Phixtral 4x2 8 Gates Poc by marcel

 ยป  All LLMs  ยป  marcel  ยป  Phixtral 4x2 8 Gates Poc   URL Share it on

  Autotrain compatible   Code Cognitivecomputations/dolphin-...   Conversational   Custom code   En   Lxuechen/phi-2-dpo   Moe   Mrm8488/phi-2-coder   Phi-msft   Region:us   Safetensors   Sharded   Tensorflow Yhyu13/phi-2-sft-dpo-gpt4 en-e...

Phixtral 4x2 8 Gates Poc Benchmarks

nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Phixtral 4x2 8 Gates Poc (marcel/phixtral-4x2_8-gates-poc)
๐ŸŒŸ Advertise your project ๐Ÿš€

Phixtral 4x2 8 Gates Poc Parameters and Internals

Model Type 
text-generation
Additional Notes 
Phixtral model uses finetuned gates for better selection of Experts and breaks symmetry of base model.
Supported Languages 
en (high)
Training Details 
Data Sources:
https://huggingface.co/datasets/teknium/openhermes
Methodology:
Finetuning with four microsoft/phi-2 models inspired by Mixtral-8x7B-v0.1 architecture.
Model Architecture:
Mixure of Experts (MoE) with finetuned gates
LLM NamePhixtral 4x2 8 Gates Poc
Repository ๐Ÿค—https://huggingface.co/marcel/phixtral-4x2_8-gates-poc 
Model Size7.8b
Required VRAM15.7 GB
Updated2025-09-17
Maintainermarcel
Model Typephi-msft
Model Files  10.0 GB: 1-of-2   5.7 GB: 2-of-2
Supported Languagesen
Model ArchitecturePhiForCausalLM
Licensemit
Model Max Length2048
Transformers Version4.35.2
Tokenizer ClassCodeGenTokenizer
Padding Token<|endoftext|>
Vocabulary Size51200
Torch Data Typefloat16
Activation Functiongelu_new

Best Alternatives to Phixtral 4x2 8 Gates Poc

Best Alternatives
Context / RAM
Downloads
Likes
Phi Delthanar2K / 15.6 GB32
Phi Stoma2K / 15.6 GB52
Phixtral 4x2 80K / 15.7 GB20209
Note: green Score (e.g. "73.2") means that the model is better than marcel/phixtral-4x2_8-gates-poc.

Rank the Phixtral 4x2 8 Gates Poc 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? 51415 in total.

Our Social Media →  
Original data from HuggingFace, OpenCompass and various public git repos.
Release v20241124