feature/onnx-to-tosa #4
Draft
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
Signed-off-by: Gong Su <[email protected]>
Signed-off-by: Sam <[email protected]> Co-authored-by: Alexandre Eichenberger <[email protected]>
) * Delay scrubbing disposable elements attribute Signed-off-by: Tung D. Le <[email protected]> --------- Signed-off-by: Tung D. Le <[email protected]>
Signed-off-by: Sunny Anand <[email protected]>
…nnx#3002) * Check the return value when mmapping a file and exit gently if failed Signed-off-by: Tung D. Le <[email protected]> * Fix a bug in z/OS code Signed-off-by: Tung D. Le <[email protected]> * Convert filename to EBCDIC when generating code for zOS Signed-off-by: Tung D. Le <[email protected]> --------- Signed-off-by: Tung D. Le <[email protected]>
Signed-off-by: Tung D. Le <[email protected]>
Do not fuse locations when normalizing constants for Add and Mul
…tatic shapes are only run if the types have a static shape
Ensure that constant propagation patterns that rely on types having static shapes are only run if the types have a static shape
Use CHECK-DAG instead of CHECK to make location test more robust
) Signed-off-by: Christopher Munoz <[email protected]>
…ng functions and misc cleanups. Also remove some unecessary semicolons and standardize to use mlir instead of llvm casts
…n because of unraked operand types
…t run because of failed shape inference
Replace deprecated use of .cast/dyn_cast/isa with the mlir freestanding functions and misc cleanups.
Make ONNX to TOSA conversion resistant to dynamic types
…lice_of_const_for_folding
…of_const_for_folding Revert "Use tensor.extract_slice of wts/bias to benefit from folders"
* Fix security vulenraibilities in requirements.txt for mnist example- CVE-2024-28219 pillow - CVE-2024-48063 torch package - CVE-2024-31584 CVE-2024-31583 CVE-2024-31580 for torch package Signed-off-by: Sunny Anand <[email protected]> * upgrade torchvision to match torch Signed-off-by: Sunny Anand <[email protected]> --------- Signed-off-by: Sunny Anand <[email protected]>
…_work_with_integer_type ONNXPadOp now also lowes to TOSA integer types
…/xilinx/onnx-mlir into chaitany.convtranspose_as_option
Chaitany.convtranspose as option
Fix lowering of onnx.Mul with dynamic shape
…ked types. - Add checks for types being ranked in various lowerings - Refactor getSplattedConst to only require a type and not a rank - Manually annotate the return type for some tosa ops to prevent type mismatches -- In some cases to result type of an op can not be inferred just from its inputs, in these cases the annotation is always necxessary -- In some cases the TOSA shape inference does not infere types even if it is possible, this is something I want to improve in another PR for MLIR.
Make ONNX to TOSA lowering more resistant to dynamic shapes and unranked types.
Fix LayerNorm recomposition by properly using provided epsilon value
* Bump various ops to opset 22, adding bf16 support Bumped ops: Acos Acosh Asin Asinh Atan Atanh AveragePool Bernoulli Conv ConvTranspose Cos Cosh DeformConv Det Dropout Elu EyeLike GRU GlobalAveragePool GlobalMaxPool HardSigmoid HardSwish InstanceNormalization LSTM LpNormalization LpPool MaxPool MaxRoiPool MaxUnpool Mish Multinomial NegativeLogLikelihoodLoss RNN RandomNormal RandomNormalLike RandomUniform RandomUniformLike RoiAlign Round Selu Sin Sinh Softplus Softsign Tan ThresholdedRelu Signed-off-by: Rickert, Jonas <[email protected]> * Add test to ensure that we can parse multiple opsets Signed-off-by: Rickert, Jonas <[email protected]> --------- Signed-off-by: Rickert, Jonas <[email protected]> Co-authored-by: Liam Fitzpatrick <[email protected]>
* Add support for TensorProto::UINT4/INT4 Signed-off-by: Rickert, Jonas <[email protected]> * Upgrade onnx.Cast to opset 21 Signed-off-by: Rickert, Jonas <[email protected]> --------- Signed-off-by: Rickert, Jonas <[email protected]>
Signed-off-by: Rickert, Jonas <[email protected]>
* feat: add verifier and shape inference for Onnx.GridSample. Signed-off-by: Rickert, Jonas <[email protected]> * Support onnx.GridSampleV22 Signed-off-by: Rickert, Jonas <[email protected]> * Check attributes in GridSample verifier Signed-off-by: Rickert, Jonas <[email protected]> --------- Signed-off-by: Rickert, Jonas <[email protected]> Co-authored-by: Tiago Trevisan Jost <[email protected]>
Cherry-pick opset updates from upstream
…ert.gridsample_22
Add support for Gridsample 20 and 22.
…e_return Fix return type of recomposed ops
Add options to specify directory with ONNX initializer files
…d_file Remove accidentally commited file
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.
DO NOT merge, just to look at our diff