Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -12,6 +12,10 @@ model = AutoModelForCausalLM.from_pretrained(model_name)
|
|
12 |
|
13 |
# Function to generate text
|
14 |
def generate_text(prompt, max_new_tokens=200):
|
|
|
|
|
|
|
|
|
15 |
# Tokenize the input prompt
|
16 |
inputs = tokenizer(prompt, return_tensors="pt", padding=True)
|
17 |
|
@@ -27,7 +31,7 @@ def generate_text(prompt, max_new_tokens=200):
|
|
27 |
top_k=50, # Consider only the top 50 tokens for sampling
|
28 |
top_p=0.95,
|
29 |
temperature=0.8, # Increase diversity slightly
|
30 |
-
pad_token_id=tokenizer.
|
31 |
)
|
32 |
|
33 |
# Decode the output text
|
|
|
12 |
|
13 |
# Function to generate text
|
14 |
def generate_text(prompt, max_new_tokens=200):
|
15 |
+
# Set pad_token_id to eos_token_id if not set
|
16 |
+
if tokenizer.pad_token_id is None:
|
17 |
+
tokenizer.pad_token = tokenizer.eos_token # Set pad token to be the eos token
|
18 |
+
|
19 |
# Tokenize the input prompt
|
20 |
inputs = tokenizer(prompt, return_tensors="pt", padding=True)
|
21 |
|
|
|
31 |
top_k=50, # Consider only the top 50 tokens for sampling
|
32 |
top_p=0.95,
|
33 |
temperature=0.8, # Increase diversity slightly
|
34 |
+
pad_token_id=tokenizer.pad_token_id # Ensure correct behavior for padding
|
35 |
)
|
36 |
|
37 |
# Decode the output text
|