### Description
This PR adds SbgemmKernel for aarch64. This includes Sbegmm kernel to
implement matrix multiplication with bfloat16 SIMD instructions (bfmmla)
and MatMul operator changes to invoke the Sbgemm kernel. To enable
Sbgemm kernel, set the following session option:
"kOrtSessionOptionsGemmFastMathMode"
The PR also adds new test cases for mlas and ort.
### Motivation and Context
This is to improve MatMul performance on aarch64 platform.
I have run the below benchmarking script (bert , roberta and gpt2 model
inference) on AWS Graviton3 based c7g.4xl instance and observed 1.2x
-1.76x performance improvement compared to sgemm (fp32) kernel
performance.
```
cd onnxruntime/python/tools/transformers
python3 benchmark.py
```
And the unit test precision results are matching to sgemm kernel
results.
`./build.sh --config RelWithDebInfo --build_shared_lib --parallel
--compile_no_warning_as_error --skip_submodule_sync `
### Description
- Adds the following session options to configure the device:
- `soc_model`: The SoC model number. Refer to the QNN SDK documentation
for valid values. Defaults to "0" (unknown).
- `htp_arch`: The minimum HTP architecture the driver will use to select
compatible QNN operators.
- `device_id`: The ID of the device to use when setting 'htp_arch'.
Defaults to "0" (for single device).
### Motivation and Context
Allow more configuration.
### Description
<!-- Describe your changes. -->
1. support causal mask in MHA cpu
2. support custom rotary_dim in rotary_emb
3. add bf16 for rotary_emb
4. fix a bug in attention rotary
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
Several changes:
1. To align with other EPs' setting of EP context configs in session
options, for example [QNN
EP](https://github.com/microsoft/onnxruntime/pull/18877), EP context
configs for TRT EP can be configured through:
1. Session Options: `ep.context_enable`, `ep.context_file_path` and
`ep.context_embed_mode`
2. Provider Options: `trt_dump_ep_context_model`,
`trt_ep_context_file_path` and `trt_dump_ep_context_embed_mode`
3. Above setting has 1:1 mapping and provider options has higher
priority over session options.
```
Please note that there are rules for using following context model related provider options:
1. In the case of dumping the context model and loading the context model,
for security reason, TRT EP doesn't allow the "ep_cache_context" node attribute of EP context node to be
the absolute path or relative path that is outside of context model directory.
It means engine cache needs to be in the same directory or sub-directory of context model.
2. In the case of dumping the context model, the engine cache path will be changed to the relative path of context model directory.
For example:
If "trt_dump_ep_context_model" is enabled and "trt_engine_cache_enable" is enabled,
if "trt_ep_context_file_path" is "./context_model_dir",
- if "trt_engine_cache_path" is "" -> the engine cache will be saved to "./context_model_dir"
- if "trt_engine_cache_path" is "engine_dir" -> the engine cache will be saved to "./context_model_dir/engine_dir"
```
2. User can decide the naming of the dumped "EP context" model by using
`trt_ep_context_file_path`, please see GetCtxModelPath() for more
details.
3. Added suggested comments from
https://github.com/microsoft/onnxruntime/pull/18217
Fix issue that the generated context cache model inputs/outputs order is not guaranteed
### Description
Currently, QNN EP generate the context cache model in Compile() method which only get access to the partitioned graph. And the inputs/outputs order for the partitioned graph is not guaranteed. And EP doesn't have the view of the input user model. Have to move the context cache model generation to a higher level in GraphPartitioner which has the view of the partitioned model.
This is also a break down of PR for multi-partition support.
https://github.com/microsoft/onnxruntime/pull/18865
### Description
This PR updates the LLaMA-2 attention fusions by adding the following.
- Loading the PyTorch model from Hugging Face with the `LlamaAttention`
class before exporting
- Updating the attention mask pattern matching to support another case
This PR also fixes [this
issue](https://github.com/microsoft/onnxruntime/issues/19040).
### Motivation and Context
Recent changes to Hugging Face's `transformers` library break the
existing pattern matching. Since the attention fusions aim to change the
graph from `LayerNorm Op --> Set of Attention Nodes --> LayerNorm Op` to
`LayerNorm Op --> Attention Op --> LayerNorm Op` per layer, ultimately
it does not matter what nodes comprise the `Set of Attention Nodes`
because they will all be removed and replaced by the `Attention Op` in
the end.
Therefore, it does not matter whether the `LlamaAttention` class or a
different attention class is used to load the PyTorch model before
exporting because the expected graphs after the attention fusions will
look identical no matter the attention class chosen. By loading the
PyTorch model with the `LlamaAttention` class instead of other attention
classes (e.g. `LlamaFlashAttention2` or `LlamaSdpaAttention`) and then
exporting it to ONNX, the existing pattern matching will continue to
work.
### Description
upgrade packages version.
```
# npm audit report
electron 23.0.0-alpha.1 - 23.3.13
Severity: moderate
ASAR Integrity bypass via filetype confusion in electron - https://github.com/advisories/GHSA-7m48-wc93-9g85
fix available via `npm audit fix --force`
Will install electron@28.1.4, which is a breaking change
node_modules/electron
get-func-name <2.0.1
Severity: high
Chaijs/get-func-name vulnerable to ReDoS - https://github.com/advisories/GHSA-4q6p-r6v2-jvc5
fix available via `npm audit fix`
node_modules/get-func-name
semver <=5.7.1 || 6.0.0 - 6.3.0 || 7.0.0 - 7.5.1
Severity: moderate
semver vulnerable to Regular Expression Denial of Service - https://github.com/advisories/GHSA-c2qf-rxjj-qqgw
semver vulnerable to Regular Expression Denial of Service - https://github.com/advisories/GHSA-c2qf-rxjj-qqgw
semver vulnerable to Regular Expression Denial of Service - https://github.com/advisories/GHSA-c2qf-rxjj-qqgw
