423A35C7/pytorch3d
1
0
Fork 0
mirror of https://github.com/facebookresearch/pytorch3d.git synced 2025-12-22 23:30:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

Move load_stats to TrainingLoop

Browse Source 
Summary:
Stats are logically connected to the training loop, not to the model. Hence, moving to the training loop.

Also removing resume_epoch from OptimizerFactory in favor of a single place - ModelFactory. This removes the need for config consistency checks etc.

Reviewed By: kjchalup

Differential Revision: D38313475

fbshipit-source-id: a1d188a63e28459df381ff98ad8acdcdb14887b7
This commit is contained in:
David Novotny
2022-08-02 15:40:53 -07:00
committed by Facebook GitHub Bot
parent 760305e044
commit c3f8dad55c
11 changed files with 259 additions and 189 deletions
Download Patch File Download Diff File Expand all files Collapse all files

46
projects/implicitron_trainer/tests/test_experiment.py
View File

@@ -5,6 +5,7 @@
# LICENSE file in the root directory of this source tree.
import os
import tempfile
import unittest
from pathlib import Path
@@ -172,3 +173,48 @@ class TestNerfRepro(unittest.TestCase):
experiment_runner = experiment.Experiment(**cfg)
experiment.dump_cfg(cfg)
experiment_runner.run()
@unittest.skip("This test checks resuming of the NeRF training.")
def test_nerf_blender_resume(self):
# Train one train batch of NeRF, then resume for one more batch.
# Set env vars BLENDER_DATASET_ROOT and BLENDER_SINGLESEQ_CLASS first!
if not interactive_testing_requested():
return
with initialize_config_dir(config_dir=str(IMPLICITRON_CONFIGS_DIR)):
with tempfile.TemporaryDirectory() as exp_dir:
cfg = compose(config_name="repro_singleseq_nerf_blender", overrides=[])
cfg.exp_dir = exp_dir
# set dataset len to 1
# fmt: off
(
cfg
.data_source_ImplicitronDataSource_args
.data_loader_map_provider_SequenceDataLoaderMapProvider_args
.dataset_length_train
) = 1
# fmt: on
# run for one epoch
cfg.training_loop_ImplicitronTrainingLoop_args.max_epochs = 1
experiment_runner = experiment.Experiment(**cfg)
experiment.dump_cfg(cfg)
experiment_runner.run()
# update num epochs + 2, let the optimizer resume
cfg.training_loop_ImplicitronTrainingLoop_args.max_epochs = 3
experiment_runner = experiment.Experiment(**cfg)
experiment_runner.run()
# start from scratch
cfg.model_factory_ImplicitronModelFactory_args.resume = False
experiment_runner = experiment.Experiment(**cfg)
experiment_runner.run()
# force resume from epoch 1
cfg.model_factory_ImplicitronModelFactory_args.resume = True
cfg.model_factory_ImplicitronModelFactory_args.force_resume = True
cfg.model_factory_ImplicitronModelFactory_args.resume_epoch = 1
experiment_runner = experiment.Experiment(**cfg)
experiment_runner.run()