[CK][CK TILE] Clean up tile_engine grouped_conv harness
(#7761)
## Motivation
Tile_engine grouped_conv contains ML heuristic validation scripts which
cause confusion to new developers. So, this PR is intended to relocate
the scripts into dispatcher/heuristic directory to maintain separation
of concern.
## Technical Details
The grouped_conv tile_engine directory is a benchmarking harness for
grouped convolution kernels; ML-heuristic content does not belong there.
- Move compare_ml_vs_oracle.py and validate_ml_vs_oracle.py from
tile_engine/ops/grouped_conv/ to
dispatcher/heuristics/validation/grouped_conv/, and rebase their
sys.path / oracle CSV / model dir lookups for the new location (CSV path
is now an --oracle-csv flag instead of a hard-coded sibling).
- Move GROUPED_CONV_HEURISTIC_REPORT.md (system-level ML report) into
dispatcher/heuristics/ where the rest of the heuristic docs live.
- Rewrite tile_engine/ops/grouped_conv/README.md as a pure benchmarking
/ dispatcher-sweep doc (kernel enumeration, JIT pipeline, CSV schema,
problem registry), in the style of tile_engine/ops/fmha/README.md. All
ML training / model-efficiency content is removed and replaced with a
pointer to dispatcher/heuristics/.
## Test Plan
Validation scripts are re-wired and tested locally
## Test Result
Tests passed on local machine.
## Submission Checklist
- [x ] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
[CK][CK TILE] Dispatcher kernel selection heuristic for grouped conv (#6327)
## Motivation
The ML heuristic in dispatcher does not support grouped-conv operator
yet. In this PR, the support for fwd, bdw-data, and bwd-weight
grouped-conv kernels have been added. A tile_engine utility has also
been added to compile and run any selected kernel configuration through
dispatcher infrastructure.
## Technical Details
1. Tile engine utility is added to benchmark each shape with all the
possible kernel+tile_size combinations here -
[https://github.com/ROCm/rocm-libraries/blob/users/yraparti/ck/dispatcher-grouped-conv-heuristics/projects/composablekernel/tile_engine/ops/grouped_conv/grouped_conv_full_benchmark.py](url)
2. New LGBM regressor models for grouped conv are added to models
directory. We have 3 separate models for fwd, bwd-data, and bwd-weights
[https://github.com/ROCm/rocm-libraries/tree/users/yraparti/ck/dispatcher-grouped-conv-heuristics/projects/composablekernel/dispatcher/heuristics/models](url)
3. Implemented lazy GPU initialization (dispatcher/python)
- **Issue**: ProcessPoolExecutor fork() + GPU context caused memory
access faults
- **Solution**: Mirror FMHA pattern - defer GPU initialization until
first run()
- **Changes**:
- setup_multiple_grouped_conv_dispatchers() returns List[Path], not
loaded libs
- GpuGroupedConvRunner.__init__() no longer calls ctypes.CDLL
- Added _ensure_initialized() method for lazy GPU loading
- GPU context created only on first run() call
- **Benefit**: Parallel compilation now works without GPU conflicts
4. Addressed few miscellaneous issues such as:
- Fixed BF16->FP16 naming bug in the dispatcher wrapper
- Added new tile sizes, and comp_v5 pipeline to the arch spec to expand
the kernel selection
- Added automatic padding support for unsupported shapes in dispatcher
runner
- Created a single source of truth between tile_engine and dispatcher
about the architecture and tile_size details
- Build a validation scripts to compare oracle_best vs ml_heuristic
comparison
## Test Plan
1. Validated fwd, bwd-data, and bwd-weight kernels with both known and
unseen data sets with up to 300 problems.
2. Ensured that test cases are added in both dispatcher and tile_engine
to validate the heuristic.
## Test Result
Results on Unseen shapes validated on gfx950
#### Forward Pass Model
- **Training Data**: 48,845 measurements across 1,372 unique problem
shapes
- **Validation Set**: 300 unseen problems from model crawler
- **Validation Performance** (vs. oracle):
- Mean Efficiency: **93.05%**
- Median Efficiency: **96.8%**
- P10 Efficiency: **79.9%**
#### Backward Data Gradient (bwd_data) Model
- **Training Data**: 18,773 measurements across 891 unique problem
shapes
- **Validation Set**: 300 unseen problems from model crawler
- **Validation Performance** (vs. oracle):
- Mean Efficiency: **93.8%**
- Median Efficiency: **96.5%**
- P10 Efficiency: **82.9%**
#### Backward Weight Gradient (bwd_weight) Model
- **Training Data**: 34,900 measurements across 1,508 unique problem
shapes
- **Validation Set**: 300 unseen problems from model crawler
- **Validation Performance** (vs. oracle):
- Mean Efficiency: **96.1%**
- Median Efficiency: **99.2%**
- P10 Efficiency: **89.4%**
## Submission Checklist
- [ x] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
---------
Co-authored-by: Vidyasagar Ananthan <vidyasagar.ananthan@amd.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Jan Patrick Lehr <JanPatrick.Lehr@amd.com>
