Support tab-separated inputs #617

snukky · 2020-04-03T15:47:15Z

Description

This PR adds support for TSV inputs in the trainer, scorer and decoder.
Examples:

marian -c train.yml --tsv -t file.ende --valid-sets 
marian-scorer -c score.yml --tsv -t file.ende > scores.txt
marian-decoder -c ape.yml --tsv < file.srcmt > file.pe

Examples for stdin (--tsv is automatically added for -t stdin):

cat file.ende | marian -c train.yml --no-shuffle -t stdin
cat file.ende | marian-scorer -c score.yml -t stdin > scores.txt

List of changes:

Added the --tsv option for marian and marian-scorer (-t), and marian-decoder (-i)
Added the --tsv-fields for setting the number of fields in the TSV file
--tsv-fields is guessed based on the model type, so usually it does not need to be set
For training with TSV, if vocab(s) does not exist, it is created from appropriate fields in the TSV file (all cases described here:

marian-dev/src/data/corpus_base.cpp

Line 98 in 1d6da8b

// Possible cases:

)
For training with TSV, if no vocab paths are given, a common vocabulary is created (this is different than providing separate training files for source and target)
For training with TSV, validation sets also have to be in TSV format
Added support for -t stdin in marian and marian-scorer
For training from stdin, --no-shuffle is required, otherwise an appropriate error message is displayed
For training from stdin, creating vocabs automatically is disabled and an appropriate error message is displayed
If -t stdin is provided, --tsv is automatically added
If --tsv is provided and -t is empty, -t stdin is assumed

Added dependencies: none

How to test

Run example commands above.
Tested manually and added several regression tests: marian-nmt/marian-regression-tests#45

Tested on Ubuntu 16.04, compiling with:

cmake .. -DUSE_SENTENCEPIECE=on -DCOMPILE_EXAMPLES=on -DCOMPILE_TESTS=on -DCOMPILE_SERVER=on -DCMAKE_BUILD_TYPE=Release -DUSE_CCACHE=on

The new option potentially does not work with (1) BERT models, (2) factored vocabs, and (3) multi-node training. I don't know how to test these properly.

Checklist

I have tested the code manually
I have run regression tests
I have read and followed CONTRIBUTING.md
I have updated CHANGELOG.md

frankseide

Thanks for doing this. I am already using TSV files with Marian, through adventurous named-pipe constructions, and this is a welcome simplification. I have a Python-based command-line tool that reads the corpus from many little zipped TSV files and randomizes over them on-the-fly. This randomization obviously requires that src and tgt are together, hence TSV. This mode greatly reduces startup times, from hours to minutes.

One thing I noticed: The TSV format does not support alignments (and weights). For this to be maximally useful, alignments should also be part of the TSV format. Otherwise I would not be able to use my randomizer tool with alignments, or would have to resort to named-pipe tricks again.

I suggest to first merge this PR without alignment support (after addressing all other feedback), and then add alignment support as the next PR.

src/common/config.cpp

src/common/config_parser.cpp

src/common/config_validator.cpp

src/data/corpus_base.cpp

src/data/corpus_base.h

src/training/scheduler.h

snukky · 2020-04-06T19:51:32Z

WIP. I will finish and respond to the comments tomorrow.

snukky · 2020-04-07T19:08:55Z

Many thanks for the review, @frankseide! I have addressed all comments, and left a few of them unresolved, where I thought you may want to take a look again.

frankseide

Looks good. A few minor things, then this is good to go in my view.

frankseide

I approve. @marcin, since you did not chip in, I presume you are OK with it.

Roman, thanks again for doing this. It will make my life easier immediately, since I already run from a pipeline (using named pipes as a workaround to be able to pipe the different TSV columns into presently different streams). I am using an external data reader that we hope to be able to open-source soon. It will be a complement to this TSV feature. For our prod system, we'd also need a version that has the alignments in the TSV file. I hope that can be next.

