[Usage]: OOM when using Llama-3.2-11B-Vision-Instruct #8879
Comments
|
To split the model across GPUs, you should set |
@DarkLight1337 If it's 11B, wouldn't it be possible to use approximately 22GB of memory? |
|
You should also consider the memory required for inference, not just the model weights. If you run into OOM issues, you may need to reduce |
DarkLight1337
changed the title
llm = LLM(
model="meta-llama/Llama-3.2-11B-Vision-Instruct",
tensor_parallel_size=2,
max_model_len=4096,
gpu_memory_utilization=0.8,
trust_remote_code=True, # !
)When �I run the above code, lines 338 to 339 of llm_engine.py if not self.model_config.embedding_mode:
self._initialize_kv_caches()An error occurs when running here. |
|
How did you install vLLM? I see in the output of |
I don't know why vllm version is not included;; I'm using |
|
What is the command you used to install vLLM? |
I created a conda virtual environment and installed vllm using the |
|
@youkaichao @dtrifiro There seems to be something wrong with |
Can you show more about the error? |
I'm continuing debugging now;; I'm still debugging so I don't know exactly where the error occurred. |
|
File "/home/heerak/miniconda3/envs/eval/lib/python3.10/site-packages/vllm/model_executor/models/mllama.py", line 1084, in forward In this part, OOM occurs as the memory continues to increase as it passes forward. |
|
@hrson-1203 You should only need to set |
Oh, it finally works. Could you also explain why we need to use |
Mostly because of its architecture.
|
thank you so much for explaining Thanks to you, I was able to test it. What parts of each model should I look at to figure out if such a setting is necessary? |
For most models, you don't need to worry about enforcing eager mode unless you need additional VRAM (This is because cuda graphs themselves also consume some memory), and if you run into OOM issue, always try lowering |
|
I'm trying to run GPU Server specifications:
vLLM Docker run command: docker run --gpus all \
-v /data/hf_cache/ \
--env "HUGGING_FACE_HUB_TOKEN=<token>" \
-p 8000:8000 \
--ipc=host \
vllm/vllm-openai:latest \
--model meta-llama/Llama-3.2-11B-Vision-Instruct \
--tensor-parallel-size 4 \
--max-model-len 4096 \
--download_dir /data/vllm_cache \
--enforce-eagerFacing similar issue. |
As mentioned above, you should limit |
|
@DarkLight1337 fix is included in #8900 |
|
Found this thread very useful. Appreciate the guidance and generally all your work on the vllm multimodal models @DarkLight1337 |
|
Thank you @DarkLight1337. This solved my issue as well. |
Your current environment
Collecting environment information...
PyTorch version: 2.4.0+cu121
Is debug build: False
CUDA used to build PyTorch: 12.1
ROCM used to build PyTorch: N/A
OS: Ubuntu 20.04.6 LTS (x86_64)
GCC version: (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0
Clang version: Could not collect
CMake version: Could not collect
Libc version: glibc-2.31
Python version: 3.10.14 (main, May 6 2024, 19:42:50) [GCC 11.2.0] (64-bit runtime)
Python platform: Linux-5.15.0-117-generic-x86_64-with-glibc2.31
Is CUDA available: True
CUDA runtime version: 12.1.66
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration:
GPU 0: NVIDIA A100 80GB PCIe
GPU 1: NVIDIA A100 80GB PCIe
Nvidia driver version: 535.171.04
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
Byte Order: Little Endian
Address sizes: 46 bits physical, 57 bits virtual
CPU(s): 64
On-line CPU(s) list: 0-63
Thread(s) per core: 2
