Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+DeepSeek-Coder-V2-Lite-Instruct-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text

DeepSeek-Coder-V2-Lite-Instruct-Q4_K_M.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38bc76f3326b49b4d81d1027d092bf7ce5b4ed2de4136d1d2e7e6347c3ec8376
+size 10364416480

README.md

@@ -0,0 +1,34 @@
+---
+title: DeepSeek Coder V2 Lite 16B
+emoji: 💻
+colorFrom: blue
+colorTo: green
+sdk: gradio
+sdk_version: 4.0.0
+app_file: app.py
+pinned: false
+---
+
+# 🚀 o87Dev - Maximum Capacity Deployment
+
+**Strategy:** Deploy the largest viable model (`DeepSeek-Coder-V2-Lite-Instruct-16B-Q4_K_M`) on Hugging Face's free CPU tier.
+
+## ⚙️ Technical Details
+- **Model:** DeepSeek-Coder-V2-Lite-Instruct-Q4_K_M.gguf (10.4GB)
+- **Quantization:** Q4_K_M (Optimal quality/size for free tier)
+- **Loader:** `llama-cpp-python` (CPU optimized)
+- **Context:** 2048 tokens (max for free tier stability)
+
+## 📊 Performance Expectations
+- **First load:** ~60-120 seconds (model loads from disk)
+- **Inference speed:** ~2-5 tokens/second on CPU
+- **Memory usage:** ~12-14GB of 16GB available
+
+## 🎯 Usage Tips
+1. First request triggers model load (be patient)
+2. Keep prompts under 500 tokens for best results
+3. Use temperature 0.7-0.9 for creative tasks
+4. Monitor memory usage in Space logs
+
+## 🔗 Integration
+This Space serves as the primary AI endpoint for the o87Dev local API server.

app.py

@@ -0,0 +1,100 @@
+import gradio as gr
+from huggingface_hub import hf_hub_download
+from llama_cpp import Llama
+import time
+import os
+
+# Configuration
+MODEL_NAME = "DeepSeek-Coder-V2-Lite-Instruct-Q4_K_M.gguf"
+MODEL_PATH = MODEL_NAME  # Since we placed it in the same directory
+
+# Initialize model (will be loaded on first use)
+llm = None
+
+def load_model():
+    """Lazy-load the model only when needed."""
+    global llm
+    if llm is None:
+        print(f"⏳ Loading model {MODEL_NAME}... This may take 1-2 minutes on first run.")
+        start_time = time.time()
+        
+        # CPU-optimized settings for free tier
+        llm = Llama(
+            model_path=MODEL_PATH,
+            n_ctx=2048,           # Context size (smaller = less memory)
+            n_threads=2,          # Use 2 CPU threads
+            n_gpu_layers=0,       # CPU only on free tier
+            verbose=False
+        )
+        
+        load_time = time.time() - start_time
+        print(f"✅ Model loaded in {load_time:.1f} seconds. Ready for inference.")
+    return llm
+
+def generate_code(prompt, max_tokens=256, temperature=0.7):
+    """Main generation function."""
+    try:
+        model = load_model()
+        
+        # Format prompt for DeepSeek-Coder Instruct models
+        formatted_prompt = f"### Instruction:\n{prompt}\n\n### Response:\n"
+        
+        # Generate
+        output = model(
+            formatted_prompt,
+            max_tokens=max_tokens,
+            temperature=temperature,
+            top_p=0.95,
+            echo=False,
+            stop=["###", "\n\n\n"]
+        )
+        
+        return output['choices'][0]['text'].strip()
+        
+    except Exception as e:
+        return f"❌ Error: {str(e)}"
+
+# Create Gradio interface
+demo = gr.Interface(
+    fn=generate_code,
+    inputs=[
+        gr.Textbox(
+            label="Code Prompt",
+            placeholder="Write a Python function to reverse a string...",
+            lines=4
+        ),
+        gr.Slider(
+            minimum=32,
+            maximum=512,
+            value=256,
+            step=32,
+            label="Max Tokens"
+        ),
+        gr.Slider(
+            minimum=0.1,
+            maximum=1.0,
+            value=0.7,
+            step=0.1,
+            label="Temperature"
+        )
+    ],
+    outputs=gr.Code(
+        label="Generated Code",
+        language="python"
+    ),
+    title="💻 DeepSeek Coder V2 Lite (16B) - o87Dev",
+    description="**CPU Deployment** - Largest viable model on Hugging Face Spaces free tier. ⚠️ **First request loads model (~1-2 min)**",
+    examples=[
+        ["Write a Python function to check if a number is prime"],
+        ["Create a React component for a login form"],
+        ["Explain binary search algorithm in Python"]
+    ]
+)
+
+# Launch with queue for better handling on free tier
+if __name__ == "__main__":
+    demo.launch(
+        server_name="0.0.0.0",
+        server_port=7860,
+        share=False  # Set to True if you want a public link
+    )

requirements.txt

@@ -0,0 +1,4 @@
+gradio>=4.0.0
+llama-cpp-python>=0.3.0  # CPU-optimized GGUF loader
+huggingface-hub>=0.20.0
+sentencepiece>=0.1.99  # Tokenizer for many models