AquilaMoE SFT by BAAI

 ยป  All LLMs  ยป  BAAI  ยป  AquilaMoE SFT   URL Share it on

  Arxiv:2202.08906   Arxiv:2212.05055   Arxiv:2401.09192   Arxiv:2408.06567   Aquilamoe   Autotrain compatible   Conversational   Custom code   Dataset:baai/infinity-instruct   En   Instruct   Moe   Region:us   Safetensors   Sharded   Tensorflow   Zh
Model Card on HF ๐Ÿค—: https://huggingface.co/BAAI/AquilaMoE-SFT 

AquilaMoE SFT Benchmarks

nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
AquilaMoE SFT (BAAI/AquilaMoE-SFT)

AquilaMoE SFT Parameters and Internals

Model Type 
Mixture of Experts (MoE), Bilingual
Use Cases 
Areas:
Research, Commercial Applications
Limitations:
No moderation mechanisms
Considerations:
Engagement with the community for guardrails
Additional Notes 
The AquilaMoE Instruct model is a demonstration of easy fine-tuning for compelling performance
Supported Languages 
en (fluent), zh (fluent)
Training Details 
Data Sources:
RedPajama-Data-V2, falcon-refinedweb, C4, Pile, WuDaoCorporaText, ChineseWebText
Data Volume:
4TB tokens
Methodology:
EfficientScale with Scale-Up and Scale-Out strategies
Context Length:
4096
LLM NameAquilaMoE SFT
Repository ๐Ÿค—https://huggingface.co/BAAI/AquilaMoE-SFT 
Required VRAM209.4 GB
Updated2025-01-24
MaintainerBAAI
Model Typeaquilamoe
Instruction-BasedYes
Model Files  4.8 GB: 1-of-43   5.0 GB: 2-of-43   5.0 GB: 3-of-43   5.0 GB: 4-of-43   5.0 GB: 5-of-43   5.0 GB: 6-of-43   5.0 GB: 7-of-43   5.0 GB: 8-of-43   5.0 GB: 9-of-43   5.0 GB: 10-of-43   5.0 GB: 11-of-43   5.0 GB: 12-of-43   5.0 GB: 13-of-43   5.0 GB: 14-of-43   5.0 GB: 15-of-43   5.0 GB: 16-of-43   4.8 GB: 17-of-43   5.0 GB: 18-of-43   5.0 GB: 19-of-43   5.0 GB: 20-of-43   5.0 GB: 21-of-43   5.0 GB: 22-of-43   5.0 GB: 23-of-43   5.0 GB: 24-of-43   5.0 GB: 25-of-43   5.0 GB: 26-of-43   5.0 GB: 27-of-43   5.0 GB: 28-of-43   5.0 GB: 29-of-43   5.0 GB: 30-of-43   5.0 GB: 31-of-43   5.0 GB: 32-of-43   5.0 GB: 33-of-43   5.0 GB: 34-of-43   5.0 GB: 35-of-43   5.0 GB: 36-of-43   5.0 GB: 37-of-43   5.0 GB: 38-of-43   5.0 GB: 39-of-43   5.0 GB: 40-of-43   4.8 GB: 41-of-43   5.0 GB: 42-of-43
Supported Languagesen zh
Model ArchitectureAquilaMoeForCausalLM
Licenseapache-2.0
Context Length4096
Model Max Length4096
Transformers Version4.40.1
Vocabulary Size151851
Torch Data Typebfloat16

Rank the AquilaMoE SFT Capabilities

๐Ÿ†˜ 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! ๐ŸŒŸ

Instruction Following and Task Automation  
Factuality and Completeness of Knowledge  
Censorship and Alignment  
Data Analysis and Insight Generation  
Text Generation  
Text Summarization and Feature Extraction  
Code Generation  
Multi-Language Support and Translation  

What open-source LLMs or SLMs are you in search of? 41817 in total.

Our Social Media →  
Original data from HuggingFace, OpenCompass and various public git repos.
Release v20241227