File size: 4,205 Bytes
f18de00 9c3d676 8081ccc 0f5897f 9c3d676 e1b0723 8846773 9c3d676 8081ccc 9c3d676 8846773 9c3d676 8846773 3ec3dc0 8846773 14e48d1 ae69077 70f3c4a 5080cee 70f3c4a 5080cee 3ec3dc0 9c3d676 8081ccc 70f3c4a 8081ccc 70f3c4a 5080cee 8081ccc 8846773 9c3d676 8846773 9c3d676 8081ccc 9c3d676 8081ccc 9c3d676 8081ccc 9c3d676 8081ccc 447ff45 8081ccc 9c3d676 8081ccc 9c3d676 447ff45 8081ccc 8846773 8081ccc 8846773 8081ccc 8846773 8081ccc 8846773 3f58090 8081ccc |
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 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 |
import gradio as gr
from huggingface_hub import hf_hub_download
from llama_cpp import Llama
import re
from datasets import load_dataset
import random
import logging
import os
import autopep8
import textwrap
# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
# Model settings
MODEL_NAME = "leetmonkey_peft__q8_0.gguf"
REPO_ID = "sugiv/leetmonkey-peft-gguf"
def download_model(model_name):
logger.info(f"Downloading model: {model_name}")
model_path = hf_hub_download(
repo_id=REPO_ID,
filename=model_name,
cache_dir="./models",
force_download=True,
resume_download=True
)
logger.info(f"Model downloaded: {model_path}")
return model_path
# Download and load the 8-bit model at startup
model_path = download_model(MODEL_NAME)
llm = Llama(
model_path=model_path,
n_ctx=1024,
n_threads=8,
n_gpu_layers=-1, # Use all available GPU layers
verbose=False,
n_batch=512,
mlock=True
)
logger.info("8-bit model loaded successfully")
# Load the dataset
dataset = load_dataset("sugiv/leetmonkey_python_dataset")
train_dataset = dataset["train"]
# Generation parameters
generation_kwargs = {
"max_tokens": 512,
"stop": ["```", "### Instruction:", "### Response:"],
"echo": False,
"temperature": 0.05,
"top_k": 10,
"top_p": 0.9,
"repeat_penalty": 1.1
}
def generate_solution(instruction):
system_prompt = "You are a Python coding assistant specialized in solving LeetCode problems. Provide only the complete implementation of the given function. Ensure proper indentation and formatting. Do not include any explanations or multiple solutions."
full_prompt = f"""### Instruction:
{system_prompt}
Implement the following function for the LeetCode problem:
{instruction}
### Response:
Here's the complete Python function implementation:
```python
"""
for chunk in llm(full_prompt, stream=True, **generation_kwargs):
yield chunk["choices"][0]["text"]
def extract_and_format_code(text):
# Extract code between triple backticks
code_match = re.search(r'```python\s*(.*?)\s*```', text, re.DOTALL)
if code_match:
code = code_match.group(1)
else:
code = text
# Dedent the code to remove any common leading whitespace
code = textwrap.dedent(code)
# Split the code into lines
lines = code.split('\n')
# Ensure proper indentation
indented_lines = []
for line in lines:
if line.strip().startswith('class') or line.strip().startswith('def'):
indented_lines.append(line) # Keep class and function definitions as is
elif line.strip(): # If the line is not empty
indented_lines.append(' ' + line) # Add 4 spaces of indentation
else:
indented_lines.append(line) # Keep empty lines as is
formatted_code = '\n'.join(indented_lines)
try:
return autopep8.fix_code(formatted_code)
except:
return formatted_code
def select_random_problem():
return random.choice(train_dataset)['instruction']
def stream_solution(problem):
logger.info("Generating solution")
generated_text = ""
for token in generate_solution(problem):
generated_text += token
yield generated_text
formatted_code = extract_and_format_code(generated_text)
logger.info("Solution generated successfully")
yield formatted_code
with gr.Blocks() as demo:
gr.Markdown("# LeetCode Problem Solver (8-bit GGUF Model)")
with gr.Row():
with gr.Column():
problem_display = gr.Textbox(label="LeetCode Problem", lines=10)
select_problem_btn = gr.Button("Select Random Problem")
with gr.Column():
solution_display = gr.Code(label="Generated Solution", language="python", lines=25)
generate_btn = gr.Button("Generate Solution")
select_problem_btn.click(select_random_problem, outputs=problem_display)
generate_btn.click(stream_solution, inputs=[problem_display], outputs=solution_display)
if __name__ == "__main__":
logger.info("Starting Gradio interface")
demo.launch(share=True)
|