fix available via `npm audit fix`
node_modules/cross-spawn/node_modules/semver
node_modules/global-agent/node_modules/semver
node_modules/semver
```
### Description
<!-- Describe your changes. -->
1. Make JBLAS codes an external module of ORT.
2. Move q4 gemm code to contrib_ops.
3. Update template kernel library to v0.1 release.
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
We found that the current LLM model performance is far below our
expectations. Here is some performance data collected on Mistral-7B
model with Xeon-8480:
8 threads | prompt length=32 past_len=32 | prompt length=1 past_len=32
-- | -- | --
ORT-main | 1220ms | 263ms
Neural-speed | 564ms | 87ms
ORT-this PR|597ms|120ms
Although `Neural-speed` and `ORT-this PR` use the same int4 kernel code,
there is a 33ms(87ms vs. 120ms) latency gap between the two frameworks.
Through some statistics analysis, the summary latency of `MatMulNBits`
is 86.7ms
The summary latency of all int4 GEMMs in `Neural-speed` is 84.8ms. So
other OPs introduce an extra 30ms latency.
The performance of MatMulNBits in this PR meets our expectations.
### Remain Issues
1. For hybrid CPUs, like core 12900K, the ONNXRuntime thread pool uses
TaskGranularityFactor to scale its number of threads. This is not
expected in our code design. It may slow down the hybrid CPU performance
by 30~40%.
2. Prepack uses a single thread which is very slow to init a session.
3. MatMulNBits with zero points will fall through to COMP_FP32 even
accuracy_level=4. Our COMP_INT8 IGemmCore with zero points process is
not optimized for now. It will be updated in the future. So, for an int4
model with zero points, whether the accuracy_level is 0 or 4 will be no
difference.
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
show warning when numThreads is set but threads is not supported.
Resolves#19148, #18933
for web: when crossOriginIsolated is false.
for node: always disable.
### Description
<!-- Describe your changes. -->
Bump up version to 1.18.0 since the release branch has been cut.
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
Co-authored-by: rachguo <rachguo@rachguos-Mini.attlocal.net>
### Description
Extends the code coverage to Entroy, Histogram and Distribution
calibration method, fix bugs while doing it.
### Motivation and Context
Bugs detected in [Olive](https://github.com/microsoft/OLive).
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
Previously building webnn ep with --disable_rtti will throw
unboundTypeError since unbound type names are illegal with RTTI disabled
in Embind API, we can fix it by adding a
-DEMSCRIPTEN_HAS_UNBOUND_TYPE_NAMES=0 flag.
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
Update DML version to 1.13.1
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
Fix build error when ENABLE_NPU_ADAPTER_ENUMERATION is defined
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
The Java `TensorInfo` object which is used to describe a tensor's shape,
along with the input and output placeholders for a model couldn't show
any symbolic/named dimensions in that tensor. Now this information is
stored in Java strings on construction and included in the toString.
### Motivation and Context
Setting symbolic dimensions required external information in Java, the
names were not discoverable from within the API.
### Description
<!-- Describe your changes. -->
### Motivation and Context
Linux_GPU_x64 job in the pipeline has been canceled due to timeout since
0112.
### Description
<!-- Describe your changes. -->
The `split` input of the Split op is int64_t. Fixing that resolves a
type mismatch build error on Windows when CoreML is enabled (for
debugging the partitioning code).
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
Fix build error
---------
Co-authored-by: Edward Chen <18449977+edgchen1@users.noreply.github.com>
### Description
The `OnnxValue` and `OrtProviderOptions` implementations now check to
see if they've been closed before accessing the native pointer, and also
before close is called.
### Motivation and Context
Before they could be closed twice which SIGSEGV'd the JVM. Fixes#19125.
### Description
This way, we will not need to update the windows images constantly and
allow more flexibility to choose the cuda version in the future.
### Description
Set default flags nvcc and do not set the flags for ROCM EP.
### Motivation and Context
1. To meet a BinSkim requirement for CUDA EP.
https://github.com/microsoft/binskim/blob/main/docs/BinSkimRules.md#rule-BA2024EnableSpectreMitigations
2. The ROCM EP's pipeline is broken since PR #19073 . Unit tests failed
to load the EP with the following error message:
Failed to load library libonnxruntime_providers_rocm.so with error:
/build/Release/libonnxruntime_providers_rocm.so: undefined symbol:
vtable for onnxruntime::InsertMaxPoolOutput .
This PR is a hot fix to bring the pipeline back. So far I don't know why
the error happened. The symbol "InsertMaxPoolOutput" is in
onnxruntime_optimizers. I don't see any EP code references it directly.
### Description
Disable ccache for all the jobs in in Windows CPU CI pipeline.
Before disabling it, the build has a warning that:
"MSIL .netmodule or module compiled with /GL found; restarting link with
/LTCG; add /LTCG to the link command line to improve linker performance"
After disabling it, the warning is gone and the build doesn't use /GL or
/LTCG.
Cache itself should not cause this difference.
### Motivation and Context
We submit kernels in a batch (a fixed number 16 is used except for the
last batch) for better performance. However, timestamp query support is
at pass level so we disable the batch execution in profiling mode in
previous implementation. Actually we can have multiple passes in a batch
so that we don't have to disable batch execution, which is the first
enhancement of this PR.
Furthermore, WebGPU has an extension to support timestamp query inside
passes, which isn't supported by all the platforms (e.g., Windows
supports it, while macOS doesn't). This is expected to have lower cost
compared with multiple passes solution. So this PR also introduce this
support when available.
This PR also refactors some implementation related to kernelInfo, and
try to unify the related kernel names.
