Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
330 commits
Select commit Hold shift + click to select a range
b3b7e6a
Merge branch 'deepmipt-dev' into clean_multitask_ready
Jul 11, 2022
7fc98e5
As crf.py fails unless this torchcrf version is used. It is default v…
dimakarp1996 Jul 11, 2022
819ddcc
Update multitask_iterator.py
dimakarp1996 Jul 11, 2022
889ded6
Update torch_multitask_bert.py
dimakarp1996 Jul 11, 2022
e5f0d01
Update multitask_iterator.py
dimakarp1996 Jul 11, 2022
046a39c
Update multitask_iterator.py
dimakarp1996 Jul 11, 2022
c3416ec
Update torch_multitask_bert.py
dimakarp1996 Jul 11, 2022
84a4231
Update torch_multitask_bert.py
dimakarp1996 Jul 11, 2022
0aea26c
Update torch_multitask_bert.py
dimakarp1996 Jul 11, 2022
f20c168
Update torch_multitask_bert.py
dimakarp1996 Jul 11, 2022
9a54e59
Update accuracy.py
dimakarp1996 Jul 11, 2022
c09f0e5
Update multitask_preprocessor.py
dimakarp1996 Jul 11, 2022
00744a5
Update requirements.txt
dimakarp1996 Jul 12, 2022
2fccd7c
Update crf.py
dimakarp1996 Jul 12, 2022
113e62b
Update torch_multitask_bert.py
dimakarp1996 Jul 12, 2022
ed3036c
Update simple_vocab.py
dimakarp1996 Jul 12, 2022
baadbf9
Update torch_transformers_preprocessor.py
dimakarp1996 Jul 12, 2022
434a1b1
Update requirements.txt
dimakarp1996 Jul 15, 2022
cc52126
Update torch_multitask_bert.py
dimakarp1996 Jul 15, 2022
d81fc20
Script is deleted untill newer version is added. It can be added in n…
dimakarp1996 Jul 15, 2022
337cfb6
Update config_glue.json
dimakarp1996 Jul 15, 2022
3e2cf40
Update config_glue.json
dimakarp1996 Jul 15, 2022
a878fdf
Paths were changed as in deeppavlov/configs/ner/ner_conll2003_bert.js…
dimakarp1996 Jul 15, 2022
4cc337d
Fixed codestyle + added configs to debug
dimakarp1996 Jul 20, 2022
26b504e
Fixed last comment, instead of debugging configs
dimakarp1996 Jul 20, 2022
428c61d
Added GLUE config
Jul 20, 2022
78bf4c1
Added SUPERGLUE config and comments also
Jul 20, 2022
d82950e
Added SUPERGLUE config and comments also
Jul 20, 2022
2489ff8
Added SUPERGLUE config and comments also
Jul 20, 2022
58e479d
Added SUPERGLUE config and comments also
Jul 20, 2022
31cf8cd
Distilbert works. SUPERGLUE in training
Jul 20, 2022
5b26230
Now json for superglue is training
Jul 21, 2022
ac1a474
minor fixes
Jul 21, 2022
185f8a3
minor fixes
Jul 21, 2022
c7bffca
huggingface reader as in dex
Jul 21, 2022
b53b3fb
huggingface reader as in dex
Jul 21, 2022
04f103a
Fixed RECORD bug
Jul 26, 2022
a0922ad
minor fix
Jul 27, 2022
3e29fa0
streamline reader code
Aug 29, 2022
73e04a5
Update basic_classification_reader.py
dimakarp1996 Sep 4, 2022
db1d20d
Update proba2labels.py
dimakarp1996 Sep 5, 2022
7c407dd
Update multitask_preprocessor.py
dimakarp1996 Sep 5, 2022
6cfee80
Update multitask_preprocessor.py
dimakarp1996 Sep 5, 2022
8dada78
Update torch_multitask_bert.py
dimakarp1996 Sep 5, 2022
