Update handler.py
Browse files- handler.py +8 -9
handler.py
CHANGED
|
@@ -39,21 +39,20 @@ alpaca_prompt = (
|
|
| 39 |
)
|
| 40 |
|
| 41 |
def initialize():
|
| 42 |
-
"""
|
| 43 |
-
This function loads the finetuned unsloth model and tokenizer.
|
| 44 |
-
Adjust the 'model_name' to point to your model repository.
|
| 45 |
-
"""
|
| 46 |
global model, tokenizer
|
| 47 |
-
|
| 48 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
model, tokenizer = FastLanguageModel.from_pretrained(
|
| 50 |
model_name=model_name,
|
| 51 |
max_seq_length=2048,
|
| 52 |
-
load_in_4bit=True
|
|
|
|
| 53 |
)
|
| 54 |
-
# Set the model in inference mode
|
| 55 |
FastLanguageModel.for_inference(model)
|
| 56 |
-
model.eval()
|
| 57 |
|
| 58 |
def preprocess(request):
|
| 59 |
"""
|
|
|
|
| 39 |
)
|
| 40 |
|
| 41 |
def initialize():
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
global model, tokenizer
|
| 43 |
+
model_name = "RichardLu/Mistral7b_AE_res"
|
| 44 |
+
|
| 45 |
+
# Read the environment variable
|
| 46 |
+
hf_auth_token = os.environ.get("HF_AUTH_TOKEN", None)
|
| 47 |
+
|
| 48 |
+
# Pass the token if it exists
|
| 49 |
model, tokenizer = FastLanguageModel.from_pretrained(
|
| 50 |
model_name=model_name,
|
| 51 |
max_seq_length=2048,
|
| 52 |
+
load_in_4bit=True,
|
| 53 |
+
use_auth_token=hf_auth_token # <--- Use your token here
|
| 54 |
)
|
|
|
|
| 55 |
FastLanguageModel.for_inference(model)
|
|
|
|
| 56 |
|
| 57 |
def preprocess(request):
|
| 58 |
"""
|