Repository navigation
Poi label names - #115
Closed
Hendrik-code wants to merge 759 commits into
Closed
Poi label names#115Hendrik-code wants to merge 759 commits into
Hendrik-code wants to merge 759 commits into
Conversation
…nto development_robert
…nto development_robert
…nto development_robert
…nto development_robert
…nto development_robert
Development robert
…nto development_robert
…nto development_robert
Added vibe instances
…nto development_robert
…nto development_robert
All sliding-window tiles share the same patch_size shape, so cuDNN can autotune the fastest conv algorithms once and reuse them across tiles and images. TF32 accelerates fp32 matmul/conv on Ampere+ GPUs with negligible accuracy impact. Gated behind a new fast_perf flag (default True, CUDA only). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…erence inference_mode disables autograd view/version tracking entirely, so it is slightly faster and uses less memory than no_grad. All tensors here are pure inference outputs (moved to CPU / converted to numpy downstream), so the stricter inference-tensor semantics are safe. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
predict_sliding_window_return_logits runs once per fold; calling torch.cuda.empty_cache() at its start and end returned allocator blocks to the driver right before the next fold reallocated them (slow cudaMalloc + sync). The pool is still cleared once per image after fold averaging, and all OOM-recovery empty_cache() calls are untouched. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
initialize_from_trained_model_folder built loaded_networks by appending the same self.network object once per fold, so every entry ended up holding the LAST fold's weights. predict_logits_from_preprocessed_data then ran that one weight set N times and averaged it, silently collapsing the N-fold ensemble to a single fold while still paying Nx compute. Separately, cache_state_dicts=False left loaded_networks as [] (not None), causing an IndexError in the predict loop. Now loaded_networks is None unless exactly one fold is present (preloaded for zero per-prediction reloads). For >1 fold it stays None and the predict loop swaps weights per fold via load_state_dict, restoring a correct ensemble using a single network's worth of GPU memory. Device-availability warnings now fire once regardless of fold count. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Every run_inference_on_file call previously reloaded the predictor from disk and re-uploaded weights to the GPU. With cache_model=True the loaded predictor is kept in a process-wide cache keyed by model path + folds + device/runtime settings, so a loop over many files reuses the in-memory model. When caching, the end-of-call del/empty_cache is skipped so the CUDA allocator pool stays warm between images. Default is False to preserve current memory semantics; the flag forwards through run_VibeSeg/run_vibeseg/run_nnunet via **kwargs. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
_run_sub processed one tile per forward pass, leaving the GPU underutilised for small patches. Tiles all share patch_size, so they batch densely: with tile_batch_size>1 they are stacked into a (B, C, *patch) tensor and run in one pass, then scattered back per tile. Mirroring/gaussian/accumulation are unchanged. tile_batch_size=1 (default) takes the original view-based path verbatim, so behaviour and memory are unchanged unless opted in. Threaded through nnUNetPredictor, load_inf_model and run_inference_on_file. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds benchmarks/ with a phase-resolved timing tool for run_inference_on_file (load / preprocess / sliding-window predict / postprocess), using CUDA synchronisation, warmup, repeats and peak-memory tracking. It instruments the pipeline by monkeypatching (no library changes) and signature-filters its kwargs so the same file runs against every commit in the optimisation range. - benchmark_nnunet_inference.py: run/compare subcommands; synthetic or real input. - bench_across_commits.sh: replays one config across baseline+each commit and prints per-commit deltas (isolates the always-on changes; the fold fix shows up as the fold_status column flipping from DUPLICATED to lazy-per-fold/distinct). - bench_flag_sweep.sh: sweeps the opt-in flags at HEAD (cache_model needs >1 call, tile_batch_size, max_folds, step_size). - README.md: usage, which tool measures which commit, and caveats. pyproject: exempt benchmarks/ from docstring rules, mirroring the test dirs. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…nto development_robert
Speed up nnUnet inference
POI.copy(), _set_inplace(), POI_Global.to_other(), and calc_poi_average() dropped the naming metadata (level_one_info, level_two_info, and with it info['label_name']), and POI_Descriptor.copy() dropped its name<->id definition. Now every copy/derive path preserves them, so names survive reorient/rescale/extract/map_labels/to_global/ average (all of which route through copy()). Verified preserved across those ops; 76 existing POI tests still pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
label_name now uses the nested form {region:{subregion:name,'name':group}} instead
of flat {'(1, 2)':'C2'}. Old flat files (and JSON string keys) are migrated on load
(normalize_label_name); saving always writes the new format (mrk.json export stays
byte-identical, fed from the nested dict). New accessors POI.label_name(region,
subregion) and POI.level_one_name(region) resolve custom names with priority over
level_two_info/level_one_info enum names (warning when a custom name diverges from a
given enum name); set_label_name/set_level_one_name write them. mrk/txt import and
join_left updated to the nested form. 89 POI tests + 8 new tests pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
unpack_poi_id now resolves a key part via level_one_info/level_two_info enum names (when set) before the descriptor's definition, so poi[idx, 'level2name'] and poi['level1name', 'level2name'] work for get/set/contains alongside ids and Enums. Using an Enum key while the matching level_*_info is still 'Any' (unset) now emits a warning (the .value is still used). 100 POI tests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Collaborator
|
Do not merge for the new release. Needs a lot of testing. |
robert-graf
force-pushed
the
poi-label-names
branch
from
August 6, 2026 12:11
3ab343d to
5ea7ea5
Compare
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
Overhauls POI naming across
POIandPOI_Globalper the requested spec. Three focused commits; full unit suite 406 passed / 4 skipped (395 baseline + 11 new POI tests). Coloring/serialization of.mrk.jsonis kept byte-identical.What changed (mapped to the requirements)
POI.copy,_set_inplace,POI_Global.to_other,calc_poi_average, andPOI_Descriptor.copynow carrylevel_one_info/level_two_info/info/definition.Since reorient/rescale/extract/map_labels/to_global/average route through
copy(), the names follow.poi[idx, "level2name"]andpoi["level1name", "level2name"]work for get/set/in(viaunpack_poi_id→level_*_info._get_id→ descriptor definition), alongside ids and Enums..value; a warning fires when the matchinglevel_*_infois stillAny(unset).poi.level_one_name(region)— returns the group (level-one) name.poi.label_name(region, subregion)— returns the point name.set_label_name/set_level_one_namewrite them.label_nameformat —{region: {subregion: name, "name": group}}instead of flat{"(1, 2)": "C2"}.normalize_label_name()migrates the old flat form (and JSON string keys) on load;save always writes the new form.
.mrk.jsonexport/import stays byte-identical, now fed from the nested dict.join_left/txt-import updated.label_namewins over thelevel_x_infoenum name, with a warning when it diverges from a given enum name.Notes for reviewers
Location.save_as_name()isFalse(subregions saved as ints), so its_get_namereturns the id string — the accessor treats that as "no enum name" and falls back to the id/custom name.Vertebra_Instanceuses real names.(Tell me if you'd like
Locationto also resolve to enum names.)__getitem__/__setitem__/__contains__;extract_region/extract_subregion/removestill take ids+Enums (not yet level names) — easy to extend.unit_tests/test_poi_label_names.py(11 tests) locking migration, accessors, round-trip, old-format load, name indexing, warnings, and metadata transfer.Commits:
18d9096fix(poi): carry level_one_info/level_two_info/info through all POI ops38be00efeat(poi): nested label_name format + name accessors + migration3ab343dfeat(poi): index POIs by level names + warn on enum without level_*_info