DRAFT: MAS (Model Analyze Script) #75
Draft
Conversation
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
parjong
force-pushed
the
draft/roofline_v2
branch
3 times, most recently
from
723e003 to
593fbc0
Compare
parjong
commented
Apr 3, 2022
| assert operator.OutputsLength() == 1 | ||
| ofm_idx = operator.Outputs(0) | ||
|
|
||
| # Q. Is it possible to enumerate defined fields in conv2d_options? |
Owner
Author
There was a problem hiding this comment.
Yes, it is possible (with minor hack). See mas.
parjong
commented
Apr 3, 2022
|
|
||
| class OperationTranslatorRegistry: | ||
| def __init__(self): | ||
| # Let's port this part to user-level scripts |
Owner
Author
There was a problem hiding this comment.
This is the key difference between rfl-model-analyze and mas.
parjong
commented
Apr 3, 2022
parjong
commented
Apr 3, 2022
parjong
commented
Apr 3, 2022
parjong
commented
Apr 3, 2022
| out = tf.transpose(inp) | ||
|
|
||
| tfgex_set_inputs(inp) | ||
| tfgex_set_outputs(out) |
Owner
Author
There was a problem hiding this comment.
It it a bit interesting that a .tflite built from this .tfgex file does not have TransposeOptions... ``)
parjong
force-pushed
the
draft/roofline_v2
branch
from
593fbc0 to
33930de
Compare
parjong
commented
Apr 3, 2022
parjong
commented
Apr 3, 2022
parjong
force-pushed
the
draft/roofline_v2
branch
2 times, most recently
from
967f720 to
9ec4931
Compare
parjong
commented
Apr 3, 2022
| rules['CONV_2D'] = conv2d | ||
| rules['DEPTHWISE_CONV_2D'] = dconv2d_ref | ||
| rules['FULLY_CONNECTED'] = fully_connected | ||
| rules['LEAKY_RELU'] = unimplemented |
parjong
commented
Apr 3, 2022
| ScalarField('AsymmetricQuantizeInputs'), | ||
| ] | ||
| opschema['LEAKY_RELU'] = [ | ||
| ScalarField('Alpha'), |
parjong
force-pushed
the
draft/roofline_v2
branch
3 times, most recently
from
70c12c1 to
99f1509
Compare
parjong
commented
Apr 3, 2022
| return self.id.src == ModelInput | ||
|
|
||
| def is_model_const(self): | ||
| return self.id.src == ModelConst |
parjong
commented
Apr 3, 2022
| wgt_size += mas.model.value(wgt_value_id).byte_size | ||
|
|
||
| info(f'Effetive weight size is {wgt_size} bytes') | ||
| mas.set_attr('basic.wsize', wgt_size) |
parjong
force-pushed
the
draft/roofline_v2
branch
8 times, most recently
from
b9c0ce7 to
8b094b1
Compare
parjong
force-pushed
the
draft/roofline_v2
branch
from
8b094b1 to
fb39b1f
Compare
parjong
force-pushed
the
draft/roofline_v2
branch
10 times, most recently
from
6dbc189 to
0843655
Compare
parjong
force-pushed
the
draft/roofline_v2
branch
3 times, most recently
from
3c8f46b to
f1268a8
Compare
Let's design and implement a NN model characteristics analysis toolkit. [ci skip] Signed-off-by: Jonghyun Park <parjong@gmail.com>
parjong
force-pushed
the
draft/roofline_v2
branch
from
f1268a8 to
b59bd70
Compare
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.
Let's design and implement a NN model characteristics analysis toolkit.
Signed-off-by: Jonghyun Park parjong@gmail.com