| title: MLflow Tracking | |
| emoji: 📈 | |
| colorFrom: blue | |
| colorTo: indigo | |
| sdk: docker | |
| app_port: 7860 | |
| pinned: false | |
| # MLflow Tracking (Mach1Corp) | |
| Docker Space that runs an [MLflow](https://mlflow.org/) tracking server for MOSS emotion LoRA experiments. | |
| ## Architecture | |
| | Layer | Resource | Visibility | | |
| | --- | --- | --- | | |
| | UI / API | this Space | public (no login) | | |
| | Persistence | [`Mach1Corp/mlflow-bucket`](https://huggingface.co/buckets/Mach1Corp/mlflow-bucket) mounted at `/data` | private | | |
| - Backend store: `sqlite:////data/mlflow.db` | |
| - Artifacts: `/data/mlartifacts` | |
| ## Client usage | |
| ```bash | |
| export MLFLOW_TRACKING_URI=https://mach1corp-mlflow.hf.space | |
| export MLFLOW_EXPERIMENT_NAME=moss-emotion-lora | |
| ``` | |
| Then train with: | |
| ```bash | |
| python train_emotion_lora.py ... --mlflow-tracking-uri "$MLFLOW_TRACKING_URI" | |
| # or rely on MLFLOW_TRACKING_URI from the environment / run_train_emotion.sh | |
| ``` | |