'_IPEXLinear' object has no attribute 'use_dnnl'

[Alok-Ranjan23]

Describe the bug

I am trying to run llm.optimize API for GPT-j and getting the following Error.
_'IPEXLinear' object has no attribute 'use_dnnl'

Please check the attached code snippet.

`import torch
import time

from transformers import AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained(
'EleutherAI/gpt-j-6b',
torchscript=True,
return_dict=False,
torch_dtype=torch.bfloat16,
low_cpu_mem_usage=True,
trust_remote_code=True,
)

vocab_size = model.config.vocab_size

step_count = 3
num_warmup = 2

with torch.inference_mode(), torch.no_grad(), torch.cpu.amp.autocast(
enabled=True
):
import intel_extension_for_pytorch as ipex
model = ipex.llm.optimize(model.eval(), dtype=torch.bfloat16, inplace=True,deployment_mode=False)

total_time = 0
batch_size=32
seq_len=384
sample_input = torch.randint(vocab_size, size=[batch_size, seq_len])
with torch.inference_mode(), torch.no_grad(), torch.cpu.amp.autocast(
enabled=True):
for i in range(int(num_warmup)):
outputs = model(sample_input)
for i in range(int(step_count)):
start_time = time.time()
outputs = model(sample_input)
end_time = time.time()
total_time += (end_time - start_time) * 1e3 # ms
time_per_step = total_time / int(step_count)
if batch_size == 1:
print("%s: Latency = %.2f ms" % ('GPT-J', time_per_step))
else:
print(
"%s: Throughput = %.2f QPS"
% ('GPT-J', (batch_size * 1000) / time_per_step)
)`

The above code file produces the following error:
_AttributeError: 'IPEXLinear' object has no attribute 'use_dnnl'

What I think about this error is that ipex.llm.optimize Convert LinearOp into IPEXLinear and ipexLINEAR does not need use_dnnl attribute.

Why is there no attribute 'use_dnnl' in _IPEXLinear Op? How can we remove this attribute from _IPEXLinear Op?

Versions

Collecting environment information...
PyTorch version: 2.2.0+cu121
PyTorch CXX11 ABI: No
IPEX version: 2.2.0+cpu
IPEX commit: 211813b
Build type: Release

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
IGC version: N/A
CMake version: version 3.29.2
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-5.15.0-97-generic-x86_64-with-glibc2.35
Is XPU available: False
DPCPP runtime version: N/A
MKL version: N/A
GPU models and configuration:

Intel OpenCL ICD version: N/A
Level Zero version: N/A

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 52 bits physical, 57 bits virtual
Byte Order: Little Endian
CPU(s): 384
On-line CPU(s) list: 0-383
CPU family: 25
Model: 17
Thread(s) per core: 2
Core(s) per socket: 96
Socket(s): 2
Stepping: 1
BogoMIPS: 4799.99
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 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local avx512_bf16 clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin cppc arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif v_spec_ctrl avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq la57 rdpid overflow_recov succor smca fsrm flush_l1d
L1d cache: 6 MiB (192 instances)
L1i cache: 6 MiB (192 instances)
L2 cache: 192 MiB (192 instances)
L3 cache: 768 MiB (24 instances)
NUMA node(s): 8
NUMA node0 CPU(s): 0-23,192-215
NUMA node1 CPU(s): 24-47,216-239
NUMA node2 CPU(s): 48-71,240-263
NUMA node3 CPU(s): 72-95,264-287
NUMA node4 CPU(s): 96-119,288-311
NUMA node5 CPU(s): 120-143,312-335
NUMA node6 CPU(s): 144-167,336-359
NUMA node7 CPU(s): 168-191,360-383
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 and seccomp
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] intel-extension-for-pytorch==2.2.0
[pip3] numpy==1.26.4
[pip3] torch==2.2.0
[conda] intel-extension-for-pytorch 2.2.0 pypi_0 pypi
[conda] numpy 1.26.4 pypi_0 pypi
[conda] torch 2.2.0 pypi_0 pypi

[feng-intel]

I can reproduce this issue. I will ask dev team for help and feedback later.
Our 2.3.0 version will be released soon. It can work with removing "torch.inference_mode(), "

[jgong5]

cc @jianan-gu

[jgong5]

Mark it as high priority due to the crash.

[feng-intel]

We have found the root cause. Before fixing, users can use

batch_size=32
seq_len=384
input_ids = torch.randint(vocab_size, size=[batch_size, seq_len])
att_mask = torch.ones_like(input_ids)
sample_input = model.prepare_inputs_for_generation(input_ids, attention_mask=att_mask)

To generate sample input

[Alok-Ranjan23]

That's fine. But Why is this optimization not working with dummy benchmark?

[feng-intel]

It will be fixed in release 2.3.1