✨[Feature] handling the `accumulate=True` for aten index_put operations.

pytorch / TensorRT

PyTorch/TorchScript/FX compiler for NVIDIA GPUs using TensorRT

https://pytorch.org/TensorRT

BSD 3-Clause "New" or "Revised" License

2.41k stars 333 forks source link

Open chohk88 opened 2 weeks ago

chohk88 commented 2 weeks ago

Is your feature request related to a problem? Please describe.

The current implementation uses a validator to handle the accumulate=True case. https://github.com/pytorch/TensorRT/pull/2880

Describe the solution you'd like

Describe alternatives you've considered

Additional context