Add first-class GTSFM Studio workflow - #1121
Open
suhanpark wants to merge 18 commits into
Open
Conversation
Remove SPZ export dependency
akshay-krishnan
marked this pull request as ready for review
September 1, 2026 15:21
Collaborator
|
@copilot summarize the changes to each file in this PR |
Contributor
Here’s a per-file summary of what changed in this PR:
|
Collaborator
There was a problem hiding this comment.
could we add a readme in this folder that explains the different workflows? what is this one for?
Collaborator
There was a problem hiding this comment.
could we add a readme in this folder to explain how this dockerfile is meant to be used?
Collaborator
Author
There was a problem hiding this comment.
I found that the Dockerfile duplicates the dependency setup already defined in modal_app.py. i removed the custom Dockerfile and its publishing path
Comment on lines
+157
to
+170
| delayed(self._save_cached_result_outputs, pure=False)(cached_graph, context.output_paths.results) | ||
| ] | ||
| gaussian_optimizer = vars(self._optimizer).get("gaussian_splatting_optimizer") | ||
| if gaussian_optimizer is not None: | ||
| import gtsfm.splat.rendering as gtsfm_rendering | ||
|
|
||
| images_graph = context.get_delayed_image_map() | ||
| splats_graph, cfg_graph = gaussian_optimizer.create_computation_graph(images_graph, cached_graph) | ||
| io_tasks.extend( | ||
| [ | ||
| delayed(gtsfm_rendering.save_splats)(context.output_paths.results, splats_graph), | ||
| delayed(gtsfm_rendering.generate_interpolated_video)( | ||
| images_graph, | ||
| cached_graph, |
Collaborator
There was a problem hiding this comment.
did you test running gaussian splatting with the cluster optimizer cacher? this feels unrelated to this PR, not sure if its intentional?
| gtsfm_data = gtsfm_data.filter_landmark_measurements( | ||
| pre_ba_max_reproj_error, min_track_length | ||
| ) | ||
| gtsfm_data = gtsfm_data.filter_landmark_measurements(pre_ba_max_reproj_error, min_track_length) |
Collaborator
There was a problem hiding this comment.
are you sure the maximum line length setting in your IDE has been applied? the formatting seems off here.
Collaborator
Author
There was a problem hiding this comment.
I’ll check my editor’s formatting settings and clean up the formatting here
| ) | ||
|
|
||
| if step == 0 or (step + 1) % 10 == 0 or (step + 1) == max_steps: | ||
| from gtsfm.splat.live import publish_training_update |
Collaborator
There was a problem hiding this comment.
is this file heavy to import? does it need some cuda kernels registered? normally we put all imports at the top unless we really want to avoid importing it for most use cases.
| from gtsfm.scene_optimizer import SceneOptimizer | ||
|
|
||
|
|
||
| def _scheduler_task_counts(dask_scheduler: Any) -> dict[str, int]: |
Collaborator
There was a problem hiding this comment.
to avoid bloating in this central and important file, can we move these functions to a separate file?
| OmegaConf.update( | ||
| main_cfg, | ||
| "cluster_optimizer.gaussian_splatting_optimizer", | ||
| snapshot_path, |
Collaborator
There was a problem hiding this comment.
"snapshot" is a bit cryptic. not obvious what this means, until someone reads and understands the code.
|
|
||
|
|
||
| class TestProcessGraphGenerator(unittest.TestCase): | ||
| @unittest.skipUnless(shutil.which("dot"), "Graphviz 'dot' is not installed") |
Collaborator
There was a problem hiding this comment.
is this not a dependency? why skip these tests?
This was referenced Sep 10, 2026
addressing file removals docker/modal-runtime.Dockerfile .github/workflows/modal-runtime-image.yml
moved splat live import to top
snapshot_path → gs_optimizer_config_path
Refactor filter_landmark_measurements call for better readability.
Remove test for modal deployment preferring prebuilt runtime.
Removed redundant information about the prebuilt runtime and clarified the setup process for Modal workspace.
Removed redundant cancellation check for deployment.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Introduce a packaged, browser-based GTSFM Studio that launches with
gtsfm runand provides a FastAPI-backed React workspace for configuring, running, monitoring, and inspecting reconstructions.Why
The existing workflow requires users to know repository paths and command-line arguments. This creates a first-class package entry point and a discoverable UI while retaining the terminal execution path.
User impact
gtsfm runValidation
python -m pytest tests/visualization/test_runtime.py tests/ui/test_optional_process_graph.py -q(36 passed)pnpm buildReview notes
This is PR 1 of a stacked draft series. Later drafts isolate the reusable Modal runtime, workspace-control improvements, and live Gaussian visualization.