e748d76
Create multitask_bert.rst
dimakarp1996 Sep 5, 2022
a54b4e4
Update multitask_bert.rst
dimakarp1996 Sep 5, 2022
92841b9
Update multitask_distilbert_conll.json
dimakarp1996 Sep 5, 2022
6677dc2
Update multitask_bert.rst
dimakarp1996 Sep 5, 2022
b7f6f5a
Update multitask_bert.rst
dimakarp1996 Sep 6, 2022
0ef57bb
pending changes
Sep 6, 2022
ace63c1
Merge branch 'clean_multitask_ready' of http://github.com/dimakarp199…
Sep 6, 2022
4c1a212
Clean multitask ready (#2)
dimakarp1996 Sep 6, 2022
ff97fb3
rst is compiled
Sep 6, 2022
e4a1e2d
rst is compiled
Sep 6, 2022
162d6c9
bug fix
Sep 7, 2022
fbe0524
bug fix
Sep 7, 2022
cb05b22
Fixed ner preprocessor bug
Sep 7, 2022
9ec12b0
config fix
Sep 8, 2022
ddce64b
Update basic_classification_reader.py
dimakarp1996 Sep 12, 2022
1df1eac
Update record_metrics.py
dimakarp1996 Sep 12, 2022
dde0c15
current
Sep 13, 2022
9f53254
minor simplification
Sep 14, 2022
310df71
Merge branch 'clean_multitask_ready' of https://github.com/dimakarp19…
Sep 14, 2022
fedeeaa
Current version
Sep 15, 2022
5b885cd
Clean multitask ready - PR (#18)
dimakarp1996 Sep 15, 2022
af0ce65
Dev (#19)
dimakarp1996 Sep 15, 2022
3b583aa
Delete russian_superglue_danetqa_rubert.json
dimakarp1996 Sep 15, 2022
509059a
Delete config_cola.json
dimakarp1996 Sep 15, 2022
fd2c27a
Delete deeppavlov/configs/russian_super_glue directory
dimakarp1996 Sep 15, 2022
bfac5d3
Update ner_case_agnostic_mdistilbert.json
dimakarp1996 Sep 15, 2022
4aa9dc9
Delete config_mnli.json
dimakarp1996 Sep 15, 2022
75e7a73
Delete config_mrpc.json
dimakarp1996 Sep 15, 2022
612ebcf
Delete config_rte.json
dimakarp1996 Sep 15, 2022
e11544b
Delete config_qnli.json
dimakarp1996 Sep 15, 2022
59b1d3f
Delete config_stsb.json
dimakarp1996 Sep 15, 2022
17bcbf8
Delete config_sst2.json
dimakarp1996 Sep 15, 2022
7df3aa0
Delete config_qqp.json
dimakarp1996 Sep 15, 2022
e311fee
Delete config_superglue_record.json
dimakarp1996 Sep 15, 2022
ca105b5
Delete multitask_distilbert_cola.json
dimakarp1996 Sep 15, 2022
3bb2a19
Delete multitask_distilbert_conll.json
dimakarp1996 Sep 15, 2022
da753df
Delete multitask_distilbert_copa.json
dimakarp1996 Sep 15, 2022
11cd148
Delete multitask_distilbert_rte.json
dimakarp1996 Sep 15, 2022
5bcb850
Delete multitask_distilbert_stsb.json
dimakarp1996 Sep 15, 2022
31dc9dc
Update ner_case_agnostic_mdistilbert.json
dimakarp1996 Sep 15, 2022
af2c01a
long to gpu
Sep 16, 2022
92b5085
gh
Sep 16, 2022
01ce359
Update torch_multitask_bert.py
dimakarp1996 Sep 19, 2022
1346a84
Update torch_multitask_bert.py
dimakarp1996 Sep 19, 2022
d21825e
new changes
Sep 25, 2022
884832a
Handle empty input in accuracy
Sep 28, 2022
d79d548
Critical fix
Sep 29, 2022
7b92107
fmeasure refactor
Sep 29, 2022
24f5c46
Added small-scale caching to speed up the similar requests on DREAM s…
dimakarp1996 Oct 11, 2022
bec2124
Fixed bug in caching.
