Initial deploy

README.md

@@ -0,0 +1,16 @@
+---
+title: GLM-4.6V-Flash API
+emoji: 🚀
+colorFrom: blue
+colorTo: indigo
+sdk: gradio
+sdk_version: 5.0.0
+app_file: app.py
+pinned: false
+license: mit
+---
+
+# GLM-4.6V-Flash API Space
+
+This Space hosts the [zai-org/GLM-4.6V-Flash](https://huggingface.co/zai-org/GLM-4.6V-Flash) model.
+It provides a Gradio interface and an API for multimodal (text + image) inference.

app.py

@@ -0,0 +1,81 @@
+import gradio as gr
+from transformers import AutoProcessor, Glm4vForConditionalGeneration
+import torch
+from PIL import Image
+
+# Configuration
+MODEL_PATH = "zai-org/GLM-4.6V-Flash"
+
+# Load Model
+print(f"Loading model: {MODEL_PATH}...")
+processor = AutoProcessor.from_pretrained(MODEL_PATH, trust_remote_code=True)
+model = Glm4vForConditionalGeneration.from_pretrained(
+    MODEL_PATH,
+    torch_dtype=torch.bfloat16,
+    low_cpu_mem_usage=True,
+    trust_remote_code=True,
+    device_map="auto" 
+)
+print("Model loaded successfully.")
+
+def predict(image, text, history):
+    if not text and not image:
+        return "Please upload an image or enter text."
+
+    messages = []
+    
+    # Add history if needed (simplified for this demo to just current turn)
+    # Ideally, you'd format history as expected by the model
+    
+    content = []
+    if image:
+        content.append({"type": "image", "image": image})
+    if text:
+        content.append({"type": "text", "text": text})
+        
+    messages.append({"role": "user", "content": content})
+
+    # Prepare inputs
+    inputs = processor.apply_chat_template(
+        messages,
+        add_generation_prompt=True,
+        tokenize=True,
+        return_dict=True,
+        return_tensors="pt"
+    ).to(model.device)
+    
+    # Generate
+    with torch.no_grad():
+        generated_ids = model.generate(**inputs, max_new_tokens=1024, do_sample=True, temperature=0.7)
+    
+    # Decode
+    output_text = processor.decode(generated_ids[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True)
+    return output_text
+
+# Create Gradio Interface
+with gr.Blocks() as demo:
+    gr.Markdown(f"# {MODEL_PATH} API")
+    
+    with gr.Row():
+        with gr.Column():
+            image_input = gr.Image(type="pil", label="Upload Image (Optional)")
+            text_input = gr.Textbox(label="Message")
+            submit_btn = gr.Button("Submit")
+        with gr.Column():
+            output = gr.Markdown(label="Response")
+
+    submit_btn.click(
+        fn=predict,
+        inputs=[image_input, text_input, "state"], # passing mock state for now
+        outputs=output
+    )
+
+    # API documentation
+    gr.Markdown("""
+    ### API Usage
+    You can use this Space as an API. 
+    Click 'Use via API' at the bottom of this page for details.
+    """)
+
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -0,0 +1,5 @@
+transformers>=4.46.0
+torch
+accelerate
+pillow
+gradio