[Bug]: Speculative decoding generate gibberish when receiving parallel requests with different seeds

[wallashss]

Your current environment

The output of `python collect_env.py`

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: Red Hat Enterprise Linux 9.4 (Plow) (x86_64)
GCC version: (GCC) 11.4.1 20231218 (Red Hat 11.4.1-3)
Clang version: Could not collect
CMake version: version 3.30.4
Libc version: glibc-2.34

Python version: 3.12.1 (main, Aug 23 2024, 00:00:00) [GCC 11.4.1 20231218 (Red Hat 11.4.1-3)] (64-bit runtime)
Python platform: Linux-4.18.0-372.46.1.el8_6.x86_64-x86_64-with-glibc2.34
Is CUDA available: True
CUDA runtime version: 12.4.131
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: 
GPU 0: NVIDIA A100-SXM4-80GB
  MIG 3g.40gb     Device  0:

Nvidia driver version: 535.104.05
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, 57 bits virtual
Byte Order:                      Little Endian
CPU(s):                          80
On-line CPU(s) list:             0-79
Vendor ID:                       GenuineIntel
Model name:                      Intel Xeon Processor (Icelake)
CPU family:                      6
Model:                           134
Thread(s) per core:              2
Core(s) per socket:              20
Socket(s):                       2
Stepping:                        0
BogoMIPS:                        5600.03
Flags:                           fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology cpuid tsc_known_freq pni pclmulqdq vmx 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 tpr_shadow vnmi flexpriority ept vpid ept_ad 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 pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq la57 rdpid fsrm md_clear arch_capabilities
Virtualization:                  VT-x
Hypervisor vendor:               KVM
Virtualization type:             full
L1d cache:                       2.5 MiB (80 instances)
L1i cache:                       2.5 MiB (80 instances)
L2 cache:                        160 MiB (40 instances)
L3 cache:                        32 MiB (2 instances)
NUMA node(s):                    2
NUMA node0 CPU(s):               0-39
NUMA node1 CPU(s):               40-79
Vulnerability Itlb multihit:     Not affected
Vulnerability L1tf:              Not affected
Vulnerability Mds:               Not affected
Vulnerability Meltdown:          Not affected
Vulnerability Mmio stale data:   Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown
Vulnerability Retbleed:          Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:        Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:        Mitigation; Enhanced IBRS, IBPB conditional, RSB filling, PBRSB-eIBRS Not affected
Vulnerability Srbds:             Not affected
Vulnerability Tsx async abort:   Not affected

Versions of relevant libraries:
[pip3] flashinfer==0.1.6+cu124torch2.4
[pip3] mypy==1.11.1
[pip3] mypy-extensions==1.0.0
[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==12.560.30
[pip3] nvidia-nccl-cu12==2.20.5
[pip3] nvidia-nvjitlink-cu12==12.6.77
[pip3] nvidia-nvtx-cu12==12.1.105
[pip3] pyzmq==26.2.0
[pip3] sentence-transformers==3.1.1
[pip3] torch==2.4.0
[pip3] torchvision==0.19.0
[pip3] transformers==4.45.2
[pip3] transformers-stream-generator==0.0.5
[pip3] triton==3.0.0
[conda] Could not collect
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: N/A (dev)
vLLM Build Flags:
CUDA Archs: 7.0 7.5 8.0 8.6 8.9 9.0+PTX; ROCm: Disabled; Neuron: Disabled
GPU Topology:
�[4mGPU0	NIC0	CPU Affinity	NUMA Affinity	GPU NUMA ID�[0m
GPU0	 X 	NODE	0-39	0		N/A
NIC0	NODE	 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_0

Model Input Dumps

No response

🐛 Describe the bug

I was doing some experiments with speculative decoding and found an strange behavior when the engine process request in parallel, there are some requests generated with gibberish.