dimakarp1996 Oct 11, 2022
8010d25
Further speeded up the multitask BERT model. Decreased cache, store i…
dimakarp1996 Oct 11, 2022
930cb02
Quicker hashing
dimakarp1996 Oct 11, 2022
82d1010
Quicker hashing, but store hash in CPU not in CUDA
dimakarp1996 Oct 11, 2022
dcdbab3
Revert commits where less cache were stored as it led to a bug. Only …
Oct 11, 2022
e36e73f
Merge branch 'clean_multitask_ready' of https://github.com/dimakarp19…
Oct 11, 2022
b225390
Revert commits where less cache were stored as it led to a bug. Only …
Oct 11, 2022
25b41cc
incremental debug
Oct 11, 2022
8f45048
Update torch_multitask_bert.py
dimakarp1996 Oct 11, 2022
97e2904
incremental debug
Oct 11, 2022
f2abdce
Update torch_multitask_bert.py
dimakarp1996 Oct 11, 2022
e3c2b7a
Update torch_multitask_bert.py
dimakarp1996 Oct 11, 2022
d3232ff
Update torch_multitask_bert.py
dimakarp1996 Oct 11, 2022
c281caf
Update torch_multitask_bert.py
dimakarp1996 Oct 11, 2022
952fd6a
g
Oct 11, 2022
09187d5
minor fix
Oct 11, 2022
9c4a44c
Version on 27 Oct
Oct 27, 2022
72271d5
Speeds up the inference for multilabel tasks
Oct 28, 2022
af04827
Bug fix
Oct 28, 2022
68390a8
Bug fix
Oct 28, 2022
3f2b1f4
Bug fix
Oct 28, 2022
7b8e92f
Removed hashing as it is unnesessary now
Oct 28, 2022
71a26b3
Removed cache as it is unnesessary now
dimakarp1996 Oct 28, 2022
b569141
Merge branch 'clean_multitask_ready' of https://github.com/dimakarp19…
Oct 28, 2022
acd5b6a
Fix for memory efficiency
Oct 28, 2022
3b2b18a
Fixed bug with CPU memory use
dimakarp1996 Oct 28, 2022
2d7a7c0
Returned CUDA cache
Oct 31, 2022
1efb042
Returned CUDA cache
Oct 31, 2022
d52f49e
bug fix
dimakarp1996 Oct 31, 2022
c7e4d0c
Update torch_transformers_preprocessor.py
dimakarp1996 Oct 31, 2022
89e2b9e
Version on 16 November
Nov 16, 2022
c6a9ecf
Add files via upload
dimakarp1996 Nov 16, 2022
a6a77bd
Update torch_multitask_bert.py
dimakarp1996 Nov 16, 2022
b1a295f
Update pal_modeling.py
dimakarp1996 Nov 16, 2022
bb12c98
Update torch_multitask_bert.py
dimakarp1996 Nov 16, 2022
3797377
Update pal_modeling.py
dimakarp1996 Nov 16, 2022
50f2829
Update torch_multitask_bert.py
dimakarp1996 Nov 16, 2022
d9bb5a8
Update torch_multitask_bert.py
dimakarp1996 Nov 16, 2022
8d51d06
Version as on 5 DEC
Dec 5, 2022
b230913
Version 07-12-2022
Dec 7, 2022
fc16fc6
Merge branch 'clean_multitask_ready' of https://github.com/dimakarp19…
Dec 7, 2022
0dea618
Merge branch 'dev' into clean_multitask_ready
dimakarp1996 Dec 7, 2022
3ddb232
Update test_quick_start.py
dimakarp1996 Dec 7, 2022
895825f
Update index.rst
dimakarp1996 Dec 7, 2022
6bc6518
Update torch_transformers_multiplechoice.py
dimakarp1996 Dec 7, 2022
2d8a354
Update metrics_registry.json
dimakarp1996 Dec 7, 2022
ff4bfd5
Update registry.json
dimakarp1996 Dec 7, 2022
3db6b1b
Update simple_vocab.py
dimakarp1996 Dec 7, 2022
682f3ee
Update fit_trainer.py
dimakarp1996 Dec 7, 2022
54988f6
Update nn_trainer.py
dimakarp1996 Dec 7, 2022
0bfd247
Update multitask_preprocessor.py
dimakarp1996 Dec 7, 2022
5c2146e
Update torch_multitask_bert.py
dimakarp1996 Dec 7, 2022
1511ac3
Fixes to pass tests (#20)
dimakarp1996 Dec 17, 2022
7c07735
Fixes to pass tests (#20)
dimakarp1996 Dec 17, 2022
5f2a278
Update fmeasure.py
dimakarp1996 Dec 26, 2022
a10ada3
Update torch_multitask_bert.py
dimakarp1996 Dec 26, 2022
04344d6
Update torch_multitask_bert.py
dimakarp1996 Dec 26, 2022
ffc42e8
Update torch_multitask_bert.py
dimakarp1996 Dec 28, 2022
a98cad5
Update torch_multitask_bert.py
dimakarp1996 Dec 28, 2022
6ce43e7
Update multitask_distilbert_example.json
dimakarp1996 Jan 31, 2023
c3711bc
Update multitask_distilbert_example.json
dimakarp1996 Jan 31, 2023
0301b8b
Delete multitask_example.json
dimakarp1996 Jan 31, 2023
7d06411
Rename multitask_distilbert_example.json to multitask_example.json
dimakarp1996 Jan 31, 2023
da785c4
Dev (#21)
