Update app.py

app.py

@@ -2,29 +2,52 @@ import gradio as gr
 from transformers import AutoTokenizer, AutoModelForCausalLM
 import torch
 
+# Model name
 model_name = "deepseek-ai/deepseek-coder-1.3b-base"
 
+# Load tokenizer
 tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+# Load model (optimized for CPU)
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
-    torch_dtype=torch.float16,
-    device_map="auto"
+    torch_dtype=torch.float32,     # CPU friendly
+    low_cpu_mem_usage=True
 )
 
+model.to("cpu")
+model.eval()
+
+# Generate function
 def generate_code(prompt):
-    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
-    outputs = model.generate(
-        **inputs,
-        max_new_tokens=200,
-        temperature=0.7
-    )
-    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+    if not prompt.strip():
+        return "Please enter a prompt."
+
+    inputs = tokenizer(prompt, return_tensors="pt")
+
+    with torch.no_grad():
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=100,     # lower = faster
+            temperature=0.5,
+            do_sample=True,
+            top_p=0.9
+        )
+
+    result = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    return result
 
+# Gradio UI
 iface = gr.Interface(
     fn=generate_code,
-    inputs="text",
-    outputs="text",
-    title="DeepSeek Coder 1.3B"
+    inputs=gr.Textbox(
+        lines=5,
+        placeholder="Write a coding prompt here..."
+    ),
+    outputs=gr.Textbox(lines=10),
+    title="DeepSeek Coder 1.3B (CPU Optimized)",
+    description="Generate code using DeepSeek Coder running on Hugging Face Spaces (CPU mode)."
 )
 
+# Launch app
 iface.launch()