Ti-Tai Wang

[ONNX] Exporter successfully export a ONNX model, but unexecutable on ONNXRUNTIME due to failed shape inference..

1

### 🐛 Describe the bug In exporter, there is a onnx.shape_inference_infer_shapes checker before exporting, but it is not using strict_mode as onnxruntime does before executing the model, so the discrepancy...

# Bug Report ### Is the issue related to model conversion? Yes, but the error happens during shape type inference on Loop(Node) https://github.com/onnx/onnx/blob/eb634addcbb23ec1baf4d548d826bf3932527533/onnx/defs/shape_inference.cc#L343 which we can track back to https://github.com/onnx/onnx/blob/eb634addcbb23ec1baf4d548d826bf3932527533/onnx/defs/shape_inference.cc#L414-L425...

torch.nn.LayerNorm mismatches in nightly.

7

**Describe the bug** torch.nn.LayerNorm mismatches in nightly, but matches in 1.12.1 **Urgency** None. **System information** Nightly torch Nightly onnxruntime **To Reproduce** ```python import io import torch from onnxruntime import InferenceSession,...

[ONNX] Support huggingface BART to ONNX

4

Add BART into transformer support, specificalyy for `BartForConditionalGeneration` **Motivation and Context** - fixes #11210 Currently, the custom op beam search is not working in nightly, this PR should be run...

Onnx team/atenlib evaluator

1

depends on onnx-script branch: https://github.com/microsoft/onnx-script/tree/torchscript_evaluator 1. Use torchscript as an example graph builder (as we don't have one yet) to demonstrate the usage of onnx-script API and atenlib 2. How...

Does antialias in Resize-18 work with empty scale?

2

The attribute `antialias` says that it is achieved by stretching the resampling filter by a factor max(1, 1 / scale), and under the rule that only one of the input...

Freeze the model

1

Hi, I am wondering if you freeze the pre-trained part of parameters, and only fine-tune on the downstream task with head? It seems like all parameters are tuned during the...

[ONNX] Skip assertion nodes

4

Continue from #125930 The PR addresses the common existence of unused assertion nodes in models. (workaround #112443) Assertion ops does not return anything in the graph, so it's not even...

Index put loop model regression with ort==1.18

1

### Describe the issue The error is only raised after 1.18. I tried 1.17.3, and it works fine. onnxruntime.capi.onnxruntime_pybind11_state.InvalidArgument: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Non-zero status code returned while...