Commit Graph

133 Commits

Author SHA1 Message Date
allegroai
bc85ddf78d Fix pytorch direct resolve replacing wheel link with directly installed version 2023-05-21 22:53:51 +03:00
allegroai
5b5fb0b8a6 Add agent.package_manager.pytorch_resolve configuration setting with pip or direct values. pip sets extra index based on cuda and lets pip resolve, direct is the previous parsing algorithm that does the matching and downloading (default pip) 2023-05-21 22:53:11 +03:00
allegroai
b6ca0fa6a5 Print error on resource monitor failure 2023-05-11 16:18:11 +03:00
allegroai
307ec9213e Fix git+ssh:// links inside installed packages not being converted properly to HTTPS authenticated and vice versa 2023-05-11 16:16:51 +03:00
allegroai
154db59ce6 Add agent.package_manager.poetry_install_extra_args configuration option 2023-03-28 14:37:48 +03:00
allegroai
afffa83063 Fix git+ssh:// links inside installed packages not being converted properly to https authenticated links 2023-03-28 14:35:51 +03:00
allegroai
667c2ced3d Fix very old pip version support (<20) 2023-03-28 14:34:19 +03:00
allegroai
8d51aed679 Protect against cache folders without permission 2023-03-05 18:05:00 +02:00
allegroai
bfc4ba38cd Fix torch inside nvidia containers to use preinstalled version (i.e. ==x.y.z.* matching) 2023-03-05 18:05:00 +02:00
allegroai
f69e16ea9d Fix clearml-agent build --docker stuck on certain containers 2022-12-13 15:24:32 +02:00
allegroai
ebdc215632 Remove " from pip commands in venv 2022-12-10 20:58:30 +02:00
allegroai
ca2791c65e Fix pip support allowing multiple pip version constraints (by default, one for <PY3.10 and one for >=PY3.10) 2022-12-07 22:09:25 +02:00
allegroai
dc5e0033c8 Remove support for kubectl run
Allow customizing pod name prefix and limit pod label
Return deleted pods from cleanup
Some refactoring
2022-12-05 11:40:19 +02:00
allegroai
53d379205f Support raise_error in get_bash_output() 2022-12-05 11:26:40 +02:00
allegroai
d2f3614ab0 Add support for CLEARML_AGENT_DOCKER_ARGS_HIDE_ENV environment variable (see agent.hide_docker_command_env_vars config option) 2022-10-23 13:04:00 +03:00
allegroai
efb06891a8 Add support for PyTorch new extra_index_url repo support. We will find the correct index url based on the cuda version, and let pip do the rest. 2022-10-23 13:04:00 +03:00
allegroai
3f7a4840cc Add support for operator != in package version (mostly for pytorch resolving) 2022-10-23 13:04:00 +03:00
allegroai
2ef5d38b32 Remove future (Python 2 is not supported for clearml-agent) 2022-10-23 13:03:59 +03:00
allegroai
0de10345f7 Moved pyhocon to internal packages 2022-10-23 13:03:59 +03:00
allegroai
f0fd62a28f Fix docker extra args showing up in configuration printout 2022-10-23 13:03:59 +03:00
allegroai
7ff6aee20c Add warning if venv cache is disabled 2022-10-07 15:23:10 +03:00
allegroai
67fc884895 Fix --gpus all not reporting GPU stats on worker machine 2022-10-07 15:22:13 +03:00
allegroai
594ee5842e Allow to pverride pytorch lookup page: "agent.package_manager.torch_page / torch_nightly_page / torch_url_template_prefix" 2022-09-15 20:16:41 +03:00
allegroai
ec216198a0 Add agent.enable_git_ask_pass to improve passing user/pass to git commands 2022-08-29 18:06:26 +03:00
allegroai
fe6adbf110 Fix package @ file:// with quoted (url style) links should not be ignored 2022-08-29 18:06:09 +03:00
allegroai
ec9d027678 Add support for MIG devices, use 0:1 for GPU 0 slice 1 (or use 0.1) 2022-08-01 18:58:42 +03:00
allegroai
9acbad28f7 Fix repository URL contains credentials even when agent.force_git_ssh_protocolagent.force_git_ssh_protocol is true 2022-06-16 23:20:53 +03:00
allegroai
f66e42ddb1 Fix optional priority packaged always compare lower case package name 2022-06-16 23:18:31 +03:00
