profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/nutsiepully/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.
Pulkit Bhuwalka nutsiepully Google Mountain View, CA

ICT4H/atomfeed 15

An ATOM implementation in Java to broadcast and consume events.

ICT4H/simplefeed 2

A simple Atom-based protocol for broadcasting a feed of events.

angshu/formidable-android-client 1

android client for Formidable

ICT4H/atomfeed-standalone 1

A server that creates and publishes events using the ATOM protocol.

ICT4H/Cordova-SQLitePlugin 1

Cordova-SQLitePlugin used by dcs-mobile

ICT4H/dcs-mobile 1

DCS mobile app

ICT4H/dcs-web 1

Forked from datawinners

angshu/formidable 0

Data collection platform for ICT4H

angshu/formidable-server 0

server app for formidable

djazayeri/openmrs-module-referenceemr 0

Configuration and setup for the OpenMRS EMR Reference Application

pull request commenttensorflow/model-optimization

[Collaborative optimization] Pruning-Clustering-preserving Quantization Aware Training

Merging was blocked by build file strict dependencies -- resubmitting with it fixed myself now, should be merged today.

Brill, thanks a lot, David.

Ruomei

comment created time in 2 days

issue closedtensorflow/model-optimization

TypeError: tf__call() got an unexpected keyword argument 'y'

Describe the bug When I use the API "tfmot.quantization.keras.quantize_model", there is a bug.I guess that some layers in mobilenetV3 do not support quantization operations, such as the lamba layer. I wonder if you have encountered it?

System information

I am using TensorFlow 2.5 with Python 3.6 I have pip-installed TF2.4 within an anaconda environment.

Code to reproduce the issue import tensorflow as tf assert float(tf.version[:3]) >= 2.3 config = tf.compat.v1.ConfigProto() config.gpu_options.allow_growth = True session = tf.compat.v1.InteractiveSession(config=config) import tensorflow.keras as keras import pathlib import numpy as np import tensorflow_model_optimization as tfmot from tensorflow.keras.preprocessing.image import ImageDataGenerator

origin_model = tf.keras.applications.MobileNetV3Small( input_shape=(224, 224, 3), alpha=1.0, include_top=True, weights='imagenet', input_tensor=None, pooling=None, classes=1000, classifier_activation='softmax' ) quant_aware_model = tfmot.quantization.keras.quantize_model(origin_model) quant_aware_model.compile(optimizer=tf.keras.optimizers.RMSprop(lr=0.00001), # 0.045, momentum=0.9, decay=0.98), loss='sparse_categorical_crossentropy', metrics=['accuracy'])

Screenshots image

thanks in advance.

closed time in 2 days

qiyangzhang0329

issue commenttensorflow/model-optimization

TypeError: tf__call() got an unexpected keyword argument 'y'

Let discuss in https://github.com/tensorflow/tensorflow/issues/50079

qiyangzhang0329

comment created time in 2 days

pull request commenttensorflow/model-optimization

[Collaborative optimization] Pruning-Clustering-preserving Quantization Aware Training

Merging was blocked by build file strict dependencies -- resubmitting with it fixed myself now, should be merged today.

On Fri, Jun 11, 2021 at 5:39 AM Ruomei Yan ***@***.***> wrote:

Hi, @daverim https://github.com/daverim and @Xhark https://github.com/Xhark, could you please also let us know if there is anything we can do to help with the failed internal checks shown in this PR? @akarmi https://github.com/akarmi @wwwind https://github.com/wwwind for visibility Thanks all!

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/tensorflow/model-optimization/pull/699#issuecomment-859043957, or unsubscribe https://github.com/notifications/unsubscribe-auth/AASV4JJOCS5IS2OIGPIZV53TSEPIDANCNFSM433XELBA .

Ruomei

comment created time in 2 days

delete branch tensorflow/model-optimization

delete branch : test_378643716

delete time in 2 days

PR merged tensorflow/model-optimization

Fix typo. cla: yes technique:pruning

Fix typo.

