Update app.py
Browse files
app.py
CHANGED
@@ -53,7 +53,7 @@ def cache_answer(question: str, answer: str):
|
|
53 |
|
54 |
|
55 |
# --- Model Setup ---
|
56 |
-
MODEL_NAME = 'distilbert/distilgpt2'#'deepseek-ai/DeepSeek-R1-Distill-Qwen-7B'#'mistralai/Mistral-7B-Instruct-v0.2'
|
57 |
|
58 |
def load_model():
|
59 |
"""Download and load the model and tokenizer."""
|
@@ -87,8 +87,19 @@ class BasicAgent:
|
|
87 |
|
88 |
def __call__(self, question: str) -> str:
|
89 |
print(f"Agent received question (first 50 chars): {question[:50]}...")
|
90 |
-
|
91 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
92 |
return answer
|
93 |
|
94 |
|
|
|
53 |
|
54 |
|
55 |
# --- Model Setup ---
|
56 |
+
MODEL_NAME = 'EleutherAI/gpt-neo-2.7B'#'distilbert/distilgpt2'#'deepseek-ai/DeepSeek-R1-Distill-Qwen-7B'#'mistralai/Mistral-7B-Instruct-v0.2'
|
57 |
|
58 |
def load_model():
|
59 |
"""Download and load the model and tokenizer."""
|
|
|
87 |
|
88 |
def __call__(self, question: str) -> str:
|
89 |
print(f"Agent received question (first 50 chars): {question[:50]}...")
|
90 |
+
|
91 |
+
encoded_input = tokenizer(question, return_tensors="pt", padding=True, truncation=True, max_length=512)
|
92 |
+
|
93 |
+
# Generate text using the model
|
94 |
+
output = model.generate(
|
95 |
+
encoded_input['input_ids'],
|
96 |
+
attention_mask=encoded_input['attention_mask'], # Pass attention mask
|
97 |
+
max_length=100,
|
98 |
+
pad_token_id=tokenizer.pad_token_id,
|
99 |
+
|
100 |
+
)
|
101 |
+
|
102 |
+
answer = tokenizer.decode(output[0], skip_special_tokens=True)
|
103 |
return answer
|
104 |
|
105 |
|