vllm
vllm copied to clipboard
vllm-project
[Usage]: Cannot run the starter code in tutorial
Your current environment
I am a new user who recently ran this starter code on my lab server:
import torch
from vllm import LLM, SamplingParams
# Clear any leftover memory from previous models
torch.cuda.set_device('cuda:3')
torch.cuda.empty_cache()
prompts = [
"Hello, my name is",
"The president of the United States is",
"The capital of France is",
"The future of AI is",
]
sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
llm = LLM(model="allenai/OLMo-1B-hf", device='cuda:3')
outputs = llm.generate(prompts, sampling_params)
# Print the outputs.
for output in outputs:
prompt = output.prompt
generated_text = output.outputs[0].text
print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
This piece of code always throws error:
---------------------------------------------------------------------------
AssertionError Traceback (most recent call last)
Cell In[9], [line 16](vscode-notebook-cell:?execution_count=9&line=16)
[8](vscode-notebook-cell:?execution_count=9&line=8) prompts = [
[9](vscode-notebook-cell:?execution_count=9&line=9) "Hello, my name is",
[10](vscode-notebook-cell:?execution_count=9&line=10) "The president of the United States is",
[11](vscode-notebook-cell:?execution_count=9&line=11) "The capital of France is",
[12](vscode-notebook-cell:?execution_count=9&line=12) "The future of AI is",
[13](vscode-notebook-cell:?execution_count=9&line=13) ]
[15](vscode-notebook-cell:?execution_count=9&line=15) sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
---> [16](vscode-notebook-cell:?execution_count=9&line=16) llm = LLM(model="allenai/OLMo-1B-hf", device='cuda:3')
[17](vscode-notebook-cell:?execution_count=9&line=17) outputs = llm.generate(prompts, sampling_params)
[19](vscode-notebook-cell:?execution_count=9&line=19) # Print the outputs.
File ~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/entrypoints/llm.py:123, in LLM.__init__(self, model, tokenizer, tokenizer_mode, skip_tokenizer_init, trust_remote_code, tensor_parallel_size, dtype, quantization, revision, tokenizer_revision, seed, gpu_memory_utilization, swap_space, enforce_eager, max_context_len_to_capture, max_seq_len_to_capture, disable_custom_all_reduce, **kwargs)
[102](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/entrypoints/llm.py:102) kwargs["disable_log_stats"] = True
[103](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/entrypoints/llm.py:103) engine_args = EngineArgs(
[104](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/entrypoints/llm.py:104) model=model,
[105](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/entrypoints/llm.py:105) tokenizer=tokenizer,
(...)
[121](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/entrypoints/llm.py:121) **kwargs,
[122](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/entrypoints/llm.py:122) )
--> [123](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/entrypoints/llm.py:123) self.llm_engine = LLMEngine.from_engine_args(
[124](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/entrypoints/llm.py:124) engine_args, usage_context=UsageContext.LLM_CLASS)
[125](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/entrypoints/llm.py:125) self.request_counter = Counter()
...
[153](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/worker/worker.py:153) num_gpu_blocks = int(
[154](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/worker/worker.py:154) (total_gpu_memory * self.cache_config.gpu_memory_utilization -
[155](https://vscode-remote+ssh-002dremote-002bdocjk-002dgpu-002d02.vscode-resource.vscode-cdn.net/home/21zz42/open-moe-llm-leaderboard/~/open-moe-llm-leaderboard/.venv/lib/python3.10/site-packages/vllm/worker/worker.py:155) peak_memory) // cache_block_size)
AssertionError: Error in memory profiling. This happens when the GPU memory was not properly cleaned up before initializing the vLLM instance.
