[feat] Make hdfs function compatible with tensorflow-io and TensorFlow version earlier than 2.6 by using FileSystem in TensorFlow Env. Also support S3 and so on now. #281
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
rhdong
reviewed
Sep 29, 2022
tensorflow_recommenders_addons/dynamic_embedding/core/kernels/lookup_impl/lookup_table_op_cpu.h
Outdated
Show resolved
Hide resolved
MoFHeka
changed the title
luliyucoordinate
previously approved these changes
Oct 8, 2022
rhdong
reviewed
Oct 9, 2022
tensorflow_recommenders_addons/dynamic_embedding/core/kernels/cuckoo_hashtable_op_gpu.cu.cc
Outdated
Show resolved
Hide resolved
rhdong
reviewed
Oct 9, 2022
tensorflow_recommenders_addons/dynamic_embedding/core/kernels/cuckoo_hashtable_op_gpu.cu.cc
Outdated
Show resolved
Hide resolved
rhdong
reviewed
Oct 9, 2022
tensorflow_recommenders_addons/dynamic_embedding/core/kernels/lookup_impl/lookup_table_op_gpu.h
Outdated
Show resolved
Hide resolved
rhdong
reviewed
Oct 9, 2022
tensorflow_recommenders_addons/dynamic_embedding/core/kernels/lookup_impl/lookup_table_op_gpu.h
Outdated
Show resolved
Hide resolved
tensorflow_recommenders_addons/dynamic_embedding/core/kernels/lookup_impl/lookup_table_op_cpu.h
Outdated
Show resolved
Hide resolved
MoFHeka
force-pushed
the
master-dev
branch
8 times, most recently
from
f278091
to
0323ee3
Compare
rhdong
reviewed
Oct 14, 2022
tensorflow_recommenders_addons/dynamic_embedding/core/kernels/lookup_impl/lookup_table_op_cpu.h
Outdated
Show resolved
Hide resolved
MoFHeka
force-pushed
the
master-dev
branch
2 times, most recently
from
9642b5f
to
a65c567
Compare
MoFHeka
force-pushed
the
master-dev
branch
9 times, most recently
from
a5fd535
to
36540ce
Compare
rhdong
reviewed
Oct 26, 2022
| virtual void get(const K* d_keys, ValueType<V>* d_vals, bool* d_status, | ||
| size_t len, ValueType<V>* d_def_val, cudaStream_t stream, | ||
| bool is_full_size_default) const {} | ||
| virtual size_t get_size(cudaStream_t stream) const { return 0; } | ||
| virtual size_t get_capacity() const { return 0; } | ||
| virtual void remove(const K* d_keys, size_t len, cudaStream_t stream) {} | ||
| virtual void clear(cudaStream_t stream) {} | ||
| virtual void remove(const K* d_keys, size_t len, cudaStream_t stream) const {} |
There was a problem hiding this comment.
Why do you change here? remove & clear should not be const.
luliyucoordinate
previously approved these changes
Oct 26, 2022
MoFHeka
force-pushed
the
master-dev
branch
2 times, most recently
from
85f7ea8
to
6a37b0c
Compare
…w version earlier than 2.6 by using FileSystem in TensorFlow Env. Change hdfs file system to all file system. Now also support S3, local and so on. Also change Redis backend function to make it more low coupling with TF. Also modify GPU/CPU hash table code(rehashifneeded/dump/cpu insert) to make them more decoupled and stable.
rhdong
reviewed
Oct 28, 2022
tensorflow_recommenders_addons/dynamic_embedding/core/kernels/lookup_impl/lookup_table_op_gpu.h
Show resolved
Hide resolved
but for unknown reasons with nvhashtable, the insert may have unknowable consequences that result in NaN values for the training parameters. Avoid unknown errors by calling cudaDeviceSynchronize() for purposes such as cache synchronization before insert kernels luanching.
|
Excellent job! Thanks, @MoFHeka ! |
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.
Description
[feat] Make hdfs function compatible with tensorflow-io and TensorFlow version earlier than 2.6 by using FileSystem in TensorFlow Env.
After TF 2.6, TF file system was migrated from TF main repo to TF io repo. So when a TF add-on need to use file system, it's better to get file system pointer from TF env for compatibility.
Also support S3 and so on now.
Also add GPU/Redis backend implementation.
Also modify GPU/CPU hash table code(rehashifneeded/dump) to make them more decoupled.
An insert is typically performed after the rehash_if_needed call, but for unknown reasons with nvhashtable, the insert may have unknowable consequences that result in NaN values for the training parameters. Avoid unknown errors by calling cudaDeviceSynchronize() for purposes such as cache synchronization before insert kernels luanching.
Type of change
Checklist:
How Has This Been Tested?
Compile and unit test it in different TF version.