Spaces:
Sleeping
Sleeping
add model debugging
Browse files
app.py
CHANGED
@@ -402,7 +402,7 @@ with gr.Blocks() as demo:
|
|
402 |
gr.Textbox("Thank you for using our interpretable attribution method, which originates from the ICLR 2024 Oral paper titled \"Less is More: Fewer Interpretable Regions via Submodular Subset Selection.\" We have now implemented this method on the multimodal ViT model and achieved promising results in explaining model predictions. A key feature of our approach is its ability to clarify the reasons behind the model's prediction errors. We invite you to try out this demo and explore its capabilities. The source code is available at https://github.com/RuoyuChen10/SMDL-Attribution.\nYou can upload an image yourself or select one from the following, then click the button Interpreting Model to get the result. The demo currently does not support selecting categories or descriptions by yourself. If you are interested, you can try it from the source code.", label="Instructions for use", interactive=False)
|
403 |
|
404 |
# 文本输入框和滑块
|
405 |
-
text_input = gr.Textbox(label="Text Input", placeholder="You can choose what you want to explain. You can enter a word (e.g., '
|
406 |
|
407 |
# 第二排:两个缩略图
|
408 |
with gr.Row():
|
|
|
402 |
gr.Textbox("Thank you for using our interpretable attribution method, which originates from the ICLR 2024 Oral paper titled \"Less is More: Fewer Interpretable Regions via Submodular Subset Selection.\" We have now implemented this method on the multimodal ViT model and achieved promising results in explaining model predictions. A key feature of our approach is its ability to clarify the reasons behind the model's prediction errors. We invite you to try out this demo and explore its capabilities. The source code is available at https://github.com/RuoyuChen10/SMDL-Attribution.\nYou can upload an image yourself or select one from the following, then click the button Interpreting Model to get the result. The demo currently does not support selecting categories or descriptions by yourself. If you are interested, you can try it from the source code.", label="Instructions for use", interactive=False)
|
403 |
|
404 |
# 文本输入框和滑块
|
405 |
+
text_input = gr.Textbox(label="Text Input", placeholder="You can choose what you want to explain. You can enter a word (e.g., 'tabby cat') or a description (e.g., 'A photo of a tabby cat'). If the input is empty, the model will explain the predicted category.")
|
406 |
|
407 |
# 第二排:两个缩略图
|
408 |
with gr.Row():
|