├── .env.example ├── .github ├── pull_request_template.md └── workflows │ └── checks.yaml ├── .gitignore ├── .pre-commit-config.yaml ├── .vscode ├── extensions.json ├── launch.json └── settings-example.json ├── ACCESS.md ├── Makefile ├── README.md ├── conftest.py ├── pyproject.toml ├── spd ├── __init__.py ├── experiments │ ├── resid_mlp │ │ ├── model_interp.py │ │ ├── models.py │ │ ├── plotting.py │ │ ├── resid_mlp_dataset.py │ │ ├── resid_mlp_decomposition.py │ │ ├── resid_mlp_sweep_config.yaml │ │ ├── resid_mlp_topk_config.yaml │ │ ├── spd_interp.py │ │ └── train_resid_mlp.py │ └── tms │ │ ├── models.py │ │ ├── spd_interp.py │ │ ├── tms_decomposition.py │ │ ├── tms_lp_config.yaml │ │ ├── tms_sweep_config.yaml │ │ ├── tms_topk_config.yaml │ │ └── train_tms.py ├── hooks.py ├── log.py ├── models │ ├── __init__.py │ ├── base.py │ └── components.py ├── module_utils.py ├── plotting.py ├── run_spd.py ├── settings.py ├── types.py ├── utils.py └── wandb_utils.py └── tests ├── test_resid_mlp.py ├── test_spd_losses.py ├── test_spd_model.py ├── test_tms.py └── test_utils.py /.env.example: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/.env.example -------------------------------------------------------------------------------- /.github/pull_request_template.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/.github/pull_request_template.md -------------------------------------------------------------------------------- /.github/workflows/checks.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/.github/workflows/checks.yaml -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/.gitignore -------------------------------------------------------------------------------- /.pre-commit-config.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/.pre-commit-config.yaml -------------------------------------------------------------------------------- /.vscode/extensions.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/.vscode/extensions.json -------------------------------------------------------------------------------- /.vscode/launch.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/.vscode/launch.json -------------------------------------------------------------------------------- /.vscode/settings-example.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/.vscode/settings-example.json -------------------------------------------------------------------------------- /ACCESS.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/ACCESS.md -------------------------------------------------------------------------------- /Makefile: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/Makefile -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/README.md -------------------------------------------------------------------------------- /conftest.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/conftest.py -------------------------------------------------------------------------------- /pyproject.toml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/pyproject.toml -------------------------------------------------------------------------------- /spd/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /spd/experiments/resid_mlp/model_interp.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/resid_mlp/model_interp.py -------------------------------------------------------------------------------- /spd/experiments/resid_mlp/models.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/resid_mlp/models.py -------------------------------------------------------------------------------- /spd/experiments/resid_mlp/plotting.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/resid_mlp/plotting.py -------------------------------------------------------------------------------- /spd/experiments/resid_mlp/resid_mlp_dataset.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/resid_mlp/resid_mlp_dataset.py -------------------------------------------------------------------------------- /spd/experiments/resid_mlp/resid_mlp_decomposition.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/resid_mlp/resid_mlp_decomposition.py -------------------------------------------------------------------------------- /spd/experiments/resid_mlp/resid_mlp_sweep_config.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/resid_mlp/resid_mlp_sweep_config.yaml -------------------------------------------------------------------------------- /spd/experiments/resid_mlp/resid_mlp_topk_config.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/resid_mlp/resid_mlp_topk_config.yaml -------------------------------------------------------------------------------- /spd/experiments/resid_mlp/spd_interp.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/resid_mlp/spd_interp.py -------------------------------------------------------------------------------- /spd/experiments/resid_mlp/train_resid_mlp.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/resid_mlp/train_resid_mlp.py -------------------------------------------------------------------------------- /spd/experiments/tms/models.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/tms/models.py -------------------------------------------------------------------------------- /spd/experiments/tms/spd_interp.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/tms/spd_interp.py -------------------------------------------------------------------------------- /spd/experiments/tms/tms_decomposition.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/tms/tms_decomposition.py -------------------------------------------------------------------------------- /spd/experiments/tms/tms_lp_config.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/tms/tms_lp_config.yaml -------------------------------------------------------------------------------- /spd/experiments/tms/tms_sweep_config.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/tms/tms_sweep_config.yaml -------------------------------------------------------------------------------- /spd/experiments/tms/tms_topk_config.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/tms/tms_topk_config.yaml -------------------------------------------------------------------------------- /spd/experiments/tms/train_tms.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/experiments/tms/train_tms.py -------------------------------------------------------------------------------- /spd/hooks.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/hooks.py -------------------------------------------------------------------------------- /spd/log.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/log.py -------------------------------------------------------------------------------- /spd/models/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /spd/models/base.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/models/base.py -------------------------------------------------------------------------------- /spd/models/components.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/models/components.py -------------------------------------------------------------------------------- /spd/module_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/module_utils.py -------------------------------------------------------------------------------- /spd/plotting.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/plotting.py -------------------------------------------------------------------------------- /spd/run_spd.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/run_spd.py -------------------------------------------------------------------------------- /spd/settings.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/settings.py -------------------------------------------------------------------------------- /spd/types.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/types.py -------------------------------------------------------------------------------- /spd/utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/utils.py -------------------------------------------------------------------------------- /spd/wandb_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/spd/wandb_utils.py -------------------------------------------------------------------------------- /tests/test_resid_mlp.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/tests/test_resid_mlp.py -------------------------------------------------------------------------------- /tests/test_spd_losses.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/tests/test_spd_losses.py -------------------------------------------------------------------------------- /tests/test_spd_model.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/tests/test_spd_model.py -------------------------------------------------------------------------------- /tests/test_tms.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/tests/test_tms.py -------------------------------------------------------------------------------- /tests/test_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ApolloResearch/apd/HEAD/tests/test_utils.py --------------------------------------------------------------------------------