Spaces:
Sleeping
Sleeping
Commit
·
e2116c0
1
Parent(s):
ad9f174
updated model.py
Browse files
model.py
CHANGED
@@ -5,7 +5,6 @@ import torch
|
|
5 |
MODEL_NAME = "bigcode/starcoderbase-1b"
|
6 |
HF_TOKEN = os.getenv("HUGGINGFACE_TOKEN")
|
7 |
|
8 |
-
# Force CPU mode
|
9 |
device = "cpu"
|
10 |
|
11 |
# Load tokenizer and model
|
@@ -18,23 +17,34 @@ if tokenizer.pad_token is None:
|
|
18 |
model = AutoModelForCausalLM.from_pretrained(
|
19 |
MODEL_NAME,
|
20 |
token=HF_TOKEN,
|
21 |
-
torch_dtype=torch.float32, #
|
22 |
trust_remote_code=True
|
23 |
-
).to(device)
|
24 |
|
25 |
def generate_code(prompt: str, max_tokens: int = 256):
|
|
|
|
|
26 |
inputs = tokenizer(
|
27 |
-
|
28 |
return_tensors="pt",
|
29 |
padding=True,
|
30 |
-
truncation=True,
|
31 |
-
max_length=1024 #
|
32 |
).to(device)
|
33 |
|
34 |
output = model.generate(
|
35 |
**inputs,
|
36 |
max_new_tokens=max_tokens,
|
37 |
-
pad_token_id=tokenizer.pad_token_id
|
|
|
|
|
|
|
38 |
)
|
39 |
|
40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
MODEL_NAME = "bigcode/starcoderbase-1b"
|
6 |
HF_TOKEN = os.getenv("HUGGINGFACE_TOKEN")
|
7 |
|
|
|
8 |
device = "cpu"
|
9 |
|
10 |
# Load tokenizer and model
|
|
|
17 |
model = AutoModelForCausalLM.from_pretrained(
|
18 |
MODEL_NAME,
|
19 |
token=HF_TOKEN,
|
20 |
+
torch_dtype=torch.float32, # Ensure compatibility with CPU
|
21 |
trust_remote_code=True
|
22 |
+
).to(device)
|
23 |
|
24 |
def generate_code(prompt: str, max_tokens: int = 256):
|
25 |
+
formatted_prompt = f"# Python\n{prompt}\n\n" # Ensure the model understands it's code
|
26 |
+
|
27 |
inputs = tokenizer(
|
28 |
+
formatted_prompt,
|
29 |
return_tensors="pt",
|
30 |
padding=True,
|
31 |
+
truncation=True,
|
32 |
+
max_length=1024 # Explicit max length to prevent issues
|
33 |
).to(device)
|
34 |
|
35 |
output = model.generate(
|
36 |
**inputs,
|
37 |
max_new_tokens=max_tokens,
|
38 |
+
pad_token_id=tokenizer.pad_token_id,
|
39 |
+
do_sample=True, # Enable randomness for better outputs
|
40 |
+
top_p=0.95, # Nucleus sampling to improve generation
|
41 |
+
temperature=0.7 # Control creativity
|
42 |
)
|
43 |
|
44 |
+
generated_code = tokenizer.decode(output[0], skip_special_tokens=True)
|
45 |
+
|
46 |
+
# Clean the output: remove the repeated prompt at the start
|
47 |
+
if generated_code.startswith(formatted_prompt):
|
48 |
+
generated_code = generated_code[len(formatted_prompt):]
|
49 |
+
|
50 |
+
return generated_code.strip()
|