Spaces:
Sleeping
Sleeping
Update myagent.py
Browse files- myagent.py +37 -20
myagent.py
CHANGED
|
@@ -61,26 +61,43 @@ class LocalLlamaModel:
|
|
| 61 |
self.device = model.device if hasattr(model, 'device') else 'cpu'
|
| 62 |
|
| 63 |
def generate(self, prompt: str, max_new_tokens=512, **kwargs):
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 84 |
def __call__(self, prompt: str, max_new_tokens=512, **kwargs):
|
| 85 |
"""Make the model callable like a function"""
|
| 86 |
return self.generate(prompt, max_new_tokens, **kwargs)
|
|
|
|
| 61 |
self.device = model.device if hasattr(model, 'device') else 'cpu'
|
| 62 |
|
| 63 |
def generate(self, prompt: str, max_new_tokens=512, **kwargs):
|
| 64 |
+
try:
|
| 65 |
+
# Generate answer using the provided prompt - following the recommended pattern
|
| 66 |
+
input_ids = self.tokenizer.apply_chat_template(
|
| 67 |
+
[{"role": "user", "content": str(prompt)}],
|
| 68 |
+
add_generation_prompt=True,
|
| 69 |
+
return_tensors="pt",
|
| 70 |
+
tokenize=True,
|
| 71 |
+
).to(self.model.device)
|
| 72 |
+
|
| 73 |
+
# Generate output - exactly as in recommended code
|
| 74 |
+
output = self.model.generate(
|
| 75 |
+
input_ids,
|
| 76 |
+
do_sample=True,
|
| 77 |
+
temperature=0.3,
|
| 78 |
+
min_p=0.15,
|
| 79 |
+
repetition_penalty=1.05,
|
| 80 |
+
max_new_tokens=max_new_tokens,
|
| 81 |
+
)
|
| 82 |
+
|
| 83 |
+
# Decode the full output - as in recommended code
|
| 84 |
+
decoded_output = self.tokenizer.decode(output[0], skip_special_tokens=False)
|
| 85 |
+
|
| 86 |
+
# Extract only the assistant's response (after the last <|im_start|>assistant)
|
| 87 |
+
if "<|im_start|>assistant" in decoded_output:
|
| 88 |
+
assistant_response = decoded_output.split("<|im_start|>assistant")[-1]
|
| 89 |
+
# Remove any trailing special tokens
|
| 90 |
+
assistant_response = assistant_response.replace("<|im_end|>", "").strip()
|
| 91 |
+
return assistant_response
|
| 92 |
+
else:
|
| 93 |
+
# Fallback: return the full decoded output
|
| 94 |
+
return decoded_output
|
| 95 |
+
|
| 96 |
+
except Exception as e:
|
| 97 |
+
print(f"Error in model generation: {e}")
|
| 98 |
+
return f"Error generating response: {str(e)}"
|
| 99 |
+
|
| 100 |
+
|
| 101 |
def __call__(self, prompt: str, max_new_tokens=512, **kwargs):
|
| 102 |
"""Make the model callable like a function"""
|
| 103 |
return self.generate(prompt, max_new_tokens, **kwargs)
|