Miqu 1 103B 3.0bpw H6 EXL2 by LoneStriker

 ยป  All LLMs  ยป  LoneStriker  ยป  Miqu 1 103B 3.0bpw H6 EXL2   URL Share it on

  Merged Model   Autotrain compatible Base model:152334h/miqu-1-70b-... Base model:finetune:152334h/mi...   Conversational   De   En   Endpoints compatible   Es   Exl2   Fr   It   Llama   Quantized   Region:us   Safetensors   Sharded   Tensorflow

Miqu 1 103B 3.0bpw H6 EXL2 Benchmarks

nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Miqu 1 103B 3.0bpw H6 EXL2 (LoneStriker/miqu-1-103b-3.0bpw-h6-exl2)

Miqu 1 103B 3.0bpw H6 EXL2 Parameters and Internals

Model Type 
transformers
Additional Notes 
This model is inspired by and created using a variety of merge and quantization methods, involving contributions from different developers and teams.
Supported Languages 
en (high), de (high), fr (high), es (high), it (high)
Training Details 
Methodology:
This is a 103b frankenmerge of miqu-1-70b created by interleaving layers of miqu-1-70b-sf with itself using mergekit.
Context Length:
32768
Model Architecture:
Layers: 120
Input Output 
Input Format:
Mistral ~~[INST] {prompt} [/INST]
Accepted Modalities:
text
LLM NameMiqu 1 103B 3.0bpw H6 EXL2
Repository ๐Ÿค—https://huggingface.co/LoneStriker/miqu-1-103b-3.0bpw-h6-exl2 
Base Model(s)  Miqu 1 70B Sf   152334H/miqu-1-70b-sf
Merged ModelYes
Model Size70b
Required VRAM39.2 GB
Updated2024-12-23
MaintainerLoneStriker
Model Typellama
Model Files  8.5 GB: 1-of-5   8.5 GB: 2-of-5   8.5 GB: 3-of-5   8.6 GB: 4-of-5   5.1 GB: 5-of-5
Supported Languagesen de fr es it
Quantization Typeexl2
Model ArchitectureLlamaForCausalLM
Context Length32768
Model Max Length32768
Transformers Version4.37.2
Tokenizer ClassLlamaTokenizer
Padding Token<unk>
Vocabulary Size32000
Torch Data Typefloat16

Best Alternatives to Miqu 1 103B 3.0bpw H6 EXL2

Best Alternatives
Context / RAM
Downloads
Likes
...B Instruct Gradient 1048K 8bit1024K / 75 GB155
...B Instruct Gradient 1048K 4bit1024K / 39.7 GB183
...B Instruct Gradient 1048K 2bit1024K / 21.9 GB122
...0B Instruct Gradient 262K 4bit256K / 39.7 GB163
...0B Instruct Gradient 262K 8bit256K / 75 GB82
...0B Instruct Gradient 262K 2bit256K / 21.9 GB151
...t Gradient 262K 4.0bpw H6 EXL2256K / 37.2 GB161
... Gradient 262K 2.25bpw H6 EXL2256K / 22.2 GB80
...t Gradient 262K 2.4bpw H6 EXL2256K / 23.5 GB120
...t Gradient 262K 3.5bpw H6 EXL2256K / 32.9 GB60
Note: green Score (e.g. "73.2") means that the model is better than LoneStriker/miqu-1-103b-3.0bpw-h6-exl2.

Rank the Miqu 1 103B 3.0bpw H6 EXL2 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? 40126 in total.

Our Social Media →  
Original data from HuggingFace, OpenCompass and various public git repos.
Release v20241217