Update README.md
Browse files
README.md
CHANGED
@@ -12,27 +12,6 @@ license: apache-2.0
|
|
12 |
|
13 |
```
|
14 |
|
15 |
-
### Script to AWQ quantization
|
16 |
-
```
|
17 |
-
from awq import AutoAWQForCausalLM
|
18 |
-
from transformers import AutoTokenizer
|
19 |
-
|
20 |
-
model_path = 'PATH_TO Poro-34B'
|
21 |
-
quant_path = 'Poro-34B-AWQ'
|
22 |
-
quant_config = { "zero_point": True, "q_group_size": 128, "w_bit": 4, "version": "GEMM" }
|
23 |
-
|
24 |
-
# Load model
|
25 |
-
model = AutoAWQForCausalLM.from_pretrained(model_path, safetensors=True)
|
26 |
-
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
|
27 |
-
|
28 |
-
# Quantize
|
29 |
-
model.quantize(tokenizer, quant_config=quant_config)
|
30 |
-
|
31 |
-
# Save quantized model
|
32 |
-
model.save_quantized(quant_path)
|
33 |
-
tokenizer.save_pretrained(quant_path)
|
34 |
-
```
|
35 |
-
|
36 |
### Generate
|
37 |
```
|
38 |
from awq import AutoAWQForCausalLM
|
|
|
12 |
|
13 |
```
|
14 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
### Generate
|
16 |
```
|
17 |
from awq import AutoAWQForCausalLM
|