allegroai
249aa006cb Make sure that if we have "setuptools" in the original required packages, we preserve the line in the pip freeze list 2022-05-12 23:31:32 +03:00
allegroai
335ef91d8e Fix git unsafe directory issue (disable check on cached vcs folder) 2022-05-05 12:22:40 +03:00
allegroai
6c7a639673 Fix broken pytorch setuptools incompatibility (force setuptools < 59 if torch is below 1.11) 2022-05-05 12:22:13 +03:00
allegroai
6572023173 Fix avoid reinstall pytorch package if the same version is already installed 2022-04-09 14:18:38 +03:00
allegroai
c901bd331c Fix git packages are installed even if commit is given and is preinstalled when using cached virtual environment 2022-03-28 18:11:46 +03:00
allegroai
341086d86a Fix vcs packages are reinstalled when same commit version is already installed 2022-03-24 22:03:25 +02:00
allegroai
8712c5e636 Fix PyTorch aarch64 and windows support 2022-03-16 17:40:21 +02:00
allegroai
531e514003 Add custom build script support
Add extra configurations when starting daemon
Propagate token to docker in case credentials are not available
2022-03-15 10:04:25 +02:00
Idan Tene
e3e6a1dda8
Fix virtualenv python interpreter used (#98)
* Add virtualenv version logging
* Force using requested python interpreter
2022-02-27 11:25:25 +02:00
allegroai
1f53c4fd1b Fix agent fails to check out code from main branch when branch/commit is not explicitly specified 2022-02-07 20:04:08 +02:00
allegroai
bfed3ccf4d Fix agent attempts to check out code when in standalone mode 2022-02-07 20:03:08 +02:00
allegroai
53eba5658f Fix conda package manager listed packages with local links (@ file://) should ignore the local package if it does not exist
Fix cuda patch version support in conda
2022-02-02 16:33:07 +02:00
allegroai
363aaeaba8 Fix symbolic links not copied from cached VCS into working copy. Windows platform will result with default copy content instead of original symbolic link (issue #89) 2022-01-23 10:42:11 +02:00
allegroai
fa1307e62c Add agent.poetry_version to specify poetry version (and force installation of poetry if missing) 2022-01-23 10:40:05 +02:00
allegroai
e7c9e9695b Fix using deprecated abc support 2022-01-23 10:39:13 +02:00
allegroai
b3e8be6296 Add agent.force_git_root_python_path configuration setting to force adding the git repository root folder to the PYTHONPATH (if set working directory is not added to the PYHTONPATH) 2022-01-07 15:11:59 +02:00
allegroai
f41ed09dc1 Add support for custom docker image resolving 2021-12-21 14:29:43 +02:00
allegroai
5a6caf6399 Fix "git+git://" requirements 2021-10-29 22:58:28 +03:00
allegroai
a890e36a36 Fix PY2.7 support for pytorch 2021-10-19 10:47:09 +03:00
allegroai
26d76f52ac Fix venv cache cannot reinstall package from git with http credentials 2021-08-13 11:00:54 +03:00
allegroai
5c5802c089 Fix python package with git+git:// links or git+ssh:// conversion 2021-08-12 13:37:10 +03:00
allegroai
0fbbe774fa Fix support for "-r requirements.txt" in installed packages 2021-08-05 19:19:54 +03:00
allegroai
5ed47d2d2c Add support for CLEARML_NO_DEFAULT_SERVER env var to prevent agent from using the demo server
Add support for FORCE_CLEARML_AGENT_REPO env var to allow installing agent from a repo url when executing a task
Implement skip venv installation on execute and allow custom binary
Fix services mode limit implementation in docker mode
2021-08-02 22:51:26 +03:00
Simon Gasse
9456e493ac Enable rewriting SSH URLs
ClearML Agent allows to force git cloning via SSH and also has a
setting to force a username. The relevant settings are:
agent.force_git_ssh_protocol: true
agent.force_git_ssh_user: "git"

