Update space

app.py

@@ -7,9 +7,19 @@ import os # Useful for environment variables if needed
 # Replace "your-model-id" with the actual ID of the model on Hugging Face Hub
 # Using pipeline is often the easiest way to start for common tasks
 try:
+    from llama_cpp import Llama
+    print("llama_cpp imported successfully")
+except ImportError:
+    print("Error: llama-cpp-python not installed. Please check requirements.txt and logs.")
+    Llama = None # Set to None if import fails
+llm = None
+if Llama is not None:
+    try:
+        model_repo_id = "mradermacher/DeepSeek-R1-Distill-Qwen-7B-Multilingual-i1-GGUF"
+        model_file_name = "deepseek-r1-distill-qwen-7b-multilingual-i1.Q4_K_M.gguf" # <<== VERIFY THIS FILENAME ON HF HUB
     # Example: Sentiment Analysis model
     # model = pipeline("sentiment-analysis", model="distilbert/distilbert-base-uncased-finetuned-sst-2-english")
-    model = AutoModel.from_pretrained("mradermacher/DeepSeek-R1-Distill-Qwen-7B-Multilingual-i1-GGUF")
+    # model = AutoModel.from_pretrained("mradermacher/DeepSeek-R1-Distill-Qwen-7B-Multilingual-i1-GGUF")
     # Or load specific model/tokenizer if pipeline isn't suitable:
     # from transformers import AutoModel, AutoTokenizer
     # tokenizer = AutoTokenizer.from_pretrained("your-model-id")
@@ -23,39 +33,51 @@ except Exception as e:
 # 2. Define the function that uses the model
 # This function takes the input from the Gradio interface
 # and returns the output that Gradio will display.
-def process_input_with_model(input_text):
-    if model is None:
-        return "Model could not be loaded. Please check logs."
+def generate_text(prompt):
+    if llm is None:
+        return "Model failed to load. Please check App Space logs."
 
-    # Example using a pipeline:
-    result = model(input_text)
-    return result[0]['label'] # Adjust based on your model's output format
+    try:
+        print(f"Generating completion for prompt: {prompt[:100]}...") # Log start of generation
+        # Use the model to generate text
+        # Adjust max_tokens, stop sequence, etc. based on your needs and the model
+        output = llm(
+            prompt,
+            max_tokens=512, # Max tokens to generate
+            stop=["Qwen:", "\n\n"], # Stop sequence examples (adjust as needed)
+            echo=False, # Don't include prompt in output
+            temperature=0.7, # Creativity level
+            top_p=0.9, # Nucleus sampling
+        )
+        print("Generation complete.")
 
-    # Example if you loaded model/tokenizer manually:
-    # inputs = tokenizer(input_text, return_tensors="pt")
-    # outputs = model(**inputs)
-    # Process outputs to get your desired result...
-    # return processed_result
+        # Extract the generated text
+        generated_text = output["choices"][0]["text"]
+
+        return generated_text
+
+    except Exception as e:
+        print(f"Error during text generation: {e}")
+        return f"An error occurred during generation: {e}"
 
 
 # 3. Define the Gradio Interface
-# Set up the input and output components and link the processing function
-if model is not None: # Only create the interface if the model loaded successfully
+if llm is not None: # Only create the interface if the model loaded successfully
     interface = gr.Interface(
-        fn=process_input_with_model,  # Your function
-        inputs=gr.Textbox(label="Enter text for analysis"), # Input component (adjust type as needed)
-        outputs=gr.Label(), # Output component (adjust type as needed)
-        title="My Hugging Face Model Test App",
-        description="Test out the sentiment analysis model."
+        fn=generate_text,        # Your new generation function
+        inputs=gr.Textbox(label="Enter your prompt", lines=5), # Text input
+        outputs=gr.Textbox(label="Generated Text", lines=10),   # Text output
+        title="DeepSeek-R1-Distill-Qwen-7B GGUF Demo",
+        description="Interact with the DeepSeek-R1-Distill-Qwen-7B Multilingual model in GGUF format."
     )
 else:
-    # Create a simple interface indicating an error if the model failed to load
+    # Interface to show error if model loading failed
      interface = gr.Interface(
-        fn=lambda x: "Application failed to load model.", # Simple function
+        fn=lambda x: "Application failed to load model. See logs for details.",
         inputs=gr.Textbox(label="Status"),
         outputs=gr.Textbox(),
         title="Application Error",
-        description="Failed to load the model. Check the logs for details."
+        description="Failed to load the GGUF model. Check the logs for details on model loading errors."
     )
 
 

requirements.txt

@@ -1,4 +1,5 @@
 huggingface_hub==0.25.2
 gradio
 transformers
-torch
+torch
+llama-cpp-python