🐛 Describe the bug

Problem: There seem to be 2 ways of using torch compile with a HF model, both of which don't work for all the ways a model inference is called, which is one of 3 possible methods: generate(), forward() and __call__().

Option 1: model = torch.compile(model)

This works if we use either forward() or the __call__() methods. But, if we try to call the .generate() method (which is the more popular API for inferencing and calls forward() internally), we notice that we DON'T seem to be using the compiled model (ex. TORCH_LOGS="dynamo" gives no output).

Simple reproducible example (custom class with generate and forward like implementations):

import torch

class MyModule(torch.nn.Module):
    def __init__(self):
        super().__init__()

    def forward(self, input):
        # return average of the inputs
        return torch.Tensor(torch.sum(input)/len(input))

    def generate(self, max_tokens, input):
        for i in range(max_tokens):
            output = self(input) # Doesn't work with either call or forward
            input = torch.cat((input, output.view(1)))
        return input

model = MyModule()
model = torch.compile(model)
input = torch.rand(4)
output = model.generate(input=input, max_tokens=3) # THIS DOES NOT WORK!!!
#output = model.forward(input=input) # THIS WORKS

or use any HF model compile followed by generate:

model = AutoModelForCausalLM.from_pretrained("facebook/opt-125m")
model = torch.compile(model)
output = model.generate(input_ids, max_new_tokens=100)

The problem is that the output of torch.compile(model) is an OptimizedModule object with the __call__() set to the compiled forward and orig_mod set to model itself.
When compiled_model.generate() is called, this accesses the generate through the __getattr__() function which gets the model's generate. That generate calls self(), which calls the original model's forward instead of the compiled forward.

Option 2: model.compile()

The other option is to use the torch.nn.Module's compile, which does an inplace modification where the compiled forward is stored in _compiled_call_impl variable and used when __call__() is done. But, this only works with the __call__() method and does NOT work with the forward() method. If the generate() internally uses call, then generate works.

model.compile()
output = model.generate(input=input, max_tokens=3) # Works
#output = model.forward(input_ids) # DOES NOT WORK

Problem is that neither of these approaches works with both generate() and forward() methods.

As an aside, I tried a couple of unsuccessful possible fixes:

• Tried if Option 1 could be fixed somehow by setting the orig_mod.forward to the compiled forward but that causes infinite recursion because of the circular dependency
• I also tried changing TorchDynamoContext.__call__() (in eval_frame.py) in the nn.Module case, to internally do model.compile instead of creating an OptimizedModule. This fixes things slightly, for ex. Option 1 works if it generate uses call instead of forward, but obviously, not really a solution.

cc: @ChanderG

Error logs

No response

Versions

Collecting environment information...
PyTorch version: 2.6.0.dev20241103+cu124
Is debug build: False
CUDA used to build PyTorch: 12.4
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.4 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect
CMake version: version 3.30.2
Libc version: glibc-2.35

Python version: 3.10.12 (main, Jul 29 2024, 16:56:48) [GCC 11.4.0] (64-bit runtime)
Python platform: Linux-5.14.0-284.73.1.el9_2.x86_64-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: 12.6.68
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: 525.105.17
cuDNN version: Probably one of the following:
/usr/lib/x86_64-linux-gnu/libcudnn.so.9.4.0
/usr/lib/x86_64-linux-gnu/libcudnn_adv.so.9.4.0
/usr/lib/x86_64-linux-gnu/libcudnn_cnn.so.9.4.0
/usr/lib/x86_64-linux-gnu/libcudnn_engines_precompiled.so.9.4.0
/usr/lib/x86_64-linux-gnu/libcudnn_engines_runtime_compiled.so.9.4.0
/usr/lib/x86_64-linux-gnu/libcudnn_graph.so.9.4.0
/usr/lib/x86_64-linux-gnu/libcudnn_heuristic.so.9.4.0
/usr/lib/x86_64-linux-gnu/libcudnn_ops.so.9.4.0
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: 48 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 256
On-line CPU(s) list: 0-255
Vendor ID: AuthenticAMD
Model name: AMD EPYC 7763 64-Core Processor
CPU family: 25
Model: 1
Thread(s) per core: 2
Core(s) per socket: 64
Socket(s): 2
Stepping: 1
Frequency boost: enabled
CPU max MHz: 3529.0520
CPU min MHz: 1500.0000
BogoMIPS: 4900.23
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 pcid sse4_1 sse4_2 x2apic 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 invpcid_single hw_pstate ssbd mba ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 invpcid cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin brs arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold v_vmsave_vmload vgif v_spec_ctrl umip pku ospke vaes vpclmulqdq rdpid overflow_recov succor smca
Virtualization: AMD-V
L1d cache: 4 MiB (128 instances)
L1i cache: 4 MiB (128 instances)
L2 cache: 64 MiB (128 instances)
L3 cache: 512 MiB (16 instances)
NUMA node(s): 2
NUMA node0 CPU(s): 0-63,128-191
NUMA node1 CPU(s): 64-127,192-255
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: Mitigation; Safe RET
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; Retpolines, IBPB conditional, IBRS_FW, STIBP always-on, RSB filling, PBRSB-eIBRS Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected

Versions of relevant libraries:
[pip3] mypy-extensions==1.0.0
[pip3] numpy==1.24.4
[pip3] nvidia-cublas-cu12==12.4.5.8
[pip3] nvidia-cuda-cupti-cu12==12.4.127
[pip3] nvidia-cuda-nvrtc-cu12==12.4.127
[pip3] nvidia-cuda-runtime-cu12==12.4.127
[pip3] nvidia-cudnn-cu12==9.1.0.70
[pip3] nvidia-cudnn-frontend==1.6.0
[pip3] nvidia-cufft-cu12==11.2.1.3
[pip3] nvidia-curand-cu12==10.3.5.147
[pip3] nvidia-cusolver-cu12==11.6.1.9
[pip3] nvidia-cusparse-cu12==12.3.1.170
[pip3] nvidia-cusparselt-cu12==0.6.2
[pip3] nvidia-nccl-cu12==2.21.5
[pip3] nvidia-nvjitlink-cu12==12.4.127
[pip3] nvidia-nvtx-cu12==12.4.127
[pip3] nvtx==0.2.5
[pip3] onnx==1.16.2
[pip3] optree==0.12.1
[pip3] pynvjitlink==0.3.0
[pip3] pytorch-triton==3.1.0+cf34004b8a
[pip3] torch==2.6.0.dev20241103+cu124
[pip3] torch_tensorrt==2.5.0a0
[pip3] torchaudio==2.5.0.dev20241103+cu124
[pip3] torchdata==0.9.0
[pip3] torchvision==0.20.0.dev20241103+cu124

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