dimakarp1996 Jan 31, 2023
e2e7196
metadata to top
dimakarp1996 Jan 31, 2023
cf881f7
Update config_superglue.json
dimakarp1996 Jan 31, 2023
33bde96
Update multitask_example.json
dimakarp1996 Jan 31, 2023
fa4cfd6
Update proba2labels.py
dimakarp1996 Jan 31, 2023
23e8314
Update torch_multitask_bert.py
dimakarp1996 Jan 31, 2023
a5834ef
Update config_superglue.json
dimakarp1996 Jan 31, 2023
de7b8d1
Update config_glue.json
dimakarp1996 Jan 31, 2023
36d00dd
Update multitask_example.json
dimakarp1996 Jan 31, 2023
e90a90a
h
Jan 31, 2023
4308af7
Update multitask_preprocessor.py
dimakarp1996 Jan 31, 2023
9149135
Update multitask_preprocessor.py
dimakarp1996 Feb 1, 2023
912e4d6
Update multitask_preprocessor.py
dimakarp1996 Feb 1, 2023
ba5b3aa
Streamlined GLUE/SUPERGLUE configs, added evaluation script
Feb 2, 2023
550ac07
Merge branch 'clean_multitask_ready' of https://github.com/dimakarp19…
Feb 2, 2023
590c43f
Update multitask_bert.rst
dimakarp1996 Feb 2, 2023
02a386f
Update evaluate_glue_superglue.py
dimakarp1996 Feb 2, 2023
3c975b8
Eval script bug fix
Feb 2, 2023
8fb16cf
Merge branch 'clean_multitask_ready' of https://github.com/dimakarp19…
Feb 2, 2023
4154a20
Current version of the inference setting. ReCoRD still has troubles
Feb 2, 2023
9f1c87d
Update multitask_bert.rst
dimakarp1996 Feb 2, 2023
0d26e7f
Update multitask_bert.rst
dimakarp1996 Feb 2, 2023
9a95f3a
superglue larger batch
Feb 2, 2023
2a13ee9
Merge branch 'clean_multitask_ready' of https://github.com/dimakarp19…
Feb 2, 2023
7c73bcf
Update config_superglue.json
dimakarp1996 Feb 3, 2023
840902b
fix
Feb 3, 2023
aecf231
s
Feb 3, 2023
0be2dcd
j
Feb 3, 2023
d738945
Older version - until newer proved to be better
dimakarp1996 Feb 3, 2023
ee6d382
Update multitask_iterator.py
dimakarp1996 Feb 13, 2023
4ba5154
Merge branch 'dev' into clean_multitask_ready1
IgnatovFedor Feb 15, 2023
ade00e9
Update evaluate_glue_superglue.py
dimakarp1996 Feb 16, 2023
1e4bf41
Update evaluate_glue_superglue.py
dimakarp1996 Feb 16, 2023
fd9b959
Update multitask_bert.rst
dimakarp1996 Feb 16, 2023
3af3a1e
Create multitask_bert.rst
dimakarp1996 Feb 16, 2023
976757f
fsd
dimakarp1996 Feb 16, 2023
18ac533
RST != markdown
dimakarp1996 Feb 16, 2023
092298d
h
dimakarp1996 Feb 16, 2023
eec1f5d
refactor: small style changes and removal of unused functions
IgnatovFedor Feb 17, 2023
29e134a
Update multitask_iterator.py
dimakarp1996 Feb 17, 2023
238bfe9
Update evaluate_glue_superglue.py
dimakarp1996 Feb 17, 2023
c3fc4e3
Update metrics_registry.json
dimakarp1996 Feb 17, 2023
89d0b96
Update simple_vocab.py
dimakarp1996 Feb 17, 2023
e285b46
Update huggingface_dataset_iterator.py
dimakarp1996 Feb 17, 2023
2dd4c5a
Update registry.json
dimakarp1996 Feb 17, 2023
8b07d75
Update registry.json
dimakarp1996 Feb 17, 2023
c044a27
Update huggingface_dataset_iterator.py
dimakarp1996 Feb 17, 2023
5ce10ca
Update requirements_registry.json
dimakarp1996 Feb 17, 2023
4bbc5cc
Update evaluate_glue_superglue.py
dimakarp1996 Feb 17, 2023
91139a2
