Commit Graph

459 Commits

Author SHA1 Message Date
allegroai
19a24c4b20 Fix default argument behavior to match argparse behavior (if value is the same as default, use default value and type) 2020-05-13 20:29:35 +03:00
allegroai
e1f6e5ec6a Fix typo in warning 2020-05-13 20:26:43 +03:00
allegroai
cb6ecb458a Version bump to v0.14.3 2020-05-10 09:49:14 +03:00
allegroai
7962b041c0 Fix code examples in docstrings 2020-05-08 22:16:20 +03:00
allegroai
5d04b21a79 Fix upload_object_via_stream() in Azure storage driver 2020-05-08 22:13:39 +03:00
allegroai
0298b84030 Refactor histogram code for PyTorch Ignite integration 2020-05-08 22:12:49 +03:00
allegroai
966cd6118a Add Model and Task type-annotations 2020-05-08 22:08:48 +03:00
allegroai
f90f8f06e2 Fix Task.get_tasks() return type to Task 2020-05-08 22:07:43 +03:00
allegroai
a5ff2ba9c8 Add Logger histogram mode (stack/group/relative) 2020-05-08 22:05:33 +03:00
allegroai
5a85d40fc7 Add Logger type-annotations 2020-05-08 21:34:47 +03:00
allegroai
3de95063ee Remove auto_connect_arg_parser constraint when ArgumentParser.parse_arg() is used before initializing task 2020-05-08 21:33:54 +03:00
allegroai
8b5377ee4f Add PEP610 support for new pip>=20.1 git reference feature 2020-05-08 21:31:34 +03:00
allegroai
6991e06e14 Fix RGBA upload with default JPEG files 2020-05-08 21:30:02 +03:00
allegroai
031d26a6cf Support '#' and '?' in image names 2020-05-08 21:28:48 +03:00
allegroai
3f52a6088a Version bump to v0.14.2 2020-04-28 15:32:33 +03:00
allegroai
3d9683f290 Fix logger creation guard 2020-04-28 10:44:40 +03:00
allegroai
7dc27b9a5e Rename utilities dir 2020-04-28 10:43:47 +03:00
allegroai
6746144803 Use UID in message if user name cannot be retrieved 2020-04-26 23:22:46 +03:00
allegroai
7085a2bb74 Report GPU memory usage per process (and sub-process) 2020-04-26 23:21:27 +03:00
allegroai
0a7a32f2eb Support caching of extracted zip artifacts 2020-04-26 23:19:42 +03:00
allegroai
6ff3cc0ee4 Fix duplicate suffix for artifacts 2020-04-26 23:17:51 +03:00
allegroai
302a8cbf75 Report memory usage only for the experiment process (and sub-processes) 2020-04-26 23:16:13 +03:00
allegroai
7e839204d1 Raise error in case of missing artifact directory 2020-04-26 23:14:33 +03:00
allegroai
617e5f7c1d Improve Jupyter support by ignoring IPython directives in converted python script (like % and ! lines) 2020-04-26 23:13:00 +03:00
allegroai
d9f2f01bbd Make sure trains is included in Jupyter requirements 2020-04-26 23:11:51 +03:00
allegroai
dbb3346332 Improve module requirements detection 2020-04-26 23:10:45 +03:00
allegroai
9726f782f2 Add requirement parsing optimization using entry point by default 2020-04-26 22:58:13 +03:00
allegroai
7e4fba9d10 Fix scikit-learn module detection 2020-04-26 22:54:49 +03:00
allegroai
13ce783fa3 Check for updates based on session version 2020-04-26 22:53:18 +03:00
allegroai
466b2859d9 Upload debug assets with series / title including dot or slash did not add iteration index to the filename 2020-04-16 16:49:21 +03:00
allegroai
f7b80a0da2 matplotlib with no logger should not use the last iteration reported by the resource monitor 2020-04-16 16:48:19 +03:00
allegroai
cb139f2d17 Fix Tensorflow V1/V2 audio support 2020-04-16 16:46:02 +03:00
allegroai
215aa32239 Detect exception in debugger session, set Task to Failed 2020-04-16 16:43:10 +03:00
allegroai
4eaa77dbdb Fix single log request exceeds max packet size, set limit to 1MB per request 2020-04-16 16:41:18 +03:00
