Upload 10-episode all-metrics verification sample with reproduction files

artifacts/results/metric_iter31_sample10_all_metrics_verify/README.md

@@ -5,3 +5,48 @@ This folder contains 10 randomly selected episodes rendered in `all_metrics`-onl
 Selection seed: `20260408`
 
 Episodes: `episode97`, `episode90`, `episode73`, `episode69`, `episode12`, `episode63`, `episode47`, `episode43`, `episode14`, `episode46`
+
+## What Is Included
+
+For each sampled episode directory:
+
+- `episodeX.dense.csv`: dense metric table from the completed 100-episode single-pass fuller logging run
+- `episodeX.keyframes.csv`: keyframe downsample for the same episode
+- `episodeX.metrics.json`: per-episode summary metrics
+- `episodeX.debug.jsonl`: fuller per-frame debug sidecar
+- `templates.json` and `templates.pkl`: fixed template bundle used for replay and rendering
+- `visualizations/episodeX_all_metrics.gif`: compact verification GIF
+- `visualizations/README.md`: per-episode visualization note
+
+This sample bundle was derived from the completed 100-episode run:
+
+- `metric_iter30_full100_single_pass_full_logging_fixed_templates_merged`
+
+## Render Mode
+
+These GIFs were rendered with the fast `all_metrics`-only path added in:
+
+- `code/scripts/render_oven_metric_frame.py`
+- `code/scripts/render_oven_metric_gifs.py`
+
+That mode skips simulator replay when only the full-metrics panel is requested, so the render depends only on:
+
+- stored RGB frames in the RLBench episode directory
+- the dense CSV row for each rendered frame
+
+## Reproduce
+
+Per episode, the render command is:
+
+```bash
+python code/scripts/render_oven_metric_gifs.py \
+  --episode-dir /workspace/data/bimanual_take_tray_out_of_oven_train_128/all_variations/episodes/episode97 \
+  --dense-csv artifacts/results/metric_iter31_sample10_all_metrics_verify/episode97/episode97.dense.csv \
+  --templates-pkl artifacts/results/metric_iter31_sample10_all_metrics_verify/episode97/templates.pkl \
+  --output-dir artifacts/results/metric_iter31_sample10_all_metrics_verify/episode97/visualizations \
+  --all-metrics-only \
+  --num-workers 8 \
+  --base-display 1500
+```
+
+The sampled episode list and seed are also recorded in `selection.json`.

artifacts/results/metric_iter31_sample10_all_metrics_verify/selection.json

@@ -1,5 +1,6 @@
 {
   "type": "sample_verification_bundle",
+  "source_run_name": "metric_iter30_full100_single_pass_full_logging_fixed_templates_merged",
   "source_result_root": "/workspace/VLAdaptorBench_upload/artifacts/results/metric_iter30_full100_single_pass_full_logging_fixed_templates_merged",
   "selection_seed": 20260408,
   "episode_indices": [
@@ -16,4 +17,4 @@
   ],
   "render_mode": "all_metrics_only",
   "created_at_epoch": 1775672903.333171
-}
+}