Leetmonkey In Action. Darn LeetMonkey these days

app.py

@@ -6,6 +6,7 @@ from datasets import load_dataset
 import random
 import autopep8
 import textwrap
+import threading
 
 # Define the model options
 gguf_models = {
@@ -15,11 +16,23 @@ gguf_models = {
     "Super Block Q6": "leetmonkey_peft_super_block_q6.gguf"
 }
 
+# Global dictionary to store loaded models
+loaded_models = {}
+
 # Function to download and load the model
 def load_model(model_name):
     model_path = hf_hub_download(repo_id="sugiv/leetmonkey-peft-gguf", filename=model_name)
     return Llama(model_path=model_path, n_ctx=2048, n_threads=4, n_gpu_layers=0, verbose=False)
 
+# Function to preload all models
+def preload_models():
+    for name, file in gguf_models.items():
+        loaded_models[name] = load_model(file)
+    print("All models loaded successfully!")
+
+# Start preloading models in a separate thread
+threading.Thread(target=preload_models, daemon=True).start()
+
 # Generation parameters
 generation_kwargs = {
     "max_tokens": 2048,
@@ -33,8 +46,7 @@ generation_kwargs = {
 
 def generate_solution(instruction, model):
     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"""<｜begin▁of▁sentence｜>
-### Instruction:
+    full_prompt = f"""### Instruction:
 {system_prompt}
 
 Implement the following function for the LeetCode problem:
@@ -89,15 +101,16 @@ def extract_and_format_code(text):
 dataset = load_dataset("sugiv/leetmonkey_python_dataset")
 val_dataset = dataset["train"].train_test_split(test_size=0.1)["test"]
 
-def gradio_interface(model_name):
-    model = load_model(gguf_models[model_name])
+def update_problem():
     sample = random.choice(val_dataset)
-    instruction = sample['instruction']
-    
-    generated_output = generate_solution(instruction, model)
-    python_code = extract_and_format_code(generated_output)
-    
-    return instruction, python_code
+    return sample['instruction']
+
+def update_solution(problem, model_name):
+    model = loaded_models.get(model_name)
+    if model is None:
+        return "Model is still loading. Please wait and try again."
+    generated_output = generate_solution(problem, model)
+    return extract_and_format_code(generated_output)
 
 with gr.Blocks() as demo:
     gr.Markdown("# LeetCode Problem Solver")
@@ -109,18 +122,9 @@ with gr.Blocks() as demo:
         
         with gr.Column():
             model_dropdown = gr.Dropdown(choices=list(gguf_models.keys()), label="Select GGUF Model", value="Exact Copy")
-            solution_display = gr.Code(label="Generated Solution", language="python")
+            solution_display = gr.Code(label="Generated Solution", language="python", lines=25)
             generate_btn = gr.Button("Generate Solution")
     
-    def update_problem():
-        sample = random.choice(val_dataset)
-        return sample['instruction']
-    
-    def update_solution(problem, model_name):
-        model = load_model(gguf_models[model_name])
-        generated_output = generate_solution(problem, model)
-        return extract_and_format_code(generated_output)
-    
     select_problem_btn.click(update_problem, outputs=problem_display)
     generate_btn.click(update_solution, inputs=[problem_display, model_dropdown], outputs=solution_display)