The fx tracer does not support tracing "real" quantized tensors
currently. A "real" quantized tensor here means a tensor that is created
using a method like `torch.quantize_per_tensor()` and carries the
quantization parameters (scale, zero_point, scheme) in the object.
However, it seems like the DQ-Q type fake quantizatation is now commonly
used as a high level representation of quantized operators and is only
lowered to native quantized ops (if available) in the respective
hardware backend. Quantization of floating point modules in PyTorch is
recently also performed as a graph transformation after
exporting/tracing the original module.
```python
# Examples of "real"/native quantization
tens = torch.randint(-127, 127, (1,), dtype=torch.int8)
torch._make_per_tensor_quantized_tensor(tens, 1, 0)
# tensor([90.], size=(1,), dtype=torch.qint8,
# quantization_scheme=torch.per_tensor_affine, scale=1.0, zero_point=0)
tens = torch.rand((1,))
torch.quantize_per_tensor(tens, 1, 0, torch.qint8)
# tensor([1.], size=(1,), dtype=torch.qint8,
# quantization_scheme=torch.per_tensor_affine, scale=1.0, zero_point=0)
# Example of DQ/Q quantization
import torch.ao.quantization.fx._decomposed
tens = torch.rand((1,))
torch.ops.quantized_decomposed.quantize_per_tensor.default(tens, 1, 0, -128, 127, torch.int8)
# tensor([1], dtype=torch.int8)
```
This means that a typical import flow for a quantized network
into/through torch-mlir would look like this:
`torch.export() -> quantization transformations on fx graph ->
fx_importer` Where the tensors in the graph are normal float/int tensors
and the quantization parameters are carried by the DQ/Q ops. These kinds
of graphs can be traced without issues.
Currently, our quantized convolution tests use the "real" quantized
tensors. This means that with the retirement of the `jit_ir_importer`,
these tests cannot be imported any longer. In summary, I see no reason
to stick to the "real" quantization in these tests, as both PyTorch 2.0
is using DQ/Q quantization and our linalg backend is also using it.
This patch updates our quantized convolution tests to use the DQ-Q
quantization with the ops from `torch.ops.quantized_decomposed`.
Note: For future reference, there seems to be an ongoing consolidation
of the ops for the DQ/Q scheme on the PyTorch side
(https://github.com/pytorch/ao/issues/986#issuecomment-2390296826).
Downstream projects don't necessarily register this C++ module. This
package removes the dependency and uses `torch.iinfo` to access the max
and min values instead.
This PR add `floordiv` to the `PY_BUILTIN_TO_TORCH_OP`. For
`aten.mul.int` and `aten.floordiv.int` ops, we add new Canonicalization
Patterns as follow:
```
%1 = torch.aten.mul.int %input, %const-5
%2 = torch.aten.mul.int %1, %const-6
```
Will be replaced by
`torch.aten.mul.int %input, %const-30`
And
```
%1 = torch.aten.mul.int %input, %const-5
%2 = torch.aten.floordiv.int %1, %const-5
```
Will directly return `%input`
This PR also relaxes the `float` type constraint in TorchToTosa for the
`AtenRsubScalarOp` conversion.
To test:
`cmake --build build --target check-torch-mlir-all`
New sympy type is introduced to represent integer infinity in upstream
PyTorch repo. Subsequently, sympy.oo is no longer used to represent
infinity upper bound for dynamic dimensions where the upper bound is
unknown. Instead `int_oo` is used to represent integer infinity. This
commit updates the `_sympy_int_to_int` utility in light of this change.
Now that the PyDev feature request pytorch/pytorch#117188 has been
completed, we can remove all the ad-hoc code that propagates sparsity
metadata and replace it with the built-int PyDev metadata for sparse
tensors. This removes a lot of code and also ensures sparsity is
consistent with the torch.sparse package for all cases.
This PR adds support to `fx_importer.py` for handling custom ops that
return an array of tensors. As long as the length of the array is
consistent across runs (determined statically), then this patch will
work. This does not require that the number of tensors returned is
determined by the op's definition.
CC @sjain-stanford
This commit adds the support for new data types: uint4, and int4 and
uint8 tensor protos. Also, it moves some tests from failing to crashing.
