Hugging Face's logo

ssmits
/
Qwen2.5-95B-Instruct

Text Generation
Safetensors
English
qwen2
chat
conversational
Eval Results
Model card Files Community
2
Qwen2.5-95B-Instruct / README.md
ssmits's picture
ssmits
Update README.md
75010bd verified
preview code
|
raw
history blame contribute delete
23.8 kB
metadata
license: other
license_name: qwen
license_link: https://huggingface.co/Qwen/Qwen2.5-72B-Instruct/blob/main/LICENSE
language:
  - en
pipeline_tag: text-generation
base_model:
  - Qwen/Qwen2.5-72B-Instruct
model-index:
  - name: Qwen2.5-95B-Instruct
    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: 84.31
            name: strict accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=ssmits/Qwen2.5-95B-Instruct
          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: 58.53
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=ssmits/Qwen2.5-95B-Instruct
          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: 6.04
            name: exact match
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=ssmits/Qwen2.5-95B-Instruct
          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: 15.21
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=ssmits/Qwen2.5-95B-Instruct
          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: 13.61
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=ssmits/Qwen2.5-95B-Instruct
          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: 46.85
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=ssmits/Qwen2.5-95B-Instruct
          name: Open LLM Leaderboard
tags:
  - chat

Qwen2.5-95B-Instruct

Qwen2.5-95B-Instruct is a Qwen/Qwen2.5-72B-Instruct self-merge made with MergeKit.

The layer ranges chosen for this merge were inspired by a rough estimate of the layer similarity analysis of ssmits/Falcon2-5.5B-multilingual. Layer similarity analysis involves examining the outputs of different layers in a neural network to determine how similar or different they are. This technique can help identify which layers contribute most significantly to the model's performance. In the context of the Falcon-11B model, layer similarity analysis across multiple languages revealed that the first half of the layers were more important for maintaining performance. Additionally, this analysis can be used to more rigidly slice and add extra layers for optimal Next Token Prediction, allowing for possibly a model architecture that's more creative and powerful.

Special thanks to Eric Hartford for both inspiring and evaluating the original model, to Charles Goddard for creating MergeKit, and to Mathieu Labonne for creating the Meta-Llama-3-120B-Instruct model that served as the main inspiration for this merge.

🔍 Applications

This model is probably good for creative writing tasks. It uses the Qwen chat template with a default context window of 128K.

The model could be quite creative and maybe even better than the 72B model at some tasks.

⚡ Quantized models

To be quantized.

  • GGUF: [Link to GGUF model]
  • EXL2: [Link to EXL2 model]
  • mlx: [Link to mlx model]

🏆 Evaluation

This model has yet to be thoroughly evaluated. It is expected to excel in creative writing and more but may have limitations in other tasks. Use it with caution and don't expect it to outperform state-of-the-art models outside of specific creative use cases.

Once the model is created and tested, this section will be updated with:

  • Links to evaluation threads on social media platforms
  • Examples of the model's performance in creative writing tasks
  • Comparisons with other large language models in various applications
  • Community feedback and use cases

We encourage users to share their experiences and evaluations to help build a comprehensive understanding of the model's capabilities and limitations.

🧩 Configuration 

slices:
- sources:
  - layer_range: [0, 10]
    model: Qwen/Qwen2.5-72B-Instruct
- sources:
  - layer_range: [5, 15]
    model: Qwen/Qwen2.5-72B-Instruct
- sources:
  - layer_range: [10, 20]
    model: Qwen/Qwen2.5-72B-Instruct
- sources:
  - layer_range: [15, 25]
    model: Qwen/Qwen2.5-72B-Instruct
- sources:
  - layer_range: [20, 30]
    model: Qwen/Qwen2.5-72B-Instruct
- sources:
  - layer_range: [25, 80]
    model: Qwen/Qwen2.5-72B-Instruct
dtype: bfloat16
merge_method: passthrough

