Update README.md
Browse files
README.md
CHANGED
@@ -7,14 +7,22 @@ tags: []
|
|
7 |
|
8 |
4-bit weight only quantization by AutoGPTQ on [yujiepan/llama-3-tiny-random](https://huggingface.co/yujiepan/llama-3-tiny-random)
|
9 |
|
10 |
-
```
|
11 |
from transformers import AutoModelForCausalLM, AutoTokenizer, GPTQConfig
|
12 |
import torch
|
13 |
|
14 |
model_id = "yujiepan/llama-3-tiny-random"
|
15 |
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
16 |
|
17 |
-
quantization_config = GPTQConfig(
|
18 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
```
|
20 |
|
|
|
7 |
|
8 |
4-bit weight only quantization by AutoGPTQ on [yujiepan/llama-3-tiny-random](https://huggingface.co/yujiepan/llama-3-tiny-random)
|
9 |
|
10 |
+
```python
|
11 |
from transformers import AutoModelForCausalLM, AutoTokenizer, GPTQConfig
|
12 |
import torch
|
13 |
|
14 |
model_id = "yujiepan/llama-3-tiny-random"
|
15 |
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
16 |
|
17 |
+
quantization_config = GPTQConfig(
|
18 |
+
bits=4, group_size=-1,
|
19 |
+
dataset="c4",
|
20 |
+
tokenizer=tokenizer,
|
21 |
+
)
|
22 |
+
model = AutoModelForCausalLM.from_pretrained(
|
23 |
+
model_id,
|
24 |
+
device_map="auto",
|
25 |
+
quantization_config=quantization_config,
|
26 |
+
)
|
27 |
```
|
28 |
|