13 Pull requests merged by 12 people

• ThreadPool: Spend less time busy waiting.

  #21545 merged Oct 2, 2024

• fix issues for e2e phi3

  #22287 merged Oct 1, 2024

• [java] Multi-LoRA support

  #22280 merged Oct 1, 2024

• Address Android warning error

  #22285 merged Oct 1, 2024

• [webgpu-native] Fix a few build errors on Linux

  #22286 merged Oct 1, 2024

• [docs] Fix fp16 mixed precision example to reflect correct input variable names

  #22250 merged Oct 1, 2024

• [Java] Add API for appending QNN EP

  #22208 merged Oct 1, 2024

• Update SAM2 benchmark for testing torch compile modes and profiling

  #22279 merged Oct 1, 2024

• remove neural-speed

  #22236 merged Oct 1, 2024

• Fix equation in MatMulNBits op spec

  #22253 merged Oct 1, 2024

• [webgpu-native] support for webgpu layernorms

  #22249 merged Oct 1, 2024

• Fixes a crash on macOS 15 when using CoreML.

  #22277 merged Oct 1, 2024

• Fix syntax for some CoreML ML Program supported operator entries

  #22268 merged Oct 1, 2024

3 Pull requests opened by 3 people

• Fix too strict assert in onnx_quantizer.py

  #22283 opened Oct 1, 2024

• Fix Memory Issue GQA CPU Rotary

  #22290 opened Oct 1, 2024

• [DML] Add support for AppendExecutionProvider_DML1 with C#.

  #22291 opened Oct 1, 2024

4 Issues closed by 4 people

• 1.19.2 Support for CUDA 12.6 and CUDNN 9.4

  #22276 closed Oct 2, 2024

• QNN EP: Context cache only works with single partition

  #18354 closed Oct 1, 2024

• [Documentation] the example does not define test_data variable

  #21822 closed Oct 1, 2024

• ONNXRuntime crashes on macOS 15 when using CoreML

  #22275 closed Oct 1, 2024

3 Issues opened by 3 people

• [Build] Using xnnpack fails with "_Float16 is not supported on this target" for iOS x86_64 simulator

  #22289 opened Oct 1, 2024

• [Mobile] QNN Execution provider mem alloc failure.

  #22288 opened Oct 1, 2024

• ORT would be crashed while loading the specific INT4 model

  #22284 opened Oct 1, 2024

20 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• [C#] Expose Multi-Lora support in C#

  #22281 commented on Oct 1, 2024 • 6 new comments

• Add MaxPool FP16 in XnnPack EP

  #22258 commented on Oct 1, 2024 • 2 new comments

• Fix MlasSgemmKernel: properly process more than 2 rows

  #22125 commented on Oct 1, 2024 • 1 new comment

• Migrate Android Java E2E tests from App Center to Browserstack

  #22117 commented on Oct 1, 2024 • 1 new comment

• Adding convInteger u8s8 support

  #22034 commented on Oct 1, 2024 • 1 new comment

• Fix matmul nbits casts

  #22265 commented on Oct 1, 2024 • 0 new comments

• enable serialize prepacked weights into data file

  #22256 commented on Oct 1, 2024 • 0 new comments

• Add microbenchmark for layer normalization and improve latency

  #22223 commented on Oct 1, 2024 • 0 new comments

• Update gradle version to 8.7

  #22206 commented on Oct 1, 2024 • 0 new comments

• [WebNN EP] Support Tile operator

  #22148 commented on Oct 2, 2024 • 0 new comments

• [VitisAI] Cache node subgraph when necessary

  #22073 commented on Oct 2, 2024 • 0 new comments

• [WIP] WebGPU EP [skip ci]

  #21904 commented on Oct 1, 2024 • 0 new comments

• Integrate onnx 1.17.0

  #21897 commented on Oct 2, 2024 • 0 new comments

• adjust orttraining binary size test

  #21741 commented on Oct 1, 2024 • 0 new comments

• Enabling c++20 on linux

  #17816 commented on Oct 2, 2024 • 0 new comments

• Memory leak after running onnx model numerous times

  #22271 commented on Oct 1, 2024 • 0 new comments

• onnxruntime.InferenceSession.run sometimes get stuck, sometimes not

  #21418 commented on Oct 1, 2024 • 0 new comments

• [bug] `quantize_dynamic` results in `AttributeError: 'NoneType' object has no attribute 'HasField'`

  #19997 commented on Oct 1, 2024 • 0 new comments

• [Feature Request] Make DirectML execution provider thread safe (allow Run() concurrency)

  #22147 commented on Oct 1, 2024 • 0 new comments

• [Performance] Very slow load of ONNX model in Windows

  #22219 commented on Oct 1, 2024 • 0 new comments