TensorRT

TensorRT copied to clipboard

## Bug Description Module triggers a runtime error at inference time when compiled with dynamo and dynamic batch sizes. Error appears to be related to using slice-based assignment (minimum example...

cpjenkins

## Bug Description when I test torch_tensorrt.compile for a simple model (convs + layernorm + zeropad), I got an error. This bug occurs, especially when nn.ZeroPad2d() and nn.LayerNorm() are used...

minimok7

🐛 [Bug] compiled model gives different outputs from torch model (used to work on torch_tensorrt 2.2.0)

1

## Bug Description my model outputs a tuple of `mu` and `logvar`. for the `mu`, there are 4 columns (features), consisting of 3 features of type A and 1 feature...

orioninthesky98

## Bug Description ``` DEBUG:torch_tensorrt.dynamo.conversion._TRTInterpreter:Adding input to in-progress INetwork: arg0_1 [shape=[10], dtype=DataType.FLOAT] DEBUG:torch_tensorrt.dynamo.conversion._TRTInterpreter:Converting node __/reshape_default (kind: aten.reshape.default, args: ('arg0_1 ', [10, 1])) DEBUG:torch_tensorrt.dynamo.conversion._TRTInterpreter:Converting node __/mul (kind: aten.mul.Tensor, args: ('[SHUFFLE]-[aten_ops.reshape.default]-[__/reshape_default]_output ',...

tchatow

There's a copy-paste error: https://github.com/pytorch/TensorRT/blob/2bba527213270aca7290baadefaa562f12b84b2c/py/torch_tensorrt/dynamo/lowering/passes/_aten_lowering_pass.py#L102

zewenli98

# Description The current version of the code example in the README.md (Option 2: Export, Step 1) results in an error: ``` SyntaxError: positional argument follows keyword argument ``` ##...

juliusgh

## TL;DR - This story focuses on identifying and resolving errors or issues that arise during the Torch Compile (Torch-TRT) process for key AI models, addressing both model-specific and common...

chohk88

🐛 [Bug] [Dynamic Shapes] Encountered bug when using Torch-TensorRT

84

## Bug Description when I use dynamic shape in trt, will raise error, ``` ERROR:torch_tensorrt [TensorRT Conversion Context]:ITensor::getDimensions: Error Code 4: Internal Error (Tensor [SLICE]-[aten_ops.expand.default]-[__/expand]_output has axis 0 with inherently...

yjjinjie

chore: Adapt CIA ops decomposition handling in upsample converters to torch 2.6

1

# Description Fixes #3226 ## Type of change - Bug fix (non-breaking change which fixes an issue) # Checklist: - [x] My code follows the style guidelines of this project...

HolyWu

skip dummy inference and run_shape_analysis

1

# Description There is two changes introduced in this PR: 1) during the compile stage: skipped dummy inference and use graph inspection instead to get the output_node.meta['val'] 2) during the...

lanluo-nvidia