This repository has been archived by the owner on Sep 18, 2024. It is now read-only.
[DO NOT SQUASH] Support nn-Meter in Retiarii framework #3938
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
ultmaster
changed the title
* Fix mutable default * LayerChoice:forward now will default run the first candidate to support trace (#3910) * New GraphConverter to parse shape info required by nn-meter (#3910) * Support model filter in Random strategy * Support latency aware search in SPOS multi-trial example * Fix for review (#3910) * Add doc for hardware-aware NAS * Fix lint python & Add nn_meter to sphinx mock * Add comments * Move LatencyFilter to examples * Move example inputs into configs * Support nested layer choice Co-authored-by: Jianyu Wei <v-wjiany@microsoft.com> Co-authored-by: kalineid <nnob@mail.ustc.edu.cn> Co-authored-by: Yuge Zhang <scottyugochang@gmail.com> Co-authored-by: Yuge Zhang <Yuge.Zhang@microsoft.com>
J-shang
reviewed
Jul 26, 2021
| @@ -654,8 +644,214 @@ def convert_module(self, script_module, module, module_name, ir_model): | |||
|
|
|||
| return ir_graph, {} | |||
|
|
|||
| def convert_module(self, script_module, module, module_name, ir_model): | |||
There was a problem hiding this comment.
why we need both convert_module and _convert_module?
There was a problem hiding this comment.
_convert_module is exposed to subclass to overwrite internal logic.
J-shang
reviewed
Jul 26, 2021
nni/retiarii/converter/graph_gen.py
Outdated
| If forward path of candidates depends on input data, then wrong path will be traced. | ||
| This will result in incomplete shape info. | ||
| """ | ||
| def convert_module(self, script_module, module, module_name, ir_model, example_inputs): |
There was a problem hiding this comment.
example_inputs -> dummy_input?
J-shang
reviewed
Jul 26, 2021
| self._trace_module(module, module_name, ir_model, example_inputs) | ||
| return ir_graph, attrs | ||
|
|
||
| def _initialize_parameters(self, ir_model: 'Model'): |
There was a problem hiding this comment.
is this a common use case of type hints?
There was a problem hiding this comment.
I think so. It's used "everywhere" in Retiarii.
liuzhe-lz
approved these changes
Jul 26, 2021
|
|
||
| Basically, this demo will select the model whose latency satisfy constraints to train. | ||
|
|
||
| To run this demo, first install nn-Meter from source code (currently we haven't released this package, so development installation is required). |
There was a problem hiding this comment.
Shall we give a link to nn-Meter here?
There was a problem hiding this comment.
I'm not sure. My suggestion is that you can do that later when you actually have nn-Meter open sourced.
J-shang
approved these changes
Jul 26, 2021
zheng-ningxin
approved these changes
Jul 27, 2021
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
No description provided.