
- spa
|
| 17 |
+
- por
|
| 18 |
+
- deu
|
| 19 |
+
- ita
|
| 20 |
+
- rus
|
| 21 |
+
- jpn
|
| 22 |
+
- kor
|
| 23 |
+
- vie
|
| 24 |
+
- tha
|
| 25 |
+
- ara
|
| 26 |
+
model-index:
|
| 27 |
+
- name: OpenThinker-32B
|
| 28 |
+
results: []
|
| 29 |
+
---
|
| 30 |
+
|
| 31 |
+
<p align="center">
|
| 32 |
+
<img src="https://huggingface.co/datasets/open-thoughts/open-thoughts-114k/resolve/main/open_thoughts.png" width="50%">
|
| 33 |
+
</p>
|
| 34 |
+
|
| 35 |
+
# OpenThinker-32B
|
| 36 |
+
|
| 37 |
+
This model is a fine-tuned version of [Qwen/Qwen2.5-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-32B-Instruct) on the
|
| 38 |
+
[OpenThoughts-114k](https://huggingface.co/datasets/open-thoughts/OpenThoughts-114k) dataset.
|
| 39 |
+
|
| 40 |
+
The dataset is derived by distilling DeepSeek-R1 using the [data pipeline available on github](https://github.com/open-thoughts/open-thoughts).
|
| 41 |
+
More info about the dataset can be found on the dataset card at [OpenThoughts-114k dataset](https://huggingface.co/datasets/open-thoughts/open-thoughts-114k).
|
| 42 |
+
|
| 43 |
+
The numbers reported in the table below are evaluated with our open-source tool [Evalchemy](https://github.com/mlfoundations/Evalchemy).
|
| 44 |
+
|
| 45 |
+
|
| 46 |
+
|Model Name|Dataset Size|AIME24 I/II|AIME25 I|MATH500|GPQA Diamond|LCBv2|
|
| 47 |
+
|---|---|---|---|---|---|---|
|
| 48 |
+
|LIMO-32B|0.8k|56.7|49.3|86.6|58.1|60.0|
|
| 49 |
+
|s1-32B|1k|36.0|25.3|84.8|50.5|40.9|
|
| 50 |
+
|s1.1-32B|1k|64.7|49.3|89.0|60.1|65.5|
|
| 51 |
+
|DeepSeek-R1-Distill-Qwen-32B|800k (closed)|**76.7**|**55.9**|89.4|57.6|**71.2**|
|
| 52 |
+
|**OpenThinker-32B**|114k|66.0|53.3|**90.6**|**61.6**|68.9|
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
We are fully open-source. Our [model weights](https://huggingface.co/open-thoughts), [datasets](https://huggingface.co/open-thoughts), [data generation code](https://github.com/open-thoughts/open-thoughts), [evaluation code](https://github.com/mlfoundations/Evalchemy), and [training code](https://github.com/hiyouga/LLaMA-Factory) are all publicly available.
|
| 56 |
+
|
| 57 |
+
| | Open Weights | Open Data | Open Code |
|
| 58 |
+
|--|--------------|-----------| --------- |
|
| 59 |
+
|OpenThinker-32B|✅|[✅](https://huggingface.co/datasets/open-thoughts/OpenThoughts-114k)|[✅](https://github.com/open-thoughts/open-thoughts) |
|
| 60 |
+
|DeepSeek-R1-Distill-Qwen-32B|✅|❌|❌|
|
| 61 |
+
|OpenAI/Gemini|❌|❌|❌|❌|
|
| 62 |
+
|
| 63 |
+
|
| 64 |
+
|
| 65 |
+
## Intended uses & limitations
|
| 66 |
+
|
| 67 |
+
Apache 2.0 License
|
| 68 |
+
|
| 69 |
+
|
| 70 |
+
## Training procedure
|
| 71 |
+
|
| 72 |
+
We finetune [Qwen2.5-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-32B-Instruct)
|
| 73 |
+
on [OpenThoughts-114k](https://huggingface.co/datasets/open-thoughts/OpenThoughts-114k) for
|
| 74 |
+
3 epochs with a 16k context length using [LlamaFactory](https://github.com/hiyouga/LLaMA-Factory).
|
| 75 |
+
Our [full training configuration](https://github.com/open-thoughts/open-thoughts/blob/main/train/OpenThinker-32B.yaml)
|
| 76 |
+
is provided in [our repository](https://github.com/open-thoughts/open-thoughts/tree/main).
|
| 77 |
+
Training the 32B model on [OpenThoughts-114k](https://huggingface.co/datasets/open-thoughts/OpenThoughts-114k)
|
| 78 |
+
was done on AWS SageMaker with 8xH100 P5 nodes. On 4 nodes, this took around 90 hours.
|
| 79 |
+
Meanwhile, for training on [OpenThoughts-Unverified-173k](https://huggingface.co/datasets/open-thoughts/OpenThoughts-Unverfied-173k),
|
| 80 |
+
we used 96 nodes of 4xA100 (64 GB per GPU), training took 30 hours, spending 11,520 A100 hours on the Leonardo Supercomputer.
|
| 81 |
+
|
| 82 |
+
### Training hyperparameters
|
| 83 |
+
|
| 84 |
+
The following hyperparameters were used during training:
|
| 85 |
+
- learning_rate: 1e-05
|
| 86 |
+
- train_batch_size: 1
|
| 87 |
+
- eval_batch_size: 8
|
| 88 |
+
- seed: 42
|
| 89 |
+
- distributed_type: multi-GPU
|
| 90 |
+
- num_devices: 32
|
| 91 |
+
- gradient_accumulation_steps: 3
|
| 92 |
+
- total_train_batch_size: 96
|
| 93 |
+
- total_eval_batch_size: 256
|
| 94 |
+
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
|
| 95 |
+
- lr_scheduler_type: cosine
|
| 96 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 97 |
+
- num_epochs: 3.0
|
| 98 |
+
|
| 99 |
+
### Framework versions
|
| 100 |
+
|
| 101 |
+
- Transformers 4.46.1
|
| 102 |
+
- Pytorch 2.3.0
|
| 103 |
+
- Datasets 3.1.0
|
| 104 |
+
- Tokenizers 0.20.3
|
| 105 |
+
|
| 106 |
+
More info can be found in our repository: [https://github.com/open-thoughts/open-thoughts](https://github.com/open-thoughts/open-thoughts).
|
| 107 |
+
|
| 108 |
+
# Citation
|
| 109 |
+
```
|
| 110 |
+
@misc{openthoughts,
|
| 111 |
+
author = {Team, OpenThoughts},
|
| 112 |
+
month = jan,
|
| 113 |
+
title = {{Open Thoughts}},
|
| 114 |
+
howpublished = {https://open-thoughts.ai},
|
| 115 |
+
year = {2025}
|
| 116 |
+
}
|
| 117 |
+
```
|
| 118 |
+
|
| 119 |
+
# Links
|
| 120 |
+
- 📊 [Open Thoughts Launch Blog Post](https://www.open-thoughts.ai/blog/launch)
|
| 121 |
+
- 📊 [Open Thoughts Measuring Reasoning with Evalchmey Blog Post](https://www.open-thoughts.ai/blog/measure)
|
| 122 |
+
- 📊 [Open Thoughts OpenThinker-32B Post](https://www.open-thoughts.ai/blog/scale)
|
| 123 |
+
- 💻 [Open Thoughts GitHub Repository](https://github.com/open-thoughts/open-thoughts)
|
| 124 |
+
- 🧠 [OpenThoughts-114k dataset](https://huggingface.co/datasets/open-thoughts/OpenThoughts-114k)
|
| 125 |
+
- 🧠 [OpenThoughts-Unverified-173k dataset](https://huggingface.co/datasets/open-thoughts/OpenThoughts-Unverified-173k)
|
| 126 |
+
- 🤖 [OpenThinker-7B model](https://huggingface.co/open-thoughts/OpenThinker-7B)
|
| 127 |
+
- 🤖 [OpenThinker-7B-Unverfied model](https://huggingface.co/open-thoughts/OpenThinker-7B-Unverified)
|
| 128 |
+
- 🤖 [OpenThinker-32B model](https://huggingface.co/open-thoughts/OpenThinker-32B) - this model
|
| 129 |
- 🤖 [OpenThinker-32B-Unverified model](https://huggingface.co/open-thoughts/OpenThinker-32B-Unverified)
|