🐛 Describe the bug

Somehow passing an empty tensor in the optimizer prevents the update of normally updatable parameters. This happens only on CUDA device.

from typing import Dict

import torch
from torch import nn
from torch.optim.adamw import AdamW

def assert_state_dict_equal(first: Dict[str, torch.Tensor], second: Dict[str, torch.Tensor]):
    for name, first_tensor in first.items():
        second_tensor = second[name]
        torch.testing.assert_close(first_tensor, second_tensor, atol=0, rtol=0)

def test_empty_tensor_in_optimizer(device: torch.device):
    model = nn.Linear(2, 3, bias=True, device=device)
    reference_model = nn.Linear(2, 3, bias=True, device=device)

    # sync parameters
    with torch.no_grad():
        model.weight.copy_(reference_model.weight)
        model.bias.copy_(reference_model.bias)

    empty_tensor = torch.tensor([], requires_grad=True, device=device)

    optimizer = AdamW(
        [
            model.weight,
            model.bias,
            empty_tensor,
        ]
    )
    reference_optimizer = AdamW(reference_model.parameters())

    random_input = torch.randn(5, 2, device=device)

    # Check that state dict are equal before optimizer step
    assert_state_dict_equal(model.state_dict(), reference_model.state_dict())

    model(random_input).sum().backward()
    empty_tensor.grad = torch.tensor([], requires_grad=False, device=device) # important
    reference_model(random_input).sum().backward()

    optimizer.step()
    reference_optimizer.step()

    # Check that state dict are equal after optimizer step
    assert_state_dict_equal(model.state_dict(), reference_model.state_dict())


def main():
    for device in [
        torch.device("cpu"),
        torch.device("cuda")
    ]:
        print(f"Testing device: {device}")
        test_empty_tensor_in_optimizer(device)
        print("Pass")


if __name__ == "__main__":
    main()

I would be expecting that adding empty parameters does not change the behaviour of other parameters passed down to the optimizer.

Versions

PyTorch version: 2.0.0+cu117
Is debug build: False
CUDA used to build PyTorch: 11.7
ROCM used to build PyTorch: N/A

OS: Ubuntu 20.04.5 LTS (x86_64)
GCC version: (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0
Clang version: Could not collect
CMake version: version 3.26.0
Libc version: glibc-2.31

Python version: 3.10.9 (main, Jan 11 2023, 15:21:40) [GCC 11.2.0] (64-bit runtime)
Python platform: Linux-5.15.0-1023-aws-x86_64-with-glibc2.31
Is CUDA available: True
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration:
GPU 0: NVIDIA A100-SXM4-80GB
GPU 1: NVIDIA A100-SXM4-80GB
GPU 2: NVIDIA A100-SXM4-80GB
GPU 3: NVIDIA A100-SXM4-80GB
GPU 4: NVIDIA A100-SXM4-80GB
GPU 5: NVIDIA A100-SXM4-80GB
GPU 6: NVIDIA A100-SXM4-80GB
GPU 7: NVIDIA A100-SXM4-80GB

Nvidia driver version: 510.73.08
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
Byte Order: Little Endian
Address sizes: 46 bits physical, 48 bits virtual
CPU(s): 96
On-line CPU(s) list: 0-95
Thread(s) per core: 2
Core(s) per socket: 24
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 85
Model name: Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz
Stepping: 7
CPU MHz: 3000.000
BogoMIPS: 6000.00
Hypervisor vendor: KVM
Virtualization type: full
L1d cache: 1.5 MiB
L1i cache: 1.5 MiB
L2 cache: 48 MiB
L3 cache: 71.5 MiB
NUMA node0 CPU(s): 0-23,48-71
NUMA node1 CPU(s): 24-47,72-95
Vulnerability Itlb multihit: KVM: Mitigation: VMX unsupported
Vulnerability L1tf: Mitigation; PTE Inversion
Vulnerability Mds: Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown
Vulnerability Meltdown: Mitigation; PTI
Vulnerability Mmio stale data: Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown
Vulnerability Retbleed: Vulnerable
Vulnerability Spec store bypass: Vulnerable
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; Retpolines, STIBP disabled, RSB filling, PBRSB-eIBRS Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single pti fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves ida arat pku ospke

Versions of relevant libraries:
[pip3] numpy==1.24.2
[pip3] torch==2.0.0
[pip3] triton==2.0.0
[conda] numpy 1.24.2 pypi_0 pypi
[conda] torch 2.0.0 pypi_0 pypi
[conda] triton 2.0.0 pypi_0 pypi

cc @ezyang @gchanan @zou3519 @vincentqb @jbschlosser @albanD @janeyx99