Here's the script to reproduce the bug:

import asyncio
import tempfile
import uuid

from tests.mq_llm_engine.utils import RemoteMQLLMEngine
from vllm import SamplingParams
from vllm.engine.arg_utils import AsyncEngineArgs
from vllm.engine.multiprocessing.client import MQLLMEngineClient

DEFAULT_SEED=0

async def generate_repeated(client : MQLLMEngineClient, seed=None):
    params = SamplingParams(temperature=0, stop=['20:'], seed=seed, max_tokens=220)
    prompt = 'Write the following phrase exactly 20 times: I will never miss any tokens.\n1: I will never miss any tokens.\n2: '
    
    async for r in client.generate(prompt=prompt,
                                    sampling_params=params,
                                    request_id=uuid.uuid4()):
        if r.finished:
            txt = r.outputs[0].text
            n = len(txt.split('I will never miss any tokens.'))
            
            print(r.request_id, f'Has {n} occurrences')
            if n != 19:
                print(txt)
            # print(txt)
                
async def generate_other(client : MQLLMEngineClient, seed: int):
    params = SamplingParams(temperature=0, seed=seed, max_tokens=200)
    prompt = "How to make pizza" 
    async for _ in client.generate(prompt=prompt,
                                    sampling_params=params,
                                    request_id=uuid.uuid4()):
        # No need to process output
        pass
                
            
            
            
async def run_tests(client):
    await client.check_health()
        
    print(">>>  Running only the repeated prompt")
    
    repeated_prompt_tasks = [asyncio.create_task(generate_repeated(client)) for _ in range(10)]
    await asyncio.gather(*repeated_prompt_tasks) 
    
    print(">>>  Running the repeated prompt with other prompts with same seed")
    
    repeated_prompt_tasks = [asyncio.create_task(generate_repeated(client, DEFAULT_SEED)) for _ in range(5)]
    other_prompt_tasks = [asyncio.create_task(generate_other(client, DEFAULT_SEED)) for _ in range(5)]
    tasks = repeated_prompt_tasks + other_prompt_tasks
    await asyncio.gather(*(tasks)) 
    
    print(">>>  Running the repeated prompt with other prompts with different seeds")
    
    repeated_prompt_tasks = [asyncio.create_task(generate_repeated(client)) for _ in range(5)]
    other_prompt_tasks = [asyncio.create_task(generate_other(client, DEFAULT_SEED)) for _ in range(5)]
    tasks = repeated_prompt_tasks + other_prompt_tasks
    
    await asyncio.gather(*(tasks))
    
    client.close()
    
async def main(ipc_path):
    
    engine_args = AsyncEngineArgs(model='meta-llama/Llama-3.1-8B-Instruct', 
                                  gpu_memory_utilization=0.95, use_v2_block_manager=True)

    print("__________RUNNING WITHOUT SPECULATING DECODING__________")

    with RemoteMQLLMEngine(engine_args=engine_args,
                           ipc_path=ipc_path) as engine:

        client = await engine.make_client()
        await run_tests(client)
        
    print("__________RUNNING WITH SPECULATING DECODING__________")
    
    engine_args = AsyncEngineArgs(model='meta-llama/Llama-3.1-8B-Instruct', 
                                  gpu_memory_utilization=0.95, use_v2_block_manager=True,
                                  max_model_len=1984, 
                                  speculative_model='ibm-fms/llama3-8b-accelerator', num_speculative_tokens=4)
    
    
    with RemoteMQLLMEngine(engine_args=engine_args,
                           ipc_path=ipc_path) as engine:

        client = await engine.make_client()
        await run_tests(client)


if __name__ == "__main__": 
    
    with tempfile.TemporaryDirectory() as td:
        socket = f"ipc://{td}/{uuid.uuid4()}"
        asyncio.run(main(socket))
        
    print("Done")

In a nutshell:

• The issue only occurs with speculative decoding
• With speculative decoding, only when you do not set a seed and the engine receives other requests with seed.