This is the GPU profile of my server using nvtop:
Device 0 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 47°C FAN N/A% POW 69 / 300 W
GPU[ 0%] MEM[|||||||||||||||||79.034Gi/80.000Gi]
Device 1 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 44°C FAN N/A% POW 67 / 300 W
GPU[ 0%] MEM[|||||||||| 22.999Gi/80.000Gi]
Device 2 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 44°C FAN N/A% POW 65 / 300 W
GPU[ 0%] MEM[|||||||||| 22.999Gi/80.000Gi]
Device 3 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 42°C FAN N/A% POW 64 / 300 W
GPU[ 0%] MEM[|||||||||||| 28.737Gi/80.000Gi]
Device 4 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 44°C FAN N/A% POW 66 / 300 W
GPU[ 0%] MEM[|||||||||| 22.999Gi/80.000Gi]
Device 5 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 46°C FAN N/A% POW 67 / 300 W
GPU[ 0%] MEM[|||||||||| 23.528Gi/80.000Gi]
Device 6 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 44°C FAN N/A% POW 63 / 300 W
GPU[ 0%] MEM[|||||||||| 22.999Gi/80.000Gi]
Device 7 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 49°C FAN N/A% POW 73 / 300 W
GPU[ 0%] MEM[||| 8.389Gi/80.000Gi]
Here is what wget https://raw.githubusercontent.com/vllm-project/vllm/main/collect_env.py && python3 collect_env.py give:
Collecting environment information...
PyTorch version: N/A
Is debug build: N/A
CUDA used to build PyTorch: N/A
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 3.22.1
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-5.15.0-88-generic-x86_64-with-glibc2.35
Is CUDA available: N/A
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration:
GPU 0: NVIDIA A100 80GB PCIe
GPU 1: NVIDIA A100 80GB PCIe
GPU 2: NVIDIA A100 80GB PCIe
GPU 3: NVIDIA A100 80GB PCIe
GPU 4: NVIDIA A100 80GB PCIe
GPU 5: NVIDIA A100 80GB PCIe
GPU 6: NVIDIA A100 80GB PCIe
GPU 7: NVIDIA A100 80GB PCIe
Nvidia driver version: 525.147.05
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: N/A
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): 96
On-line CPU(s) list: 0-95
Vendor ID: GenuineIntel
Model name: Intel(R) Xeon(R) Platinum 8268 CPU @ 2.90GHz
CPU family: 6
Model: 85
Thread(s) per core: 2
Core(s) per socket: 24
Socket(s): 2
Stepping: 7
CPU max MHz: 3900.0000
CPU min MHz: 1200.0000
BogoMIPS: 5800.00
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 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 cdp_l3 invpcid_single intel_ppin 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 mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts pku ospke avx512_vnni md_clear flush_l1d arch_capabilities
Virtualization: VT-x
L1d cache: 1.5 MiB (48 instances)
L1i cache: 1.5 MiB (48 instances)
L2 cache: 48 MiB (48 instances)
L3 cache: 71.5 MiB (2 instances)
NUMA node(s): 2
NUMA node0 CPU(s): 0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62,64,66,68,70,72,74,76,78,80,82,84,86,88,90,92,94
NUMA node1 CPU(s): 1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49,51,53,55,57,59,61,63,65,67,69,71,73,75,77,79,81,83,85,87,89,91,93,95
Vulnerability Gather data sampling: Mitigation; Microcode
Vulnerability Itlb multihit: KVM: Mitigation: VMX disabled
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Mmio stale data: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Retbleed: Mitigation; Enhanced IBRS
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: Mitigation; TSX disabled
Versions of relevant libraries:
[pip3] numpy==1.26.1
[conda] Could not collectROCM 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:
GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 CPU Affinity NUMA Affinity
GPU0 X PIX NODE NODE SYS SYS SYS SYS 0,2,4,6,8,10 0
GPU1 PIX X NODE NODE SYS SYS SYS SYS 0,2,4,6,8,10 0
GPU2 NODE NODE X PIX SYS SYS SYS SYS 0,2,4,6,8,10 0
GPU3 NODE NODE PIX X SYS SYS SYS SYS 0,2,4,6,8,10 0
GPU4 SYS SYS SYS SYS X PIX NODE NODE 1,3,5,7,9,11 1
GPU5 SYS SYS SYS SYS PIX X NODE NODE 1,3,5,7,9,11 1
GPU6 SYS SYS SYS SYS NODE NODE X PIX 1,3,5,7,9,11 1
GPU7 SYS SYS SYS SYS NODE NODE PIX X 1,3,5,7,9,11 1
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
How should I proceed?