Thanks!

frankseide · 2020-04-08T23:59:37Z

Roman, could you please confirm that the test suite passes, and only then Merge it?

snukky · 2020-04-09T16:28:48Z

Yes, I am running the tests again. Thanks for the review! I will continue with supporting alignment and weights in TSV inputs.

@emjotde Do you have any comments with the PR? (@marcin from Warsaw has to be getting a lot of unrelated mentions because of his username...)

emjotde · 2020-04-10T19:23:40Z

Looking now.

emjotde · 2020-04-10T20:01:31Z

@snukky Great work here. We will extensively test this by using it a lot now :)

* Add basic support for TSV inputs * Fix mini-batch-fit for TSV inputs * Abort if shuffling data from stdin * Fix terminating training with data from STDIN * Allow creating vocabs from TSV files * Add comments; clean creation of vocabs from TSV files * Guess --tsv-size based on the model type * Add shortcut for STDIN inputs * Rename --tsv-size to --tsv-fields * Allow only one 'stdin' in --train-sets * Properly create separate vocabularies from a TSV file * Clearer logging message * Add error message for wrong number of valid sets if --tsv is used * Use --no-shuffle instead of --shuffle in the error message * Fix continuing training from STDIN * Update CHANGELOG * Support both 'stdin' and '-' * Guess --tsv-fields from dim-vocabs if special:model.yml available * Update error messages * Move variable outside the loop * Refactorize utils::splitTsv; add unit tests * Support '-' as stdin; refactorize; add comments * Abort if excessive field(s) in the TSV input * Add a TODO on passing one vocab with fully-tied embeddings * Remove the unit test with excessive tab-separated fields

