aws-neuron/aws-neuron-sdk issues and pull requests

#676 - Question on TRN Randomness

Issue - State: open - Opened by xanderdunn over 1 year ago - 5 comments
Labels: bug, Trn1, pytorch

#675 - Inferentia pod with sidecar can't start with 'UnexpectedAdmissionError' error.

Issue - State: open - Opened by everpeace over 1 year ago - 4 comments
Labels: Inf1, inference, runtime

#674 - [torch-neuronx] Compiling Error from PyTorch Sagemaker Training Job on trn1

Issue - State: open - Opened by rileyhun over 1 year ago - 4 comments
Labels: bug, model, training, Trn1

#673 - SIGSEGV Calling nrt_execute

Issue - State: open - Opened by xanderdunn over 1 year ago - 8 comments
Labels: aws-neuronx-tools, runtime

#672 - Getting unexpected results when tracing the seq-to-seq model with the neuronx runtime

Issue - State: open - Opened by hhhhzy over 1 year ago - 3 comments
Labels: bug, inference, torch-neuronx, pytorch

#671 - Clarification on neuron tracing support on seq-to-seq model

Issue - State: closed - Opened by ndenStanford over 1 year ago - 1 comment
Labels: documentation

#670 - SIGSEGV Calling nrt_execute

Issue - State: closed - Opened by xanderdunn over 1 year ago - 5 comments

#669 - Trace TF model inside a Google Colab environment

Issue - State: closed - Opened by jc-louis over 1 year ago - 3 comments

#668 - Minor doc updates for 2.10.0 Release Notes

Pull Request - State: closed - Opened by awsjoshir over 1 year ago

#667 - Training accuracy issue when using `bf16` for multi-class classification with > 10 labels on Trainium

Issue - State: open - Opened by philschmid over 1 year ago - 3 comments
Labels: training, Trn1, huggingface

#666 - Pull request 2.10.0

Pull Request - State: closed - Opened by awsjoshir over 1 year ago

#665 - penguin.py segfaults compiling Transformer XLA HLO .pb

Issue - State: open - Opened by xanderdunn over 1 year ago - 4 comments
Labels: bug, Trn1, compiler, neuronx-cc

#664 - neuronx-cc Compiler Internal Error on Transfomer XLA HLO .pb

Issue - State: closed - Opened by xanderdunn over 1 year ago - 4 comments

#663 - neuronx-cc Compilation Time is a Function of Input Size / (16000, 16000) Square Matrices Fail Compilation

Issue - State: open - Opened by xanderdunn over 1 year ago - 4 comments
Labels: training, compiler, neuronx-cc

#662 - Error compiling PyTorch model

Issue - State: open - Opened by andreadasilvabaudet over 1 year ago - 2 comments
Labels: bug, Inf1, inference, compiler, neuron-cc

#661 - Error while installing aws-neuron-dkms 2.x

Issue - State: closed - Opened by Mihir-Gajera1 over 1 year ago - 6 comments

#660 - Neuron Runtime compatibility issue when loading a compiled model

Issue - State: closed - Opened by g-laz77 over 1 year ago - 7 comments

#659 - neuronx-cc compile crashes on `tril(x)`

Issue - State: open - Opened by xanderdunn over 1 year ago - 2 comments
Labels: bug, compiler, neuronx-cc

#658 - Causal Language Model from Huggingface does not compile

Issue - State: open - Opened by junoriosity over 1 year ago - 4 comments
Labels: bug, inference, neuronx-cc, pytorch, neuron-cc

#657 - Cannot compile MT5 model

Issue - State: open - Opened by yzGao22 over 1 year ago - 1 comment
Labels: bug, inference, torch-neuronx

#656 - Cannot compile a mixed model with Transformers and normal Pytorch content

Issue - State: open - Opened by junoriosity over 1 year ago - 2 comments
Labels: Inf1, inference, compiler

#655 - Compatibility issues on DLAMI with aws-neuron-dkms

Issue - State: open - Opened by junoriosity over 1 year ago - 6 comments
Labels: good first issue

#654 - Add cxx11 ABI compatibility to troubleshooting docs

Issue - State: open - Opened by jsleight over 1 year ago - 2 comments
Labels: documentation

#653 - How to compile seq2seq model using torch_neuronx

Issue - State: open - Opened by yzGao22 over 1 year ago - 5 comments
Labels: inference, torch-neuronx, pytorch

#652 - How to run neuron model in ECS using Inf1 EC2 instances

Issue - State: closed - Opened by YuryShchanouskiTR over 1 year ago - 12 comments

#651 - Inconsistent output with EfficientNet in TF 2.x (inf1)