Core(s) per socket: 16
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 106
Model name: Intel(R) Xeon(R) Gold 6326 CPU @ 2.90GHz
Stepping: 6
CPU MHz: 806.789
CPU max MHz: 3500.0000
CPU min MHz: 800.0000
BogoMIPS: 5800.00
Virtualization: VT-x
L1d cache: 1.5 MiB
L1i cache: 1 MiB
L2 cache: 40 MiB
L3 cache: 48 MiB
NUMA node0 CPU(s): 0-15,32-47
NUMA node1 CPU(s): 16-31,48-63
Vulnerability Gather data sampling: Mitigation; Microcode
Vulnerability Itlb multihit: Not affected
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Mmio stale data: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Reg file data sampling: 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 / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI SW loop, KVM SW loop
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected
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 pni pclmulqdq dtes64 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 invpcid_single 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 wbnoinvd dtherm ida arat pln pts avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid fsrm md_clear pconfig flush_l1d arch_capabilities
Versions of relevant libraries:
[pip3] numpy==1.26.4
[pip3] nvidia-cublas-cu12==12.1.3.1
[pip3] nvidia-cuda-cupti-cu12==12.1.105
[pip3] nvidia-cuda-nvrtc-cu12==12.1.105
[pip3] nvidia-cuda-runtime-cu12==12.1.105
[pip3] nvidia-cudnn-cu12==9.1.0.70
[pip3] nvidia-cufft-cu12==11.0.2.54
[pip3] nvidia-curand-cu12==10.3.2.106
[pip3] nvidia-cusolver-cu12==11.4.5.107
[pip3] nvidia-cusparse-cu12==12.1.0.106
[pip3] nvidia-ml-py==12.535.161
[pip3] nvidia-nccl-cu12==2.20.5
[pip3] nvidia-nvjitlink-cu12==12.6.68
[pip3] nvidia-nvtx-cu12==12.1.105
[pip3] pyzmq==25.1.2
[pip3] torch==2.4.0
[pip3] torchvision==0.19.0
[pip3] transformers==4.45.0
[pip3] triton==3.0.0
[conda] numpy 1.26.4 pypi_0 pypi
[conda] nvidia-cublas-cu12 12.1.3.1 pypi_0 pypi
[conda] nvidia-cuda-cupti-cu12 12.1.105 pypi_0 pypi
[conda] nvidia-cuda-nvrtc-cu12 12.1.105 pypi_0 pypi
[conda] nvidia-cuda-runtime-cu12 12.1.105 pypi_0 pypi
[conda] nvidia-cudnn-cu12 9.1.0.70 pypi_0 pypi
[conda] nvidia-cufft-cu12 11.0.2.54 pypi_0 pypi
[conda] nvidia-curand-cu12 10.3.2.106 pypi_0 pypi
[conda] nvidia-cusolver-cu12 11.4.5.107 pypi_0 pypi
[conda] nvidia-cusparse-cu12 12.1.0.106 pypi_0 pypi
[conda] nvidia-ml-py 12.535.161 pypi_0 pypi
[conda] nvidia-nccl-cu12 2.20.5 pypi_0 pypi
[conda] nvidia-nvjitlink-cu12 12.6.68 pypi_0 pypi
[conda] nvidia-nvtx-cu12 12.1.105 pypi_0 pypi
[conda] pyzmq 26.2.0 pypi_0 pypi
[conda] torch 2.4.0 pypi_0 pypi
[conda] torchvision 0.19.0 pypi_0 pypi
[conda] transformers 4.45.0 pypi_0 pypi
[conda] triton 3.0.0 pypi_0 pypi
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: N/A
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
�[4mGPU0 GPU1 CPU Affinity NUMA Affinity GPU NUMA ID�[0m
GPU0 X NV12 0-15,32-47 0 N/A
GPU1 NV12 X 0-15,32-47 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
How would you like to use vllm
I want to run inference of a meta-llama/Llama-3.2-11B-Vision-Instruct.
I tried to load the multi-modal model into vllm and proceed with inference.
However, even with two A100s, an OOM error occurred while loading the 11B model.
The error message below indicates that only A100 was used.
Even if both are used, the same OOM occurs.
How can I load the Llama-3.2-11B-Vision-Instruct model with vllm?
Before submitting a new issue...
The text was updated successfully, but these errors were encountered: