Llama 3 1 Nemotron 51B Instruct by nvidia

 ยป  All LLMs  ยป  nvidia  ยป  Llama 3 1 Nemotron 51B Instruct   URL Share it on

  Arxiv:2009.03300   Arxiv:2110.14168   Arxiv:2306.05685   Arxiv:2404.05993   Autotrain compatible   Conversational   Custom code   En   Instruct   Llama-3   Nemotron-nas   Nvidia   Pytorch   Region:us   Safetensors   Sharded   Tensorflow

Llama 3 1 Nemotron 51B Instruct Benchmarks

nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Llama 3 1 Nemotron 51B Instruct (nvidia/Llama-3_1-Nemotron-51B-Instruct)

Llama 3 1 Nemotron 51B Instruct Parameters and Internals

Model Type 
text-generation
Use Cases 
Areas:
General purpose chat
Applications:
English and coding language conversations
Primary Use Cases:
Chat and dialogue generation
Limitations:
May produce biased, inaccurate, or undesirable text., Weak against alignment-breaking attacks.
Considerations:
Use guardrails to prevent potentially harmful outputs.
Additional Notes 
Optimized for a single H100-80GB GPU using novel NAS approach.
Supported Languages 
en (proficient), coding languages (supported)
Training Details 
Data Sources:
FineWeb, Buzz-V1.2, Dolma
Data Volume:
40 billion tokens
Methodology:
block-wise distillation
Context Length:
8192
Training Time:
August to September 2024
Model Architecture:
Transformer Decoder (auto-regressive language model)
Safety Evaluation 
Methodologies:
Garak, AEGIS, Human Content Red Teaming
Risk Categories:
toxic language, unsafe content, societal biases
Ethical Considerations:
Developers responsible for ensuring model meets industry use case requirements.
Responsible Ai Considerations 
Fairness:
Model trained on biased data, potential to amplify biases.
Accountability:
Responsible AI development and shared trustworthiness responsibility.
Mitigation Strategies:
Deploy guardrails to prevent harmful outputs.
Input Output 
Input Format:
Text only
Accepted Modalities:
text
Output Format:
Text only
Performance Tips:
High throughput and efficiency using the NAS approach.
LLM NameLlama 3 1 Nemotron 51B Instruct
Repository ๐Ÿค—https://huggingface.co/nvidia/Llama-3_1-Nemotron-51B-Instruct 
Model Size51b
Required VRAM103.4 GB
Updated2024-12-22
Maintainernvidia
Model Typenemotron-nas
Instruction-BasedYes
Model Files  5.0 GB: 1-of-22   4.9 GB: 2-of-22   4.9 GB: 3-of-22   4.6 GB: 4-of-22   5.0 GB: 5-of-22   5.0 GB: 6-of-22   4.7 GB: 7-of-22   4.7 GB: 8-of-22   4.7 GB: 9-of-22   5.0 GB: 10-of-22   5.0 GB: 11-of-22   4.7 GB: 12-of-22   4.7 GB: 13-of-22   4.7 GB: 14-of-22   4.9 GB: 15-of-22   5.0 GB: 16-of-22   4.9 GB: 17-of-22   4.7 GB: 18-of-22   4.7 GB: 19-of-22   4.7 GB: 20-of-22   4.8 GB: 21-of-22   2.1 GB: 22-of-22
Supported Languagesen
Model ArchitectureDeciLMForCausalLM
Licenseother
Context Length131072
Model Max Length131072
Transformers Version4.44.2
Tokenizer ClassPreTrainedTokenizerFast
Vocabulary Size128256
Torch Data Typebfloat16

Rank the Llama 3 1 Nemotron 51B Instruct 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? 40066 in total.

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