Issue - State: closed - Opened by Askannz over 1 year ago - 3 comments
Labels: bug

#650 - Inconsistent results between Torch/JIT and Neuron

Issue - State: open - Opened by RobinFrcd over 1 year ago - 7 comments
Labels: inference, neuron-cc

#649 - Unable to Add Neuron Apt Repo on Ubuntu 22.04

Issue - State: open - Opened by xanderdunn over 1 year ago - 3 comments
Labels: runtime

#648 - Unused memory on Inf1

Issue - State: closed - Opened by RobinFrcd over 1 year ago - 4 comments

#647 - Issue installing aws-neuronx-dkms on Ubuntu 22.04

Issue - State: closed - Opened by IsaacRodgzb over 1 year ago - 3 comments
Labels: runtime

#646 - neuronx-cc compilation incredibly slow due to only using single-thread/core

Issue - State: closed - Opened by DanielRWhite over 1 year ago - 7 comments

#645 - [Hugging Face] neuron-cc fails on tracing Distilbert model on multiple-choice task on INF1

Issue - State: closed - Opened by JingyaHuang over 1 year ago - 2 comments
Labels: huggingface

#644 - [Optimum Neuron] Pegasus-X compilation fails

Issue - State: open - Opened by michaelbenayoun over 1 year ago - 3 comments
Labels: bug, compiler, huggingface

#643 - [Optimum Neuron] Compilation error when using a label smoothing factor

Issue - State: open - Opened by michaelbenayoun over 1 year ago - 5 comments
Labels: huggingface

#642 - [Hugging Face] neuron compiler fails on tracing DeBERTa v1 and v2 models on INF1

Issue - State: closed - Opened by JingyaHuang over 1 year ago - 3 comments
Labels: huggingface

#641 - [Hugging Face] neuronx compiler unusual behaviors on ConvBERT / XLM / FlauBERT inference

Issue - State: closed - Opened by JingyaHuang over 1 year ago - 9 comments
Labels: huggingface

#640 - Torch.neuron.trace() -> c10::Error

Issue - State: closed - Opened by sammystevens1983 over 1 year ago - 1 comment

#639 - Internal Compiler Error when compiling GPT2

Issue - State: closed - Opened by gnawpaul over 1 year ago - 4 comments

#638 - Inferentia outperforms GPU on ResNet while showing worse performance on MobileNet

Issue - State: closed - Opened by pEAceATLast53 over 1 year ago - 2 comments

#637 - Issue on page /frameworks/torch/torch-neuronx/tutorials/training/bert.html

Issue - State: closed - Opened by rsindreu over 1 year ago - 1 comment

#636 - Update torch module version to 1.13.1

Issue - State: closed - Opened by YuryShchanouskiTR over 1 year ago - 4 comments
Labels: torch-neuron, torch-neuronx, pytorch

#635 - `neuron_parallel_compile` fails but the original command line works

Issue - State: open - Opened by michaelbenayoun over 1 year ago - 4 comments
Labels: training, Trn1, huggingface

#634 - Custom model using torch.flip does compile, but cannot infer

Issue - State: closed - Opened by BayMinimum over 1 year ago - 3 comments

#633 - Getting no neuron devices available on an inf1 instance using aws provided containers

Issue - State: closed - Opened by sanjay23singh over 1 year ago - 7 comments
Labels: Inf1, runtime

#632 - neuronx-cc fails during fine-tuning attempt for pre-trained microsoft/layoutlm-base-uncased when using torchrun

Issue - State: open - Opened by vprecup over 1 year ago - 7 comments
Labels: bug, training, Trn1

#631 - Unable to reproduce seq2seq example from docs

Issue - State: closed - Opened by mathcass over 1 year ago - 5 comments
Labels: Inf1, inference

#630 - Unable to trace swinv2 on inferentia

Issue - State: open - Opened by Varghese-Kuruvilla over 1 year ago - 8 comments
Labels: bug, Inf1, inference

#629 - Bump ipython from 7.26.0 to 8.10.0

Pull Request - State: open - Opened by dependabot[bot] over 1 year ago
Labels: dependencies

#628 - Failed converting wav2vec2 model

Issue - State: closed - Opened by piuy11 almost 2 years ago - 4 comments
Labels: bug, Inf1, inference, compiler

#627 - Update bert.rst

Pull Request - State: closed - Opened by jyang-aws almost 2 years ago

#626 - [torch-neuron] Is Swin (hugging face) supported in inferentia?

Issue - State: closed - Opened by heylamourding almost 2 years ago - 12 comments
Labels: bug, Inf1, torch-neuron, compiler

#625 - torch.jit.load sometimes fails due to an allocation error

Issue - State: closed - Opened by DHdroid almost 2 years ago - 10 comments
Labels: Inf1, torch-neuron, runtime

#624 - Is it possible to limit neuron memory usage per neuronCore or model ?

