Spaces:
Running
Running
File size: 1,239 Bytes
d9ded01 ea0af80 d9ded01 d36dc81 d9ded01 d36dc81 ea0af80 d9ded01 ea0af80 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 |
import os
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
# Correct model name
MODEL_NAME = "bigcode/starcoder"
# Ensure the token is provided
HF_TOKEN = os.getenv("HUGGINGFACE_TOKEN")
if not HF_TOKEN:
raise ValueError("Missing Hugging Face token. Set HUGGINGFACE_TOKEN as an environment variable.")
# Set device
device = "cuda" if torch.cuda.is_available() else "cpu"
# Load tokenizer with authentication
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, token=HF_TOKEN)
# Load model with optimizations
model = AutoModelForCausalLM.from_pretrained(
MODEL_NAME,
token=HF_TOKEN,
torch_dtype=torch.float16, # Reduce memory usage
low_cpu_mem_usage=True, # Optimize loading
device_map="auto", # Automatic device placement
offload_folder="offload" # Offload to disk if needed
).to(device)
def generate_code(prompt: str, max_tokens: int = 256):
"""Generates code based on the input prompt."""
if not prompt.strip():
return "Error: Empty prompt provided."
inputs = tokenizer(prompt, return_tensors="pt").to(device)
output = model.generate(**inputs, max_new_tokens=max_tokens)
return tokenizer.decode(output[0], skip_special_tokens=True)
|