@emjotde

commit 3c1656d Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Sun Aug 23 23:41:07 2020 +0100 Fixed comment in scheduler.h. commit bdda7bd Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Sun Aug 23 23:33:50 2020 +0100 Updated CHANGELOG.md. commit 3bcd5f8 Merge: 5d80ab4 a21e48f Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Sun Aug 23 23:29:44 2020 +0100 Merge remote-tracking branch 'origin' into ug-graceful-shutdown commit 5d80ab4 Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Wed Aug 19 13:39:11 2020 +0100 Configurable signal handlers for SIGTERM - Adds custom signal handler option for SIGTERM (--sigterm) to marian training. - default behavior: 'graceful' => graceful exit (save model, then exit) - alternative: 'immediate' => exit immediately without saving model commit 9ab0be5 Merge: 0a0b83b d4102cb Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Tue Aug 18 19:39:51 2020 +0100 Merge branch 'ug-graceful-shutdown' of https://github.com/marian-nmt/marian-dev into ug-graceful-shutdown commit 0a0b83b Merge: 75459e3 3aed914 Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Tue Aug 18 18:12:17 2020 +0100 Merge branch 'master' into ug-graceful-shutdown commit d4102cb Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Wed Aug 5 16:33:30 2020 +0100 Update comments in signal_handling.h commit 521560b Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Tue Aug 4 01:58:19 2020 +0100 Update signal_handling.h Edited comments. commit 0a406d8 Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Tue Aug 4 00:41:30 2020 +0100 Update batch_generator.h Frank doesn't like curly braces. commit 286b980 Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Tue Aug 4 00:07:22 2020 +0100 Update batch_generator.h Return empty deque<BatchPtr> after SIGTERM. This makes no practical difference whatsoever but was requested by the code reviewers. commit 75459e3 Merge: 27cba8f 3dc0795 Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Sun Aug 2 21:41:52 2020 +0100 Merge branch 'ug-graceful-shutdown' of https://github.com/marian-nmt/marian-dev into ug-graceful-shutdown commit 27cba8f Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Sun Aug 2 21:38:36 2020 +0100 Fix space after comma. commit fdc519a Merge: f2d9f1e c944633 Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Sun Aug 2 21:29:09 2020 +0100 Merge branch 'master' into ug-graceful-shutdown commit 3dc0795 Merge: 9a7e3a0 4475787 Author: Roman Grundkiewicz <rgrundkiewicz@gmail.com> Date: Sun Jul 26 15:01:33 2020 +0100 Merge branch 'master' into ug-graceful-shutdown commit 9a7e3a0 Merge: 69f8192 b28905a Author: Ulrich Germann <ulrich.germann@gmail.com> Date: Sun Jul 26 09:59:18 2020 +0100 Merge branch 'master' into ug-graceful-shutdown commit 69f8192 Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Sun Jul 26 09:12:59 2020 +0100 Update training.h Insert missing space in line 84, responding to @emjotde's 'nit'. commit f2d9f1e Merge: d2d3563 ae1dd47 Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Wed Jun 17 16:02:08 2020 +0100 Merge branch 'ug-graceful-shutdown' of https://github.com/marian-nmt/marian-dev into ug-graceful-shutdown commit d2d3563 Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Fri May 22 15:46:13 2020 +0100 Post-rebase fixes. commit 2f2a00b Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Mon Apr 27 10:34:10 2020 +0100 Update Simple-WebSocket-Server and move it to submodules (#639) * Fix server build with current boost, move simple-websocket-server to submodule * Change submodule to marian-nmt/Simple-WebSocket-Server * Update submodule simple-websocket-server Co-authored-by: Gleb Tv <glebtv@gmail.com> commit 1312c18 Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Mon Apr 13 17:31:06 2020 -0700 update changelog and version commit 65c9c44 Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Sun Apr 12 18:56:11 2020 +0100 Support relative paths in shortlist and sqlite options (#612) * Refactorize processPaths * Fix relative paths for shortlist and sqlite options * Rename InterpolateEnvVars to interpolateEnvVars * Update CHANGELOG commit 17167dd Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Sat Apr 11 09:45:57 2020 -0700 Fix 0 * nan behavior due to using -O3 instead of -OFast (#630) * fix 0 * nan behavior in concatention * bump patch * change epsilon to margin commit 709522c Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Sat Apr 11 16:04:20 2020 +0100 Fix TSV training with mini-batch-fit after the last merge commit 5ce67c6 Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Fri Apr 10 15:27:34 2020 -0700 use float values for catch::Approx commit b06531d Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Fri Apr 10 13:53:21 2020 -0700 actually save the merge file commit 98dff9d Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Fri Apr 10 21:01:56 2020 +0100 Support tab-separated inputs (#617) * Add basic support for TSV inputs * Fix mini-batch-fit for TSV inputs * Abort if shuffling data from stdin * Fix terminating training with data from STDIN * Allow creating vocabs from TSV files * Add comments; clean