torch.utils.cpp_extension.load recompiling every time

[lahavlipson]

🐛 Describe the bug

When I run torch.utils.cpp_extension.load, it will recompile every single time, regardless of whether or not the source files actually changed. For example, when I run

myext = torch.utils.cpp_extension.load(
    name='myext',
    sources='./images2triplanes.cu',
    build_directory='build/',
    with_cuda=True,
    extra_cuda_cflags=['-O3'],
    keep_intermediates=True,
    verbose=True
)

The stdout will print Emitting ninja build file build/build.ninja... every single time, even though a build/myext.so file is already present and I did not touch the source files. In principle, load(...) should not re-compile the files if there is no change to the source code and the .so is already present.

The likely cause is that the JIT_EXTENSION_VERSIONER object which stores the existing version numbers for pytorch extensions is just a local variable which obviously does not persist between runs. As a result this if-statement will always be triggered on the first cpp_extension.load(...) call because old_version will be None

I observed this issue in Pytorch nightly, Pytorch 2.2, and 2.1

Versions

Collecting environment information... PyTorch version: 2.4.0.dev20240418 Is debug build: False CUDA used to build PyTorch: 12.1 ROCM used to build PyTorch: N/A

OS: Pop!_OS 22.04 LTS (x86_64) GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 Clang version: Could not collect CMake version: version 3.28.3 Libc version: glibc-2.35

Python version: 3.10.14 (main, Mar 21 2024, 16:24:04) [GCC 11.2.0] (64-bit runtime) Python platform: Linux-6.6.10-76060610-generic-x86_64-with-glibc2.35 Is CUDA available: True CUDA runtime version: 12.1.66 CUDA_MODULE_LOADING set to: LAZY GPU models and configuration: GPU 0: NVIDIA GeForce RTX 3090 Nvidia driver version: 545.29.06 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: 46 bits physical, 48 bits virtual Byte Order: Little Endian CPU(s): 20 On-line CPU(s) list: 0-19 Vendor ID: GenuineIntel Model name: 12th Gen Intel(R) Core(TM) i7-12700K CPU family: 6 Model: 151 Thread(s) per core: 2 Core(s) per socket: 12 Socket(s): 1 Stepping: 2 CPU max MHz: 5000.0000 CPU min MHz: 800.0000 BogoMIPS: 7219.20 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 vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l2 cdp_l2 ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdt_a rdseed adx smap clflushopt clwb intel_pt sha_ni xsaveopt xsavec xgetbv1 xsaves split_lock_detect user_shstk avx_vnni dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req hfi vnmi umip pku ospke waitpkg gfni vaes vpclmulqdq tme rdpid movdiri movdir64b fsrm md_clear serialize pconfig arch_lbr ibt flush_l1d arch_capabilities Virtualization: VT-x L1d cache: 512 KiB (12 instances) L1i cache: 512 KiB (12 instances) L2 cache: 12 MiB (9 instances) L3 cache: 25 MiB (1 instance) NUMA node(s): 1 NUMA node0 CPU(s): 0-19 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 Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected 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; Enhanced / Automatic IBRS, IBPB conditional, RSB filling, PBRSB-eIBRS SW sequence Vulnerability Srbds: Not affected Vulnerability Tsx async abort: Not affected

Versions of relevant libraries: [pip3] numpy==1.26.4 [pip3] torch==2.4.0.dev20240418 [pip3] torchaudio==2.2.0.dev20240418 [pip3] torchvision==0.19.0.dev20240418 [pip3] triton==3.0.0 [conda] blas 1.0 mkl
[conda] libjpeg-turbo 2.0.0 h9bf148f_0 pytorch-nightly [conda] mkl 2023.1.0 h213fc3f_46344
[conda] mkl-service 2.4.0 py310h5eee18b_1
[conda] mkl_fft 1.3.8 py310h5eee18b_0
[conda] mkl_random 1.2.4 py310hdb19cb5_0
[conda] numpy 1.26.4 py310h5f9d8c6_0
[conda] numpy-base 1.26.4 py310hb5e798b_0
[conda] pytorch 2.4.0.dev20240418 py3.10_cuda12.1_cudnn8.9.2_0 pytorch-nightly [conda] pytorch-cuda 12.1 ha16c6d3_5 pytorch-nightly [conda] pytorch-mutex 1.0 cuda pytorch-nightly [conda] torchaudio 2.2.0.dev20240418 py310_cu121 pytorch-nightly [conda] torchtriton 3.0.0+989adb9a29 py310 pytorch-nightly [conda] torchvision 0.19.0.dev20240418 py310_cu121 pytorch-nightly

cc @malfet @zou3519

[vadimkantorov]

Maybe related:

• https://github.com/pytorch/pytorch/issues/68066#issuecomment-1610211293

[bdhirsh]

cc @drisspg @zou3519 maybe, for cpp extensions modernization?

[LamForest]

I had the same problem. This is quite anonying, I have to wast ~5 minutes for each run.

My workaround :

[HGGshiwo]

use this:

from torch.utils.cpp_extension import load, _import_module_from_library

def _load(name, sources):
    module_path = os.path.dirname(__file__)
    build_directory = os.path.join(module_path, "build", name)
    os.makedirs(build_directory, exist_ok=True)

    try:
        module = _import_module_from_library(name, build_directory, True)
    except ImportError:
        sources = [os.path.join(module_path, source) for source in sources]  
        module = load(
            name,
            sources,
            build_directory=build_directory,
            verbose=False,
            with_cuda=True,
            extra_cflags=["-O3"],
        )
    return module