| Name | rockmate JSON |
| Version |
2.0.0
JSON |
| download |
| home_page | None |
| Summary | Efficient and Automatic Rematerialization for Pytorch training |
| upload_time | 2024-12-09 16:25:54 |
| maintainer | None |
| docs_url | None |
| author | Théotime Le Hellard, Xunyi Zhao, Julia Gusak, Li Zhe, Olivier Beaumont, Lionel Eyraud-Dubois |
| requires_python | >=3.8 |
| license | MIT License Copyright (c) 2021-present Inria Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. |
| keywords |
rematerialization
training
pytorch
memory
|
| VCS |
 |
| bugtrack_url |
|
| requirements |
No requirements were recorded.
|
| Travis-CI |
No Travis.
|
| coveralls test coverage |
No coveralls.
|
# Rockmate
The `Rockmate` framework is designed for training a PyTorch neural network within a given GPU budget
constraint using automatic re-materialization (activation checkpointing) technique.
Given a PyTorch model, a sample input, and a GPU memory budget, `Rockmate` builds a new
`torch.nn.Module`, which performs forward and backward pass keeping activations under the given
budget.
- The new model produces the same outputs and gradients as the original one.
- Model training with a budget constraint, which is lower than the one required by PyTorch Autodiff,
is achieved by re-computing some of the activations instead of storing them for gradient
calculation.
- Depending on the budget, `Rockmate` defines automatically which activations should be recomputed.
More information on [our repository](https://github.com/topal-team/rockmate).
Raw data
{
"_id": null,
"home_page": null,
"name": "rockmate",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.8",
"maintainer_email": "Lionel Eyraud-Dubois <lionel.eyraud-dubois@inria.fr>",
"keywords": "rematerialization, training, pytorch, memory",
"author": "Th\u00e9otime Le Hellard, Xunyi Zhao, Julia Gusak, Li Zhe, Olivier Beaumont, Lionel Eyraud-Dubois",
"author_email": null,
"download_url": "https://files.pythonhosted.org/packages/6c/8f/7052feecb918599323be3fad744b3ac53acea483bb577a907ba9241a6f84/rockmate-2.0.0.tar.gz",
"platform": null,
"description": "# Rockmate\n\nThe `Rockmate` framework is designed for training a PyTorch neural network within a given GPU budget\nconstraint using automatic re-materialization (activation checkpointing) technique.\n\nGiven a PyTorch model, a sample input, and a GPU memory budget, `Rockmate` builds a new\n`torch.nn.Module`, which performs forward and backward pass keeping activations under the given\nbudget.\n\n- The new model produces the same outputs and gradients as the original one.\n- Model training with a budget constraint, which is lower than the one required by PyTorch Autodiff,\n is achieved by re-computing some of the activations instead of storing them for gradient\n calculation.\n- Depending on the budget, `Rockmate` defines automatically which activations should be recomputed.\n\nMore information on [our repository](https://github.com/topal-team/rockmate).\n",
"bugtrack_url": null,
"license": "MIT License Copyright (c) 2021-present Inria Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the \"Software\"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. ",
"summary": "Efficient and Automatic Rematerialization for Pytorch training",
"version": "2.0.0",
"project_urls": {
"Repository": "https://github.com/topal-team/rockmate.git"
},
"split_keywords": [
"rematerialization",
" training",
" pytorch",
" memory"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "6c8f7052feecb918599323be3fad744b3ac53acea483bb577a907ba9241a6f84",
"md5": "0bc63c8746da0fe09a3254063c8c5653",
"sha256": "68c41a21858121c569ea6dbe973c0208b14579c4c5a360fdbcf00e8d7b1ffbe6"
},
"downloads": -1,
"filename": "rockmate-2.0.0.tar.gz",
"has_sig": false,
"md5_digest": "0bc63c8746da0fe09a3254063c8c5653",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8",
"size": 78804,
"upload_time": "2024-12-09T16:25:54",
"upload_time_iso_8601": "2024-12-09T16:25:54.673209Z",
"url": "https://files.pythonhosted.org/packages/6c/8f/7052feecb918599323be3fad744b3ac53acea483bb577a907ba9241a6f84/rockmate-2.0.0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-12-09 16:25:54",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "topal-team",
"github_project": "rockmate",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "rockmate"
}