creation of vocabs from TSV files * Guess --tsv-size based on the model type * Add shortcut for STDIN inputs * Rename --tsv-size to --tsv-fields * Allow only one 'stdin' in --train-sets * Properly create separate vocabularies from a TSV file * Clearer logging message * Add error message for wrong number of valid sets if --tsv is used * Use --no-shuffle instead of --shuffle in the error message * Fix continuing training from STDIN * Update CHANGELOG * Support both 'stdin' and '-' * Guess --tsv-fields from dim-vocabs if special:model.yml available * Update error messages * Move variable outside the loop * Refactorize utils::splitTsv; add unit tests * Support '-' as stdin; refactorize; add comments * Abort if excessive field(s) in the TSV input * Add a TODO on passing one vocab with fully-tied embeddings * Remove the unit test with excessive tab-separated fields commit 128e1fc Author: Young Jin Kim <youki@microsoft.com> Date: Fri Mar 27 21:44:31 2020 +0000 Merged PR 12243: For int8 quantized model, use int8 quantization for encoders as well For int8 quantized model, use int8 quantization for encoders as well. The quality difference between fp16 encoder and int8 encoder is small, but they have quite amount of speed difference. commit 63006db Author: Young Jin Kim <youki@microsoft.com> Date: Wed Mar 25 02:52:17 2020 +0000 Merged PR 11831: Change the weight matrix quantization to use 7-bit min/max quantization to avoid overflow 1. Change the weight matrix quantization to use 7-bit min/max quantization -> This resolves all the overflow issue, because weight and activations are quantized by min/max range. 2. Clip fp16 quantization to avoid overflow 3. Fix windows build errors (cmake options, vcproj file) 4. int8 pack model (encoder -> fp16) commit 9cd1623 Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Mon Apr 6 12:49:23 2020 +0100 Bug fix: better handling of SIGTERM for graceful shutdown during training. Prior to this bug fix, BatchGenerator::fetchBatches, which runs in a separate thread, would ignore SIGTERM during training (training uses a custom signal handler for SIGTERM, which simply sets a global flag, to enable graceful shutdown (i.e., save models and current state of training before shutting down). The changes in this commit also facilitate custom handling of other signals in the future by providing a general singal handler for all signals with a signal number below 32 (setSignalFlag) and a generic flag checking function (getSignalFlag(sig)) for checking such flags. commit ae1dd47 Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Sun May 17 11:34:18 2020 +0100 Update submodule regression-tests commit 1603d2f Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Thu May 14 08:00:41 2020 -0700 update version commit 9ae1951 Author: Nikolay Bogoychev <nheart@gmail.com> Date: Thu May 14 15:55:27 2020 +0100 Batched gemm (#633) * Use cblas_sgemm_batch when available * Merge with master, add comments and describe contribution commit 3f7b459 Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Mon Apr 27 10:34:10 2020 +0100 Update Simple-WebSocket-Server and move it to submodules (#639) * Fix server build with current boost, move simple-websocket-server to submodule * Change submodule to marian-nmt/Simple-WebSocket-Server * Update submodule simple-websocket-server Co-authored-by: Gleb Tv <glebtv@gmail.com> commit 342db58 Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Sun Apr 26 16:43:36 2020 +0100 Update submodule regression-tests commit 59dad14 Author: Kenneth Heafield <kpu@users.noreply.github.com> Date: Thu Apr 16 11:15:42 2020 +0100 python3 shebang from #620 (#621) * python3 shebang from #620 * Add changelog entry for python3 change commit ce94fe9 Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Mon Apr 13 17:31:06 2020 -0700 update changelog and version commit bc8b6fa Author: Martin Junczys-Dowmunt <Marcin.JunczysDowmunt@microsoft.com> Date: Tue Apr 14 00:28:44 2020 +0000 Merged PR 12442: cherry pick a few improvements/fixes from Frank's branch Cherry pick a few improvements/fixes from Frank's branch * Adds Frank's fix for label-based mini-batch sizing from Frank's current experimental branch. * Also copies minor improvements and a few comments. commit 34bc47c Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Sun Apr 12 19:14:03 2020 +0100 Dump version commit 7bf486a Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Sun Apr 12 18:58:33 2020 +0100 Fix Iris example on CPU (#623) commit 733cb50 Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Sun Apr 12 18:56:11 2020 +0100 Support relative paths in shortlist and sqlite options (#612) * Refactorize processPaths * Fix relative paths for shortlist and sqlite options * Rename InterpolateEnvVars to interpolateEnvVars * Update CHANGELOG commit 93a27dc Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Sat Apr 11 18:47:17 2020 +0100 Update submodule regression-tests commit 0ba438c Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Sat Apr 11 09:45:57 2020 -0700 Fix 0 * nan behavior due to using -O3 instead of -OFast (#630) * fix 0 * nan behavior in