Update evaluate_glue_superglue.py
dimakarp1996 Feb 17, 2023
bb4134b
Update evaluate_glue_superglue.py
dimakarp1996 Feb 17, 2023
401860a
Update torch_multitask_bert.py
dimakarp1996 Feb 17, 2023
5f948bd
Update basic_classification_reader.py
dimakarp1996 Feb 17, 2023
90d5d81
Update multitask_preprocessor.py
dimakarp1996 Feb 17, 2023
3836c56
Update multitask_reader.py
dimakarp1996 Feb 17, 2023
32391bb
Update multitask_iterator.py
dimakarp1996 Feb 17, 2023
71a7b59
Update evaluate_glue_superglue.py
dimakarp1996 Feb 17, 2023
bfd5214
For CONLL not to break down while getting other parameters as common …
dimakarp1996 Feb 17, 2023
0cc5a83
Update multitask_iterator.py
dimakarp1996 Feb 17, 2023
74aeae4
Update multitask_reader.py
dimakarp1996 Feb 17, 2023
8e21a29
Update evaluate_glue_superglue.py
dimakarp1996 Feb 17, 2023
cdd16ab
Making the code more robust to the errors when input somehow is given…
dimakarp1996 Feb 17, 2023
f7db669
Update multitask_reader.py
dimakarp1996 Feb 17, 2023
8550bf9
Update train.py
dimakarp1996 Feb 17, 2023
1e2c8f6
Update evaluate_glue_superglue.py
dimakarp1996 Feb 17, 2023
05869e8
Update evaluate_glue_superglue.py
dimakarp1996 Feb 17, 2023
81810ea
Update evaluate_glue_superglue.py
dimakarp1996 Feb 17, 2023
0999592
Update evaluate_glue_superglue.py
dimakarp1996 Feb 17, 2023
6552ac6
Update evaluate_glue_superglue.py
dimakarp1996 Feb 17, 2023
00b2a32
Update evaluate_glue_superglue.py
dimakarp1996 Feb 18, 2023
dd0d84c
Update evaluate_glue_superglue.py
dimakarp1996 Feb 18, 2023
2862b12
remove: f1_weighted_with_confusion_matrix function
IgnatovFedor Feb 18, 2023
3ebccfd
Update multitask_iterator.py
dimakarp1996 Feb 19, 2023
b95a480
Update config_glue.json
dimakarp1996 Feb 20, 2023
1f7b3c7
Update config_glue.json
dimakarp1996 Feb 20, 2023
457bf87
Update multitask_example.json
dimakarp1996 Feb 20, 2023
88f6448
Update multitask_example.json
dimakarp1996 Feb 20, 2023
db4f347
Update multitask_example.json
dimakarp1996 Feb 20, 2023
54f1492
Update config_glue.json
dimakarp1996 Feb 20, 2023
73deb9e
remove: config_superglue, multitask_input_splitter
IgnatovFedor Mar 1, 2023
5530383
Bug fix
dimakarp1996 Mar 1, 2023
ff7ec6d
refactor: multitask reader
IgnatovFedor Mar 2, 2023
7e5edf5
refactor: multitask iterator
IgnatovFedor Mar 2, 2023
6531541
refactor: multitask_bert to transformer
IgnatovFedor Mar 3, 2023
281b953
refactor: renamed dir_path to data_path in squad iterator
IgnatovFedor Mar 9, 2023
1561862
fix: crossval new iterator creatin error
IgnatovFedor Mar 9, 2023
94731fb
refactor: NumpyArrayEncoder dataclass handling
IgnatovFedor Mar 9, 2023
b7c49f2
refactor: multitask configs prettified
IgnatovFedor Mar 9, 2023
c6d2815
docs: update
IgnatovFedor Mar 12, 2023
aeacbb5
remove: evaluate_glue_superblue
IgnatovFedor Mar 12, 2023
fdf5a55
revert: changes in basic_classification_reader and fmeasure
IgnatovFedor Mar 13, 2023
1731894
refactor: small code changes
IgnatovFedor Mar 13, 2023
03f1f19
update: release version
IgnatovFedor Mar 14, 2023
8531003
