Torch Profiler with_stack=True, 'utf-8' codec can't decode

[anthonytec2]

🐛 Describe the bug

I have been trying to use the pytorch profiler recently, under both the tensorboard profiler extensions analysis backend I received an error: UnicodeDecodeError: 'utf-8' codec can't decode byte 0xf0 in position 38423127: invalid continuation byte. I have tried this multiple times and receive the same error, and under the Holistic trace analysis backend: UnicodeDecodeError: 'utf-8' codec can't decode byte 0xf0 in position 38423161: invalid continuation byte. Think this could be related to: #64345, so I tried a nightly build which should in theory have the patch, although the issue persists.

Versions

Collecting environment information...
PyTorch version: 2.2.0.dev20231112+cu118
Is debug build: False
CUDA used to build PyTorch: 11.8
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.3 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect
CMake version: version 3.24.1
Libc version: glibc-2.35

Python version: 3.10.13 | packaged by conda-forge | (main, Oct 26 2023, 18:07:37) [GCC 12.3.0] (64-bit runtime)
Python platform: Linux-6.0.12-76060012-generic-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: 11.7.99
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration:
GPU 0: NVIDIA RTX A5000
GPU 1: NVIDIA GeForce RTX 4090

Nvidia driver version: 535.113.01
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 43 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 48
On-line CPU(s) list: 0-47
Vendor ID: AuthenticAMD
Model name: AMD Ryzen Threadripper 3960X 24-Core Processor
CPU family: 23
Model: 49
Thread(s) per core: 2
Core(s) per socket: 24
Socket(s): 1
Stepping: 0
Frequency boost: enabled
CPU max MHz: 3800.0000
CPU min MHz: 2200.0000
BogoMIPS: 7585.73
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif v_spec_ctrl umip rdpid overflow_recov succor smca sev sev_es
Virtualization: AMD-V
L1d cache: 768 KiB (24 instances)
L1i cache: 768 KiB (24 instances)
L2 cache: 12 MiB (24 instances)
L3 cache: 128 MiB (8 instances)
NUMA node(s): 1
NUMA node0 CPU(s): 0-47
Vulnerability Itlb multihit: Not affected
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Mmio stale data: Not affected
Vulnerability Retbleed: Mitigation; untrained return thunk; SMT enabled with STIBP protection
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; Retpolines, IBPB conditional, STIBP always-on, RSB filling, PBRSB-eIBRS Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected

Versions of relevant libraries:
[pip3] backpack-for-pytorch==1.6.0
[pip3] numpy==1.24.1
[pip3] pytorch-lightning==2.1.0
[pip3] pytorch-triton==2.1.0+6e4932cda8
[pip3] torch==2.2.0.dev20231112+cu118
[pip3] torch-tb-profiler==0.4.3
[pip3] torchaudio==2.2.0.dev20231112+cu118
[pip3] torchmetrics==1.2.0
[pip3] torchvision==0.17.0.dev20231112+cu118
[pip3] triton==2.1.0
[conda] backpack-for-pytorch 1.6.0 pypi_0 pypi
[conda] blas 2.16 mkl conda-forge
[conda] ffmpeg 4.3 hf484d3e_0 pytorch
[conda] libblas 3.8.0 16_mkl conda-forge
[conda] libcblas 3.8.0 16_mkl conda-forge
[conda] libjpeg-turbo 2.0.0 h9bf148f_0 pytorch
[conda] liblapack 3.8.0 16_mkl conda-forge
[conda] liblapacke 3.8.0 16_mkl conda-forge
[conda] mkl 2020.2 256
[conda] numpy 1.24.1 pypi_0 pypi
[conda] pytorch-cuda 11.8 h7e8668a_5 pytorch
[conda] pytorch-lightning 2.1.0 pypi_0 pypi
[conda] pytorch-mutex 1.0 cuda pytorch
[conda] pytorch-triton 2.1.0+6e4932cda8 pypi_0 pypi
[conda] torch 2.2.0.dev20231112+cu118 pypi_0 pypi
[conda] torch-tb-profiler 0.4.3 pypi_0 pypi
[conda] torchaudio 2.2.0.dev20231112+cu118 pypi_0 pypi
[conda] torchmetrics 1.2.0 pypi_0 pypi
[conda] torchtriton 2.1.0 py310 pytorch
[conda] torchvision 0.17.0.dev20231112+cu118 pypi_0 pypi

[anthonytec2]

Problem appears to be fixed when downgrading to python3.8.

[agkphysics]

@anthonytec2 I think this issue should be reopened because the underlying issue is still there. I am encountering the same issue where the exported JSON trace file is sometimes invalid due to function names being corrupted as per #64345. I don't think downgrading the Python version is an acceptable solution.