search

vllm-project / vllm

A high-throughput and memory-efficient inference and serving engine for LLMs
https://docs.vllm.ai
Apache License 2.0
26.56k stars 3.89k forks source link

[Bug]: DeepSeek-V2-Lite quantized model raise model shape Error #7075

Open s-natsubori opened 1 month ago

s-natsubori commented 1 month ago

Your current environment

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 3.27.6
Libc version: glibc-2.35

Python version: 3.10.12 (main, Jun 11 2023, 05:26:28) [GCC 11.4.0] (64-bit runtime)
Python platform: Linux-5.15.0-113-generic-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: 12.2.140
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: 
GPU 0: NVIDIA H100 PCIe
  MIG 2g.20gb     Device  0:

Nvidia driver version: 535.183.01
cuDNN version: Probably one of the following:
/usr/lib/x86_64-linux-gnu/libcudnn.so.8.9.5
/usr/lib/x86_64-linux-gnu/libcudnn_adv_infer.so.8.9.5
/usr/lib/x86_64-linux-gnu/libcudnn_adv_train.so.8.9.5
/usr/lib/x86_64-linux-gnu/libcudnn_cnn_infer.so.8.9.5
/usr/lib/x86_64-linux-gnu/libcudnn_cnn_train.so.8.9.5
/usr/lib/x86_64-linux-gnu/libcudnn_ops_infer.so.8.9.5
/usr/lib/x86_64-linux-gnu/libcudnn_ops_train.so.8.9.5
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):                             64
On-line CPU(s) list:                0-63
Vendor ID:                          GenuineIntel
Model name:                         Intel(R) Xeon(R) Gold 6448Y
CPU family:                         6
Model:                              143
Thread(s) per core:                 1
Core(s) per socket:                 32
Socket(s):                          2
Stepping:                           8
Frequency boost:                    enabled
CPU max MHz:                        2101.0000
CPU min MHz:                        800.0000
BogoMIPS:                           4200.00
Flags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cat_l2 cdp_l3 invpcid_single intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities
Virtualization:                     VT-x
L1d cache:                          3 MiB (64 instances)
L1i cache:                          2 MiB (64 instances)
L2 cache:                           128 MiB (64 instances)
L3 cache:                           120 MiB (2 instances)
NUMA node(s):                       2
NUMA node0 CPU(s):                  0-31
NUMA node1 CPU(s):                  32-63
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: Not affected
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; Enhanced IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                Not affected
Vulnerability Tsx async abort:      Not affected

Versions of relevant libraries:
[pip3] flashinfer==0.0.8+cu121torch2.3
[pip3] mypy-extensions==1.0.0
[pip3] numpy==1.24.4
[pip3] nvidia-nccl-cu12==2.20.5
[pip3] onnx==1.14.0
[pip3] onnxruntime==1.18.1
[pip3] pytorch-quantization==2.1.2
[pip3] sentence-transformers==3.0.1
[pip3] torch==2.3.1
[pip3] torch-tensorrt==0.0.0
[pip3] torchdata==0.7.0a0
[pip3] torchtext==0.16.0a0
[pip3] torchvision==0.18.1
[pip3] transformers==4.43.0
[pip3] triton==2.3.1
[conda] Could not collect
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.5.3.post1
vLLM Build Flags:
CUDA Archs: 5.2 6.0 6.1 7.0 7.2 7.5 8.0 8.6 8.7 9.0+PTX; ROCm: Disabled; Neuron: Disabled
GPU Topology:
GPU0    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X      0-31    0               N/A

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

🐛 Describe the bug

DeepSeek-V2-Lite-gptq-4bit and DeepSeek-Coder-V2-Lite-Instruct-AWQ raise model shape Error.

Repro

from vllm.engine.arg_utils import AsyncEngineArgs
from vllm.engine.async_llm_engine import AsyncLLMEngine

engine_args = AsyncEngineArgs(
            model="ModelCloud/DeepSeek-V2-Lite-gptq-4bit",
            # model="TechxGenus/DeepSeek-Coder-V2-Lite-Instruct-AWQ",
            tensor_parallel_size=1,
            dtype=torch.float16,
            gpu_memory_utilization=0.9,
            swap_space=16,
            disable_log_requests=True,
            quantization=quantization,
            max_model_len=163840,
            tokenizer_mode="auto",
            trust_remote_code=True,
            enforce_eager=True,
        )
engine = AsyncLLMEngine.from_engine_args(engine_args)