+1 -1

0 comment

1 changed file

copybara-service[bot]

pr closed time in 2 days

push eventtensorflow/model-optimization

Frederic Rechtenstein

commit sha ec43970c86ed16c231931f089c3ba29390ec21b5

Fix typo. PiperOrigin-RevId: 378785927

view details

push time in 2 days

push eventtensorflow/model-optimization

Frederic Rechtenstein

commit sha ec43970c86ed16c231931f089c3ba29390ec21b5

Fix typo. PiperOrigin-RevId: 378785927

view details

push time in 2 days

push eventtensorflow/model-optimization

Frederic Rechtenstein

commit sha 3f1de651188d4ab7a7bcdac800e0c12b30bc0892

Fix typo. PiperOrigin-RevId: 378643716

view details

push time in 2 days

pull request commenttensorflow/model-optimization

[Collaborative optimization] Pruning-Clustering-preserving Quantization Aware Training

Hi, @daverim and @Xhark, could you please also let us know if there is anything we can do to help with the failed internal checks shown in this PR? @akarmi @wwwind for visibility Thanks all!

Ruomei

comment created time in 3 days

pull request commenttensorflow/model-optimization

[Collaborative optimization] Pruning-Clustering-preserving Quantization Aware Training

Thanks, @Xhark, the number is now updated.

Ruomei

comment created time in 3 days

pull request commenttensorflow/model-optimization

[Collaborative optimization] Pruning-Clustering-preserving Quantization Aware Training

Hi @Xhark Yes, this is the mistake, last digit is missing. The compression ratio was around 2.3 in our experiments. Thanks for noticing.

Ruomei

comment created time in 3 days

pull request commenttensorflow/model-optimization

[Collaborative optimization] Pruning-Clustering-preserving Quantization Aware Training

Hi, Just curious why Pruned_Clustered Model - Mobilenet_v1 (ImageNet) - INT8 .tflite gzip compression (bytes) is so small? PCQAT model is larger than PC model? or missed a digit for this case?

Ruomei

comment created time in 3 days

issue commenttensorflow/model-optimization

Quantization not supported for tensorflow.python.keras.layers.wrappers.Bidirectional

Why/how the initial bug would be solved?

ericqu

comment created time in 3 days

issue commenttensorflow/model-optimization

Quantization not supported for tensorflow.python.keras.layers.wrappers.Bidirectional

it's not solved, no. I am no longer working on this project though! Thanks.

ericqu

comment created time in 3 days

push eventtensorflow/model-optimization

Frederic Rechtenstein

commit sha 8e58a90b80381d68047381c5ecc0bcdf86f8158b

Fix typo. PiperOrigin-RevId: 378643716

view details

push time in 3 days

create barnchtensorflow/model-optimization

branch : test_378643716

created branch time in 3 days

PR opened tensorflow/model-optimization

Fix typo.

Fix typo.

+1 -1

0 comment

1 changed file

pr created time in 3 days

issue openedtensorflow/model-optimization

Support for Multiply layer

System information

  • TensorFlow version (you are using): 2.4.1
  • Are you willing to contribute it (Yes/No): no

Motivation

The implementation of some models (e.g. SENet) requires the use of the Multiply layer

Describe the feature

I tried to quantize a model including a Squeeze-Excitation block but I got an error: Layer multiply:<class 'tensorflow.python.keras.layers.merge.Multiply'> is not supported. You can quantize this layer by passing a tfmot.quantization.keras.QuantizeConfig instance to the quantize_annotate_layer API. It would be very useful to have this layer supported

Describe how the feature helps achieve the use case It would be possible to fully quantize models including this layer

created time in 3 days

issue commenttensorflow/model-optimization

After QAT and TFLite converter, the type of input and output of averagepooling2d node is not same

I also could not reproduce the different behaviour when converting directly vs converting after loading quantized weights. However there seem to be improperly quantized nodes in the converted TFLite model (regardless of loading quantized weights).

Conv2d_not_quantized

Codelab to reproduce.

guls999

comment created time in 3 days

delete branch tensorflow/model-optimization

delete branch : test_378100361

delete time in 4 days

PR merged tensorflow/model-optimization

Clarify that currently 8-bit quantization is the only supported deployment path. cla: yes technique:qat

Clarify that currently 8-bit quantization is the only supported deployment path.

