divvun-tts/tts-evaluation

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

Saving responses to a private Hugging Face dataset

This project can optionally push saved responses to a private dataset on the Hugging Face Hub. The feature is disabled by default and only active when the following environment variables are set:

• —HF_TOKEN — a Hugging Face access token with dataset write permissions.
• —HF_DATASET_ID — repo id like your-username/my-eval-responses.

Setup:

1. 1.Create a token on https://huggingface.co/settings/tokens and grant it write permissions for datasets/repos.
2. 2.Export the variables locally before running the app:

export HF_TOKEN="hf_...your_token..."
export HF_DATASET_ID="your-username/my-eval-responses"

1. 1.Install dependencies and run the app:

pip install -r requirements.txt
python app.py
# open http://localhost:7860 and submit ratings via the UI

Notes:

• —If the datasets or huggingface_hub packages are not installed, HF pushing is skipped gracefully and responses are still written to responses.csv.
• —Pushing audio files will use Git LFS on the Hub; monitor your account storage and LFS quota.
• —Each UI save triggers a small push (single-record commit). For heavy usage consider batching pushes instead.

There is also a small test script included to programmatically verify pushing a single sample row to your dataset. See hf_push_test.py.