meta-llama/Meta-Llama-3-8B - 4b_2n4m_128bs Compression

This is a compressed model using deltazip.

Paper, Compression Tool, Inference Engine (Soon).

Compression Configuration

  • Base Model: meta-llama/Meta-Llama-3-8B
  • 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|>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 who made significant contributions to the development of computer science, artificial intelligence, and cryptography. He is widely considered one of the most important figures in the history of computer science and artificial intelligence.

Turing's work had a profound impact on the development of modern computing and artificial intelligence. Some of his most notable contributions include:

1. The Turing Machine: Turing proposed the Turing machine, a theoretical model for a computer that can perform calculations and manipulate data. The

Evaluation

Downloads last month
1
Inference Providers NEW
This model is not currently available via any of the supported third-party Inference Providers, and HF Inference API was unable to determine this model’s pipeline type.

Model tree for deltazip/meta-llama.Meta-Llama-3-8B-Instruct.4b_2n4m_128bs

Finetuned
(387)
this model

Dataset used to train deltazip/meta-llama.Meta-Llama-3-8B-Instruct.4b_2n4m_128bs

Collection including deltazip/meta-llama.Meta-Llama-3-8B-Instruct.4b_2n4m_128bs