zhicao commited on
Commit
c44e95a
·
verified ·
1 Parent(s): f559dcb

Scheduled Commit

Browse files
ablation_runs/tools/splice_artifact.py CHANGED
@@ -10,7 +10,7 @@ from make_ablation_section import EXTRA_CSS, build_section # noqa: E402
10
  SAVED = Path(
11
  "/home/zhicao/.claude/projects/-scratch1-home-zhicao/"
12
  "79307854-97d6-42b0-a223-99c8f638fe96/tool-results/"
13
- "artifact-56a2b6c9-1787172588-0a8a.html"
14
  )
15
  OUT = Path("/scratch1/home/zhicao/dreamzero/ablation_runs/track_force_cascade.html")
16
 
 
10
  SAVED = Path(
11
  "/home/zhicao/.claude/projects/-scratch1-home-zhicao/"
12
  "79307854-97d6-42b0-a223-99c8f638fe96/tool-results/"
13
+ "artifact-56a2b6c9-1787175554-8e0c.html"
14
  )
15
  OUT = Path("/scratch1/home/zhicao/dreamzero/ablation_runs/track_force_cascade.html")
16
 
ablation_runs/track_force_cascade.html CHANGED
@@ -1596,15 +1596,15 @@
1596
  <tr><th>Task</th><th class="num">Diffusion Policy</th><th class="num">GR00T-N1-2B</th><th class="num">Ours</th></tr>
1597
  </thead>
1598
  <tbody>
1599
- <tr><td>Tray Lift</td><td class="num">5.0</td><td class="num">5.8</td><td class="num"><span class="pending">pending</span></td></tr>
1600
- <tr><td>Box Cleanup</td><td class="num">30.0</td><td class="num">33.3</td><td class="num"><span class="pending">pending</span></td></tr>
1601
- <tr><td>Drawer Cleanup</td><td class="num">1.7</td><td class="num">10.8</td><td class="num"><span class="pending">pending</span></td></tr>
1602
- <tr class="mean"><td><b>Mean</b></td><td class="num"><b>12.2</b></td><td class="num"><b>16.6</b></td><td class="num">&mdash;</td></tr>
1603
  </tbody>
1604
  </table>
1605
  </div>
1606
  <div class="prose">
1607
- <p style="font-size:13px; color:var(--muted);">Success rate (%). All policies are fine-tuned on 30 demonstrations per task and evaluated closed-loop over 100 rollouts per task. Ours: evaluation in progress.</p>
1608
  </div>
1609
  </section>
1610
 
@@ -2000,3 +2000,4 @@
2000
 
2001
  </section>
2002
  </div>
 
 
1596
  <tr><th>Task</th><th class="num">Diffusion Policy</th><th class="num">GR00T-N1-2B</th><th class="num">Ours</th></tr>
1597
  </thead>
1598
  <tbody>
1599
+ <tr><td>Tray Lift</td><td class="num">5.0</td><td class="num">5.8</td><td class="num"><b>10.0</b></td></tr>
1600
+ <tr><td>Box Cleanup</td><td class="num">30.0</td><td class="num">33.3</td><td class="num"><b>45.0</b></td></tr>
1601
+ <tr><td>Drawer Cleanup</td><td class="num">1.7</td><td class="num"><b>10.8</b></td><td class="num">10.0</td></tr>
1602
+ <tr class="mean"><td><b>Mean</b></td><td class="num">12.2</td><td class="num">16.6</td><td class="num"><b>20.0</b></td></tr>
1603
  </tbody>
1604
  </table>
1605
  </div>
1606
  <div class="prose">
1607
+ <p style="font-size:13px; color:var(--muted);">Success rate (%). All policies are fine-tuned on 30 demonstrations per task and evaluated closed-loop over 100 rollouts per task. Bold marks the best result per row.</p>
1608
  </div>
1609
  </section>
1610
 
 
2000
 
2001
  </section>
2002
  </div>
2003
+