Ethanfel/ComfyUI-SelVA
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: bump eval sample to 25 ODE steps (was 8)

Browse Source 
Inference is fast on RTX PRO 6000 — 8 steps was washing out quality
differences between experiments.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Ethanfel
2026-04-08 10:32:27 +02:00
parent a7923d5fb7
commit 1663b39833
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
nodes/selva_lora_trainer.py
+1 -1
View File
@@ -88,7 +88,7 @@ def _load_npz(path: Path) -> dict:
# --------------------------------------------------------------------------- # ---------------------------------------------------------------------------
def _eval_sample(generator, feature_utils_orig, dataset, seq_cfg, device, dtype, def _eval_sample(generator, feature_utils_orig, dataset, seq_cfg, device, dtype,
num_steps: int = 8, seed: int = 0): num_steps: int = 25, seed: int = 0):
"""Run a quick no-CFG inference pass on a fixed training clip. """Run a quick no-CFG inference pass on a fixed training clip.
Always uses dataset[0] and a fixed noise seed so samples across checkpoints Always uses dataset[0] and a fixed noise seed so samples across checkpoints