allegroai
7f4b100042 Fix text encoding utf-8 and pr_curve broken in Tensorboard support 2020-04-16 16:40:14 +03:00
allegroai
4bb17ca420 Fix renaming/deleting model file right after saving will break async upload (trains/issues#123) 2020-04-13 19:03:15 +03:00
allegroai
12659307a8 Fix update_weights() to use model upload target file when passed 2020-04-13 19:00:35 +03:00
allegroai
4b9c5c235c Update docstrings 2020-04-13 18:58:39 +03:00
allegroai
648779380c Add media (audio) support for both Logger and Tensorboard bind 2020-04-09 13:14:14 +03:00
allegroai
7ac7e088a1 Add trace feature 2020-04-09 13:12:50 +03:00
allegroai
0df3d38862 Fix self references in configuration when environment variables exist 2020-04-09 13:11:21 +03:00
allegroai
3ac7dbdb49 Refactor shutdown sequence 2020-04-09 13:10:29 +03:00
allegroai
7dae058359 Optimize locking for TaskHandler, avoid lock when shutting down 2020-04-09 13:08:46 +03:00
allegroai
d9aa83380f Stop resource monitoring before signaling task stop 2020-04-09 13:07:26 +03:00
allegroai
ab263bb59f Raise ValueError if Task.get_logger() is called after task was closed 2020-04-09 13:06:06 +03:00
allegroai
3c4925d605 Fix resource monitor and check if task is valid 2020-04-09 13:00:24 +03:00
allegroai
7f00e45d6c Do not recreate logger after Task was closed/exited 2020-04-09 12:59:00 +03:00
allegroai
3f6fb5379a Revert fork patching as signal is not enough and is not called from forked processes 2020-04-09 12:57:50 +03:00
allegroai
5eb4ae6600 Use a daemon thread for the log 2020-04-09 12:56:55 +03:00
allegroai
1b901b7d13 Fix logger in case a packet was dropped before it was overwritten 2020-04-09 12:56:02 +03:00
allegroai
aa737e6b5b Make sure task is marked as started in remote execution (just in case) 2020-04-09 12:53:43 +03:00
allegroai
f61cbdeb39 Check if join was successful when waiting for std flush pool 2020-04-09 12:51:34 +03:00
allegroai
2f395cc76b Use sub-process fork signal hooks instead of os._exit patch 2020-04-09 12:50:04 +03:00
allegroai
004f925454 ThreadPool should be terminated, not closed, otherwise it might hang 2020-04-09 12:47:38 +03:00
allegroai
9916c93ce0 Add 10sec timeout for stdout/stderr flush at end of process 2020-04-09 12:46:30 +03:00
allegroai
1718aa20d4 Add thread_waited_join waited join for Thread/Process Pools 2020-04-09 12:45:06 +03:00
allegroai
23bd6097a8 Add nicer stdout log flush 2020-04-09 12:42:45 +03:00
allegroai
9a0a84a83e Do not wait for logs if we are aborting the task manually (i.e. ctrl-C) 2020-04-09 12:41:10 +03:00
allegroai
98ce0bbe43 Change TaskHandler.close() wait default to False as it should not wait for logs to flush 2020-04-09 12:39:09 +03:00
allegroai
b3c9872a3f Intercept SystemExit and do nothing so we could kill the thread 2020-04-09 12:33:16 +03:00
allegroai
5ec4d80493 Disconnect stdout/stderr logger on exit 2020-04-09 12:31:43 +03:00
allegroai
de9c88bc2d Do not try to wait for Lock 2020-04-09 12:30:42 +03:00
allegroai
337e60a376 Kill repo/package detection thread on exit 2020-04-09 12:28:57 +03:00
allegroai
b2c2002c40 Create dev task manually when constructing the Task 2020-04-09 12:27:13 +03:00
allegroai
11420adce7 Log reports at the end of the task 2020-04-09 12:24:37 +03:00
allegroai
ffedb219d5 Local modules (except trains) imported from a folder inside the git project should not be logged as "local packages", they should be ignored 2020-04-09 12:21:37 +03:00
allegroai
07daf8f5e6 Fix logger sometimes getting stuck at end of experiment 2020-04-09 12:05:56 +03:00
