_allow_non_fake_inputs parameter of make_fx has no effect

[ttrouwen-dmatrix]

🐛 Describe the bug

Running make_fx with _allow_non_fake_inputs=True will throw Exception that the inputs are non fake:

Exception: Please convert all Tensors to FakeTensors first or instantiate FakeTensorMode with 'allow_non_fake_inputs'. Found in aten.t.default(Parameter containing:
tensor([...], size=(64, 64), requires_grad=True))

Notice that this issue does not reproduce on Torch 2.0 but does occur on Torch 2.1-2.3
This is a full reproducer:

import torch
import torch.nn as nn
from torch.fx.experimental.proxy_tensor import make_fx

class M(nn.Module):
    def __init__(self, *args, **kwargs) -> None:
        super().__init__(*args, **kwargs)
        self.l = nn.Linear(64, 64)
    
    def forward(self, inp):
        return self.l(inp)

m = M()
inp = torch.zeros(64, 64)

def backend(gm, inps):
    input_cloned = [i.clone().detach() for i in inps][0]
    # inputs_cloned = [i.clone() if isinstance(i, torch.Tensor) else i for i in inputs]
    gm = make_fx(
        gm,
        tracing_mode="fake",
        _allow_non_fake_inputs=True,
        decomposition_table=None,
    )(input_cloned)

    return gm

fn = torch.compile(m, backend=backend)
fn(inp)

Versions

PyTorch version: 2.3.0+cu121
Is debug build: False
CUDA used to build PyTorch: 12.1
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: 14.0.0-1ubuntu1.1
CMake version: version 3.29.2
Libc version: glibc-2.35

Python version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime)
Python platform: Linux-6.2.0-1018-azure-x86_64-with-glibc2.35
Is CUDA available: False
CUDA runtime version: No CUDA
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
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, 57 bits virtual
Byte Order:                         Little Endian
CPU(s):                             8
On-line CPU(s) list:                0-7
Vendor ID:                          GenuineIntel
Model name:                         Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
CPU family:                         6
Model:                              106
Thread(s) per core:                 2
Core(s) per socket:                 4
Socket(s):                          1
Stepping:                           6
CPU max MHz:                        2800.0000
CPU min MHz:                        800.0000
BogoMIPS:                           5586.87
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 rep_good nopl xtopology tsc_reliable nonstop_tsc cpuid aperfmperf pni pclmulqdq vmx 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 tpr_shadow vnmi ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq la57 rdpid fsrm arch_capabilities
Virtualization:                     VT-x
Hypervisor vendor:                  Microsoft
Virtualization type:                full
L1d cache:                          192 KiB (4 instances)
L1i cache:                          128 KiB (4 instances)
L2 cache:                           5 MiB (4 instances)
L3 cache:                           48 MiB (1 instance)
NUMA node(s):                       1
NUMA node0 CPU(s):                  0-7
Vulnerability Gather data sampling: Unknown: Dependent on hypervisor status
Vulnerability Itlb multihit:        Not affected
Vulnerability L1tf:                 Not affected
Vulnerability Mds:                  Not affected
Vulnerability Meltdown:             Not affected
Vulnerability Mmio stale data:      Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown
Vulnerability Retbleed:             Vulnerable
Vulnerability Spec rstack overflow: Not affected
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

Versions of relevant libraries:
[pip3] numpy==1.26.4
[pip3] torch==2.3.0
[pip3] triton==2.3.0
[conda] Could not collect

cc @ezyang @SherlockNoMad @EikanWang @jgong5 @wenzhe-nrv @msaroufim @bdhirsh @anijain2305 @chauhang @eellison

[ttrouwen-dmatrix]

This occurs because a call to detect_fake_mode was added to make_fx. If a fake mode is detected, the _allow_non_fake_inputs argument is ignored.

A hack to circumvent the issue is by changing some lines in make_fx to the following:

elif tracing_mode == "fake":
    import torch._dynamo
    fake_tensor_mode = torch._dynamo.utils.detect_fake_mode(args)
    if fake_tensor_mode is None:
        # ...
    else:  # This line and the next line should be inserted after line 1193
        fake_tensor_mode.allow_non_fake_inputs = _allow_non_fake_inputs