Llama 2 70B AWQ by TheBloke

 ยป  All LLMs  ยป  TheBloke  ยป  Llama 2 70B AWQ   URL Share it on

  Arxiv:2307.09288   4-bit   Autotrain compatible   Awq Base model:meta-llama/llama-2-... Base model:quantized:meta-llam...   En   Facebook   Llama   Llama2   Meta   Pytorch   Quantized   Region:us   Safetensors   Sharded   Tensorflow

Llama 2 70B AWQ Benchmarks

nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Llama 2 70B AWQ (TheBloke/Llama-2-70B-AWQ)

Llama 2 70B AWQ Parameters and Internals

Model Type 
text-generation
Use Cases 
Areas:
Commercial, Research
Primary Use Cases:
Assistant-like chat
Limitations:
English language only
Considerations:
Intended for assistant-like chat; developers should perform safety testing.
Additional Notes 
AWQ supports 4-bit quantization for efficient and fast inference.
Supported Languages 
en (High)
Training Details 
Data Sources:
A new mix of publicly available online data
Data Volume:
2 trillion tokens
Methodology:
Supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF)
Context Length:
4096
Hardware Used:
A100-80GB GPUs (350-400W TDP)
Model Architecture:
Transformer architecture with Grouped-Query Attention (GQA) for 70B model
Responsible Ai Considerations 
Fairness:
Testing conducted to date has been in English, and possibilities of inaccurate or biased responses may exist.
Mitigation Strategies:
Developers are advised to perform safety testing and tuning tailored to specific applications.
Input Output 
Input Format:
Models input text only.
Accepted Modalities:
text
Output Format:
Models generate text only.
LLM NameLlama 2 70B AWQ
Repository ๐Ÿค—https://huggingface.co/TheBloke/Llama-2-70B-AWQ 
Model NameLlama 2 70B
Model CreatorMeta Llama 2
Base Model(s)  Llama 2 70B Hf   meta-llama/Llama-2-70b-hf
Model Size70b
Required VRAM36.6 GB
Updated2024-12-23
MaintainerTheBloke
Model Typellama
Model Files  9.9 GB: 1-of-4   9.9 GB: 2-of-4   9.9 GB: 3-of-4   6.9 GB: 4-of-4
Supported Languagesen
AWQ QuantizationYes
Quantization Typeawq
Model ArchitectureLlamaForCausalLM
Licensellama2
Context Length4096
Model Max Length4096
Transformers Version4.32.0.dev0
Tokenizer ClassLlamaTokenizer
Beginning of Sentence Token<s>
End of Sentence Token</s>
Unk Token<unk>
Vocabulary Size32000
Torch Data Typefloat16

Best Alternatives to Llama 2 70B AWQ

Best Alternatives
Context / RAM
Downloads
Likes
...0B Instruct Gradient 1048K AWQ1024K / 39.9 GB121
...70B Instruct Gradient 262K AWQ256K / 39.9 GB110
Llama 3.3 70B Instruct AWQ128K / 39.9 GB2078018
...lama 3.3 70B Instruct AWQ INT4128K / 39.9 GB756110
... SauerkrautLM 70B Instruct AWQ128K / 39.9 GB16034
MultiVerse 70B AWQ32K / 41.3 GB922
Opus V1.2 70B AWQ32K / 36.7 GB231
QuartetAnemoi 70B T0.0001 AWQ31K / 36.7 GB131
Senku 70B AWQ 4bit GEMM31K / 36.7 GB121
Kiqu 70B AWQ31K / 36.7 GB141
Note: green Score (e.g. "73.2") means that the model is better than TheBloke/Llama-2-70B-AWQ.

Rank the Llama 2 70B AWQ 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? 40123 in total.

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