Error(gptq)

INFO 08-02 08:40:49 llm_engine.py:176] Initializing an LLM engine (v0.5.3.post1) with config: model='/usr/local/models/llm', speculative_config=None, tokenizer='/usr/local/models/llm', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, rope_scaling=None, rope_theta=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.float16, max_seq_len=32768, download_dir=None, load_format=LoadFormat.AUTO, tensor_parallel_size=1, pipeline_parallel_size=1, disable_custom_all_reduce=False, quantization=gptq, enforce_eager=True, kv_cache_dtype=auto, quantization_param_path=None, device_config=cuda, decoding_config=DecodingConfig(guided_decoding_backend='outlines'), observability_config=ObservabilityConfig(otlp_traces_endpoint=None), seed=0, served_model_name=/usr/local/models/llm, use_v2_block_manager=False, enable_prefix_caching=False)
INFO 08-02 08:40:49 model_runner.py:680] Starting to load model /usr/local/models/llm...
Cache shape torch.Size([163840, 64])
[2024-08-02 08:40:50,110] [WARN] /usr/local/api/chat_router.py(115):__init__: The input size is not aligned with the quantized weight shape. This can be caused by too large tensor parallel size.
[2024-08-02 08:40:50,111] [WARN] /usr/local/api/chat_router.py(116):__init__: Traceback (most recent call last):
  File "/usr/local/api/chat_router.py", line 107, in __init__
    self.chat_client = ChatLocalVLLM.from_pretraind(model_path=llm_dir, NL="\n"
  File "/usr/local/api/chat_models/chat_local_vllm.py", line 73, in from_pretraind
    engine = cls._prepare_vllm(model_path, tensor_parallel_size
  File "/usr/local/api/chat_models/chat_local_vllm.py", line 123, in _prepare_vllm
    engine = AsyncLLMEngine.from_engine_args(engine_args)
  File "/usr/local/lib/python3.10/dist-packages/vllm/engine/async_llm_engine.py", line 466, in from_engine_args
    engine = cls(
  File "/usr/local/lib/python3.10/dist-packages/vllm/engine/async_llm_engine.py", line 380, in __init__
    self.engine = self._init_engine(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/vllm/engine/async_llm_engine.py", line 547, in _init_engine
    return engine_class(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/vllm/engine/llm_engine.py", line 251, in __init__
    self.model_executor = executor_class(
  File "/usr/local/lib/python3.10/dist-packages/vllm/executor/executor_base.py", line 47, in __init__
    self._init_executor()
  File "/usr/local/lib/python3.10/dist-packages/vllm/executor/gpu_executor.py", line 36, in _init_executor
    self.driver_worker.load_model()
  File "/usr/local/lib/python3.10/dist-packages/vllm/worker/worker.py", line 139, in load_model
    self.model_runner.load_model()
  File "/usr/local/lib/python3.10/dist-packages/vllm/worker/model_runner.py", line 682, in load_model
    self.model = get_model(model_config=self.model_config,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/model_loader/__init__.py", line 21, in get_model
    return loader.load_model(model_config=model_config,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/model_loader/loader.py", line 280, in load_model
    model = _initialize_model(model_config, self.load_config,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/model_loader/loader.py", line 111, in _initialize_model
    return model_class(config=model_config.hf_config,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/models/deepseek_v2.py", line 439, in __init__
    self.model = DeepseekV2Model(config, cache_config, quant_config)
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/models/deepseek_v2.py", line 401, in __init__
    self.layers = nn.ModuleList([
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/models/deepseek_v2.py", line 402, in <listcomp>
    DeepseekV2DecoderLayer(config,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/models/deepseek_v2.py", line 343, in __init__
    self.mlp = DeepseekV2MLP(
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/models/deepseek_v2.py", line 68, in __init__
    self.down_proj = RowParallelLinear(intermediate_size,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/layers/linear.py", line 728, in __init__
    self.quant_method.create_weights(
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/layers/quantization/gptq.py", line 112, in create_weights
    raise ValueError(
ValueError: The input size is not aligned with the quantized weight shape. This can be caused by too large tensor parallel size.