# Clear any leftover memory from previous models
torch.cuda.set_device('cuda:3')
torch.cuda.empty_cache()
Try to remove these code, and run in a fresh new python script like python test.py with export CUDA_VISIBLE_DEVICES=3 ?
# Clear any leftover memory from previous models torch.cuda.set_device('cuda:3') torch.cuda.empty_cache()Try to remove these code, and run in a fresh new python script like
python test.pywithexport CUDA_VISIBLE_DEVICES=3?
Thanks for your reply. I did it with what you suggested, but this time it throws OutOfMemoryError: CUDA out of memory. Tried to allocate 4.43 GiB. GPU, here is my GPU profiling:
Device 0 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 49°C FAN N/A% POW 75 / 300 W
GPU[|| 5%] MEM[|||||||||||||||||79.020Gi/80.000Gi]
Device 1 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 48°C FAN N/A% POW 70 / 300 W
GPU[|| 5%] MEM[|||||||| 18.719Gi/80.000Gi]
Device 2 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 8.789 MiB/s
GPU 1410MHz MEM 1512MHz TEMP 46°C FAN N/A% POW 67 / 300 W
GPU[|| 6%] MEM[|||||||| 19.594Gi/80.000Gi]
Device 3 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 3.906 MiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 44°C FAN N/A% POW 103 / 300 W
GPU[|| 6%] MEM[|||||||||| 22.547Gi/80.000Gi]
Device 4 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 48°C FAN N/A% POW 80 / 300 W
GPU[|| 6%] MEM[||||||| 18.026Gi/80.000Gi]
Device 5 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 1.953 MiB/s
GPU 1410MHz MEM 1512MHz TEMP 46°C FAN N/A% POW 65 / 300 W
GPU[|| 6%] MEM[|||||||| 18.645Gi/80.000Gi]
Device 6 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 41.99 MiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 46°C FAN N/A% POW 91 / 300 W
GPU[| 4%] MEM[|||||||| 19.895Gi/80.000Gi]
Device 7 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s
GPU 1410MHz MEM 1512MHz TEMP 50°C FAN N/A% POW 75 / 300 W
GPU[||| 9%] MEM[|||||||||||| 28.717Gi/80.000Gi]
I also set export CUDA_VISIBLE_DEVICES=5 and changed code to llm = LLM(model="allenai/OLMo-1B-hf", device='cuda:5'), but it still throws: AssertionError: Error in memory profiling. This happens when the GPU memory was not properly cleaned up before initializing the vLLM instance.
I am sure this server is shared across different students in our lab, but it is still weird that I cannot use any GPU for this starter code...
when you have export CUDA_VISIBLE_DEVICES=5 , you don't need to set the device arg in LLM .
export CUDA_VISIBLE_DEVICES=5
If I set export CUDA_VISIBLE_DEVICES=5 and use the following code:
from vllm import LLM, SamplingParams
prompts = [
"Hello, my name is",
"The president of the United States is",
"The capital of France is",
"The future of AI is",
]
sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
llm = LLM(model="allenai/OLMo-1B-hf")
outputs = llm.generate(prompts, sampling_params)
# Print the outputs.
for output in outputs:
prompt = output.prompt
generated_text = output.outputs[0].text
print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
Then it would throw OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 MiB. GPU error.
I think only device 0 is full in memory, am I right?