💻 Usage 

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "ssmits/Qwen2.5-95B-Instruct"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 37.43
IFEval (0-Shot) 84.31
BBH (3-Shot) 58.53
MATH Lvl 5 (4-Shot) 6.04
GPQA (0-shot) 15.21
MuSR (0-shot) 13.61
MMLU-PRO (5-shot) 46.85
Key 72b Result 95b Result Difference Which is Higher Multiplier
leaderboard_musr.acc_norm,none 0.419 0.427 0.008 95b 1.02x
leaderboard_bbh_sports_understanding.acc_norm,none 0.892 0.876 0.016 72b 0.98x
leaderboard_bbh_logical_deduction_three_objects.acc_norm,none 0.94 0.928 0.012 72b 0.99x
leaderboard_math_geometry_hard.exact_match,none 0 0.008 0.008 95b 0.00x
leaderboard_gpqa.acc_norm,none 0.375 0.364 0.011 72b 0.97x
leaderboard_math_hard.exact_match,none 0.012 0.06 0.048 95b 5.00x
leaderboard.exact_match,none 0.012 0.06 0.048 95b 5.00x
leaderboard.prompt_level_loose_acc,none 0.861 0.839 0.022 72b 0.97x
leaderboard.prompt_level_strict_acc,none 0.839 0.813 0.026 72b 0.97x
leaderboard.inst_level_loose_acc,none 0.904 0.891 0.013 72b 0.99x
leaderboard.acc_norm,none 0.641 0.622 0.02 72b 0.97x
leaderboard.inst_level_strict_acc,none 0.888 0.873 0.016 72b 0.98x
leaderboard.acc,none 0.563 0.522 0.041 72b 0.93x
leaderboard_bbh_causal_judgement.acc_norm,none 0.668 0.663 0.005 72b 0.99x
leaderboard_bbh_salient_translation_error_detection.acc_norm,none 0.668 0.588 0.08 72b 0.88x
leaderboard_gpqa_extended.acc_norm,none 0.372 0.364 0.007 72b 0.98x
leaderboard_math_prealgebra_hard.exact_match,none 0.047 0.155 0.109 95b 3.33x
leaderboard_math_algebra_hard.exact_match,none 0.02 0.114 0.094 95b 5.83x
leaderboard_bbh_boolean_expressions.acc_norm,none 0.936 0.92 0.016 72b 0.98x
leaderboard_math_num_theory_hard.exact_match,none 0 0.058 0.058 95b 0.00x
leaderboard_bbh_movie_recommendation.acc_norm,none 0.768 0.78 0.012 95b 1.02x
leaderboard_math_counting_and_prob_hard.exact_match,none 0 0.024 0.024 95b 0.00x
leaderboard_math_intermediate_algebra_hard.exact_match,none 0 0.004 0.004 95b 0.00x
leaderboard_ifeval.prompt_level_strict_acc,none 0.839 0.813 0.026 72b 0.97x
leaderboard_ifeval.inst_level_strict_acc,none 0.888 0.873 0.016 72b 0.98x
leaderboard_ifeval.inst_level_loose_acc,none 0.904 0.891 0.013 72b 0.99x
leaderboard_ifeval.prompt_level_loose_acc,none 0.861 0.839 0.022 72b 0.97x
leaderboard_bbh_snarks.acc_norm,none 0.927 0.904 0.022 72b 0.98x
leaderboard_bbh_web_of_lies.acc_norm,none 0.676 0.616 0.06 72b 0.91x
leaderboard_bbh_penguins_in_a_table.acc_norm,none 0.719 0.767 0.048 95b 1.07x
leaderboard_bbh_hyperbaton.acc_norm,none 0.892 0.9 0.008 95b 1.01x
leaderboard_bbh_object_counting.acc_norm,none 0.612 0.544 0.068 72b 0.89x
