dynamo (re)compilation issues: shape (1,1), nn.Parameter, mark_dynamic

🐛 Describe the bug

There are following issues:

1. passing Parameter as argument instead of Tensor causes recompilation
2. setting torch._dynamo.mark_dynamic on Parameter has no effect
3. passing tensor of size (1,1) causes recompilation with a strange guard failure - 2 <= L['x'].size()[0]
4. using torch._dynamo.mark_dynamic(x, 0, min=0) causes error not simple sympy type <class 'NoneType'>
5. using torch._dynamo.mark_dynamic(x, 0, min=0, max=65536) causes error CONSTRAINTS_VIOLATED

I've checked on 2.4.0 and on nightly

import torch
import torch.nn as nn
from torch import Tensor

import logging
torch._logging.set_logs(recompiles=True, recompiles_verbose = True)


def f(x):
    return x.sin() + 1.0

# def g(x):
#     return x**2

compile_args = dict(fullgraph=True, dynamic = True, backend = "inductor", options={'force_same_precision':True, 'disable_cpp_codegen':False, 'trace.graph_diagram':True, "triton.cudagraphs": False})

@torch.compile(**compile_args)
def foo(x):
    y = f(x)
    # z = g(y)
    return y

shapes = [(2,2), (3,3), (1,1)]

for P in [False, True]:
    for k, shape in enumerate(shapes):
        print(f'Shape {k}: {shape}')
        x = torch.ones(shape)
        if P:
            x = torch.nn.Parameter(x)
        for d in range(x.dim()):
            torch._dynamo.mark_dynamic(x, d) # setting min=0 errors
        z = foo(x)

Shape 0: (2, 2)
Shape 1: (3, 3)
Shape 2: (1, 1)
V0903 12:36:59.429000 140550974269248 torch/_dynamo/guards.py:2609] [0/1] [__recompiles_verbose] Recompiling function foo in /mnt/datagrid/personal/shekhovt/quant/debug/test_compile1.py:17
V0903 12:36:59.429000 140550974269248 torch/_dynamo/guards.py:2609] [0/1] [__recompiles_verbose]     triggered by the following guard failure(s):
V0903 12:36:59.429000 140550974269248 torch/_dynamo/guards.py:2609] [0/1] [__recompiles_verbose]     guard 0 failures:
V0903 12:36:59.429000 140550974269248 torch/_dynamo/guards.py:2609] [0/1] [__recompiles_verbose]     - 2 <= L['x'].size()[0]                                         # _dynamo/output_graph.py:452 in init_ambient_guards
V0903 12:36:59.429000 140550974269248 torch/_dynamo/guards.py:2609] [0/1] [__recompiles_verbose]     - 2 <= L['x'].size()[1]                                         # _dynamo/output_graph.py:452 in init_ambient_guards
Shape 0: (2, 2)
V0903 12:37:02.287000 140550974269248 torch/_dynamo/guards.py:2609] [0/2] [__recompiles_verbose] Recompiling function foo in /mnt/datagrid/personal/shekhovt/quant/debug/test_compile1.py:17
V0903 12:37:02.287000 140550974269248 torch/_dynamo/guards.py:2609] [0/2] [__recompiles_verbose]     triggered by the following guard failure(s):
V0903 12:37:02.287000 140550974269248 torch/_dynamo/guards.py:2609] [0/2] [__recompiles_verbose]     guard 0 failures:
V0903 12:37:02.287000 140550974269248 torch/_dynamo/guards.py:2609] [0/2] [__recompiles_verbose]     - expected type of 'L['x']' to be a tensor type, ' but found <class 'torch.nn.parameter.Parameter'>
V0903 12:37:02.287000 140550974269248 torch/_dynamo/guards.py:2609] [0/2] [__recompiles_verbose] 
V0903 12:37:02.287000 140550974269248 torch/_dynamo/guards.py:2609] [0/2] [__recompiles_verbose]     guard 1 failures:
V0903 12:37:02.287000 140550974269248 torch/_dynamo/guards.py:2609] [0/2] [__recompiles_verbose]     - expected type of 'L['x']' to be a tensor type, ' but found <class 'torch.nn.parameter.Parameter'>
Shape 1: (3, 3)
V0903 12:37:05.634000 140550974269248 torch/_dynamo/guards.py:2609] [0/3] [__recompiles_verbose] Recompiling function foo in /mnt/datagrid/personal/shekhovt/quant/debug/test_compile1.py:17
V0903 12:37:05.634000 140550974269248 torch/_dynamo/guards.py:2609] [0/3] [__recompiles_verbose]     triggered by the following guard failure(s):
V0903 12:37:05.634000 140550974269248 torch/_dynamo/guards.py:2609] [0/3] [__recompiles_verbose]     guard 0 failures:
V0903 12:37:05.634000 140550974269248 torch/_dynamo/guards.py:2609] [0/3] [__recompiles_verbose]     - tensor 'L['x']' size mismatch at index 0. expected 2, actual 3
V0903 12:37:05.634000 140550974269248 torch/_dynamo/guards.py:2609] [0/3] [__recompiles_verbose] 
V0903 12:37:05.634000 140550974269248 torch/_dynamo/guards.py:2609] [0/3] [__recompiles_verbose]     guard 1 failures:
V0903 12:37:05.634000 140550974269248 torch/_dynamo/guards.py:2609] [0/3] [__recompiles_verbose]     - expected type of 'L['x']' to be a tensor type, ' but found <class 'torch.nn.parameter.Parameter'>
V0903 12:37:05.634000 140550974269248 torch/_dynamo/guards.py:2609] [0/3] [__recompiles_verbose] 
V0903 12:37:05.634000 140550974269248 torch/_dynamo/guards.py:2609] [0/3] [__recompiles_verbose]     guard 2 failures:
V0903 12:37:05.634000 140550974269248 torch/_dynamo/guards.py:2609] [0/3] [__recompiles_verbose]     - expected type of 'L['x']' to be a tensor type, ' but found <class 'torch.nn.parameter.Parameter'>
Shape 2: (1, 1)
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose] Recompiling function foo in /mnt/datagrid/personal/shekhovt/quant/debug/test_compile1.py:17
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose]     triggered by the following guard failure(s):
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose]     guard 0 failures:
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose]     - tensor 'L['x']' size mismatch at index 0. expected 3, actual 1
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose] 
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose]     guard 1 failures:
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose]     - tensor 'L['x']' size mismatch at index 0. expected 2, actual 1
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose] 
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose]     guard 2 failures:
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose]     - expected type of 'L['x']' to be a tensor type, ' but found <class 'torch.nn.parameter.Parameter'>
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose] 
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose]     guard 3 failures:
V0903 12:37:08.674000 140550974269248 torch/_dynamo/guards.py:2609] [0/4] [__recompiles_verbose]     - expected type of 'L['x']' to be a tensor type, ' but found <class 'torch.nn.parameter.Parameter'>

