Give every tab a headline; restore tab 3's section rule; drop the conditional
Browse filesTabs 2 and 3 each opened with a claim; tab 1 opened with instructions, so the
proposition the other two tabs attack was never stated. Tab 1 now opens with "A
working model that explains itself", making the sequence assert, undermine, what
survives.
Restores the horizontal rule before the conformal section -- present until v3.2,
where it sat where the chart needed to go and was not put back. After a figure, a
table and a blockquote, an h3 alone does not read as a section boundary.
Tab 3 label drops the conditional: "What We'd Show an Auditor" -> "What We Show an
Auditor".
Mirrors llm-wiki commit 18dcbd6.
app.py
CHANGED
|
@@ -172,6 +172,7 @@ with gr.Blocks(title="Explainable & Auditable Alloy Defect Detection",
|
|
| 172 |
)
|
| 173 |
with gr.Tab("1 - Classify & Explain"):
|
| 174 |
gr.Markdown(
|
|
|
|
| 175 |
"Upload a micrograph (or pick an example below). The model predicts the "
|
| 176 |
"most likely defect class; the heatmap shows which pixels drove that "
|
| 177 |
"prediction (**Grad-CAM**, implemented from first principles with PyTorch "
|
|
@@ -206,7 +207,7 @@ with gr.Blocks(title="Explainable & Auditable Alloy Defect Detection",
|
|
| 206 |
)
|
| 207 |
gr.Markdown(SPURIOUS_MARKDOWN)
|
| 208 |
|
| 209 |
-
with gr.Tab("3 - What We
|
| 210 |
gr.Markdown(
|
| 211 |
"""
|
| 212 |
### First: can we even tell whether the heatmap is faithful?
|
|
@@ -240,6 +241,8 @@ with gr.Blocks(title="Explainable & Auditable Alloy Defect Detection",
|
|
| 240 |
> why we treat this curve as supporting evidence rather than proof, and why the
|
| 241 |
> guarantee below is what we would actually put in front of an auditor.
|
| 242 |
|
|
|
|
|
|
|
| 243 |
### So what survives? Conformal Risk Control
|
| 244 |
|
| 245 |
**Not covered anywhere in the course** — taken from Shen & Liu,
|
|
|
|
| 172 |
)
|
| 173 |
with gr.Tab("1 - Classify & Explain"):
|
| 174 |
gr.Markdown(
|
| 175 |
+
"### A working model that explains itself\n\n"
|
| 176 |
"Upload a micrograph (or pick an example below). The model predicts the "
|
| 177 |
"most likely defect class; the heatmap shows which pixels drove that "
|
| 178 |
"prediction (**Grad-CAM**, implemented from first principles with PyTorch "
|
|
|
|
| 207 |
)
|
| 208 |
gr.Markdown(SPURIOUS_MARKDOWN)
|
| 209 |
|
| 210 |
+
with gr.Tab("3 - What We Show an Auditor"):
|
| 211 |
gr.Markdown(
|
| 212 |
"""
|
| 213 |
### First: can we even tell whether the heatmap is faithful?
|
|
|
|
| 241 |
> why we treat this curve as supporting evidence rather than proof, and why the
|
| 242 |
> guarantee below is what we would actually put in front of an auditor.
|
| 243 |
|
| 244 |
+
---
|
| 245 |
+
|
| 246 |
### So what survives? Conformal Risk Control
|
| 247 |
|
| 248 |
**Not covered anywhere in the course** — taken from Shen & Liu,
|