copy from other space

README.md

@@ -1,5 +1,5 @@
 ---
-title: Idefics2 8b Vdpoed Playground
+title: Idefics2 8b VDPOed Playground
 emoji: 🚀
 colorFrom: red
 colorTo: pink

app.py

@@ -0,0 +1,34 @@
+import gradio as gr
+from transformers import AutoModelForVision2Seq, AutoProcessor, AutoModelForVision2Seq, BitsAndBytesConfig
+import torch
+
+quantization_config = BitsAndBytesConfig(
+    load_in_4bit=True, bnb_4bit_quant_type="nf4", bnb_4bit_use_double_quant=True, bnb_4bit_compute_dtype=torch.float16
+)
+
+processor = AutoProcessor.from_pretrained("HuggingFaceH4/idefics2-8b-vdpoed")
+model = AutoModelForVision2Seq.from_pretrained(
+    "HuggingFaceH4/idefics2-8b-vdpoed",
+    torch_dtype=torch.float16,
+    quantization_config=quantization_config,
+)
+
+
+def respond(multimodal_input):
+    images = multimodal_input["files"]
+    content = [{"type": "image"} for _ in images]
+    content.append({"type": "text", "text": multimodal_input["text"]})
+    messages = [{"role": "user", "content": content}]
+    prompt = processor.apply_chat_template(messages, add_generation_prompt=True)
+    inputs = processor(text=prompt, images=[images], return_tensors="pt")
+    inputs = {k: v.to(model.device) for k, v in inputs.items()}
+    num_tokens = len(inputs["input_ids"][0])
+    with torch.inference_mode():
+        generated_ids = model.generate(**inputs, max_new_tokens=500)
+
+    new_tokens = generated_ids[:, num_tokens:]
+    generated_text = processor.batch_decode(new_tokens, skip_special_tokens=True)[0]
+    return generated_text
+
+
+gr.Interface(respond, inputs=[gr.MultimodalTextbox(file_types=["image"], show_label=False)], outputs="text").launch()

requirements.txt

@@ -0,0 +1,4 @@
+transformers
+torch
+bitsandbytes
+accelerate