Merge branch 'dev' into clean_multitask_ready1
IgnatovFedor Mar 14, 2023
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion deeppavlov/_meta.py
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
__version__ = '1.1.0'
__version__ = '1.1.1'
__author__ = 'Neural Networks and Deep Learning lab, MIPT'
__description__ = 'An open source library for building end-to-end dialog systems and training chatbots.'
__keywords__ = ['NLP', 'NER', 'SQUAD', 'Intents', 'Chatbot']
Expand Down
287 changes: 287 additions & 0 deletions deeppavlov/configs/multitask/mt_glue.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,287 @@
{
"dataset_reader": {
"class_name": "multitask_reader",
"task_defaults": {
"class_name": "huggingface_dataset_reader",
"path": "glue",
"train": "train",
"valid": "validation"
},
"tasks": {
"cola": {"name": "cola"},
"sst2": {"name": "sst2"},
"qqp": {"name": "qqp"},
"mrpc": {"name": "mrpc"},
"rte": {"name": "rte"},
"mnli": {
"name": "mnli",
"valid": "validation_matched"
},
"qnli": {"name": "qnli"},
"stsb": {"name": "stsb"}
}
},
"dataset_iterator": {
"class_name": "multitask_iterator",
"num_train_epochs": "{NUM_TRAIN_EPOCHS}",
"gradient_accumulation_steps": "{GRADIENT_ACC_STEPS}",
"seed": 42,
"task_defaults": {
"class_name": "huggingface_dataset_iterator",
"label": "label",
"use_label_name": false,
"seed": 42
},
"tasks": {
"cola": {
"features": ["sentence"]
},
"sst2": {
"features": ["sentence"]
},
"qqp": {
"features": ["question1", "question2"]
},
"mrpc": {
"features": ["sentence1", "sentence2"]
},
"rte": {
"features": ["sentence1", "sentence2"]
},
"mnli": {
"features": ["premise", "hypothesis"]
},
"qnli": {
"features": ["question", "sentence"]
},
"stsb": {
"features": ["sentence1", "sentence2"]
}
}
},
"chainer": {
"in": ["x_cola", "x_sst2", "x_qqp", "x_mrpc", "x_rte", "x_mnli", "x_qnli", "x_stsb"],
"in_y": ["y_cola", "y_sst2", "y_qqp", "y_mrpc", "y_rte", "y_mnli", "y_qnli", "y_stsb"
],
"pipe": [
{
"class_name": "multitask_pipeline_preprocessor",
"possible_keys_to_extract": [0, 1],
"preprocessor": "TorchTransformersPreprocessor",
"vocab_file": "{BACKBONE}",
"max_seq_length": 128,
"do_lower_case": true,
"n_task": 8,
"in": ["x_cola", "x_sst2", "x_qqp", "x_mrpc", "x_rte", "x_mnli", "x_qnli", "x_stsb"],
"out": [
"bert_features_cola",
"bert_features_sst2",
"bert_features_qqp",
"bert_features_mrpc",
"bert_features_rte",
"bert_features_mnli",
"bert_features_qnli",
"bert_features_stsb"
]
},
{
"id": "multitask_transformer",
"class_name": "multitask_transformer",
"optimizer_parameters": {"lr": 2e-5},
"gradient_accumulation_steps": "{GRADIENT_ACC_STEPS}",
"learning_rate_drop_patience": 2,
"learning_rate_drop_div": 2.0,
"return_probas": true,
"backbone_model": "{BACKBONE}",
"save_path": "{MODEL_PATH}",
"load_path": "{MODEL_PATH}",
"tasks": {
"cola": {
"type": "classification",
"options": 2
},
"sst2": {
"type": "classification",
"options": 2
},
"qqp": {
"type": "classification",
"options": 2
},
"mrpc": {
"type": "classification",
"options": 2
},
"rte": {