Versions

Collecting environment information... PyTorch version: 2.5.0.dev20240902+cu118 Is debug build: False CUDA used to build PyTorch: 11.8 ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.4 LTS (x86_64) GCC version: (GCC) 13.2.0 Clang version: Could not collect CMake version: version 3.22.1 Libc version: glibc-2.35

Python version: 3.11.5 (main, Oct 2 2023, 09:22:39) [GCC 13.2.0] (64-bit runtime) Python platform: Linux-5.15.0-1062-nvidia-x86_64-with-glibc2.35 Is CUDA available: True CUDA runtime version: 12.2.140 CUDA_MODULE_LOADING set to: LAZY GPU models and configuration: GPU 0: NVIDIA A100-SXM4-40GB GPU 1: NVIDIA A100-SXM4-40GB GPU 2: NVIDIA A100-SXM4-40GB GPU 3: NVIDIA DGX Display GPU 4: NVIDIA A100-SXM4-40GB

Nvidia driver version: 535.183.01 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): 128 On-line CPU(s) list: 0-127 Vendor ID: AuthenticAMD Model name: AMD EPYC 7742 64-Core Processor CPU family: 23 Model: 49 Thread(s) per core: 2 Core(s) per socket: 64 Socket(s): 1 Stepping: 0 Frequency boost: enabled CPU max MHz: 2250,0000 CPU min MHz: 1500,0000 BogoMIPS: 4491.73 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 ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 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 sme sev sev_es Virtualization: AMD-V L1d cache: 2 MiB (64 instances) L1i cache: 2 MiB (64 instances) L2 cache: 32 MiB (64 instances) L3 cache: 256 MiB (16 instances) NUMA node(s): 1 NUMA node0 CPU(s): 0-127 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: Mitigation; untrained return thunk; SMT enabled with STIBP protection Vulnerability Spec rstack overflow: Mitigation; safe RET 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; Retpolines; IBPB conditional; STIBP always-on; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected Vulnerability Srbds: Not affected Vulnerability Tsx async abort: Not affected

Versions of relevant libraries: [pip3] numpy==1.26.4 [pip3] pytorch-triton==3.0.0+dedb7bdf33 [pip3] torch==2.5.0.dev20240902+cu118 [pip3] torchaudio==2.5.0.dev20240902+cu118 [pip3] torchvision==0.20.0.dev20240902+cu118 [conda] Could not collect

cc @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @kadeng @amjames @Lucaskabela @ezyang @bobrenjc93