Add Pipeline class #2140
Closed
Add Pipeline class #2140
Conversation
This file contains 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
github-actions
bot
added
NNCF PTQ
github-actions
bot
added
NNCF PT
andrey-churkin
force-pushed
the
ac/pipelines
branch
2 times, most recently
from
8041e5d
to
c0ce9ae
Compare
andrey-churkin
force-pushed
the
ac/pipelines
branch
5 times, most recently
from
750ad55
to
55fcd6c
Compare
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## develop #2140 +/- ##
===========================================
+ Coverage 36.24% 36.29% +0.04%
===========================================
Files 477 478 +1
Lines 42570 42617 +47
===========================================
+ Hits 15428 15466 +38
- Misses 27142 27151 +9
|
KodiaqQ
reviewed
Sep 19, 2023
| @@ -116,8 +116,7 @@ def native_quantize_impl( | |||
| advanced_parameters=advanced_parameters, | |||
| ) | |||
|
|
|||
| graph = GraphConverter.create_nncf_graph(model) | |||
| quantized_model = quantization_algorithm.apply(model, graph, dataset=calibration_dataset) | |||
| quantized_model = quantization_algorithm.run(model, calibration_dataset) | |||
There was a problem hiding this comment.
quantization_pipeline.run? For the consistency.
| calibration_dataset: Dataset, | ||
| validation_fn: Callable[[Any, Iterable[Any]], Tuple[float, Union[None, List[float], List[List[TTensor]]]]], | ||
| subset_size: int, | ||
| initial_metric_results: MetricResults, | ||
| quantized_metric_results: MetricResults, | ||
| ): | ||
| """ | ||
| :param algorithm_cls: Class of algorithm. | ||
| :param pipeline_cls: Class of pipeline. |
There was a problem hiding this comment.
Pipeline or StepwisePipeline after all?
|
|
||
| algorithm = self._algorithms[best_combination_key] | ||
| result_model = algorithm.apply(model, initial_graph, self._statistic_points) | ||
| # TODO(andrey-churkin): Show final best settings |
There was a problem hiding this comment.
How do the final best_settings apply to the model?
| if bias_correction_params.threshold is not None: | ||
| threshold = bias_correction_params.threshold | ||
|
|
||
| pipeline_steps[-1].append( |
| PipelineStep = List[Algorithm] | ||
|
|
||
|
|
||
| def get_statistic_points(pipeline_step: PipelineStep, model: TModel, graph: NNCFGraph) -> StatisticPointsContainer: |
There was a problem hiding this comment.
Why not keep these methods in the StepwisePipeline class?
| # Run current pipeline step | ||
| step_model = run_pipeline_step(pipeline_step, step_statistics, step_model, step_graph) | ||
|
|
||
| step_graph = None # We should rebuild the graph for the next pipeline step |
There was a problem hiding this comment.
Why can't we drop it and rebuild NNCFGraph in each step? Any optimizations here?
Comment on lines
126
to
129
| for algorithm in pipeline_step[:-1]: | ||
| current_model = algorithm.apply(current_model, current_graph, pipeline_step_statistics) | ||
| current_graph = NNCFGraphFactory.create(current_model) | ||
| current_model = pipeline_step[-1].apply(current_model, current_graph, pipeline_step_statistics) |
There was a problem hiding this comment.
I'm not sure that determining of the major algorithms such as MinMax quantization based on their position is a good idea. We should always keep in mind the position of all algorithms, even outside of the PostTrainingQuantization.
| algo = PostTrainingQuantization(target_device=target_device) | ||
| min_max_algo = algo.algorithms[0] | ||
| pipelines = PostTrainingQuantization(target_device=target_device) | ||
| min_max_algo = pipelines.pipeline_steps[-1][0] |
There was a problem hiding this comment.
This index hardcode doesn't explain clearly, how the MinMax algorithm should be get.
A good example of the increased pipeline complexity and the need for the MinMax as the independent unit.
cc @alexsu52
|
run openvino pre-commit tests |
andrey-churkin
force-pushed
the
ac/pipelines
branch
from
55fcd6c
to
75d81f1
Compare
alexsu52
requested changes
Sep 19, 2023
| @@ -175,7 +180,7 @@ def find_best_combination( | |||
| return best_combination_key | |||
|
|
|||
|
|
|||
| class HyperparameterTuner: | |||
| class HyperparameterTuner(Pipeline): | |||
There was a problem hiding this comment.
I couldn't find which interface I should implement to support my algorithm with the HyperparameterTuner algorithm.
andrey-churkin
force-pushed
the
ac/pipelines
branch
from
75d81f1
to
69732b9
Compare
andrey-churkin
force-pushed
the
ac/pipelines
branch
from
71b7f84
to
96e8d76
Compare
andrey-churkin
force-pushed
the
ac/pipelines
branch
from
f56b9f7
to
d3d90a1
Compare
andrey-churkin
force-pushed
the
ac/pipelines
branch
from
d3d90a1
to
37c8666
Compare
andrey-churkin
changed the title
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.
Changes
The primary goals of that pull request are to
Reason for changes
Related tickets
Ref: 117471
Tests