meta-llama/Llama-3.2-3B - 4b_2n4m_128bs Compression

This is a compressed model using deltazip.

Paper, Compression Tool, Inference Engine (Soon).

Compression Configuration

  • Base Model: meta-llama/Llama-3.2-3B
  • Compression Scheme: 4b_2n4m_128bs
  • Dataset: HuggingFaceH4/ultrachat_200k
  • Dataset Split: train_sft
  • Max Sequence Length: 2048
  • Number of Samples: 256

Sample Output

Prompt:

[{'role': 'user', 'content': 'Who is Alan Turing?'}]

Output:

<|begin_of_text|><|start_header_id|>system<|end_header_id|>

Cutting Knowledge Date: December 2023
Today Date: 05 Dec 2024

<|eot_id|><|start_header_id|>user<|end_header_id|>

Who is Alan Turing?<|eot_id|><|start_header_id|>assistant<|end_header_id|>

Alan Turing (1912-1954) was a British mathematician, computer scientist, logician, and philosopher. He is widely regarded as one of the most influential figures in the development of computer science, artificial intelligence, and cryptography.

Turing was born in London, England, and studied mathematics at Cambridge University. During World War II, he worked at the Government Code and Cypher School (GC&CS) at Bletch

Evaluation

Downloads last month
19
Inference API
Unable to determine this model’s pipeline type. Check the docs .

Model tree for deltazip/meta-llama.Llama-3.2-3B-Instruct.4b_2n4m_128bs

Finetuned
(84)
this model

Dataset used to train deltazip/meta-llama.Llama-3.2-3B-Instruct.4b_2n4m_128bs

Collection including deltazip/meta-llama.Llama-3.2-3B-Instruct.4b_2n4m_128bs