A possible output

/opt/vllm/lib64/python3.12/site-packages/vllm/connections.py:8: RuntimeWarning: Failed to read commit hash:
No module named 'vllm._version'
  from vllm.version import __version__ as VLLM_VERSION
__________RUNNING WITHOUT SPECULATING DECODING__________
/opt/vllm/lib64/python3.12/site-packages/vllm/connections.py:8: RuntimeWarning: Failed to read commit hash:
No module named 'vllm._version'
  from vllm.version import __version__ as VLLM_VERSION
WARNING 10-17 02:06:17 arg_utils.py:953] Chunked prefill is enabled by default for models with max_model_len > 32K. Currently, chunked prefill might not work with some features or models. If you encounter any issues, please disable chunked prefill by setting --enable-chunked-prefill=False.
INFO 10-17 02:06:17 config.py:1005] Chunked prefill is enabled with max_num_batched_tokens=512.
WARNING 10-17 02:06:20 arg_utils.py:953] Chunked prefill is enabled by default for models with max_model_len > 32K. Currently, chunked prefill might not work with some features or models. If you encounter any issues, please disable chunked prefill by setting --enable-chunked-prefill=False.
INFO 10-17 02:06:20 config.py:1005] Chunked prefill is enabled with max_num_batched_tokens=512.
INFO 10-17 02:06:20 llm_engine.py:237] Initializing an LLM engine (vdev) with config: model='meta-llama/Llama-3.1-8B-Instruct', speculative_config=None, tokenizer='meta-llama/Llama-3.1-8B-Instruct', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, override_neuron_config=None, rope_scaling=None, rope_theta=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=131072, download_dir=None, load_format=LoadFormat.AUTO, tensor_parallel_size=1, pipeline_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=False, 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, collect_model_forward_time=False, collect_model_execute_time=False), seed=0, served_model_name=meta-llama/Llama-3.1-8B-Instruct, use_v2_block_manager=True, num_scheduler_steps=1, chunked_prefill_enabled=True multi_step_stream_outputs=True, enable_prefix_caching=False, use_async_output_proc=True, use_cached_outputs=True, mm_processor_kwargs=None)
INFO 10-17 02:06:21 model_runner.py:1060] Starting to load model meta-llama/Llama-3.1-8B-Instruct...
INFO 10-17 02:06:22 weight_utils.py:243] Using model weights format ['*.safetensors']

Loading safetensors checkpoint shards:   0% Completed | 0/4 [00:00<?, ?it/s]

Loading safetensors checkpoint shards:  25% Completed | 1/4 [00:00<00:02,  1.29it/s]

Loading safetensors checkpoint shards:  50% Completed | 2/4 [00:01<00:01,  1.12it/s]

Loading safetensors checkpoint shards:  75% Completed | 3/4 [00:02<00:00,  1.10it/s]

Loading safetensors checkpoint shards: 100% Completed | 4/4 [00:03<00:00,  1.47it/s]

Loading safetensors checkpoint shards: 100% Completed | 4/4 [00:03<00:00,  1.33it/s]

