WIP: Simplify DeviceSerialized and usage thereof
#268
Closed
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
As `"dask"` serialization already converts a CUDA object into headers and frames that Dask is able to work with, drop code that tries to serialize frames on host further (as they are already as simple as they can be). Cuts a fair bit of boilerplate from the spilling path, which should simplify things a bit.
jakirkham
changed the title
DeviceSerialized and usage thereofDeviceSerialized and usage thereof
|
Just to be clear, this PR still has issues. So should not be merged until addressed. In particular we are seeing exceptions like this one, which requires PR ( dask/distributed#3639 ) to fix or at least some other solution here. distributed.worker - ERROR - too many values to unpack (expected 1)
Traceback (most recent call last):
File "/conda/envs/gdf/lib/python3.7/site-packages/distributed/worker.py", line 2169, in release_key
if key in self.data and key not in self.dep_state:
File "/conda/envs/gdf/lib/python3.7/_collections_abc.py", line 666, in __contains__
self[key]
File "/var/lib/jenkins/workspace/rapidsai/gpuci/dask-cuda/prb/dask-cuda-gpu-build/dask_cuda/device_host_file.py", line 125, in __getitem__
return self.device_buffer[key]
File "/conda/envs/gdf/lib/python3.7/site-packages/zict/buffer.py", line 78, in __getitem__
return self.slow_to_fast(key)
File "/conda/envs/gdf/lib/python3.7/site-packages/zict/buffer.py", line 65, in slow_to_fast
value = self.slow[key]
File "/conda/envs/gdf/lib/python3.7/site-packages/zict/func.py", line 38, in __getitem__
return self.load(self.d[key])
File "/var/lib/jenkins/workspace/rapidsai/gpuci/dask-cuda/prb/dask-cuda-gpu-build/dask_cuda/device_host_file.py", line 60, in host_to_device
return deserialize(s.header, s.frames)
File "/conda/envs/gdf/lib/python3.7/site-packages/distributed/protocol/serialize.py", line 269, in deserialize
return loads(header, frames)
File "/conda/envs/gdf/lib/python3.7/site-packages/distributed/protocol/serialize.py", line 51, in dask_loads
return loads(header, frames)
File "/conda/envs/gdf/lib/python3.7/site-packages/distributed/protocol/cupy.py", line 83, in dask_deserialize_cupy_ndarray
frames = [dask_deserialize_cuda_buffer(header, frames)]
File "/conda/envs/gdf/lib/python3.7/site-packages/distributed/protocol/rmm.py", line 38, in dask_deserialize_rmm_device_buffer
(frame,) = frames
ValueError: too many values to unpack (expected 1) |
|
Closing for now to cleanup things. May revisit in the future. |
|
Captured the main features of this in PR ( #307 ). |
|
The rest of the changes here are handled in PR ( #309 ). |
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.
As
"dask"serialization already converts a CUDA object into headers and frames that Dask is able to work with, drop code that tries to serialize frames on host further (as they are already as simple as they can be). Cuts a fair bit of boilerplate from the spilling path, which should simplify things a bit.