"type": "classification",
"options": 2
},
"mnli": {
"type": "classification",
"options": 3
},
"qnli": {
"type": "classification",
"options": 2
},
"stsb": {
"type": "regression",
"options": 1
}
},
"in": [
"bert_features_cola",
"bert_features_sst2",
"bert_features_qqp",
"bert_features_mrpc",
"bert_features_rte",
"bert_features_mnli",
"bert_features_qnli",
"bert_features_stsb"
],
"in_y": ["y_cola", "y_sst2", "y_qqp", "y_mrpc", "y_rte", "y_mnli", "y_qnli", "y_stsb"],
"out": [
"y_cola_pred_probas",
"y_sst2_pred_probas",
"y_qqp_pred_probas",
"y_mrpc_pred_probas",
"y_rte_pred_probas",
"y_mnli_pred_probas",
"y_qnli_pred_probas",
"y_stsb_pred"
]
},
{
"in": [
"y_cola_pred_probas",
"y_sst2_pred_probas",
"y_qqp_pred_probas",
"y_mrpc_pred_probas",
"y_rte_pred_probas",
"y_mnli_pred_probas",
"y_qnli_pred_probas"
],
"out": [
"y_cola_pred_ids",
"y_sst2_pred_ids",
"y_qqp_pred_ids",
"y_mrpc_pred_ids",
"y_rte_pred_ids",
"y_mnli_pred_ids",
"y_qnli_pred_ids"
],
"class_name": "proba2labels",
"max_proba": true
}
],
"out": [
"y_cola_pred_probas",
"y_sst2_pred_probas",
"y_qqp_pred_probas",
"y_mrpc_pred_probas",
"y_rte_pred_probas",
"y_mnli_pred_probas",
"y_qnli_pred_probas",
"y_stsb_pred",
"y_cola_pred_ids",
"y_sst2_pred_ids",
"y_qqp_pred_ids",
"y_mrpc_pred_ids",
"y_rte_pred_ids",
"y_mnli_pred_ids",
"y_qnli_pred_ids",
"y_stsb_pred"
]
},
"train": {
"epochs": "{NUM_TRAIN_EPOCHS}",
"batch_size": 32,
"metrics": [
{
"name": "multitask_accuracy",
"inputs": [
"y_rte",
"y_mnli",
"y_qnli",
"y_mrpc",
"y_cola",
"y_sst2",
"y_qqp",
"y_rte_pred_ids",
"y_mnli_pred_ids",
"y_qnli_pred_ids",
"y_mrpc_pred_ids",
"y_cola_pred_ids",
"y_sst2_pred_ids",
"y_qqp_pred_ids"
]
},
{
"name": "accuracy",
"alias": "accuracy_mrpc",
"inputs": ["y_mrpc", "y_mrpc_pred_ids"]
},
{
"name": "accuracy",
"alias": "accuracy_rte",
"inputs": ["y_rte", "y_rte_pred_ids"]
},
{
"name": "accuracy",
"alias": "accuracy_mnli",
"inputs": ["y_mnli", "y_mnli_pred_ids"]
},
{
"name": "accuracy",
"alias": "accuracy_qnli",
"inputs": ["y_qnli", "y_qnli_pred_ids"]
},
{
"name": "accuracy",
"alias": "accuracy_sst",
"inputs": ["y_sst2", "y_sst2_pred_ids"]
},
{
"name": "accuracy",
"alias": "accuracy_cola",
"inputs": ["y_cola", "y_cola_pred_ids"]
},
{
"name": "accuracy",
"alias": "accuracy_qqp",
"inputs": ["y_qqp", "y_qqp_pred_ids"]
},
{
"name": "pearson_correlation",
"alias": "pearson_correlation_stsb",
"inputs": ["y_stsb", "y_stsb_pred"]
},
{
"name": "spearman_correlation",
"alias": "spearman_correlation_stsb",
"inputs": ["y_stsb", "y_stsb_pred"]
}
],
"validation_patience": 3,
"val_every_n_epochs": 1,
"log_every_n_epochs": 1,
"show_examples": false,
"evaluation_targets": ["valid"],
"class_name": "torch_trainer"
},
"metadata": {
"variables": {
"BACKBONE": "bert-base-uncased",
"MODELS_PATH": "~/.deeppavlov/models/glue",
"MODEL_PATH": "{MODELS_PATH}/8task",
"NUM_TRAIN_EPOCHS": 5,
"GRADIENT_ACC_STEPS": 1
},
"download": [
{
"url": "http://files.deeppavlov.ai/deeppavlov_data/multitask/glue.tar.gz",
"subdir": "{MODELS_PATH}"
}
]
}
}
Loading