Commit
•
05364df
1
Parent(s):
3b20ca9
Update README.md
Browse files
README.md
CHANGED
@@ -30,6 +30,17 @@ model = AutoModelForCausalLM.from_pretrained("VishalMysore/cookgptlama")
|
|
30 |
tokenizer = AutoTokenizer.from_pretrained("VishalMysore/cookgptlama")
|
31 |
```
|
32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
Then use pipeline to query and interact
|
35 |
|
|
|
30 |
tokenizer = AutoTokenizer.from_pretrained("VishalMysore/cookgptlama")
|
31 |
```
|
32 |
|
33 |
+
or you can load it in 8bit precision
|
34 |
+
|
35 |
+
```
|
36 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
37 |
+
|
38 |
+
model_id = "VishalMysore/cookgptlama"
|
39 |
+
|
40 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
41 |
+
model_8bit = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto",load_in_8bit=True)
|
42 |
+
print(model_8bit.get_memory_footprint())
|
43 |
+
```
|
44 |
|
45 |
Then use pipeline to query and interact
|
46 |
|