Upload folder using huggingface_hub

app.py

@@ -0,0 +1,54 @@
+import os
+import torch
+import spaces
+import subprocess
+import gradio as gr
+from huggingface_hub import login
+from transformers import AutoProcessor, PaliGemmaForConditionalGeneration
+
+subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
+login(os.environ.get("HF_TOKEN"))
+
+model_id = "google/paligemma-3b-mix-448"
+model = PaliGemmaForConditionalGeneration.from_pretrained(
+    model_id, device_map={"": 0},
+    attn_implementation="flash_attention_2",
+    torch_dtype=torch.bfloat16,
+)
+processor = AutoProcessor.from_pretrained(model_id)
+model.eval()
+
+
+@spaces.GPU()
+def answer_question(image, prompt):
+    model_inputs = processor(text=prompt, images=image, return_tensors="pt").to("cuda")
+    input_len = model_inputs["input_ids"].shape[-1]
+
+    with torch.inference_mode():
+        generation = model.generate(**model_inputs, max_new_tokens=100, do_sample=False)
+        generation = generation[0][input_len:]
+        decoded = processor.decode(generation, skip_special_tokens=True)
+        
+    return decoded
+
+
+with gr.Blocks() as demo:
+    gr.Markdown(
+        """
+        # PaliGemma
+        Lightweight open vision-language model (VLM). [Model card](https://huggingface.co/google/paligemma-3b-mix-448)
+        """
+    )
+    
+    with gr.Row():
+        prompt = gr.Textbox(label="Input", value="Describe this image.", scale=4)
+        submit = gr.Button("Submit")
+    
+    with gr.Row():
+        image = gr.Image(type="pil", label="Upload an Image")
+        output = gr.TextArea(label="Response")
+    
+    submit.click(answer_question, [image, prompt], output)
+    prompt.submit(answer_question, [image, prompt], output)
+
+demo.queue().launch()

requirements.txt

@@ -0,0 +1,3 @@
+git+https://github.com/huggingface/transformers.git 
+accelerate
+torch