|
Download baselines/README.md from Tinghong-Ye22/code-Psi0-yth-v21: direct link, hf CLI and curl.
- Browser
- Download file 901 Bytes
-
https://huggingface.co/Tinghong-Ye22/code-Psi0-yth-v21/resolve/main/baselines/README.md
- Command line
-
hf download hf://Tinghong-Ye22/code-Psi0-yth-v21/baselines/README.md
-
curl -L -o README.md https://huggingface.co/Tinghong-Ye22/code-Psi0-yth-v21/resolve/main/baselines/README.md
901 Bytes
Environment Management
We try to maintain a central code repository for $\Psi_0$ as well as all the baselines.
Key considerations are:
- We want to train each baseline with minimum code change, so we create separate isolated environment for each baseline.
- At the same time, we want all the baseline models, dataloaders, etc can be mixed in the same environment.
So we manage the environments as illuated below:
To conclude:
- $\Psi_0$ is managed using the default
pyproject.toml, and depdendency groups includebase,psi,vizandserve. - Each baseline install
src/package and base dependencies, so all the baseline models, dataloaders etc., can be used in the envrionment. - Each baseline maintain a seperate
requirements.txtwhich include the baseline'storchandtransformersversions.