Spaces:
Sleeping
Sleeping
|
Download models/adaptformer_delhi/README.md from coderuday21/satdetect-dev: direct link, hf CLI and curl.
- Browser
- Download file 537 Bytes
-
https://huggingface.co/spaces/coderuday21/satdetect-dev/resolve/main/models/adaptformer_delhi/README.md
- Command line
-
hf download hf://spaces/coderuday21/satdetect-dev/models/adaptformer_delhi/README.md
-
curl -L -o README.md https://huggingface.co/spaces/coderuday21/satdetect-dev/resolve/main/models/adaptformer_delhi/README.md
537 Bytes
| # AdaptFormer Delhi fine-tune (Day 6+) | |
| Export the best checkpoint from a finetune run: | |
| ```powershell | |
| python scripts/export_adaptformer_delhi.py | |
| # or | |
| python scripts/export_adaptformer_delhi.py --src runs/day5_full/<run_id>/best | |
| ``` | |
| This creates: | |
| - `best/` — HuggingFace `save_pretrained` layout (used by the app) | |
| - `best.pt` — state_dict (plan deliverable name; optional) | |
| Point the app at the weights: | |
| ``` | |
| ADAPTFORMER_WEIGHTS=models/adaptformer_delhi/best | |
| ``` | |
| Weights are gitignored (large). Re-export after each training run. | |