concatention * bump patch * change epsilon to margin commit c18fc71 Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Sat Apr 11 09:23:56 2020 -0700 fix 0 * nan behavior in concatention commit 855c94a Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Sat Apr 11 16:06:34 2020 +0100 Update submodule regression-tests commit 4d12ffa Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Sat Apr 11 16:04:20 2020 +0100 Fix TSV training with mini-batch-fit after the last merge commit 09904e0 Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Fri Apr 10 15:27:34 2020 -0700 use float values for catch::Approx commit 71cc43a Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Fri Apr 10 13:53:21 2020 -0700 actually save the merge file commit c95676e Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Fri Apr 10 13:50:22 2020 -0700 bump version commit 71e0f0b Author: Roman Grundkiewicz <rgrundki@exseed.ed.ac.uk> Date: Fri Apr 10 21:01:56 2020 +0100 Support tab-separated inputs (#617) * Add basic support for TSV inputs * Fix mini-batch-fit for TSV inputs * Abort if shuffling data from stdin * Fix terminating training with data from STDIN * Allow creating vocabs from TSV files * Add comments; clean creation of vocabs from TSV files * Guess --tsv-size based on the model type * Add shortcut for STDIN inputs * Rename --tsv-size to --tsv-fields * Allow only one 'stdin' in --train-sets * Properly create separate vocabularies from a TSV file * Clearer logging message * Add error message for wrong number of valid sets if --tsv is used * Use --no-shuffle instead of --shuffle in the error message * Fix continuing training from STDIN * Update CHANGELOG * Support both 'stdin' and '-' * Guess --tsv-fields from dim-vocabs if special:model.yml available * Update error messages * Move variable outside the loop * Refactorize utils::splitTsv; add unit tests * Support '-' as stdin; refactorize; add comments * Abort if excessive field(s) in the TSV input * Add a TODO on passing one vocab with fully-tied embeddings * Remove the unit test with excessive tab-separated fields commit d0fa14e Author: Young Jin Kim <youki@microsoft.com> Date: Fri Mar 27 21:44:31 2020 +0000 Merged PR 12243: For int8 quantized model, use int8 quantization for encoders as well For int8 quantized model, use int8 quantization for encoders as well. The quality difference between fp16 encoder and int8 encoder is small, but they have quite amount of speed difference. commit 68581a6 Author: Young Jin Kim <youki@microsoft.com> Date: Wed Mar 25 02:52:17 2020 +0000 Merged PR 11831: Change the weight matrix quantization to use 7-bit min/max quantization to avoid overflow 1. Change the weight matrix quantization to use 7-bit min/max quantization -> This resolves all the overflow issue, because weight and activations are quantized by min/max range. 2. Clip fp16 quantization to avoid overflow 3. Fix windows build errors (cmake options, vcproj file) 4. int8 pack model (encoder -> fp16) commit dd06542 Author: Marcin Junczys-Dowmunt <marcinjd@microsoft.com> Date: Sat Mar 14 09:53:54 2020 -0700 bump version commit 66711b5 Author: Martin Junczys-Dowmunt <Marcin.JunczysDowmunt@microsoft.com> Date: Sat Mar 14 00:07:37 2020 +0000 Merged PR 11929: Move around code to make later comparison with FP16 code easier This does not introduce any new functionality, just moves code around, so that future PRs are easier to compare. Moving old GraphGroup code to training/deprecated. Once it is clear there is nothing in there that's worth saving, this will be deleted. Replace -Ofast with -O3 and make sure ffinite-math is turned off. commit 2586af7 Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Mon Apr 6 12:49:23 2020 +0100 Bug fix: better handling of SIGTERM for graceful shutdown during training. Prior to this bug fix, BatchGenerator::fetchBatches, which runs in a separate thread, would ignore SIGTERM during training (training uses a custom signal handler for SIGTERM, which simply sets a global flag, to enable graceful shutdown (i.e., save models and current state of training before shutting down). The changes in this commit also facilitate custom handling of other signals in the future by providing a general singal handler for all signals with a signal number below 32 (setSignalFlag) and a generic flag checking function (getSignalFlag(sig)) for checking such flags. commit 8a44759 Merge: 95c65bb 653b13d Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Fri Apr 3 12:37:13 2020 +0100 Merge branch 'ug-graceful-shutdown' of https://github.com/marian-nmt/marian-dev into ug-graceful-shutdown commit 653b13d Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Fri Nov 22 22:59:38 2019 +0000 Added explanatory comment about exiting marian_train with non-zero status after SIGTERM. commit 73bdb1f Author: Ulrich Germann <ugermann@inf.ed.ac.uk> Date: Tue Nov 19 22:49:25 2019 +0000 Return exit code 15 (SIGTERM) after SIGTERM. When marian receives signal SIGTERM and exits gracefully (save model & exit), it should then exit with a non-zero exit code, to signal to any parent process that it did not exit "naturally".

