8000 [Intel GPU][XPU] Slow DDP training using oneCCL backend · Issue #153438 · pytorch/pytorch · GitHub
[go: up one dir, main page]
Skip to content

[Intel GPU][XPU] Slow DDP training using oneCCL backend #153438

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
yozhijk opened this issue May 13, 2025 · 4 comments
Open

[Intel GPU][XPU] Slow DDP training using oneCCL backend #153438

yozhijk opened this issue May 13, 2025 · 4 comments
Assignees
@gujinghui
Labels
module: xpu Intel XPU related issues triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module

Comments

@yozhijk
Copy link
yozhijk commented May 13, 2025 •
edited by pytorch-bot bot
Loading

Summary

While training a model on a multi-GPU system (8 × Intel® Data Center GPU Max 1550), epoch performance is strong, but each epoch is preceded by excessive logging outputs in the following format (full log is attached):

[W513 00:22:03.632630396 OperatorEntry.cpp:154] Warning: Warning only once for all operators,  other operators may also be overridden.
  Overriding a previously registered kernel for the same operator and the same dispatch key
  operator: aten::geometric_(Tensor(a!) self, float p, *, Generator? generator=None) -> Tensor(a!)
    registered at /pytorch/build/aten/src/ATen/RegisterSchema.cpp:6
  dispatch key: XPU
  previous kernel: registered at /pytorch/aten/src/ATen/VmapModeRegistrations.cpp:37
       new kernel: registered at /build/intel-pytorch-extension/build/Release/csrc/gpu/csrc/gpu/xpu/ATen/RegisterXPU_0.cpp:186 (function operator())

These logging operations introduce significant overhead (sometimes 2-3 minutes), creating a training bottleneck when epoch durations are short.

xpu-smi discovery -d 0 output

Device ID Device Information Details
0 Device Type GPU
Device Name Intel(R) Data Center GPU Max 1550
PCI Device ID 0xbd5
Vendor Name Intel(R) Corporation
SOC UUID 00000000-0000-001a-0000-002f0bd58086
Serial Number unknown
Core Clock Rate 1600 MHz
Stepping B4
SKU Type N/A
Driver Version I915_25.1.17_PSB_250113.16
Driver Package Version 1.25.1.17.250113.16+i20-1
Kernel Version 5.15.0-138-generic
GFX Firmware Name GFX
GFX Firmware Version unknown
GFX Firmware Status normal
PCI BDF Address 0000:1a:00.0
PCI Slot N/A
PCIe Generation -1
PCIe Max Link Width -1
OAM Socket ID 0x1
Memory Physical Size 131072.00 MiB
Max Mem Alloc Size 124488.00 MiB
ECC State enabled
Number of Memory Channels 32
Memory Bus Width 128
Max Hardware Contexts 65536
Max Command Queue Priority 0
Number of EUs 1024
Number of Tiles 2
Number of Slices 2
Number of Sub Slices per Slice 64
Number of Threads per EU 8
Physical EU SIMD Width 16
Number of Media Engines 0
Number of Media Enhancement Engines 0
Number of Xe Link ports N/A
Max Tx/Rx Speed per Xe Link port N/A
Number of Lanes per Xe Link port N/A
Xe Link Calibration Date Not Calibrated

Versions

PyTorch version: N/A
Is debug build: N/A
CUDA used to build PyTorch: N/A
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.5 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect
CMake version: Could not collect
Libc version: glibc-2.35

Python version: 3.12.8 | packaged by conda-forge | (main, Dec 5 2024, 14:24:40) [GCC 13.3.0] (64-bit runtime)
Python platform: Linux-5.15.0-138-generic-x86_64-with-glibc2.35
Is CUDA available: N/A
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration: Could not collect
Nvidia driver version: Could not collect
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: N/A

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 52 bits physical, 57 bits virtual
Byte Order: Little Endian
CPU(s): 192
On-line CPU(s) list: 0-191
Vendor ID: GenuineIntel
Model name: Intel(R) Xeon(R) Platinum 8468V
CPU family: 6
Model: 143
Thread(s) per core: 2
Core(s) per socket: 48
Socket(s): 2
Stepping: 8
Frequency boost: enabled
CPU max MHz: 2401.0000
CPU min MHz: 800.0000
BogoMIPS: 4800.00
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cat_l2 cdp_l3 invpcid_single intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities
L1d cache: 4.5 MiB (96 instances)
L1i cache: 3 MiB (96 instances)
L2 cache: 192 MiB (96 instances)
L3 cache: 195 MiB (2 instances)
NUMA node(s): 2
NUMA node0 CPU(s): 0-47,96-143
NUMA node1 CPU(s): 48-95,144-191
Vulnerability Gather data sampling: Not affected
Vulnerability Itlb multihit: Not affected
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Mmio stale data: Not affected
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed: Not affected
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected

Versions of relevant libraries:
[pip3] numpy==2.1.2
[pip3] pytorch-triton-xpu==3.3.0
[pip3] torchaudio==2.7.0+xpu
[pip3] torchvision==0.22.0+xpu
[conda] No relevant packages

cc @gujinghui @EikanWang @fengyuan14 @guangyey
@yozhijk
Copy link
Author
8000
yozhijk commented May 13, 2025

train_output.log

@yozhijk yozhijk changed the title [Intel GPU][XPU] Slow DDP using oneCCL backend [Intel GPU][XPU] Slow DDP training using oneCCL backend May 13, 2025
@colesbury colesbury added triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module module: xpu Intel XPU related issues labels May 13, 2025
@EikanWang
Copy link
Collaborator

@yozhijk , The redundant logs are due to the Intel Extension for PyTorch(IPEX) overriding the aten::geometric_ of the stock PyTorch. The IPEX needs to remove this kernel from its repo. cc @gujinghui , @fengyuan14 ,

@EikanWang
Copy link
Collaborator

@zhangxiaoli73 FYI

@ZhaoqiongZ
Copy link
Contributor

Hi @yozhijk , seems you are using IPEX, could you try latest nightly PyTorch XPU wheels only, it is already has XCCL enabled

pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/xpu

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@gujinghui gujinghui

Labels
module: xpu Intel XPU related issues triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module
Projects
PyTorch Intel
Status: No status
Milestone
No milestone
Development

No branches or pull requests
5 participants
@colesbury @yozhijk @gujinghui @EikanWang @ZhaoqiongZ
0