...

README.md

@@ -1,13 +1,15 @@
 ---
-title: Moondream2
-emoji: 🐢
+title: moondream1
+emoji: 🌔
 colorFrom: indigo
-colorTo: indigo
+colorTo: blue
 sdk: gradio
 sdk_version: 4.19.2
 app_file: app.py
 pinned: false
 license: apache-2.0
+preload_from_hub:
+  - vikhyatk/moondream2
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,60 @@
+import argparse
+import torch
+import re
+import gradio as gr
+from threading import Thread
+from transformers import TextIteratorStreamer, AutoTokenizer, AutoModelForCausalLM
+
+parser = argparse.ArgumentParser()
+
+if torch.cuda.is_available():
+    device, dtype = "cuda", torch.float16
+else:
+    device, dtype = "cpu", torch.float32
+
+model_id = "vikhyatk/moondream2"
+tokenizer = AutoTokenizer.from_pretrained(model_id, revision="2024-03-04")
+moondream = AutoModelForCausalLM.from_pretrained(
+    model_id, trust_remote_code=True, revision="2024-03-04"
+).to(device=device, dtype=dtype)
+moondream.eval()
+
+
+def answer_question(img, prompt):
+    image_embeds = moondream.encode_image(img)
+    streamer = TextIteratorStreamer(tokenizer, skip_special_tokens=True)
+    thread = Thread(
+        target=moondream.answer_question,
+        kwargs={
+            "image_embeds": image_embeds,
+            "question": prompt,
+            "tokenizer": tokenizer,
+            "streamer": streamer,
+        },
+    )
+    thread.start()
+
+    buffer = ""
+    for new_text in streamer:
+        clean_text = re.sub("<$|<END$", "", new_text)
+        buffer += clean_text
+        yield buffer
+
+
+with gr.Blocks() as demo:
+    gr.Markdown(
+        """
+        # 🌔 moondream2
+        A tiny vision language model. [GitHub](https://github.com/vikhyat/moondream)
+        """
+    )
+    with gr.Row():
+        prompt = gr.Textbox(label="Input", placeholder="Type here...", scale=4)
+        submit = gr.Button("Submit")
+    with gr.Row():
+        img = gr.Image(type="pil", label="Upload an Image")
+        output = gr.TextArea(label="Response")
+    submit.click(answer_question, [img, prompt], output)
+    prompt.submit(answer_question, [img, prompt], output)
+
+demo.queue().launch()

requirements.txt

@@ -0,0 +1,3 @@
+timm==0.9.12
+transformers==4.36.2
+einops==0.7.0