Lin-K76 commited on
Commit
3d039b6
1 Parent(s): f53748a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -1
README.md CHANGED
@@ -18,6 +18,7 @@ license: apache-2.0
18
  - **Out-of-scope:** Use in any manner that violates applicable laws or regulations (including trade compliance laws). Use in languages other than English.
19
  - **Release Date:** 6/14/2024
20
  - **Version:** 1.0
 
21
  - **Model Developers:** Neural Magic
22
 
23
  Quantized version of [Qwen2-7B-Instruct](https://huggingface.co/Qwen/Qwen2-7B-Instruct).
@@ -28,7 +29,7 @@ It achieves an average score of 69.44 on the [OpenLLM](https://huggingface.co/sp
28
  This model was obtained by quantizing the weights and activations of [Qwen2-7B-Instruct](https://huggingface.co/Qwen/Qwen2-7B-Instruct) to FP8 data type, ready for inference with vLLM >= 0.5.0.
29
  This optimization reduces the number of bits per parameter from 16 to 8, reducing the disk size and GPU memory requirements by approximately 50%.
30
 
31
- Only the weights and activations of the linear operators within transformers blocks are quantized. Symmetric per-tensor quantization is applied, in which a linear scaling per output dimension maps the FP8 representations of the quantized weights and activations.
32
  [AutoFP8](https://github.com/neuralmagic/AutoFP8) is used for quantization with 512 sequences of UltraChat.
33
 
34
  ## Deployment
 
18
  - **Out-of-scope:** Use in any manner that violates applicable laws or regulations (including trade compliance laws). Use in languages other than English.
19
  - **Release Date:** 6/14/2024
20
  - **Version:** 1.0
21
+ - **License(s):** [apache-2.0](https://huggingface.co/datasets/choosealicense/licenses/blob/main/markdown/apache-2.0.md)
22
  - **Model Developers:** Neural Magic
23
 
24
  Quantized version of [Qwen2-7B-Instruct](https://huggingface.co/Qwen/Qwen2-7B-Instruct).
 
29
  This model was obtained by quantizing the weights and activations of [Qwen2-7B-Instruct](https://huggingface.co/Qwen/Qwen2-7B-Instruct) to FP8 data type, ready for inference with vLLM >= 0.5.0.
30
  This optimization reduces the number of bits per parameter from 16 to 8, reducing the disk size and GPU memory requirements by approximately 50%.
31
 
32
+ Only the weights and activations of the linear operators within transformers blocks are quantized. Symmetric per-tensor quantization is applied, in which a single linear scaling maps the FP8 representations of the quantized weights and activations.
33
  [AutoFP8](https://github.com/neuralmagic/AutoFP8) is used for quantization with 512 sequences of UltraChat.
34
 
35
  ## Deployment