Fixes https://github.com/llvm/torch-mlir/issues/3507
Signed-Off By: Vivek Khandelwal <vivekkhandelwal1424@gmail.com>
This bump triggered an upstream assert. Includes a WAR for #3506.
Also includes several things I needed to do to repro:
* When TORCH_MLIR_TEST_CONCURRENCY=1, test runs will be printed.
* Added TORCH_MLIR_TEST_VERBOSE=1 handling to enable verbose mode
(useful on CI).
---------
Co-authored-by: Stella Laurenzo <stellaraccident@gmail.com>
Resolves#3384.
Many ONNX operators are defined by functions and therefore could be
expanded into simpler ONNX operations during importing, avoiding the
need for tools downstream to support these operators directly.
This commit adds this capability to onnx_importer.py. When importing a
node, the schema for the node's operator is retrieved. If the schema
provides a function for the operator, a specialized version for the
node's types and attributes will be created and imported as an MLIR
function with private visibility. An MLIR function call will then be
emitted, instead of a normal operator node. Caching is used to avoid
generating redundant functions within the same module.
In order to avoid a disruptive change to the importer output for a
large number of operators that already have TorchOnnxToTorch support,
an allowlist strategy is used by default. With this commit, only one
operator is allowlisted for expansion, MeanVarianceNormalization.
However, many other operators can be correctly expanded by the current
code, so hopefully the allowlist can be gradually extended. It is
possible to disable the allowlist in the configuration, in which case
all functions are expanded (useful for testing).
Tools downstream of the importer may now need to do inlining when
consuming the output of the importer, e.g.:
cat imported.mlir | torch-mlir-opt --inline --convert-onnx-to-torch
Explanations for subtle code changes:
- Looking up the correct schema and function for an operator requires
knowing the opset version. NodeImporter retrieves this from the
opset imports on the ModelProto retained by the GraphInfo. Previously,
the model_proto field on GraphInfo was None when importing a subgraph
in import_regions, but this conflicts with the new need for opset
version info. Since the apparent purpose of setting it to None was to
control how GraphInfo generates its input map, a new flag is added to
GraphInfo (is_subgraph) to control this behavior, so that the actual
ModelProto can now be provided without breaking this. This also turned
out to be useful for getting the Config via ModelInfo via GraphInfo.
- Some operators' functions are context-dependent, which means the
function definition depends on the types of the inputs. Therefore node
importing now needs to look up the types of a node's inputs, not just
its outputs as was the case previously. Consequently the operand to
find_type_proto_for_name() may now be a graph input or initializer in
some cases, so it has to be updated.
Adds the following arguments:
- "--clear-domain": enabling this flag (default False) will delete the
domain attribute from each node in the onnx model before importing.
Shape inference does not seem to work for onnx ops in custom domains. In
the rare case when these ops have a corresponding counterpart in base
onnx, enabling this flag might allow shape inference to work properly.
- "--opset-version": allows setting the opset version manually. This
will cause the importer to attempt to update the opset_version of the
onnx model before importing. Newer opset versions sometimes have more
robust shape inference patterns.
Tests the basic constructs of registering a custom op and its abstract
implementations (with FakeTensors) in python, going through TorchDynamo
export, followed by importing the shape expressions in the Torch
dialect.
Also fixes the importer were previously the symbolic bind op insertion
was not gated in one place.
Resolving `bool` literals can result in a type change to uint8. This
needs to be converted back to the expected type before returning to the
wrapped `torch` operators.
Discord Thread:
https://discord.com/channels/636084430946959380/1238330633328005243
## Context:
[This](https://github.com/llvm/torch-mlir/blob/main/python/torch_mlir/fx.py#L61)
was updated to support e2e tests for the TorchDynamo frontend in
Torch-MLIR, where we run FX decompositions and import the FX IR to
generate Torch dialect, followed by
`torch-function-to-torch-backend-pipeline`, skipping only the shape/type
refinement for now. However, we should be able to skip many of the torch
simplification passes, as depicted in the [frontend
roadmap](https://github.com/llvm/torch-mlir/blob/main/docs/images/roadmap_frontend.png).