However, forcing a specific username or port only worked so far if the
agent translated either from https->ssh or from ssh->https. A given
ssh URL was not rewritten.

This commit adds a helper function and includes it in `_set_ssh_url`
to allow rewriting ssh URLs with the username and/or port given in the
config `agent.force_git_ssh_user`.
If neither username nor port are forced in the config, the URL is not
touched.

This is somewhat related to issue #42.
Note that rewriting https->https is not covered in this commit.
2021-07-31 23:34:27 +03:00
allegroai
aaa6b32f9f Fix support for "-r requirements.txt" inside "installed packages" 2021-06-24 19:26:35 +03:00
allegroai
742cbf5767 Add docker environment arguments log masking support (issue #67) 2021-05-25 19:31:45 +03:00
allegroai
e93384b99b Fix --stop with dynamic gpus 2021-05-20 10:58:46 +03:00
allegroai
ae3d034531 Protect against None in execution.repository 2021-05-12 15:45:31 +03:00
allegroai
e60a6f9d14 Fix --stop support for dynamic gpus 2021-04-25 10:46:43 +03:00
allegroai
82c5e55fe4 Fix usage of not_set in k8s template merge 2021-04-07 21:30:13 +03:00
allegroai
c59d268995 Fix venv cache crash on bad symbolic links 2021-04-07 18:44:11 +03:00
allegroai
9e9fcb0ba9 Add dynamic mode terminate dockers on sig_term 2021-04-07 18:43:44 +03:00
allegroai
14ac584577 Support k8s glue container env vars merging 2021-03-31 23:53:58 +03:00
allegroai
92a1e07b33 Fix local path replace back when using cache 2021-03-26 12:16:05 +03:00
allegroai
cb6bdece39 Fix cuda version from driver does not return minor version 2021-03-18 10:07:59 +02:00
allegroai
91eec99563 Add conda debug prints (--debug) 2021-03-18 10:07:58 +02:00
allegroai
f019905720 Fix venv cache support for local folders 2021-02-28 19:47:09 +02:00
allegroai
2006ab20dd Fix conda support for git+http links 2021-02-23 12:46:06 +02:00
allegroai
0caf31719c Fix venv caching always reinstall git repositories and local repositories 2021-02-23 12:45:34 +02:00
allegroai
58cb344ee6 Upgrade pynvml add detect CUDA version from driver level 2021-02-17 00:03:16 +02:00
allegroai
22d5892b12 Use shared git cache between multiple agents on the same machine 2021-02-14 13:49:29 +02:00
allegroai
ca242424ab Fix service-mode support for venvs
Fix --services-mode with venvs
2021-02-14 13:45:17 +02:00
allegroai
407deb84e9 Fix multi instances on Windows 2021-02-14 13:44:39 +02:00
allegroai
1260e3d942 Update cache entries on conda package manager 2021-02-11 14:47:26 +02:00
allegroai
b22d926d94 Fix cache to take cuda version into account 2021-02-11 14:47:05 +02:00
allegroai
410cc8c7be Add --dynamic-gpus and limit in --services-mode 2021-02-11 14:46:37 +02:00
allegroai
784c676f5b Fix "from clearml" runtime diff patching (make sure we move it to after all the __future__ imports) include handling triple quotes in comments 2021-02-11 14:46:06 +02:00
allegroai
296f7970df Fix file not found error (no 2) interpreted as aborted (i.e. ctrl-c) 2021-02-11 14:44:54 +02:00
allegroai
cd59933c9c Remove unused packages 2021-02-11 14:44:35 +02:00
allegroai
b95d3f5300 Add venv caching with docker mode support 2021-02-11 14:44:19 +02:00
allegroai
5d8bbde434 Fix applying git diff on new added file 2021-01-24 08:46:42 +02:00
allegroai
161993f66f Add agent.force_git_ssh_user configuration value (issue #42)
Change default docker to nvidia/cuda:10.1-cudnn7-runtime-ubuntu18.04
2021-01-10 12:38:45 +02:00
allegroai
428781af86 Fix support for Windows pip and Conda requirements.txt 2020-12-25 02:06:40 +02:00
allegroai
b4d143812e initial clearml-agent v0.17.0 2020-12-22 23:00:57 +02:00
allegroai
6e1f74402e Rename trains-agent -> clearml-agent 2020-12-22 21:21:29 +02:00