George S

Results 43 issues of George S

# Context In models with segmentation and fallback, it is oftentimes the case that the subgraphs therein have the same substructure, input shapes, and data types. In such cases, it...

feature request

## TL;DR First iteration of Windows support task tracking ## Tasks ```[tasklist] ### Tasks - [ ] https://github.com/pytorch/TensorRT/issues/2371 - [ ] https://github.com/pytorch/TensorRT/issues/2247 - [ ] https://github.com/pytorch/TensorRT/issues/2489 ```

Story

Add support for `torch.ops.aten.glu.default`, documented [here](https://pytorch.org/docs/stable/generated/torch.nn.functional.glu.html) and [here](https://github.com/pytorch/pytorch/blob/f3dd2a544c787aa11bcb84837a50195864f73cc9/aten/src/ATen/native/native_functions.yaml#L11564). See #2197 for usecase.

feature request
component: converters

# Description - Adds mirror functions and testing in C++ to control the `MULTI_DEVICE_SAFE_MODE` flag Fixes #2813 ## Type of change Please delete options that are not relevant and/or add...

component: tests
component: core
component: runtime
cla signed

# Description - As is added in #2806 for C++ builds, enable `--release` flag for builds in `main` to enable correct nightly versioning on PyTorch index ## Type of change...

cla signed
needs-release-cherrypick

# Description - Cherry-pick of #2821 ## Type of change - Cherry-pick # Checklist: - [ x ] My code follows the style guidelines of this project (You can use...

cla signed

- Add converter or lowering pass for `torch.ops.aten.baddbmm.default`

hf

Remove/replace: https://github.com/pytorch/TensorRT/blob/992ad70804a4967e60158095bd86c47bbe0ee1ae/py/torch_tensorrt/dynamo/lowering/passes/fuse_prims_broadcast.py#L22 Results in the following error on HF models: ```python File "/root/.pyenv/versions/3.10.14/lib/python3.10/site-packages/torch/_refs/__init__.py", line 3714, in _reshape_view_helper raise ValueError(msg) ValueError: Cannot view a tensor with shape torch.Size([256, 9, 32, 128])...

bug
hf

## Bug Description Currently, Torch-TRT displays the following error when compiling Llama-2-7B in FP16 on a 4090 GPU: ``` [05/08/2024-20:47:56] [TRT] [E] 1: [defaultAllocator.cpp::allocate::19] Error Code 1: Cuda Runtime (out...

bug
hf