leaderboard_musr_object_placements.acc_norm,none 0.258 0.285 0.027 95b 1.11x
leaderboard_bbh_logical_deduction_five_objects.acc_norm,none 0.704 0.592 0.112 72b 0.84x
leaderboard_musr_team_allocation.acc_norm,none 0.456 0.396 0.06 72b 0.87x
leaderboard_bbh_navigate.acc_norm,none 0.832 0.788 0.044 72b 0.95x
leaderboard_bbh_tracking_shuffled_objects_seven_objects.acc_norm,none 0.34 0.304 0.036 72b 0.89x
leaderboard_bbh_formal_fallacies.acc_norm,none 0.776 0.756 0.02 72b 0.97x
all.leaderboard_musr.acc_norm,none 0.419 0.427 0.008 95b 1.02x
all.leaderboard_bbh_sports_understanding.acc_norm,none 0.892 0.876 0.016 72b 0.98x
all.leaderboard_bbh_logical_deduction_three_objects.acc_norm,none 0.94 0.928 0.012 72b 0.99x
all.leaderboard_math_geometry_hard.exact_match,none 0 0.008 0.008 95b 0.00x
all.leaderboard_gpqa.acc_norm,none 0.375 0.364 0.011 72b 0.97x
all.leaderboard_math_hard.exact_match,none 0.012 0.06 0.048 95b 5.00x
all.leaderboard.exact_match,none 0.012 0.06 0.048 95b 5.00x
all.leaderboard.prompt_level_loose_acc,none 0.861 0.839 0.022 72b 0.97x
all.leaderboard.prompt_level_strict_acc,none 0.839 0.813 0.026 72b 0.97x
all.leaderboard.inst_level_loose_acc,none 0.904 0.891 0.013 72b 0.99x
all.leaderboard.acc_norm,none 0.641 0.622 0.02 72b 0.97x
all.leaderboard.inst_level_strict_acc,none 0.888 0.873 0.016 72b 0.98x
all.leaderboard.acc,none 0.563 0.522 0.041 72b 0.93x
all.leaderboard_bbh_causal_judgement.acc_norm,none 0.668 0.663 0.005 72b 0.99x
all.leaderboard_bbh_salient_translation_error_detection.acc_norm,none 0.668 0.588 0.08 72b 0.88x
all.leaderboard_gpqa_extended.acc_norm,none 0.372 0.364 0.007 72b 0.98x
all.leaderboard_math_prealgebra_hard.exact_match,none 0.047 0.155 0.109 95b 3.33x
all.leaderboard_math_algebra_hard.exact_match,none 0.02 0.114 0.094 95b 5.83x
all.leaderboard_bbh_boolean_expressions.acc_norm,none 0.936 0.92 0.016 72b 0.98x
all.leaderboard_math_num_theory_hard.exact_match,none 0 0.058 0.058 95b 0.00x
all.leaderboard_bbh_movie_recommendation.acc_norm,none 0.768 0.78 0.012 95b 1.02x
all.leaderboard_math_counting_and_prob_hard.exact_match,none 0 0.024 0.024 95b 0.00x
all.leaderboard_math_intermediate_algebra_hard.exact_match,none 0 0.004 0.004 95b 0.00x
all.leaderboard_ifeval.prompt_level_strict_acc,none 0.839 0.813 0.026 72b 0.97x
all.leaderboard_ifeval.inst_level_strict_acc,none 0.888 0.873 0.016 72b 0.98x
all.leaderboard_ifeval.inst_level_loose_acc,none 0.904 0.891 0.013 72b 0.99x
all.leaderboard_ifeval.prompt_level_loose_acc,none 0.861 0.839 0.022 72b 0.97x
all.leaderboard_bbh_snarks.acc_norm,none 0.927 0.904 0.022 72b 0.98x
