[Paddle Backend] Add mixed precision training for se_e2_a_mixed_prec #3030
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Support virial forward run on gpu with cpu kernel
Deepmd in Paddle for example, just 'water_se_a' model
…a_cpu fix error in cpu mode
…load support jist save load
Following issues fixed: 1. Removed @paddle.jit.to_static decorator. Model will be converted to static graph at save time. 2. Manually set InputSpec for "Ener" model with "se_a" descriptor 3. Due to lack of support for "double" datatype at inference time, default training precision was set to float (low precision)
[Paddle] Fixed model save issues with Ener model
Detected functional regression between CUDA 10.1 and CUDA 11.2 Therefore force 3 custom ops namely, "env_mat", "force_se_a" and "virial_se_a" to fallback on CPU Minor changes to save_model function in "trainer.py" to suppress dynamic-to-static warnings
Force env_mat force_se_a virial_se_a to fallback on CPU
This reverts commit 156c0d3.
Revert "Force env_mat force_se_a virial_se_a to fallback on CPU"
for more information, see https://pre-commit.ci
…ogenSulfate/deepmd-kit into add_ddle_backend_polish_ver
njzjz
pushed a commit
that referenced
this pull request
Dec 30, 2023
…revert code format) (#3096) Code formatting for #3030 --------- Signed-off-by: HydrogenSulfate <[email protected]> Co-authored-by: zhouwei25 <[email protected]> Co-authored-by: JiabinYang <[email protected]> Co-authored-by: Han Wang <[email protected]> Co-authored-by: Zhanlue Yang <[email protected]> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Add mixed precision training mode for se_e2_a_mixed_prec. Time&mem cost below
(GPU)Training metric with
water(se_e2_a_mixed_prec)(reference(fp64):