Error(awq)

INFO 08-02 08:16:01 llm_engine.py:176] Initializing an LLM engine (v0.5.3.post1) with config: model='/usr/local/models/llm', speculative_config=None, tokenizer='/usr/local/models/llm', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, rope_scaling=None, rope_theta=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.float16, max_seq_len=32768, download_dir=None, load_format=LoadFormat.AUTO, tensor_parallel_size=1, pipeline_parallel_size=1, disable_custom_all_reduce=False, quantization=awq_marlin, enforce_eager=True, kv_cache_dtype=auto, quantization_param_path=None, device_config=cuda, decoding_config=DecodingConfig(guided_decoding_backend='outlines'), observability_config=ObservabilityConfig(otlp_traces_endpoint=None), seed=0, served_model_name=/usr/local/models/llm, use_v2_block_manager=False, enable_prefix_caching=False)
INFO 08-02 08:16:01 model_runner.py:680] Starting to load model /usr/local/models/llm...
Cache shape torch.Size([163840, 64])
[2024-08-02 08:16:02,424] [WARN] /usr/local/api/chat_router.py(114):__init__: ERROR ChatModel not working
[2024-08-02 08:16:02,424] [WARN] /usr/local/api/chat_router.py(115):__init__: Weight input_size_per_partition = 10944 is not divisible by min_thread_k = 128. Consider reducing tensor_parallel_size or running with --quantization gptq.
[2024-08-02 08:16:02,426] [WARN] /usr/local/api/chat_router.py(116):__init__: Traceback (most recent call last):
  File "/usr/local/api/chat_router.py", line 107, in __init__
    self.chat_client = ChatLocalVLLM.from_pretraind(model_path=llm_dir, NL="\n"
  File "/usr/local/api/chat_models/chat_local_vllm.py", line 73, in from_pretraind
    engine = cls._prepare_vllm(model_path, tensor_parallel_size
  File "/usr/local/api/chat_models/chat_local_vllm.py", line 123, in _prepare_vllm
    engine = AsyncLLMEngine.from_engine_args(engine_args)
  File "/usr/local/lib/python3.10/dist-packages/vllm/engine/async_llm_engine.py", line 466, in from_engine_args
    engine = cls(
  File "/usr/local/lib/python3.10/dist-packages/vllm/engine/async_llm_engine.py", line 380, in __init__
    self.engine = self._init_engine(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/vllm/engine/async_llm_engine.py", line 547, in _init_engine
    return engine_class(*args, **kwargs)
  File "/usr/local/lib/python3.10/dist-packages/vllm/engine/llm_engine.py", line 251, in __init__
    self.model_executor = executor_class(
  File "/usr/local/lib/python3.10/dist-packages/vllm/executor/executor_base.py", line 47, in __init__
    self._init_executor()
  File "/usr/local/lib/python3.10/dist-packages/vllm/executor/gpu_executor.py", line 36, in _init_executor
    self.driver_worker.load_model()
  File "/usr/local/lib/python3.10/dist-packages/vllm/worker/worker.py", line 139, in load_model
    self.model_runner.load_model()
  File "/usr/local/lib/python3.10/dist-packages/vllm/worker/model_runner.py", line 682, in load_model
    self.model = get_model(model_config=self.model_config,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/model_loader/__init__.py", line 21, in get_model
    return loader.load_model(model_config=model_config,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/model_loader/loader.py", line 280, in load_model
    model = _initialize_model(model_config, self.load_config,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/model_loader/loader.py", line 111, in _initialize_model
    return model_class(config=model_config.hf_config,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/models/deepseek_v2.py", line 439, in __init__
    self.model = DeepseekV2Model(config, cache_config, quant_config)
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/models/deepseek_v2.py", line 401, in __init__
    self.layers = nn.ModuleList([
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/models/deepseek_v2.py", line 402, in <listcomp>
    DeepseekV2DecoderLayer(config,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/models/deepseek_v2.py", line 343, in __init__
    self.mlp = DeepseekV2MLP(
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/models/deepseek_v2.py", line 68, in __init__
    self.down_proj = RowParallelLinear(intermediate_size,
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/layers/linear.py", line 728, in __init__
    self.quant_method.create_weights(
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/layers/quantization/awq_marlin.py", line 148, in create_weights
    verify_marlin_supports_shape(
  File "/usr/local/lib/python3.10/dist-packages/vllm/model_executor/layers/quantization/utils/marlin_utils.py", line 106, in verify_marlin_supports_shape
    raise ValueError(f"Weight input_size_per_partition = "
ValueError: Weight input_size_per_partition = 10944 is not divisible by min_thread_k = 128. Consider reducing tensor_parallel_size or running with --quantization gptq.
chocoHunter commented 1 month ago

Same problem, don't know how to fix it.

jli943 commented 2 weeks ago

+1.