[Bug]: ModelRegistry.load_model_cls() circular import error on llama-llava

[datta-nimmaturi]

Your current environment

datta0@naillm:~/kserve/python/huggingfaceserver$ python collect_env.py 
Collecting environment information...
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: Could not collect
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-5.15.0-101-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 A100-PCIE-40GB
Nvidia driver version: 525.60.13
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:                      46 bits physical, 48 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) Gold 6354 CPU @ 3.00GHz
CPU family:                         15
Model:                              6
Thread(s) per core:                 1
Core(s) per socket:                 1
Socket(s):                          8
Stepping:                           1
BogoMIPS:                           6000.01
Flags:                              fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss syscall nx pdpe1gb rdtscp lm constant_tsc nopl xtopology cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch cpuid_fault invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid fsrm md_clear arch_capabilities
Hypervisor vendor:                  KVM
Virtualization type:                full
L1d cache:                          256 KiB (8 instances)
L1i cache:                          256 KiB (8 instances)
L2 cache:                           32 MiB (8 instances)
L3 cache:                           128 MiB (8 instances)
NUMA node(s):                       1
NUMA node0 CPU(s):                  0-7
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:      Unknown: No mitigations
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
Vulnerability Srbds:                Not affected
Vulnerability Tsx async abort:      Not affected

Versions of relevant libraries:
[pip3] flake8==7.0.0
[pip3] mypy-extensions==1.0.0
[pip3] numpy==1.26.4
[pip3] nvidia-nccl-cu12==2.20.5
[pip3] torch==2.3.0
[pip3] torchvision==0.16.2
[pip3] triton==2.3.0
[pip3] vllm_nccl_cu12==2.18.1.0.4.0
[conda] Could not collectROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.4.2
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
GPU0    CPU Affinity    NUMA Affinity
GPU0     X      0-7             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

Steps to reproduce:

1. Install vllm 0.4.2 (works fine on 0.4.0)
2. Run the following code

from vllm.model_executor.models import ModelRegistry
architecture = 'MistralForCausalLM' # can also use LlamaForCausalLM
model_cls = ModelRegistry.load_model_cls(architecture)

This throws the error (click to expand)

ImportError: cannot import name 'LlamaModel' from partially initialized module 'vllm.model_executor.models.llama' (most likely due to a circular import) (/home/datta0/.pyenv/lib/python3.10/site-packages/vllm/model_executor/models/llama.py)"

``` { "name": "ImportError", "message": "cannot import name 'LlamaModel' from partially initialized module 'vllm.model_executor.models.llama' (most likely due to a circular import) (/home/datta0/.pyenv/lib/python3.10/site-packages/vllm/model_executor/models/llama.py)", "stack": "--------------------------------------------------------------------------- ImportError Traceback (most recent call last) Cell In[1], line 3 1 from vllm.model_executor.models import ModelRegistry 2 architecture = 'MistralForCausalLM' ----> 3 model_cls = ModelRegistry.load_model_cls(architecture)

File ~/.pyenv/lib/python3.10/site-packages/vllm/model_executor/models/init.py:98, in ModelRegistry.load_model_cls(model_arch)
93 logger.warning(
94 "Model architecture %s is partially supported by ROCm: %s",
95 model_arch, _ROCM_PARTIALLY_SUPPORTED_MODELS[model_arch])
97 module_name, model_cls_name = _MODELS[model_arch]
---> 98 module = importlib.import_module(
99 f"vllm.model_executor.models.{module_name}")
100 return getattr(module, model_cls_name, None)

File /usr/lib/python3.10/importlib/init.py:126, in import_module(name, package)
124 break
125 level += 1
--> 126 return _bootstrap._gcd_import(name[level:], package, level)

File :1050, in _gcd_import(name, package, level)

File :1027, in find_and_load(name, import)

File :1006, in find_and_load_unlocked(name, import)

File :688, in _load_unlocked(spec)

File :883, in exec_module(self, module)

File :241, in _call_with_frames_removed(f, *args, **kwds)

File ~/.pyenv/lib/python3.10/site-packages/vllm/model_executor/models/llama.py:46
43 from vllm.model_executor.layers.sampler import Sampler
44 from vllm.model_executor.layers.vocab_parallel_embedding import (
45 DEFAULT_VOCAB_PADDING_SIZE, ParallelLMHead, VocabParallelEmbedding)
---> 46 from vllm.model_executor.model_loader.weight_utils import (
47 default_weight_loader, kv_cache_scales_loader)
48 from vllm.model_executor.sampling_metadata import SamplingMetadata
49 from vllm.sequence import SamplerOutput

File ~/.pyenv/lib/python3.10/site-packages/vllm/model_executor/model_loader/init.py:7
3 from torch import nn
5 from vllm.config import (DeviceConfig, LoadConfig, LoRAConfig, ModelConfig,
6 ParallelConfig, SchedulerConfig, VisionLanguageConfig)
----> 7 from vllm.model_executor.model_loader.loader import (BaseModelLoader,
8 get_model_loader)
9 from vllm.model_executor.model_loader.utils import (
10 get_architecture_class_name, get_model_architecture)
13 def get_model(
14 *, model_config: ModelConfig, load_config: LoadConfig,
15 device_config: DeviceConfig, parallel_config: ParallelConfig,
16 scheduler_config: SchedulerConfig, lora_config: Optional[LoRAConfig],
17 vision_language_config: Optional[VisionLanguageConfig]) -> nn.Module:

File ~/.pyenv/lib/python3.10/site-packages/vllm/model_executor/model_loader/loader.py:28
22 from vllm.model_executor.model_loader.utils import (get_model_architecture,
23 set_default_torch_dtype)
24 from vllm.model_executor.model_loader.weight_utils import (
25 download_weights_from_hf, filter_files_not_needed_for_inference,
26 get_quant_config, initialize_dummy_weights, np_cache_weights_iterator,
27 pt_weights_iterator, safetensors_weights_iterator)
---> 28 from vllm.model_executor.models.llava import LlavaForConditionalGeneration
30 _VISION_MODEL_CLASSES = [
31 LlavaForConditionalGeneration,
32 ]
34 logger = init_logger(name)

File ~/.pyenv/lib/python3.10/site-packages/vllm/model_executor/models/llava.py:18
16 from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
17 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
---> 18 from vllm.model_executor.models.llama import LlamaModel
19 from vllm.model_executor.sampling_metadata import SamplingMetadata
20 from vllm.sequence import SamplerOutput

ImportError: cannot import name 'LlamaModel' from partially initialized module 'vllm.model_executor.models.llama' (most likely due to a circular import) (/home/datta0/.pyenv/lib/python3.10/site-packages/vllm/model_executor/models/llama.py)"
}```

Cause of the issue:

Mistral (LlaMA) executor imports model_loader whose imports called BaseModelLoader which in turn uses LlavaForConditionalGeneration that depends on llama creating a circular dependency.