Issue - State: closed - Opened by paaksing almost 2 years ago - 12 comments

#623 - Internal Compiler Error when compiling a BERT-based model with neuron-cc and TF 2.8

Issue - State: closed - Opened by florianlaws almost 2 years ago - 10 comments
Labels: Inf1, compiler, tensorflow, tensorflow-neuron

#622 - A model compiled with --extract-weights option is not using NeuronCores

Issue - State: open - Opened by DHdroid almost 2 years ago - 4 comments
Labels: bug, Inf1, tensorflow-neuron

#621 - How to know "OnNeuronRatio" in TF2

Issue - State: closed - Opened by workdd almost 2 years ago - 5 comments
Labels: enhancement, Inf1, tensorflow-neuron

#620 - torch.neuron.DataParallel returns incomplete result for model with dict input when split_size is less than default

Issue - State: closed - Opened by BayMinimum almost 2 years ago - 3 comments
Labels: bug, Inf1, torch-neuron

#619 - No operations were successfully partitioned and compiled to neuron for this model

Issue - State: closed - Opened by RobinFrcd almost 2 years ago - 7 comments
Labels: bug, Inf1, compiler, pytorch

#618 - The requested number of neuroncore-pipeline-cores (4) may not be suitable for this network

Issue - State: closed - Opened by jestiny0 almost 2 years ago - 10 comments
Labels: bug, Inf1, compiler

#617 - Improved docker support

Issue - State: closed - Opened by Limess almost 2 years ago - 8 comments
Labels: runtime

#616 - Segmentation fault(core dump) while running inference

Issue - State: closed - Opened by jestiny0 almost 2 years ago - 3 comments

#615 - Doing diffusers on Inferentia

Issue - State: closed - Opened by vrobot almost 2 years ago - 7 comments
Labels: enhancement, Inf1, torch-neuron

#614 - Update nlp_data.csv

Pull Request - State: closed - Opened by aws-trsharma almost 2 years ago

#612 - outputs are all "nan" when compiled model loaded from checkpoint

Issue - State: closed - Opened by jestiny0 almost 2 years ago - 3 comments

#611 - Different input batch sizes in compile cause different outputs

Issue - State: closed - Opened by ukus04 almost 2 years ago - 3 comments

#610 - TorchScript model - NotImplementedError

Issue - State: closed - Opened by altansnl almost 2 years ago - 3 comments
Labels: Inf1, inference

#608 - Integration with Neo-DLR

Issue - State: closed - Opened by michaelhagel almost 2 years ago - 2 comments
Labels: runtime

#607 - HAL:aws_hal_tpb_pooling_write_profile failed programming the engine

Issue - State: closed - Opened by paaksing almost 2 years ago - 10 comments
Labels: runtime

#606 - [torch-neuron] Shape must be rank 2 but is rank 1 for 'Linear_59/aten_linear/MatMul'

Issue - State: closed - Opened by Saief1999 almost 2 years ago - 3 comments
Labels: bug, Inf1, torch-neuron, pytorch

#604 - Significant rounding error in FP32 matrix multiplication

Issue - State: closed - Opened by xuanqing94 almost 2 years ago - 4 comments
Labels: bug, Trn1, torch-neuronx, compiler

#600 - BIR verification failed - Access pattern did not start at parition 0 or 64. Starts at partition 48

Issue - State: closed - Opened by DanCorvesor almost 2 years ago - 3 comments
Labels: bug, Inf1, compiler, pytorch

#599 - Update neuroncores-arch.rst

Pull Request - State: closed - Opened by radbarros almost 2 years ago - 1 comment

#597 - Yolov7-Pose Only a few ops to be compatible

Issue - State: closed - Opened by josebenitezg almost 2 years ago - 3 comments
Labels: enhancement, Inf1, torch-neuron

#593 - NaNs seen with transformers version >= 4.21.0 when running HF BERT fine-tuning with XLA_USE_BF16=1

Issue - State: closed - Opened by jeffhataws almost 2 years ago - 1 comment
Labels: bug, Trn1, torch-neuronx

#589 - Issue on page /frameworks/torch/torch-neuron/tutorials/tutorial-libtorch.html

Issue - State: closed - Opened by mpetri almost 2 years ago - 2 comments
Labels: documentation, Inf1

#587 - Fix typo :)

Pull Request - State: closed - Opened by julien-c about 2 years ago

#578 - Move mxnet tutorial file

Pull Request - State: closed - Opened by aws-donc about 2 years ago

#575 - Bump ipython from 7.26.0 to 7.31.1

Pull Request - State: closed - Opened by dependabot[bot] about 2 years ago - 1 comment
Labels: dependencies

