Fix roberta conversion bugs #964
Merged
+25
−7
Conversation
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
narendasan
reviewed
Apr 6, 2022
|
@Njuapp Can you sign your commits? |
narendasan
requested changes
Apr 11, 2022
core/util/trt_util.cpp
Outdated
| @@ -238,6 +238,7 @@ const std::unordered_map<at::ScalarType, nvinfer1::DataType>& get_at_trt_type_ma | |||
| {at::kFloat, nvinfer1::DataType::kFLOAT}, | |||
| {at::kHalf, nvinfer1::DataType::kHALF}, | |||
| {at::kInt, nvinfer1::DataType::kINT32}, | |||
| {at::kLong, nvinfer1::DataType::kINT32}, | |||
There was a problem hiding this comment.
I dont want to just associate all instances of long with kINT32. If there are cases of kLong we should explicity handle these cases and explain what we want to do
There was a problem hiding this comment.
So in this case in the evaluator, we can add a code path for Long in the to evaluator (when truncate is enabled) and print a warning
Njuapp
force-pushed
the
roberta_fix
branch
2 times, most recently
from
eb69bf2 to
4153778
Compare
Signed-off-by: Cheng Hang <[email protected]>
|
I have done cpp_lint with clang-format-9, but it still fails in CI/CD as above. I don't understand the reasons. Maybe you can help check cpp_lint too. |
|
I can fix the linting |
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Outdated suggestions cannot be applied.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
Fix bugs encountered when converting a RoBERTa model.
Fixes #963
Specifically, it fixes three things:
aten::ne.Scalar(Tensor self, Scalar other) -> (Tensor), the datatype of Scalar is by default initialized to be of type float. This should be int32 in this case.aten::cumsum(Tensor self, int dim, *, int? dtype=None) -> (Tensor), the zeroValue which stores runningSum, is by default initialized to be of type float. This should be int32 in this case.aten::to.dtype(Tensor self, int dtype, bool non_blocking=False, bool copy=False, int? memory_format=None) -> (Tensor), it could be casting tensor tolongdatatype, but this cannot be processed in our code. We simply add an entry inget_at_trt_type_map, whereat::kLongwould be mapped tonvinfer1::DataType::kINT32since tensorRT could not supportlong.Type of change
Please delete options that are not relevant and/or add your own.
Checklist: