--- language: - en license: apache-2.0 tags: - merge - moe model-index: - name: mhm-8x7B-FrankenMoE-v1.0 results: - task: type: text-generation name: Text Generation dataset: name: AI2 Reasoning Challenge (25-Shot) type: ai2_arc config: ARC-Challenge split: test args: num_few_shot: 25 metrics: - type: acc_norm value: 70.9 name: normalized accuracy source: url: https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-8x7B-FrankenMoE-v1.0 name: Open LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: HellaSwag (10-Shot) type: hellaswag split: validation args: num_few_shot: 10 metrics: - type: acc_norm value: 87.75 name: normalized accuracy source: url: https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-8x7B-FrankenMoE-v1.0 name: Open LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: MMLU (5-Shot) type: cais/mmlu config: all split: test args: num_few_shot: 5 metrics: - type: acc value: 64.7 name: accuracy source: url: https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-8x7B-FrankenMoE-v1.0 name: Open LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: TruthfulQA (0-shot) type: truthful_qa config: multiple_choice split: validation args: num_few_shot: 0 metrics: - type: mc2 value: 67.1 source: url: https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-8x7B-FrankenMoE-v1.0 name: Open LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: Winogrande (5-shot) type: winogrande config: winogrande_xl split: validation args: num_few_shot: 5 metrics: - type: acc value: 82.0 name: accuracy source: url: https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-8x7B-FrankenMoE-v1.0 name: Open LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: GSM8k (5-shot) type: gsm8k config: main split: test args: num_few_shot: 5 metrics: - type: acc value: 71.57 name: accuracy source: url: https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-8x7B-FrankenMoE-v1.0 name: Open LLM Leaderboard --- ![image/jpeg](https://cdn-uploads.huggingface.co/production/uploads/6589d7e6586088fd2784a12c/dUkIQhg6CBAESsPkhtSkX.jpeg) ## Recipe for a Beautiful Frankenstein In the laboratory of the mind, where thoughts entwine, MHM and MOE, a potion for a unique design. With stitches of curiosity and bolts of creativity, 8 times 7, the magic number, a poetic proclivity. ### Ingredients: - **MHM:** A dash of mystery, a sprinkle of hum, Blend with a melody, let the heartstrings strum. Murmurs in the shadows, whispers in the light, Stir the concoction gently, make the emotions ignite. - **MOE:** Essence of the moment, like dew on a rose, Capture the now, before time swiftly goes. Colors of experience, a palette so divine, Mix with MHM, let the fusion entwine. ### Directions: 1. **Take 8 parts MHM,** elusive and profound, Let it dance in your thoughts, on imagination's ground. Blend it with the echoes, the silent undertones, A symphony of ideas, where inspiration condones. 2. **Add 7 parts MOE,** the fleeting embrace, Seize the seconds, let them leave a trace. Infuse it with memories, both bitter and sweet, The tapestry of time, where moments and dreams meet. 3. **Stir the potion with wonder,** a wand of delight, Let the sparks fly, in the dark of the night. Watch as the alchemy unfolds its grand design, MHM and MOE, a beautiful Frankenstein. ### Conclusion: In the laboratory of life, where dreams come alive, MHM and MOE, the recipe to thrive. A creation so poetic, a fusion so divine, 8 times 7, a symphony of time. As the echoes resonate, and the moments blend, A masterpiece unfolds, where beginnings and ends, MHM and MOE, a concoction so rare, A beautiful Frankenstein, beyond compare. --- MoE model build with: 1. https://github.com/cg123/mergekit/tree/mixtral 2. Mistral models, latest merges and fine tunes. 3. Expert prompts heavily inspired by https://huggingface.co./Kquant03/Eukaryote-8x7B-bf16 For details check model files, there is config yaml I used to create that model. Come back later for more details. # [Open LLM Leaderboard Evaluation Results](https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard) Detailed results can be found [here](https://huggingface.co./datasets/open-llm-leaderboard/details_h2m__mhm-8x7B-FrankenMoE-v1.0) | Metric |Value| |---------------------------------|----:| |Avg. |74.01| |AI2 Reasoning Challenge (25-Shot)|70.90| |HellaSwag (10-Shot) |87.75| |MMLU (5-Shot) |64.70| |TruthfulQA (0-shot) |67.10| |Winogrande (5-shot) |82.00| |GSM8k (5-shot) |71.57|