You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This file contains hidden or 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
pcard-86321
The Dit/LargeDit model sample a large amount of random data during its training. Especially when running in a distributed environment, these sampled random data also need to be sharded into different devices to reduce memory usage.
But at the beginning of mix2dist_pass design in PIR mode, it only supported the shard of input data and model parameter, and did not support the shard of other data. this PR will expand this capability.
importpaddleimportpaddle.distributedasdistprocess_mesh=dist.ProcessMesh([0, 1], dim_names=['mp'])
# before this PR, only supportlinear1=paddle.nn.Linear(100, 200)
linear1.weight=dist.shard_tensor(linear1.weight, process_mesh, [dist.Shard(1)]) # can shard input data and model parameternoise=paddle.randn(x.shape)
noise=dist.shard_tensor(noise, process_mesh, [dist.Replicate()) # In other cases, placements must be all Replicate, otherwise an error will be reported# after this PR, also supporttimesteps=paddle.randint(0, self.num_timesteps, (x.shape[0],))
timesteps=dist.shard_tensor(timesteps, process_mesh, [dist.Shard(1)) # can shard randomly sampled data(int)noise=paddle.randn(x.shape)
noise=dist.shard_tensor(noise, process_mesh, [dist.Shard(1)) # can shard randomly sampled data(float)
你的PR提交成功,感谢你对开源项目的贡献!
请关注后续CI自动化测试结果,详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
PR Category
Auto Parallel
PR Types
New features
Description
pcard-86321
The Dit/LargeDit model sample a large amount of random data during its training. Especially when running in a distributed environment, these sampled random data also need to be sharded into different devices to reduce memory usage.
But at the beginning of
mix2dist_pass
design in PIR mode, it only supported the shard of input data and model parameter, and did not support the shard of other data. this PR will expand this capability.