[skills]: address Gemini + Copilot review on PR #1338

add-model-02-parity:
- Use .item() before f-string formatting so the snippet portably works on
  older PyTorch versions that lack `__format__` on 0-dim tensors.
- Switch p99 to torch.quantile (cleaner, no off-by-one risk for small N
  vs. kthvalue(int(0.99 * N))).
- Document why both changes matter.

seed-ssim-references:
- .png artefact: correct trigger is `workload_type.value.endswith("2i")`
  (per `VideoGenerator._is_image_workload`), not "num_frames=1". On
  current main the SSIM helpers (`_find_reference_video`,
  `output_video_name`) hardcode .mp4; consuming .png references requires
  the `media_extension` parameter introduced in PR #1321.
- T2I gotcha: add `mkdir -p` before `cp` so the destination is created.
- .gitignore: clarify that `reference_videos_cli.py upload` uses
  `huggingface_hub.upload_folder` (filesystem upload, not git), so
  `git add -f` only matters for local commits. Note that the
  `fastvideo/tests/ssim/reference_videos/**` catch-all near `.gitignore:94`
  currently overrides the earlier `*.mp4` negation, so every reference
  file requires `git add -f` regardless of extension.

Author: Mister-Raggs

.agents/skills/add-model-02-parity/SKILL.md

@@ -212,11 +212,22 @@ and assert that the *distribution* — not just the max — looks healthy:
 
 ```python
 abs_diff = (hf_out.float().cpu() - fv_out.float().cpu()).abs()
-print(f"max_diff={abs_diff.max():.4f}  mean_diff={abs_diff.mean():.4f}  "
-      f"median_diff={abs_diff.median():.4f}  "
-      f"p99_diff={abs_diff.flatten().kthvalue(int(0.99 * abs_diff.numel())).values:.4f}")
+max_diff = abs_diff.max().item()
+mean_diff = abs_diff.mean().item()
+median_diff = abs_diff.median().item()
+p99_diff = torch.quantile(abs_diff.flatten(), 0.99).item()
+print(
+    f"max_diff={max_diff:.4f}  mean_diff={mean_diff:.4f}  "
+    f"median_diff={median_diff:.4f}  p99_diff={p99_diff:.4f}"
+)
 ```
 
+Call `.item()` before formatting (older PyTorch versions do not implement
+`__format__` on 0-dim tensors), and prefer `torch.quantile` over
+`kthvalue(int(0.99 * N))` — `kthvalue` is 1-indexed and degenerates for
+very small `N`, while `quantile` handles the distribution boundary
+cleanly.
+
 Healthy bf16-tail signature (FLUX, 57 layers, observed on A40):
 `max=0.5, mean=0.04, median=0, p99=0.25`. Median near zero and mean ≪ atol
 prove the bulk of elements match — only the tail diverges due to accumulation.

.agents/skills/seed-ssim-references/SKILL.md

@@ -19,10 +19,15 @@ side-by-side per `(model_id, backend, prompt)`:
   metadata + `slice_spec` + `format_version`) for tests that call
   `run_text_to_latent_similarity_test` in `latent_similarity_utils.py`.
   Compared via cosine distance on the slice and the full tensor.
-- **`.png`** — pixel ground-truth for **T2I** tests (`num_frames=1`) that
-  reuse `run_text_to_video_similarity_test` but produce a single frame. The
-  helper writes a `.png` instead of a `.mp4` when the output has no time
-  dimension. Compared via SSIM the same way.
+- **`.png`** — pixel ground-truth for **T2I / I2I** tests. `VideoGenerator`
+  emits a `.png` (not a `.mp4`) whenever `workload_type.value.endswith("2i")`
+  is true (`fastvideo/entrypoints/video_generator.py::_is_image_workload`).
+  **The SSIM helpers in `inference_similarity_utils.py` on current main
+  hardcode `.mp4`** (`_find_reference_video` filter, `output_video_name`
+  format). To consume `.png` references through the SSIM path, the test
+  must call a helper variant that accepts a `media_extension` parameter
+  (added in PR #1321). Until that lands on main, T2I SSIM coverage is
+  blocked at the helper layer.
 
 This skill:
 
@@ -115,22 +120,31 @@ and 6 are artefact-type-agnostic — `_iter_reference_files`,
 `copy_generated_to_reference`, and `upload_reference_videos` already walk
 both `.mp4` and `.pt` (see `reference_videos_cli.py`).
 
-**T2I gotcha:** when the helper produces a `.png` (T2I test with
-`num_frames=1`), `reference_videos_cli.py copy-local` currently walks
-`.mp4`/`.pt` only and reports `0 copied files` without erroring. After step 5,
-verify the destination contains the seeded file; if it's empty, copy the PNG
-manually:
+**T2I / I2I gotcha:** when the helper produces a `.png` (any workload
+whose `workload_type.value` ends in `2i`), `reference_videos_cli.py copy-local`
+currently walks `.mp4` / `.pt` only and reports `0 copied files` without
+erroring. After step 5, verify the destination contains the seeded file;
+if it's empty, copy the PNG manually:
 
 ```bash
+mkdir -p fastvideo/tests/ssim/reference_videos/default/L40S_reference_videos/<model_id>/<backend>/
 cp ./generated_videos_modal/default/generated_videos/L40S_reference_videos/<model_id>/<backend>/*.png \
    fastvideo/tests/ssim/reference_videos/default/L40S_reference_videos/<model_id>/<backend>/
 ```
 
-Then `git add -f` the PNG: the repo `.gitignore` has a broad `*.png` rule, and
-the `reference_videos/**` negation only applies to extensions explicitly
-re-allowed *after* the catch-all (see `.gitignore` near the
-`reference_videos/**` block — add `!fastvideo/tests/ssim/reference_videos/**/*.png`
-there if the negation is missing).
+`git add -f` is **not required for the HF upload path**:
+`reference_videos_cli.py upload` calls `huggingface_hub.upload_folder` on
+the filesystem directly (`reference_videos_cli.py:311`) — git tracking
+status is irrelevant for `pytest fastvideo/tests/ssim/` reference sync.
+
+If you *also* want to commit the PNG locally, note that
+`fastvideo/tests/ssim/reference_videos/**` (catch-all near `.gitignore:94`)
+overrides the earlier `!fastvideo/tests/ssim/reference_videos/**/*.mp4`
+negation at `.gitignore:73`. Every reference file in this dir currently
+requires `git add -f`, regardless of extension. Fixing this properly
+means moving the `!...reference_videos/**/*.{mp4,pt,png}` negations
+*after* the catch-all (or removing the catch-all). Until then, `git add -f`
+is the standing workaround for committing references locally.
 
 If either check fails, stop and tell the user what's wrong.