allegroai
e6f29428eb Add StorageManager 2020-04-09 12:03:41 +03:00
allegroai
e1fc9b3dc8 ThreadPool should be terminated, not closed, otherwise it might hang 2020-04-09 11:39:03 +03:00
allegroai
070fd8149a Store the version that matching the Session API so we do not reload every time 2020-04-09 11:35:51 +03:00
allegroai
a425a70fc6 Add api.ssl_error_count_verbosity and make sure SSL retries are taken care by the session 2020-04-09 11:33:55 +03:00
allegroai
101e5393d1 Fix TRAINS_VCS_ROOT path conversion 2020-04-01 19:06:30 +03:00
allegroai
41ca1a2e49 Fix requirements detection to make sure trains is detected even if we execute without actually being installed 2020-04-01 19:04:57 +03:00
allegroai
01772430d6 Ignore virtual-environment folder that might be inside the project's directory 2020-04-01 19:02:54 +03:00
allegroai
6de3d4b6fd Ignore local modules imported from a folder inside the git project 2020-04-01 19:01:21 +03:00
allegroai
172ed62d41 Add Task.get_tasks() filtering support 2020-04-01 18:54:16 +03:00
allegroai
581edf1098 Version bump to v0.14.1 2020-03-24 20:36:57 +02:00
allegroai
c4719f2e2f Add type annotations and fix docstrings 2020-03-23 23:26:46 +02:00
allegroai
766c8ab24f Add Task.models property 2020-03-23 23:25:55 +02:00
allegroai
0211d233d4 Deprecate Task.set_model_config(), Task.get_model_config_text() and Task.get_model_config_dict() 2020-03-23 23:25:16 +02:00
allegroai
023f1721c1 Add Task.get_models() retrieving stored models on previously executed tasks 2020-03-22 18:19:07 +02:00
allegroai
332e9e2f63 Fix Tensorflow direct V2.1 multiple FileWriters 2020-03-22 18:17:16 +02:00
allegroai
493cce443a Reuse Model objects if we are storing local files (reduce clutter) 2020-03-22 18:15:32 +02:00
allegroai
4e2564cd3a Support reusing Models. Use trains.Model as general purpose registered Model. 2020-03-22 18:13:56 +02:00
allegroai
63507c82f7 Fix Model.download_model_weights() to reuse previously downloaded file 2020-03-22 18:11:30 +02:00
allegroai
477665ee33 Fix storage_uri handling in Model.update() 2020-03-22 18:05:05 +02:00
allegroai
abc9b512f7 Fix logging typos 2020-03-22 18:03:25 +02:00
allegroai
7817ef5cda Fix joblib binding 2020-03-20 10:30:13 +02:00
allegroai
5db53ba643 Support multiple EventWriter in TensorFlow eager mode (TF 2.0+) 2020-03-20 10:29:18 +02:00
allegroai
b4050ecf25 Fix TensorFlow NaN/Inf values support 2020-03-20 10:27:52 +02:00
allegroai
babaf9f1ce Add OpenMPI/Slurm support 2020-03-20 10:23:00 +02:00
allegroai
0adbd79975 Fix StorageHelper upload on shutdown 2020-03-20 10:20:44 +02:00
allegroai
dc915d0241 Fix support for Task init/close multiple times 2020-03-20 10:20:06 +02:00
allegroai
667ddcab88 Fix import for services that do not exist in old versions 2020-03-20 10:16:48 +02:00
allegroai
3b1d2d3258 Version bump to v0.14.0 2020-03-12 19:42:48 +02:00
allegroai
afad6a42ea Add initial slurm support (multiple nodes sharing the same task id) 2020-03-12 18:12:16 +02:00
allegroai
5b29aa194c Make sure artifact temporary files names are valid file names 2020-03-12 18:10:03 +02:00
allegroai
84a34428b6 Add trains-init support for config file env override (as well as argument) 2020-03-12 18:09:03 +02:00
allegroai
b3dff9a4eb Support setting task initial iteration for continuing previous runs 2020-03-12 17:40:29 +02:00
allegroai
f3531c1af2 Allow Task.set_credentials() to override configuration file in dev mode 2020-03-12 17:22:09 +02:00