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

chore: lower default warmup_steps from 500 to 100

Browse Source 
500 warmup steps is 25% of a 2000-step run — too long. 100 steps lets
the full lr kick in much earlier without sacrificing stability.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Ethanfel
2026-04-05 22:51:27 +02:00
parent 16b3eb11cc
commit 4806daa4ca
3 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
LORA_TRAINING.md
+1 -1
View File
@@ -106,7 +106,7 @@ The script will:
| `--target` | `attn.qkv` | Which layers to adapt. Add `linear1` for post-attention projections |
| `--lr` | `1e-4` | Learning rate |
| `--steps` | `2000` | Total training steps |
| `--warmup_steps` | `500` | Linear LR warmup steps |
| `--warmup_steps` | `100` | Linear LR warmup steps |
| `--grad_accum` | `4` | Gradient accumulation steps (effective batch = grad_accum × 1) |
| `--save_every` | `500` | Save a checkpoint every N steps |
| `--resume` | `None` | Path to a step checkpoint to resume from (e.g. `lora_output/adapter_step01000.pt`) |