INFO 10-17 02:06:25 model_runner.py:1071] Loading model weights took 14.9888 GB
INFO 10-17 02:06:26 gpu_executor.py:122] # GPU blocks: 10754, # CPU blocks: 2048
INFO 10-17 02:06:26 gpu_executor.py:126] Maximum concurrency for 131072 tokens per request: 1.31x
INFO 10-17 02:06:27 model_runner.py:1402] Capturing the model for CUDA graphs. This may lead to unexpected consequences if the model is not static. To run the model in eager mode, set 'enforce_eager=True' or use '--enforce-eager' in the CLI.
INFO 10-17 02:06:27 model_runner.py:1406] CUDA graphs can take additional 1~3 GiB memory per GPU. If you are running out of memory, consider decreasing `gpu_memory_utilization` or enforcing eager mode. You can also reduce the `max_num_seqs` as needed to decrease memory usage.
INFO 10-17 02:06:38 model_runner.py:1530] Graph capturing finished in 11 secs.
INFO 10-17 02:06:38 engine.py:292] Added request fe289733-219d-48ee-96b1-5ce7f9f73874.
INFO 10-17 02:06:38 engine.py:292] Added request 8ab9d577-f6c5-4027-812f-c590248d4980.
INFO 10-17 02:06:38 engine.py:292] Added request bf23ac40-77d7-477b-8fb5-808a755a26d5.
INFO 10-17 02:06:38 engine.py:292] Added request 8f45a0bf-f4ef-4fc5-b799-8ced0e720f4e.
INFO 10-17 02:06:38 engine.py:292] Added request 7e0c750b-92f5-424a-9a96-665926da88b2.
INFO 10-17 02:06:38 engine.py:292] Added request cfc1df95-0e83-4b12-8a76-2f91d02dae56.
INFO 10-17 02:06:38 engine.py:292] Added request 35462188-01a7-4137-ab94-2a4739015207.
INFO 10-17 02:06:38 engine.py:292] Added request 644fdc5b-18e6-44e9-b7f6-bc42dfe4b9aa.
INFO 10-17 02:06:38 engine.py:292] Added request 3c859be7-67b7-4e99-96a9-29ca027b62bb.
INFO 10-17 02:06:38 engine.py:292] Added request 2b8afcac-9a76-45b4-8368-a371eb585291.
INFO 10-17 02:06:42 engine.py:292] Added request 7af075eb-e930-4084-a465-cbc4dc180f96.
INFO 10-17 02:06:42 engine.py:292] Added request 5d35b575-4ae6-4187-82a6-372a28fb7ee7.
INFO 10-17 02:06:42 engine.py:292] Added request f341e245-5bf6-4ced-83c8-7524876b3fb9.
INFO 10-17 02:06:42 engine.py:292] Added request 325a0f4f-cb5b-44a4-b208-28250b2e19e4.
INFO 10-17 02:06:42 engine.py:292] Added request fb0530a1-0f15-498a-b83c-d57f7e07f0e3.
INFO 10-17 02:06:42 engine.py:292] Added request b4605ebc-c856-4586-98bc-52cca74424cd.
INFO 10-17 02:06:42 engine.py:292] Added request 5f7fd3e1-ea75-442c-8505-ca1cd65be9d7.
INFO 10-17 02:06:42 engine.py:292] Added request a99964a9-bcde-45b3-bb41-2f8dfcff4a87.
INFO 10-17 02:06:42 engine.py:292] Added request c348c777-5c73-4b8a-8eaa-034e5ce8507b.
INFO 10-17 02:06:42 engine.py:292] Added request bfba3621-947d-407f-9765-62472e8aefa5.
INFO 10-17 02:06:43 metrics.py:345] Avg prompt throughput: 91.9 tokens/s, Avg generation throughput: 419.7 tokens/s, Running: 10 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.3%, CPU KV cache usage: 0.0%.
INFO 10-17 02:06:47 engine.py:292] Added request 2b5d3959-a07f-4b81-b431-4e02703b3e4d.
INFO 10-17 02:06:47 engine.py:292] Added request a9d0f590-9272-4292-a986-1b65e3d7b1cd.
INFO 10-17 02:06:47 engine.py:292] Added request 6fe01def-ba33-495b-8c52-90f96bb22208.
INFO 10-17 02:06:47 engine.py:292] Added request c7f2492d-0dc3-4294-aa63-620e1d573b19.
INFO 10-17 02:06:47 engine.py:292] Added request 5211e0e9-455a-4ef3-a81e-00d51ef3421a.
INFO 10-17 02:06:47 engine.py:292] Added request 759676c4-64d7-4817-81e7-ca079ceca838.
INFO 10-17 02:06:47 engine.py:292] Added request 6cf4fec8-3f60-4fbb-8d2f-49e4154b5a30.
INFO 10-17 02:06:47 engine.py:292] Added request f4b4b79f-0792-464a-95fe-445725c3bfec.
INFO 10-17 02:06:47 engine.py:292] Added request e7b4eb9e-cd4c-44bc-aaa5-4353b55b23e1.
INFO 10-17 02:06:47 engine.py:292] Added request 6b5a101f-f64c-443b-ac96-7ee580abd89a.
