Autotrain compatible Base model:merge:mistralai/mis... Base model:mistralai/mistral-7... Base model:mistralai/mistral-7... Conversational Endpoints compatible Instruct Merge Mergekit Mixtral Model-index Moe Region:us Safetensors Sharded Tensorflow
Model Card on HF 🤗: https://huggingface.co/osanseviero/mistral-instruct-moe-experimental
Mistral Instruct MoE Experimental Benchmarks
ARC: 61.01 vs 96.7 (so35)-36.9%
HellaSwag: 81.55 vs 95.3 (gpt4)-14.4%
MMLU: 58.22 vs 88.3 (so35)-34.1%
TruthfulQA: 60.4 vs 59 (gpt4)2.4%
WinoGrande: 76.09 vs 87.5 (gpt4)-13%
GSM8K: 31.08 vs 96.4 (so35)-67.8%
LLME Score: 0.18844
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
What is the LLM Explorer Rank (Score)
Mistral Instruct MoE Experimental Parameters and Internals
| Model Type |
| |
| Additional Notes |
|
| LLM Name | Mistral Instruct MoE Experimental |
| Repository 🤗 | https://huggingface.co/osanseviero/mistral-instruct-moe-experimental |
| Base Model(s) | |
| Model Size | 12.9b |
| Required VRAM | 25.8 GB |
| Updated | 2024-12-14 |
| Maintainer | osanseviero |
| Model Type | mixtral |
| Instruction-Based | Yes |
| Model Files | |
| Model Architecture | MixtralForCausalLM |
| License | apache-2.0 |
| Context Length | 32768 |
| Model Max Length | 32768 |
| Transformers Version | 4.36.2 |
| Tokenizer Class | LlamaTokenizer |
| Padding Token | <s> |
| Vocabulary Size | 32000 |
| Torch Data Type | bfloat16 |
Best Alternatives to Mistral Instruct MoE Experimental
Best Alternatives |
Context / RAM |
Downloads |
Likes |
|---|---|---|---|
| MergedExpert 2x8b | 32K / 25.8 GB | 5 | 0 |
| MergedExperts 2x8b | 32K / 25.8 GB | 6 | 0 |
| NearalMistral 2x7B | 32K / 25.8 GB | 87 | 1 |
| Megatron V3 2x7B | 32K / 25.8 GB | 76 | 3 |
| MistarlingMaid 2x7B Base | 32K / 25.8 GB | 79 | 0 |
| Orthogonal 2x7B Base | 32K / 25.8 GB | 1208 | 0 |
| ...afted Hermetic Platypus C 2x7B | 32K / 25.8 GB | 81 | 0 |
| ...tral 7B Instruct V0.2 2x7B MoE | 32K / 25.8 GB | 1131 | 4 |
| Mistral 2x7b V0.1 | 32K / 25.8 GB | 13 | 1 |
| Mistral Math 2x7b Mix | 32K / 25.8 GB | 80 | 4 |
Note: green Score (e.g. "73.2") means that the model is better than osanseviero/mistral-instruct-moe-experimental .
Rank the Mistral Instruct MoE Experimental Capabilities
🆘 Have you tried this model? Rate its performance. This feedback would greatly assist ML community in identifying the most suitable model for their needs. Your contribution really does make a difference! 🌟
What open-source LLMs or SLMs are you in search of? 39237 in total.
Original data from HuggingFace, OpenCompass and various public git repos.
Release v20241124