search

vllm-project / vllm

A high-throughput and memory-efficient inference and serving engine for LLMs
https://docs.vllm.ai
Apache License 2.0
30.17k stars 4.56k forks source link

[Bug]: failed when run Qwen2-54B-A14B-GPTQ-Int4(MOE) #6465

Open weiminw opened 4 months ago

weiminw commented 4 months ago

Your current environment

Collecting environment information...
PyTorch version: 2.3.1+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: Could not collect
CMake version: version 3.30.0
Libc version: glibc-2.35

Python version: 3.10.12 (main, Mar 22 2024, 16:50:05) [GCC 11.4.0] (64-bit runtime)
Python platform: Linux-5.15.0-91-generic-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: GPU 0: NVIDIA RTX 6000 Ada Generation
Nvidia driver version: 545.23.08
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
Address sizes:                      52 bits physical, 57 bits virtual
Byte Order:                         Little Endian
CPU(s):                             128
On-line CPU(s) list:                0-127
Vendor ID:                          AuthenticAMD
Model name:                         AMD EPYC 9354 32-Core Processor
CPU family:                         25
Model:                              17
Thread(s) per core:                 2
Core(s) per socket:                 32
Socket(s):                          2
Stepping:                           1
Frequency boost:                    enabled
CPU max MHz:                        3799.0720
CPU min MHz:                        1500.0000
BogoMIPS:                           6500.47
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
Virtualization:                     AMD-V
L1d cache:                          2 MiB (64 instances)
L1i cache:                          2 MiB (64 instances)
L2 cache:                           64 MiB (64 instances)
L3 cache:                           512 MiB (16 instances)
NUMA node(s):                       2
NUMA node0 CPU(s):                  0-31,64-95
NUMA node1 CPU(s):                  32-63,96-127
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 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] numpy==1.26.4
[pip3] nvidia-nccl-cu12==2.20.5
[pip3] torch==2.3.1
[pip3] torchvision==0.18.1
[pip3] transformers==4.42.4
[pip3] triton==2.3.1
[conda] Could not collect
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.5.2
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
GPU0    NIC0    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X  SYS 32-63,96-127    1       N/A
NIC0    SYS  X

Legend:

  X    = Self
  SYS  = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
  NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
  PHB  = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
  PXB  = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
  PIX  = Connection traversing at most a single PCIe bridge
  NV#  = Connection traversing a bonded set of # NVLinks

NIC Legend:

  NIC0: mlx5_bond_0

🐛 Describe the bug

got the following error:

[rank0]:     self.layers = nn.ModuleList([
[rank0]:   File "/workspace/heliumos-env/lib/python3.10/site-packages/vllm/model_executor/models/qwen2_moe.py", line 328, in <listcomp>
[rank0]:     Qwen2MoeDecoderLayer(config,
[rank0]:   File "/workspace/heliumos-env/lib/python3.10/site-packages/vllm/model_executor/models/qwen2_moe.py", line 268, in __init__
[rank0]:     self.mlp = Qwen2MoeSparseMoeBlock(config=config,
[rank0]:   File "/workspace/heliumos-env/lib/python3.10/site-packages/vllm/model_executor/models/qwen2_moe.py", line 103, in __init__
[rank0]:     self.experts = FusedMoE(num_experts=config.num_experts,
[rank0]:   File "/workspace/heliumos-env/lib/python3.10/site-packages/vllm/model_executor/layers/fused_moe/layer.py", line 145, in __init__
[rank0]:     assert self.quant_method is not None
[rank0]: AssertionError
robertgshaw2-neuralmagic commented 4 months ago

GPTQ is not yet supported for Qwen MoE. We are working on it

weiminw commented 4 months ago

GPTQ is not yet supported for Qwen MoE. We are working on it

so, what kind of quantization MOE model of qwen can vllm support in 0.5.2? could you recommend me the quantization MOE of qwen2 model ?

robertgshaw2-neuralmagic commented 4 months ago

We currently support fp16 and fp8 for qwen MoE

fp8 requires hopper GPUs

akai-shuuichi commented 3 months ago

This PR may solve it https://github.com/vllm-project/vllm/pull/6502 And I created a WHL for self testing using this branch: https://github.com/akai-shuuichi/vllm/releases/download/v5/vllm-0.5.2-cp310-cp310-manylinux1_x86_64.whl

Xu-Chen commented 3 months ago

GPUs

@akai-shuuichi DeepSeek V2 Support？

akai-shuuichi commented 3 months ago

GPUs

@akai-shuuichi DeepSeek V2 Support？ I only tested Qwen, I don't have enough GPU to run deepseekV2 moe :(

Xu-Chen commented 3 months ago

GPUs

@akai-shuuichi DeepSeek V2 Support？ I only tested Qwen, I don't have enough GPU to run deepseekV2 moe :( Thank you. You can also try this model

akai-shuuichi commented 3 months ago

GPUs

@akai-shuuichi DeepSeek V2 Support？ I only tested Qwen, I don't have enough GPU to run deepseekV2 moe :( Thank you. You can also try this model

sorry, model has error: Traceback (most recent call last): File "/vllm-workspace/v1Server.py", line 50, in <module> generation_config, tokenizer, stop_word, engine = load_vllm() File "/vllm-workspace/v1Server.py", line 23, in load_vllm generation_config = GenerationConfig.from_pretrained(model_dir, trust_remote_code=True) File "/usr/local/lib/python3.10/dist-packages/transformers/generation/configuration_utils.py", line 915, in from_pretrained resolved_config_file = cached_file( File "/usr/local/lib/python3.10/dist-packages/transformers/utils/hub.py", line 373, in cached_file raise EnvironmentError( OSError: /vllm-workspace/DeepSeek-V2-Lite-gptq-4bit does not appear to have a file named generation_config.json.

z18256199275 commented 3 months ago

@akai-shuuichi Hi, When I reason about the qwen-moe-gptq-int4 model, it always prompts triton.runtime.errors.OutOfResources: out of resource: shared memory, Error, how to solve it

zmlpillow commented 2 months ago

@akai-shuuichi Hi, When I reason about the qwen-moe-gptq-int4 model, it always prompts triton.runtime.errors.OutOfResources: out of resource: shared memory, Error, how to solve it

hi，I also meet this error，do you solve it now？

liangzelang commented 2 weeks ago

This PR https://github.com/vllm-project/vllm/pull/8973/files should have fixed your issue, but it is still not sufficient to run the quantized DeepSeek V2 model. I got other errors when try DeepSeek V2 AWQ-int4 with lateset vllm.