snukky added 16 commits March 26, 2020 13:59

Add basic support for TSV inputs

33b8a26

Fix mini-batch-fit for TSV inputs

4abea30

Abort if shuffling data from stdin

12f427b

Fix terminating training with data from STDIN

1e1fbf3

Allow creating vocabs from TSV files

2cdee18

Add comments; clean creation of vocabs from TSV files

3d3953c

Guess --tsv-size based on the model type

da7676a

Add shortcut for STDIN inputs

ba6f50b

Rename --tsv-size to --tsv-fields

33f67a1

Allow only one 'stdin' in --train-sets

81558ef

Properly create separate vocabularies from a TSV file

d2e8b09

Clearer logging message

3c449cb

Add error message for wrong number of valid sets if --tsv is used

cc1ca18

Use --no-shuffle instead of --shuffle in the error message

be0e431

Fix continuing training from STDIN

1d6da8b

Update CHANGELOG

08a2900

frankseide requested changes Apr 3, 2020

View reviewed changes

snukky added 6 commits April 6, 2020 11:14

Support both 'stdin' and '-'

baebd8f

Guess --tsv-fields from dim-vocabs if special:model.yml available

b90b140

Update error messages

8352c94

Move variable outside the loop

74d7b74

Refactorize utils::splitTsv; add unit tests

6da35a2

Support '-' as stdin; refactorize; add comments

2fccd8f

snukky requested a review from frankseide April 7, 2020 19:09

frankseide requested changes Apr 7, 2020

View reviewed changes

snukky added 2 commits April 8, 2020 17:49

Abort if excessive field(s) in the TSV input

92cc1f8

Add a TODO on passing one vocab with fully-tied embeddings

614a9fe

frankseide approved these changes Apr 8, 2020

View reviewed changes

Remove the unit test with excessive tab-separated fields

6b4281a

emjotde self-requested a review April 10, 2020 20:00

emjotde merged commit 696bb44 into master Apr 10, 2020

snukky mentioned this pull request Apr 18, 2020

Enable multi-source input in marian-server #505

Closed

marianminion mentioned this pull request Apr 26, 2020

Jenkins marian-dev-cpu-avx2 #22 failed #642

Closed

snukky deleted the input-tsv branch April 12, 2021 11:25

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Support tab-separated inputs #617

Support tab-separated inputs #617

snukky commented Apr 3, 2020

frankseide left a comment •

edited

Loading

snukky commented Apr 6, 2020

snukky commented Apr 7, 2020

frankseide left a comment

frankseide left a comment

frankseide commented Apr 8, 2020

snukky commented Apr 9, 2020

emjotde commented Apr 10, 2020

emjotde commented Apr 10, 2020

Support tab-separated inputs #617

Support tab-separated inputs #617

Conversation

snukky commented Apr 3, 2020

Description

How to test

Checklist

frankseide left a comment • edited Loading

Choose a reason for hiding this comment

snukky commented Apr 6, 2020

snukky commented Apr 7, 2020

frankseide left a comment

Choose a reason for hiding this comment

frankseide left a comment

Choose a reason for hiding this comment

frankseide commented Apr 8, 2020

snukky commented Apr 9, 2020

emjotde commented Apr 10, 2020

emjotde commented Apr 10, 2020

frankseide left a comment •

edited

Loading