v0.6

app.py

@@ -9,7 +9,7 @@ from PIL import Image
 from typing import Union
 import os
 
-device = "cuda"
+device = "cuda" if torch.cuda.is_available() else "cpu"
 print(f"Using {device}" if device != "cpu" else "Using CPU")
 
 def _load_model():
@@ -32,6 +32,8 @@ class MoonDream():
             yield res
         return
 
+md = MoonDream()
+
 def _respond_one(question, img):
   txt = ""
   yield (txt := txt + MoonDream()(question, [img]))
@@ -97,13 +99,21 @@ def xform_image_description(img, inst):
   res = tog.completions.create(prompt=prompt, model="meta-llama/Meta-Llama-3-70B", stop=["```"], max_tokens=1024)
   return res.choices[0].text[len(prompt):].split("```")[0]
 
-with gr.Blocks() as demo:
-  with gr.Row(visible=True):
+def simple_desc(img, prompt):
+  return md(img, prompt)
+
+with gr.Blocks() as if_simple_description:
+  with gr.Row():
     with gr.Column():
-      img = gr.Image(label="images", type='pil')
+      simple_img = gr.Image(type="pil")
     with gr.Column():
-      otpt = gr.Textbox(label="output", lines=3, interactive=True)
-      btn = gr.Button("submit")
+      simple_text = gr.Textbox(label="prompt ( Shift+Enter sends )")
+      simple_btn = gr.Button("Prompt & Image 2 Text")
+  simple_btn.click(simple_desc, inputs=[simple_img, simple_desc], outputs=[simple_desc])
+
+"""
+with gr.Blocks() as demo:
+  
   with gr.Row():
     with gr.Column():
       im1 = gr.Image(label="image 1", type='pil')
@@ -131,4 +141,7 @@ with gr.Blocks() as demo:
   btn.click(dual_images, inputs=[img], outputs=[otpt])
   btn_scd.click(merge_descriptions_to_prompt, inputs=[minst, otp2, otp3], outputs=[otp4])
 
-  demo.launch(debug=True, share=True)
+  demo.launch(debug=True, share=True)
+  """
+with gr.TabbedInterface(if_simple_description) as ifc:
+  ifc.launch(share=False, server_host="0.0.0.0", server_port=7860)

requirements.txt

@@ -6,3 +6,4 @@ pillow==10.3.0
 together==1.1.5
 torch
 torchvision
+accelerate==0.30.1