Your system driver or any other program takes too much memory:
Device 0 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 49°C FAN N/A% POW 75 / 300 W GPU[|| 5%] MEM[|||||||||||||||||79.020Gi/80.000Gi] Device 1 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 48°C FAN N/A% POW 70 / 300 W GPU[|| 5%] MEM[|||||||| 18.719Gi/80.000Gi] Device 2 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 8.789 MiB/s GPU 1410MHz MEM 1512MHz TEMP 46°C FAN N/A% POW 67 / 300 W GPU[|| 6%] MEM[|||||||| 19.594Gi/80.000Gi] Device 3 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 3.906 MiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 44°C FAN N/A% POW 103 / 300 W GPU[|| 6%] MEM[|||||||||| 22.547Gi/80.000Gi] Device 4 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 48°C FAN N/A% POW 80 / 300 W GPU[|| 6%] MEM[||||||| 18.026Gi/80.000Gi] Device 5 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 1.953 MiB/s GPU 1410MHz MEM 1512MHz TEMP 46°C FAN N/A% POW 65 / 300 W GPU[|| 6%] MEM[|||||||| 18.645Gi/80.000Gi] Device 6 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 41.99 MiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 46°C FAN N/A% POW 91 / 300 W GPU[| 4%] MEM[|||||||| 19.895Gi/80.000Gi] Device 7 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 50°C FAN N/A% POW 75 / 300 W GPU[||| 9%] MEM[|||||||||||| 28.717Gi/80.000Gi]
You need to set gpu_memory_utilization to a lower value, e.g. 0.7:
llm = LLM(model="allenai/OLMo-1B-hf", gpu_memory_utilization=0.7)
Your system driver or any other program takes too much memory:
Device 0 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 49°C FAN N/A% POW 75 / 300 W GPU[|| 5%] MEM[|||||||||||||||||79.020Gi/80.000Gi] Device 1 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 48°C FAN N/A% POW 70 / 300 W GPU[|| 5%] MEM[|||||||| 18.719Gi/80.000Gi] Device 2 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 8.789 MiB/s GPU 1410MHz MEM 1512MHz TEMP 46°C FAN N/A% POW 67 / 300 W GPU[|| 6%] MEM[|||||||| 19.594Gi/80.000Gi] Device 3 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 3.906 MiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 44°C FAN N/A% POW 103 / 300 W GPU[|| 6%] MEM[|||||||||| 22.547Gi/80.000Gi] Device 4 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 48°C FAN N/A% POW 80 / 300 W GPU[|| 6%] MEM[||||||| 18.026Gi/80.000Gi] Device 5 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 1.953 MiB/s GPU 1410MHz MEM 1512MHz TEMP 46°C FAN N/A% POW 65 / 300 W GPU[|| 6%] MEM[|||||||| 18.645Gi/80.000Gi] Device 6 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 41.99 MiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 46°C FAN N/A% POW 91 / 300 W GPU[| 4%] MEM[|||||||| 19.895Gi/80.000Gi] Device 7 [NVIDIA A100 80GB PCIe] PCIe GEN 3@16x RX: 0.000 KiB/s TX: 0.000 KiB/s GPU 1410MHz MEM 1512MHz TEMP 50°C FAN N/A% POW 75 / 300 W GPU[||| 9%] MEM[|||||||||||| 28.717Gi/80.000Gi]
You need to set
gpu_memory_utilizationto a lower value, e.g.0.7:llm = LLM(model="allenai/OLMo-1B-hf", gpu_memory_utilization=0.7)
I tried with a series of values from 0.05 to 0.9, but it always throws OutOfMemoryError: CUDA out of memory. Tried to allocate 64.00 MiB. GPU
Something wrong with my other configuration? I doubt I am still using device 0...
I suggest asking for your admin of the server, to see if there is any unusual limit on your server, or ask admin for how to restrict your program in one specific GPU.
Hi there,
This is the latest error I am facing with using allenai/OLMo-1B-hf
I using the code is from the Quick Start Guide. I thought of creating a new issue. But, found this. Hence, replying to it.
Do you try any other supported model? As said, the error is that this model is not supported.
Ok, sir. Currently I am trying mistralai/Mixtral-8x7B-v0.1.
But, is giving access to gated repository is restricted error.
After getting the access, I am exporting access token from HF using environment variable HF_TOKEN.
But, I am still having the same issue.
Would you please help me with this?
Then you need to check you huggingface configuration. Sorry but I cannot help with this. You can reach out to huggingface community.
Maybe start from a very small model (e.g. facebook/opt-125m) and see if it will run? And also, run nvidia-smi first to make sure the GPU you are using is not occupied.
