# tinker_cookbook.weights.publish_to_hf_hub

### [**tinker_cookbook.weights.publish_to_hf_hub**](https://github.com/thinking-machines-lab/tinker-cookbook/blob/main/tinker_cookbook/weights/_publish.py#L15)(*model_path*, *repo_id*, *private*, *token*, *model_card*)

Push a model or adapter directory to HuggingFace Hub.

Works with outputs from [`build_hf_model`](https://tinker-docs.thinkingmachines.ai/cookbook/api-reference/weights/build_hf_model/index.md), `build_lora_adapter`, or any HuggingFace-compatible model directory.

**Parameters:**

- [**model_path**](https://github.com/thinking-machines-lab/tinker-cookbook/blob/main/tinker_cookbook/weights/_publish.py#L17) (*str*) – Local path to the model or adapter directory to upload.
- [**repo_id**](https://github.com/thinking-machines-lab/tinker-cookbook/blob/main/tinker_cookbook/weights/_publish.py#L18) (*str*) – HuggingFace Hub repository ID (e.g. `"user/my-model"`).
- [**private**](https://github.com/thinking-machines-lab/tinker-cookbook/blob/main/tinker_cookbook/weights/_publish.py#L19) (*bool*) – Whether the repository should be private. Defaults to `True` for safety.
- [**token**](https://github.com/thinking-machines-lab/tinker-cookbook/blob/main/tinker_cookbook/weights/_publish.py#L20) (*str | None*) – HuggingFace API token. If `None` (default), uses the `HF_TOKEN` environment variable or cached login from `hf auth login`.
- [**model_card**](https://github.com/thinking-machines-lab/tinker-cookbook/blob/main/tinker_cookbook/weights/_publish.py#L21) (*ModelCardConfig | None*) – Optional model card configuration. When provided, a `README.md` is auto-generated and included in the upload. If the directory already contains a `README.md`, the existing file is preserved and a warning is logged.

**Returns:** URL of the published repository.

```python
from tinker_cookbook import weights
url = weights.publish_to_hf_hub(
model_path="./merged_model",
repo_id="my-org/my-model",
)
print(url)
```