README.md

metadata

license: apache-2.0
library_name: transformers
base_model:
  - openchat/openchat-3.5-0106
datasets:
  - Yukang/LongAlpaca-12k
model-index:
  - name: OpenChat-3.5-0106_32K-PoSE
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: IFEval (0-Shot)
          type: HuggingFaceH4/ifeval
          args:
            num_few_shot: 0
        metrics:
          - type: inst_level_strict_acc and prompt_level_strict_acc
            value: 39.69
            name: strict accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Pretergeek/OpenChat-3.5-0106_32K-PoSE
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: BBH (3-Shot)
          type: BBH
          args:
            num_few_shot: 3
        metrics:
          - type: acc_norm
            value: 8.83
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Pretergeek/OpenChat-3.5-0106_32K-PoSE
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MATH Lvl 5 (4-Shot)
          type: hendrycks/competition_math
          args:
            num_few_shot: 4
        metrics:
          - type: exact_match
            value: 1.44
            name: exact match
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Pretergeek/OpenChat-3.5-0106_32K-PoSE
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: GPQA (0-shot)
          type: Idavidrein/gpqa
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 3.47
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Pretergeek/OpenChat-3.5-0106_32K-PoSE
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MuSR (0-shot)
          type: TAUR-Lab/MuSR
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 11.33
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Pretergeek/OpenChat-3.5-0106_32K-PoSE
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MMLU-PRO (5-shot)
          type: TIGER-Lab/MMLU-Pro
          config: main
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 11.46
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Pretergeek/OpenChat-3.5-0106_32K-PoSE
          name: Open LLM Leaderboard

Buy me a Ko-Fi • Support my work using Patreon

OpenChat-3.5-0106_32K-PoSE

Description

This model is Openchat-3.5-0106 with the context length extended from 8192 tokens to 32768 tokens using PoSE.

The model was fine-tuned using Rank-Stabilized LoRA and the LongAlpaca-12K dataset. I hope to continue extending the context in future versions and then apply the same methods to my upscaled versions of OpenChat-3.5 that were created using Block Expansion instead of Depth UP Scaling.

After fine-tuning, the model was tested using passkey retrieval and achieved a score of 100%. Below you can also find the results of the Open LLM Leaderboard evaluations and I am a bit disappointed with those. The model ended up with a significant reduction in performance compared to the original model in all but one test (MUSR). I expected it to do better than the original model on MUSR since that test benefits from long context understanding but I didn't expect such a negative impact on the other tasks. Anyway, I will be addressing this on a future version. I have been running the same benchmarks from the Open LLM Leaderboard locally, using the code from their own github repo, and so far the results below are lower than the ones I am getting and the model seems to performs very close to the original. I used the LongAlpaca-12K dataset because it is small and I have limited computational resources but I might have to try a larger dataset for the next attempt. If you would like to help me, there are links on the top of the model card for my Patreon and Ko-Fi.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

These may or may not be correct, local benchmarks show performance close or identical to the original model. I will be redoing them locally following the same setup from the leaderboard. I would appreciate if someone else interested would also test and give me feedback.

Metric	Value
Avg.	12.70
IFEval (0-Shot)	39.69
BBH (3-Shot)	8.83
MATH Lvl 5 (4-Shot)	1.44
GPQA (0-shot)	3.47
MuSR (0-shot)	11.33
MMLU-PRO (5-shot)	11.46

Citation

@misc{zhu2024poseefficientcontextwindow,
      title={PoSE: Efficient Context Window Extension of LLMs via Positional Skip-wise Training}, 
      author={Dawei Zhu and Nan Yang and Liang Wang and Yifan Song and Wenhao Wu and Furu Wei and Sujian Li},
      year={2024},
      eprint={2309.10400},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2309.10400}, 
}