#574 - [torch-neuron] Vision Transformers Models - Training support on Inf1

Issue - State: open - Opened by aws-rxgupta about 2 years ago
Labels: Inf1, inference, torch-neuron, models, pytorch

#571 - [torch-neuronx] LAMB optimizer support for training on Trn1

Issue - State: closed - Opened by aws-rxgupta about 2 years ago - 1 comment
Labels: training, Trn1, torch-neuronx, pytorch

#524 - [tensorboard-plugin-neuronx] torch-neuronx Profiling support for Training on Trn1

Issue - State: closed - Opened by aws-rxgupta about 2 years ago - 1 comment
Labels: training, Trn1, torch-neuronx, pytorch, tools

#504 - [tensorflow-neuronx] Training support on Trn1

Issue - State: open - Opened by aws-rxgupta about 2 years ago - 1 comment
Labels: training, Trn1, tensorflow, tensorflow-neuronx

#502 - [torch-neuronx] FSDP support - Distributed Training on Trn1

Issue - State: open - Opened by aws-rxgupta about 2 years ago - 3 comments
Labels: training, Trn1, torch-neuronx, distributed, pytorch

#495 - Unable to compile roberta-base from huggingface using torch SDK

Issue - State: closed - Opened by parakalan about 2 years ago - 5 comments
Labels: Inf1, compiler

#494 - [torch-neuron] Large Graph Support - Remove/Mitigate protobuf limitation on Inf1

Issue - State: closed - Opened by aws-maens about 2 years ago - 1 comment
Labels: Inf1, inference, torch-neuron, pytorch

#492 - Can't convert stable diffusion model

Issue - State: closed - Opened by Dong-Ki-Lee about 2 years ago - 15 comments
Labels: enhancement, Inf1, torch-neuron

#475 - [torch-neuron] aten::_convolution_mode operator support on Inf1

Issue - State: closed - Opened by aws-maens about 2 years ago - 1 comment
Labels: Inf1, inference, torch-neuron, pytorch

#474 - [torch-neuron] Solve NaN issue when using transformers>=4.20

Issue - State: closed - Opened by aws-maens about 2 years ago - 1 comment
Labels: Inf1, inference, torch-neuron, pytorch

#463 - Dynamic batching is also supported in Pytorch Neuron

Pull Request - State: closed - Opened by hadilou over 2 years ago - 1 comment

#445 - Correct a typo of the inf1 instance type

Pull Request - State: closed - Opened by Joldnine over 2 years ago - 1 comment

#443 - Template for Inferentia T5

Issue - State: closed - Opened by ierezell over 2 years ago - 4 comments

#442 - Compilation error for 🤗 Detr model: TVMError: Check failed: pb->value != 0 (0 vs. 0) : Divide by zero

Issue - State: closed - Opened by cotrane over 2 years ago - 4 comments
Labels: bug, Inf1, compiler, pytorch

#437 - YoloR neuron model has different accuracy compared to pytorch model

Issue - State: closed - Opened by alejoGT1202 over 2 years ago - 15 comments
Labels: Inf1, torch-neuron

#428 - [torch-neuron] Optimization - Reduce memory used by models

Issue - State: closed - Opened by aws-maens over 2 years ago - 1 comment
Labels: Inf1, inference, torch-neuron, pytorch

#419 - SwinIR Neuron Compilation Failing for PyTorch

Issue - State: closed - Opened by guptajayesh over 2 years ago - 8 comments
Labels: Inf1, compiler, pytorch

#406 - [torch-neuron] torch.nn.InstanceNorm2d operator support on Inf1

Issue - State: closed - Opened by aws-maens over 2 years ago - 1 comment
Labels: Inf1, inference

#307 - [torch-neuron] RegNet model inference support on Inf1

Issue - State: open - Opened by aws-maens about 3 years ago - 1 comment
Labels: Inf1, inference, torch-neuron, models, pytorch

#306 - [torch-neuron] Wav2Vec2 model inference support on Inf1

Issue - State: closed - Opened by aws-maens about 3 years ago - 1 comment
Labels: Inf1, inference, torch-neuron, models, pytorch

#211 - [torch-neuron] GPT-2 model inference support on Inf1

Issue - State: open - Opened by AWSGH almost 4 years ago - 9 comments
Labels: enhancement, Inf1, inference, torch-neuron, models, pytorch

#107 - Loading ONNX neuron-compiled models

Issue - State: closed - Opened by RobertLucian over 4 years ago - 7 comments

GitHub / aws-neuron/aws-neuron-sdk issues and pull requests