Based on IREE's TorchDynamo
[pipeline](https://github.com/iree-org/iree/blob/main/compiler/plugins/input/Torch/InputConversion/Passes.cpp#L29),
the only two passes we seem to require are: `ReduceOpVariantsPass` and
`DecomposeComplexOpsPass`. This is inline with our findings as well
based on initial exploration.
This PR creates a dedicated frontend simplification pipeline for
TorchDynamo / FX Importer which calls only `ReduceOpVariantsPass` and
`DecomposeComplexOpsPass`. We rely on the e2e fx_importer tests to
ensure we're not regressing by removing many of the passes that were
historically needed for TorchScript.
One notable change here is that we do not call the
`LowerToBackendContractPass` anymore, which used to call
`TorchSimplificationPipeline` iteratively until VerifyBackendContract
was clean. Some of this was required for the shape/type refinement to
converge, which seems a non-issue for Dynamo frontend. Do we anticipate
this (the iterative invocation of TorchSimplificationPipeline followed
by VerifyBackendContract) to be worth retaining in the Dynamo frontend
pipeline? If so, I can make those changes, PLMK.
* Enables assume_strict_symbolic_shapes on fx_importer imported
programs, indicating strict shape semantics.
* Reworks the view->reshape lowering to take advantage of strict mode
and do one of:
* Collapse to 0D
* Flatten/Unflatten when there is an inferred dim.
* Fallback to tensor.reshape
* Splits some test cases up and adds an attribute to control the old
pattern (so new corners can be tested in strict mode in isolation).
* Dynamic inferred mode needs upstream work to generalize expand_shape
(so that case is suppressed here).
* Deletes the assert from the existing tensor.reshape lowering if strict
shape mode is enabled (since the condition it is dynamically asserting
cannot happen).
This is probably a decent PR for learning about blocks and regions.
If you're here to learn about that, consider also looking at
lib/Conversion/TorchToSCF/TorchToSCF.cpp
While this doesn't include an e2e test, it is tested downstream in
https://github.com/nod-ai/SHARK-TestSuite/blob/main/e2eshark/onnx/operators/If/model.py
---------
Co-authored-by: Xida Ren <xida.ren.dev@gmail.com>
This is a large change because prior to this point, Python files in the
project were not consistently formatted. This reformats them all with
black defaults.
Based on experience with prior projects, if you have a dev/long-term
branch with Python patches, you can minimize merge conflicts prior to
rebasing to include this commit by running `black` on your modified
Python files, squashing, and then rebasing/merging.
In the prior state when I supported mutation of user inputs by treating
them as mutable-tensor SSA values, I had left the case of buffer
mutation only vaguely implemented until a concrete use emerged.
This patch reworks this buffer mutation support by assuming that buffers
must be resolved via the hooks symbolically and treated with load/store
semantics. This is implied in the structure since we have no SSA value
that represents a buffer and we already assume that reading parameters
happens via such a mechanism.
* Also adds the basic scaffolding for handling more of these, which will
be needed for cond, while, etc.
* Refactors some of the support in the generic OpOverload emitter so it
can be shared with these other special forms.
This has been on my list for a while, but it just so happens that as
part of upgrading to PyTorch 2.3 and a pure upstream flow in Turbine, we
were using a feature that required integration with auto_functionalized.
This is perhaps the "weirdest" of the higher-order ops and a poor place
to start, but needs must. We have testing for this in Turbine.
Full support in Turbine has an entire custom ops facility. I've reduced
this down to a unit test in torch-mlir.
At some point, this op became kwarg-only instead of arg/kwarg.
Discovered when upgrading to PyTorch 2.3.
Also adds a test as this was untested in-tree (was caught out of tree).
Felix Schneider
Christopher McGirr
zjgarvey
Ze Zhang
penguin_wwy
Vimal
Aart Bik
Vivek Khandelwal
Jiawei Wu
Matthew Francis-Landau
Sambhav Jain
Yuanqiang Liu
Andrea 🦈
Wu Yuan
Rob Suderman
Stella Laurenzo
Xida Ren (Cedar)