all.leaderboard_bbh_web_of_lies.acc_norm,none 0.676 0.616 0.06 72b 0.91x
all.leaderboard_bbh_penguins_in_a_table.acc_norm,none 0.719 0.767 0.048 95b 1.07x
all.leaderboard_bbh_hyperbaton.acc_norm,none 0.892 0.9 0.008 95b 1.01x
all.leaderboard_bbh_object_counting.acc_norm,none 0.612 0.544 0.068 72b 0.89x
all.leaderboard_musr_object_placements.acc_norm,none 0.258 0.285 0.027 95b 1.11x
all.leaderboard_bbh_logical_deduction_five_objects.acc_norm,none 0.704 0.592 0.112 72b 0.84x
all.leaderboard_musr_team_allocation.acc_norm,none 0.456 0.396 0.06 72b 0.87x
all.leaderboard_bbh_navigate.acc_norm,none 0.832 0.788 0.044 72b 0.95x
all.leaderboard_bbh_tracking_shuffled_objects_seven_objects.acc_norm,none 0.34 0.304 0.036 72b 0.89x
all.leaderboard_bbh_formal_fallacies.acc_norm,none 0.776 0.756 0.02 72b 0.97x
all.leaderboard_gpqa_main.acc_norm,none 0.375 0.355 0.02 72b 0.95x
all.leaderboard_bbh_disambiguation_qa.acc_norm,none 0.744 0.772 0.028 95b 1.04x
all.leaderboard_bbh_tracking_shuffled_objects_five_objects.acc_norm,none 0.32 0.284 0.036 72b 0.89x
all.leaderboard_bbh_date_understanding.acc_norm,none 0.784 0.764 0.02 72b 0.97x
all.leaderboard_bbh_geometric_shapes.acc_norm,none 0.464 0.412 0.052 72b 0.89x
all.leaderboard_bbh_reasoning_about_colored_objects.acc_norm,none 0.864 0.84 0.024 72b 0.97x
all.leaderboard_musr_murder_mysteries.acc_norm,none 0.548 0.604 0.056 95b 1.10x
all.leaderboard_bbh_ruin_names.acc_norm,none 0.888 0.86 0.028 72b 0.97x
all.leaderboard_bbh_logical_deduction_seven_objects.acc_norm,none 0.644 0.664 0.02 95b 1.03x
all.leaderboard_bbh.acc_norm,none 0.726 0.701 0.025 72b 0.97x
all.leaderboard_bbh_temporal_sequences.acc_norm,none 0.996 0.968 0.028 72b 0.97x
all.leaderboard_mmlu_pro.acc,none 0.563 0.522 0.041 72b 0.93x
leaderboard_gpqa_main.acc_norm,none 0.375 0.355 0.02 72b 0.95x
leaderboard_bbh_disambiguation_qa.acc_norm,none 0.744 0.772 0.028 95b 1.04x
leaderboard_bbh_tracking_shuffled_objects_five_objects.acc_norm,none 0.32 0.284 0.036 72b 0.89x
leaderboard_bbh_date_understanding.acc_norm,none 0.784 0.764 0.02 72b 0.97x
leaderboard_bbh_geometric_shapes.acc_norm,none 0.464 0.412 0.052 72b 0.89x
leaderboard_bbh_reasoning_about_colored_objects.acc_norm,none 0.864 0.84 0.024 72b 0.97x
leaderboard_musr_murder_mysteries.acc_norm,none 0.548 0.604 0.056 95b 1.10x
leaderboard_bbh_ruin_names.acc_norm,none 0.888 0.86 0.028 72b 0.97x
leaderboard_bbh_logical_deduction_seven_objects.acc_norm,none 0.644 0.664 0.02 95b 1.03x
leaderboard_bbh.acc_norm,none 0.726 0.701 0.025 72b 0.97x
leaderboard_bbh_temporal_sequences.acc_norm,none 0.996 0.968 0.028 72b 0.97x
leaderboard_mmlu_pro.acc,none 0.563 0.522 0.041 72b 0.93x