+7 -4

1 comment

2 changed files

copybara-service[bot]

pr closed time in 4 days

push eventtensorflow/model-optimization

Frederic Rechtenstein

commit sha e9c1d33359b78840154e200fac23a14a74e9f98b

Clarify that currently 8-bit quantization is the only supported deployment path. PiperOrigin-RevId: 378332718

view details

push time in 4 days

push eventtensorflow/model-optimization

Frederic Rechtenstein

commit sha e9c1d33359b78840154e200fac23a14a74e9f98b

Clarify that currently 8-bit quantization is the only supported deployment path. PiperOrigin-RevId: 378332718

view details

push time in 4 days

push eventtensorflow/model-optimization

Frederic Rechtenstein

commit sha 038ed945a29c9c1ff0d37aba2593d963b4ec728a

Clarify that currently 8-bit quantization is the only supported deployment path. PiperOrigin-RevId: 378100361

view details

push time in 4 days

issue commenttensorflow/model-optimization

Puring cannot reduce tflite model size

When you refer the guide: https://www.tensorflow.org/model_optimization/guide/pruning/pruning_with_keras#create_a_10x_smaller_model_from_combining_pruning_and_quantization

The pruning reduces the model size after appropriate compression (e.g, gzip).

lxzheng

comment created time in 5 days

issue openedtensorflow/model-optimization

Structural (filter) pruning for convolutional layers

System information

  • TensorFlow version (you are using): 2.5.0
  • Are you willing to contribute it (Yes/No): Yes

Motivation

Deciding on where to have high filter/channel counts in convnets can be difficult, and smarter reductions in these numbers can lead to faster inference time across all devices.

Pruning is currently not very useful on GPU, since sparse operations are much slower than dense operations, so it would be useful to have a method of pruning that results in a reduced dense representation.

The current implementation I have (that isn't finished) doesn't require many additional components, since it works similarly to block sparsity and can reuse much of this code.

Describe the feature Add an option to prune_low_magnitude for "filter pruning" (alternatively "structural pruning") that restricts pruning of supported layers to blocks of the weights at a time. For convolutional layers these blocks represent the output channels of the layer.

In addition, an option is added to strip_pruning to restructure the layers that have been pruned in this manner, with fewer output channels than the original layers. The change in shape needs to be propagated forwards to future layers.

Describe how the feature helps achieve the use case With these two additions, models can be pruned in a way that is meaningful when running on GPU, saving memory and compute. It is also possible to find a reasonable layout for the number of output channels in each layer without hyperparameter tuning.

This feature makes pruning useful on GPU, where it currently is not so useful.

Describe how existing APIs don't satisfy your use case Using tfmot.python.core.sparsity.keras.prune.prune_low_magnitude on a convolutional layer will consider each element of the weights variable on its own, and very rarely leads to pruning that can be useful for reducing inference time on the GPU.

In addition, tfmot.python.core.sparsity.keras.prune.strip_pruning will always leave weights with zeros in them, even if a reduction in the size of the layer would be beneficial. If the outputs of a filter in the kernel of a convolutional layer are all zero, strip_pruning will leave restructuring as a step for the runtime.

created time in 5 days

push eventtensorflow/model-optimization

Frederic Rechtenstein

commit sha 5a24091a8f8a1a44c91137cdd9b673662fbb43d5

Clarify that currently 8-bit quantization is the only supported deployment path. PiperOrigin-RevId: 378100361

view details

push time in 5 days

pull request commenttensorflow/model-optimization

Clarify that currently 8-bit quantization is the only supported deployment path.

Check out this pull request on  <a href="https://app.reviewnb.com/tensorflow/model-optimization/pull/731"><img align="absmiddle" alt="ReviewNB" height="28" class="BotMessageButtonImage" src="https://raw.githubusercontent.com/ReviewNB/support/master/images/button_reviewnb.png"/></a>

See visual diffs & provide feedback on Jupyter Notebooks.


<i>Powered by <a href='https://www.reviewnb.com/?utm_source=gh'>ReviewNB</a></i>

copybara-service[bot]

comment created time in 5 days