| """ |
| upload_artifacts.py — run THIS locally before launching the Colab notebook. |
| |
| Publishes the model code + trained tokenizer + (optionally) your local |
| checkpoint to HuggingFace Hub so the Colab script can pull them. |
| |
| set HF_TOKEN=your_huggingface_token |
| python upload_artifacts.py # code + tokenizer |
| python upload_artifacts.py --with-ckpt # also upload latest local checkpoint |
| """ |
| import os, sys, argparse |
| from huggingface_hub import HfApi |
|
|
| ROOT = r"C:\Users\User\CalcGPU\clankerDiffusion" |
| CODE_REPO = "clankerDiffusion/base" |
| CKPT_REPO = "clankerDiffusion/checkpoints" |
|
|
|
|
| def main(): |
| ap = argparse.ArgumentParser() |
| ap.add_argument("--with-ckpt", action="store_true") |
| a = ap.parse_args() |
|
|
| token = os.environ.get("HF_TOKEN") |
| if not token: |
| sys.exit("set HF_TOKEN first: set HF_TOKEN=hf_xxx") |
| api = HfApi(token=token) |
|
|
| api.create_repo(CODE_REPO, repo_type="model", exist_ok=True) |
| code_files = ["model.py", "tokenizer.py", "train.py", "prep.py", "infer.py", |
| os.path.join("data", "tokenizer.json"), |
| os.path.join("data", "tokenizer.json.meta.json")] |
| for rel in code_files: |
| p = os.path.join(ROOT, rel) |
| if os.path.exists(p): |
| api.upload_file(repo_id=CODE_REPO, |
| path_in_repo=os.path.basename(p), |
| path_or_fileobj=p) |
| print("uploaded", rel) |
| else: |
| print("skip (missing)", rel) |
|
|
| if a.with_ckpt: |
| api.create_repo(CKPT_REPO, repo_type="model", exist_ok=True) |
| ckpt_dir = os.path.join(ROOT, "checkpoints") |
| if os.path.isdir(ckpt_dir): |
| pts = sorted(f for f in os.listdir(ckpt_dir) if f.endswith(".pt")) |
| if pts: |
| p = os.path.join(ckpt_dir, pts[-1]) |
| api.upload_file(repo_id=CKPT_REPO, |
| path_in_repo=os.path.basename(p), |
| path_or_fileobj=p) |
| print("uploaded checkpoint", pts[-1]) |
| print("DONE. Colab CODE_REPO =", CODE_REPO) |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
