| --- |
| license: mit |
| --- |
| |
| # experiment_pel |
| |
| A simple **PyTorch checkpoint** published for reproducible experiments. |
| This repository exposes a single file `model.ckpt` (raw PyTorch/Lightning-style checkpoint). It is **not** a Transformers model; load it with `torch.load`. |
| |
| ## Usage |
| |
| ```python |
| import torch |
| from huggingface_hub import hf_hub_download |
|
|
| # Download the checkpoint into a local folder (or use the default cache) |
| p = hf_hub_download( |
| repo_id="jeanq1/experiment_pel", |
| filename="model.ckpt", |
| local_dir="./weights", # optional: place a copy/symlink here |
| local_dir_use_symlinks=False # set to True to keep a symlink instead of a copy |
| ) |
| |
| # Load like any local PyTorch file |
| ckpt = torch.load(p, map_location="cpu", weights_only=False) |
| # Example: |
| # model = YourModelClass(...) |
| # model.load_state_dict(ckpt["state_dict"] if "state_dict" in ckpt else ckpt) |
|
|
