Hugging Face's logo

ISTA-DASLab
/
Llama-2-7b-AQLM-2Bit-8x8-hf

Text Generation
Transformers
Safetensors
llama
text-generation-inference
Inference Endpoints
8-bit precision
aqlm
Model card Files Community
2

Update README.md

#1
by SpiridonSunRotator - opened
base: refs/heads/main
←
from: refs/pr/1
Discussion Files changed
+7
-3
Files changed (1) hide show
  1. README.md +7 -3
README.md CHANGED
@@ -1,3 +1,7 @@
 
 
 
 
1
  Official [AQLM](https://arxiv.org/abs/2401.06118) quantization of `meta-llama/Llama-2-7b-hf`.
2
 
3
  For this quantization, we used 2 codebooks of 8 bits.
@@ -6,9 +10,9 @@ Selected evaluation results for this and other models:
6
 
7
  | Model | AQLM scheme | WikiText 2 PPL | Model size, Gb | Hub link |
8
  |------------|-------------|----------------|----------------|--------------------------------------------------------------------------|
9
- | Llama-2-7b | 1x16 | 6.31 | 2.4 | [Link](https://huggingface.co/BlackSamorez/Llama-2-7b-AQLM-2Bit-1x16-hf) |
10
- | Llama-2-7b | 2x8 | 7.98 | 2.2 | [Link](https://huggingface.co/BlackSamorez/Llama-2-7b-AQLM-2Bit-2x8-hf) |
11
- | Llama-2-7b (THIS) | 8x8 | 7.83 | 2.2 | [Link](https://huggingface.co/BlackSamorez/Llama-2-7b-AQLM-2Bit-8x8-hf) |
12
  | Llama-2-13b| 1x16 | 5.41 | 4.1 | [Link](https://huggingface.co/BlackSamorez/Llama-2-13b-AQLM-2Bit-1x16-hf)|
13
  | Llama-2-70b| 1x16 | 3.96 | 18.8 | [Link](https://huggingface.co/BlackSamorez/Llama-2-70b-AQLM-2Bit-1x16-hf)|
14
  | Llama-2-70b| 2x8 | 4.83 | 18.2 | [Link](https://huggingface.co/BlackSamorez/Llama-2-70b-AQLM-2Bit-2x8-hf) |
 
1
+ ---
2
+ metrics:
3
+ - perplexity
4
+ ---
5
  Official [AQLM](https://arxiv.org/abs/2401.06118) quantization of `meta-llama/Llama-2-7b-hf`.
6
 
7
  For this quantization, we used 2 codebooks of 8 bits.
 
10
 
11
  | Model | AQLM scheme | WikiText 2 PPL | Model size, Gb | Hub link |
12
  |------------|-------------|----------------|----------------|--------------------------------------------------------------------------|
13
+ | Llama-2-7b | 1x16 | 5.92 | 2.4 | [Link](https://huggingface.co/BlackSamorez/Llama-2-7b-AQLM-2Bit-1x16-hf) |
14
+ | Llama-2-7b | 2x8 | 6.69 | 2.2 | [Link](https://huggingface.co/BlackSamorez/Llama-2-7b-AQLM-2Bit-2x8-hf) |
15
+ | Llama-2-7b (THIS) | 8x8 | 6.61 | 2.2 | [Link](https://huggingface.co/BlackSamorez/Llama-2-7b-AQLM-2Bit-8x8-hf) |
16
  | Llama-2-13b| 1x16 | 5.41 | 4.1 | [Link](https://huggingface.co/BlackSamorez/Llama-2-13b-AQLM-2Bit-1x16-hf)|
17
  | Llama-2-70b| 1x16 | 3.96 | 18.8 | [Link](https://huggingface.co/BlackSamorez/Llama-2-70b-AQLM-2Bit-1x16-hf)|
18
  | Llama-2-70b| 2x8 | 4.83 | 18.2 | [Link](https://huggingface.co/BlackSamorez/Llama-2-70b-AQLM-2Bit-2x8-hf) |