-
Notifications
You must be signed in to change notification settings - Fork 325
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Lsflow develop #463
Merged
Merged
Lsflow develop #463
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
problem : LinearOp::forward is getting cublashandle without checking if context is built. LinearOp::backward is checking if the context is built before getting cublashandle. solution: Modify LinearOp::forward to check if context is built before getting cublashandle.
* add split head for beam search * alter checkin
hexisyztem
requested review from
neopro12,
godweiyang and
Taka152
as code owners
February 16, 2023 05:36
hexisyztem
added a commit
that referenced
this pull request
Mar 2, 2023
* multi device develop * change build.sh * cuda tools move into kernels * multi device develop * multi device develop * multi device develop * solve multi device develop * solve arm compile error * arm print_vec * fix regress error * solve cuda compile error * solve multi device print vec * split node class wit its derived class * lightseq x86 unit test * add x86 unit test * Canonical Namespace * add pybind compile * Lsflow develop (#463) * Fix new arch context build check(#441) problem : LinearOp::forward is getting cublashandle without checking if context is built. LinearOp::backward is checking if the context is built before getting cublashandle. solution: Modify LinearOp::forward to check if context is built before getting cublashandle. * fix config reference bug (#453) * developing lsflow * add split_head op and its test (#454) * lsflow develop * format note message * lsflow tune * add notes for context class * add note for lsflow * op example * Update CODEOWNERS (#457) * Gpt infer (#456) * add split head for beam search * alter checkin * make launch_transform_0213 more clear (#459) * fix operator compile * make launch_transform_0213 more clear (#460) * change max_shape to max_shape_size * correct bias shape in split_head (#461) * add unit test for x86 cpu kernel --------- Co-authored-by: Kangmo Kim <kangmo.kim@gmail.com> Co-authored-by: Ying Xiong <xiongying.taka@bytedance.com> Co-authored-by: Xiaohui Wang <wangxiaohui.neo@bytedance.com> * fix sys.path * fix sys.path (#466) * jit build support pure cpu machine * robust builder for x86 and cuda * fix compile error * develop test_ls_layer * format * fix training compile problem * add mkl gemm for f32 and s8 (#470) * test for encoder layer (#471) * test encoder layer * fix cuda free error * lightseq multi device develop * fix crf op error * avoid import training directories * fix strided_batch_gemm config data type * add debug message * convert shape from vector<int> to vector<size_t> * change debug log format * format * remove useless dropout * add sdpa layer into multi head attention layer * fix conflict parameter: is_post_ln and pre_or_po... * fix CMakeLists.txt compile * multi kernel develop * lightseq transformer.cu fix * fix linear col/row major * fix compile error * fix lightseq post_ln network structure * add transformer example & print error message * add shape message * format * fix concat error * add shape message * fix beam search bug --------- Co-authored-by: Kangmo Kim <kangmo.kim@gmail.com> Co-authored-by: Ying Xiong <xiongying.taka@bytedance.com> Co-authored-by: Xiaohui Wang <wangxiaohui.neo@bytedance.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.
No description provided.