microsoft
/
NeuronBlocks
зеркало из
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность
118 коммитов 24 Ветки 4 Теги 19 MiB
Граф коммитов

21 Коммитов

Автор SHA1 Сообщение Дата
yaze@microsoft.com e30ed2384d add model usage, citation, reference 2020-03-27 15:37:14 +08:00
Ze Yang 7d855626aa add pre-trained model for WSDM 2020 paper TMKD 2019-11-14 19:10:17 +08:00
Ze Yang ef84b5b02c initialize 2019-11-14 13:48:51 +08:00
Ze Yang ee31cc8b88 release TMKD 2019-11-14 13:38:52 +08:00
Flyer Cheng 58ad563a23 Add encoding cache and lazy-train mechanism (#50) 
* Add new config about knowledge distillation for query binary classifier

* remove inferenced result in knowledge distillation for query binary classifier

* Add AUC.py in tools folder

* Add test_data_path into conf_kdqbc_bilstmattn_cnn.json

* Modify AUC.py

* Rename AUC.py into calculate_AUC.py

* Modify test&calculate AUC commands for Knowledge Distillation for Query Binary Classifier

* Add cpu_thread_num parameter in conf.training_params

* Rename cpu_thread_num into cpu_num_workers

* update comments in ModelConf.py

* Add cup_num_workers in model_zoo/advanced/conf.json

* Add the description of cpu_num_workers in Tutorial.md

* Update inference speed of compressed model

* Add ProcessorsScheduler Class

* Add license in ProcessorScheduler.py

* use lazy loading instead of one-off loading

* Remove Debug Info in problem.py

* use open instead of codecs.open

* update the inference of build dictionary for classification

* add md5 function in common_utils.py

* add merge_encode_* function

* update typo

* update typo

* reorg the logical flow in train.py

* remove dummy comments in problem.py

* add encoding cache mechanism

* add lazy-load mechanism for training phase

* enumerate problem types in problem.py

* remove data_encoding.py

* add lazy load train logic

* Modify comment and remove debug code

* Judge if test_path exists

* fix parameter missing when use char embedding

* merge master

* add file_column_num in problem.py

* merge add_encoding_cache branch

* add SST-2 in .gitignore

* merge master

* use steps_per_validation instead of valid_times_per_epoch

* Fix Learning Rate decay logic bug

* add log of calculating md5 of training data

* fix multi-gpu char_emb OOM problem & add char leval fix_lengths

* Modify batch_num_to_show_results in multi-gpu

* Modify batch_num_to_show_results

* delete deepcopy in get_batches

* add new parameters chunk_size and max_building_lines in conf and update tutorials
 2019-08-02 20:59:00 +08:00
boshining 1b805656ef Dev/shining (#75) 
* add text matching compression & delete regression

* modify the Tutorial.md

* modify the Tutorial.md

* add teacher model name

* modify tutorial.md

* fix transform params2tensors problem

* add softmax output layer for slot tagging

* add slot_tagging metrics

* modify make word emb matrix

* Delete dev.tsv

* Delete test.tsv

* Delete train.tsv

* delate conll data

* Update Contributing.md

* Update tools

* Update README.md

* Update Contributing.md

* Update README.md

* Update autotest.sh

* update get_results.py

* fix sequence tagging workflow

* add model type judgement for optimizer

* delete full atis dataset and unuseful config filee

* add slot_tagging sample data

* fix load embedding slow problem

* fix Char embedding CNN problem

* add lower token when load embedding matrix

* add word level length for char emb

* Update Conv

* Add ARCI & ARCII module and Modify Conv block

* Update to the same as master

* update Linear layer

* Add block - Calculate Distance of Two Vectors

* update tutorial_zh_CN
 2019-07-19 06:26:29 +08:00
adolphk-yk 676c8283bd add interactive prediction and register block (#70) 2019-06-28 16:58:33 +08:00
adolphk-yk 3f69dac1ff add CRF related block and logic (#67) 2019-06-19 14:44:07 +08:00
ssdf93 9b2adb9a2f Fix concat in interaction layer. (#57) 
* Add Support for MatchPyramid model

* Modified interaction layer

* change 2d to 2D

* Flatten conflicts solved.

* fix conv2d and pooling2d

* add MatchPyramid score report

* add MatchPyramid score record

* Fix concat in interaction layer.

* merge

* Delete Conv2d.py

* Delete Pooling2d.py

* Delete test.json

* Update .gitignore
 2019-06-14 14:15:11 +08:00
adolphk-yk a1041f463d update sequence tagging task workflow (#62) 2019-06-05 19:55:43 +08:00
ssdf93 17d99a6a7c Add Support for MatchPyramid model (#41) 2019-05-23 13:32:05 +08:00
fareise e4fbee2483 Add Chinese Text Mathcing example. 2019-05-14 23:01:11 +08:00
ericwtlin 1d3f31e5ea Add new model for question answer matching (#35) 2019-05-10 15:53:06 +08:00
woailaosang ef366cd8bf add Chinese support (#18) 2019-05-08 19:07:52 +08:00
woailaosang c9da56a5f0 fix "softmax" problem of model's output layer. (#22) 
* distinguish logits, logits_softmax

* modity last layer softmax config for model_zoo

* fix 'prediction_scores' problem for regression task.

* fix wrongly written words

* fix logits_softmax problem for regression.

* fix logits_softmax problem for slot tagging task.
 2019-05-07 23:23:20 +08:00
adolphk-yk 87b2c253b4 add text matching compression 
* add text matching compression & delete regression

* modify the Tutorial.md

* modify the Tutorial.md

* add teacher model name

* modify tutorial.md
 2019-04-30 17:15:03 +08:00
woailaosang bc305b6dcd Dev/zj (#10) 
* Update Tutorial.md

* fix CPU train --> GPU test, GPU train --> CPU test. But CPU train --> multi-GPU test will report error.

* remove ujson

* remove glove for model_zoo/demo/conf.json

* modify glove for Tutorial.md

* add nltk.download('punkt') in problem.py
 2019-04-30 14:06:17 +08:00
Flyer Cheng 9e01e6e16f Dev/fecheng (#9) 
Add the thread count support
 2019-04-29 21:18:06 +08:00
L.J. SHOU eaef9265b7
add code for query binary classifier knowledge distillation (#3) 
* Add new config about knowledge distillation for query binary classifier
 2019-04-28 18:12:20 +08:00
Zhijie Sang 9ff1846825 add test_data_path for some configs 2019-04-26 08:58:38 -07:00
L.J. SHOU 122417816e Code commit 
Co-Authored-By: yiming1013 <yiming1013@users.noreply.github.com>
 2019-04-20 19:17:30 +08:00