423A35C7/pytorch3d
1
0
Fork 0
mirror of https://github.com/facebookresearch/pytorch3d.git synced 2025-08-02 03:42:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

IT readme updates

Browse Source 
Summary: Updates for recent replaceables.

Reviewed By: kjchalup

Differential Revision: D38437370

fbshipit-source-id: 00d600aa451e5849ba48107cd7a4319e9fc8549f
This commit is contained in:
Jeremy Reizenstein 2022-08-09 20:48:51 -07:00 committed by Facebook GitHub Bot
parent a39cad40f4
commit af48430ec0
1 changed files with 29 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

46
projects/implicitron_trainer/README.md
View File

@ -203,14 +203,29 @@ to replace the implementation and potentially override the parameters.
# Code and config structure # Code and config structure
As per above, the config structure is parsed automatically from the module hierarchy. The main object for this trainer loop is `Experiment`. It has four top-level replaceable components.
In particular, model parameters are contained in `generic_model_args` node, and dataset parameters in `data_source_args` node.
Here is the class structure (single-line edges show aggregation, while double lines show available implementations): * `data_source`: This is a `DataSourceBase` which defaults to `ImplicitronDataSource`.
It constructs the data sets and dataloaders.
* `model_factory`: This is a `ModelFactoryBase` which defaults to `ImplicitronModelFactory`.
It constructs the model, which is usually an instance of implicitron's main `GenericModel` class, and can load its weights from a checkpoint.
* `optimizer_factory`: This is an `OptimizerFactoryBase` which defaults to `ImplicitronOptimizerFactory`.
It constructs the optimizer and can load its weights from a checkpoint.
* `training_loop`: This is a `TrainingLoopBase` which defaults to `ImplicitronTrainingLoop` and defines the main training loop.
As per above, the config structure is parsed automatically from the module hierarchy.
In particular, for ImplicitronModelFactory with generic model, model parameters are contained in the `model_factory_ImplicitronModelFactory_args.model_GenericModel_args` node, and dataset parameters in `data_source_ImplicitronDataSource_args` node.
Here is the class structure of GenericModel (single-line edges show aggregation, while double lines show available implementations):
``` ```
generic_model_args: GenericModel model_GenericModel_args: GenericModel
└-- sequence_autodecoder_args: Autodecoder └-- global_encoder_*_args: GlobalEncoderBase
└-- raysampler_args: RaySampler ╘== SequenceAutodecoder
└-- autodecoder_args: Autodecoder
╘== HarmonicTimeEncoder
└-- raysampler_*_args: RaySampler
╘== AdaptiveRaysampler
╘== NearFarRaysampler
└-- renderer_*_args: BaseRenderer └-- renderer_*_args: BaseRenderer
╘== MultiPassEmissionAbsorptionRenderer ╘== MultiPassEmissionAbsorptionRenderer
╘== LSTMRenderer ╘== LSTMRenderer
@ -228,19 +243,16 @@ generic_model_args: GenericModel
╘== IdrFeatureField ╘== IdrFeatureField
└-- image_feature_extractor_*_args: FeatureExtractorBase └-- image_feature_extractor_*_args: FeatureExtractorBase
╘== ResNetFeatureExtractor ╘== ResNetFeatureExtractor
└-- view_sampler_args: ViewSampler └-- view_pooler_args: ViewPooler
└-- feature_aggregator_*_args: FeatureAggregatorBase └-- view_sampler_args: ViewSampler
╘== IdentityFeatureAggregator └-- feature_aggregator_*_args: FeatureAggregatorBase
╘== AngleWeightedIdentityFeatureAggregator ╘== IdentityFeatureAggregator
╘== AngleWeightedReductionFeatureAggregator ╘== AngleWeightedIdentityFeatureAggregator
╘== ReductionFeatureAggregator ╘== AngleWeightedReductionFeatureAggregator
solver_args: init_optimizer ╘== ReductionFeatureAggregator
data_source_args: ImplicitronDataSource
└-- dataset_map_provider_*_args
└-- data_loader_map_provider_*_args
``` ```
Please look at the annotations of the respective classes or functions for the lists of hyperparameters. Please look at the annotations of the respective classes or functions for the lists of hyperparameters. `tests/experiment.yaml` shows every possible option if you have no user-defined classes.
# Reproducing CO3D experiments # Reproducing CO3D experiments