INFO 10-17 02:06:48 metrics.py:345] Avg prompt throughput: 33.9 tokens/s, Avg generation throughput: 418.2 tokens/s, Running: 10 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.4%, CPU KV cache usage: 0.0%.
>>>  Running only the repeated prompt
fe289733-219d-48ee-96b1-5ce7f9f73874 Has 19 occurrences
8ab9d577-f6c5-4027-812f-c590248d4980 Has 19 occurrences
bf23ac40-77d7-477b-8fb5-808a755a26d5 Has 19 occurrences
8f45a0bf-f4ef-4fc5-b799-8ced0e720f4e Has 19 occurrences
7e0c750b-92f5-424a-9a96-665926da88b2 Has 19 occurrences
cfc1df95-0e83-4b12-8a76-2f91d02dae56 Has 19 occurrences
35462188-01a7-4137-ab94-2a4739015207 Has 19 occurrences
644fdc5b-18e6-44e9-b7f6-bc42dfe4b9aa Has 19 occurrences
3c859be7-67b7-4e99-96a9-29ca027b62bb Has 19 occurrences
2b8afcac-9a76-45b4-8368-a371eb585291 Has 19 occurrences
>>>  Running the repeated prompt with other prompts with same seed
7af075eb-e930-4084-a465-cbc4dc180f96 Has 19 occurrences
5d35b575-4ae6-4187-82a6-372a28fb7ee7 Has 19 occurrences
f341e245-5bf6-4ced-83c8-7524876b3fb9 Has 19 occurrences
325a0f4f-cb5b-44a4-b208-28250b2e19e4 Has 19 occurrences
fb0530a1-0f15-498a-b83c-d57f7e07f0e3 Has 19 occurrences
>>>  Running the repeated prompt with other prompts with different seeds
2b5d3959-a07f-4b81-b431-4e02703b3e4d Has 19 occurrences
a9d0f590-9272-4292-a986-1b65e3d7b1cd Has 19 occurrences
6fe01def-ba33-495b-8c52-90f96bb22208 Has 19 occurrences
c7f2492d-0dc3-4294-aa63-620e1d573b19 Has 19 occurrences
5211e0e9-455a-4ef3-a81e-00d51ef3421a Has 19 occurrences
__________RUNNING WITH SPECULATING DECODING__________
WARNING 10-17 02:06:52 config.py:1674] Casting torch.float16 to torch.bfloat16.
/opt/vllm/lib64/python3.12/site-packages/vllm/connections.py:8: RuntimeWarning: Failed to read commit hash:
No module named 'vllm._version'
  from vllm.version import __version__ as VLLM_VERSION
WARNING 10-17 02:06:56 config.py:395] Async output processing is not supported with speculative decoding currently.
WARNING 10-17 02:07:00 config.py:1674] Casting torch.float16 to torch.bfloat16.
WARNING 10-17 02:07:04 config.py:395] Async output processing is not supported with speculative decoding currently.
INFO 10-17 02:07:04 llm_engine.py:237] Initializing an LLM engine (vdev) with config: model='meta-llama/Llama-3.1-8B-Instruct', speculative_config=SpeculativeConfig(draft_model='ibm-fms/llama3-8b-accelerator', num_spec_tokens=4), tokenizer='meta-llama/Llama-3.1-8B-Instruct', skip_tokenizer_init=False, tokenizer_mode=auto, revision=None, override_neuron_config=None, rope_scaling=None, rope_theta=None, tokenizer_revision=None, trust_remote_code=False, dtype=torch.bfloat16, max_seq_len=1984, download_dir=None, load_format=LoadFormat.AUTO, tensor_parallel_size=1, pipeline_parallel_size=1, disable_custom_all_reduce=False, quantization=None, enforce_eager=False, 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, collect_model_forward_time=False, collect_model_execute_time=False), seed=0, served_model_name=meta-llama/Llama-3.1-8B-Instruct, use_v2_block_manager=True, num_scheduler_steps=1, chunked_prefill_enabled=False multi_step_stream_outputs=True, enable_prefix_caching=False, use_async_output_proc=False, use_cached_outputs=True, mm_processor_kwargs=None)
INFO 10-17 02:07:05 spec_decode_worker.py:169] Configuring SpecDecodeWorker with proposer=<class 'vllm.spec_decode.mlp_speculator_worker.MLPSpeculatorWorker'>
INFO 10-17 02:07:05 rejection_sampler.py:57] Use pytorch for rejection sampling.
INFO 10-17 02:07:05 spec_decode_worker.py:181] [Speculative Decoding] Configuring SpecDecodeWorker with sampler=<class 'vllm.model_executor.layers.rejection_sampler.RejectionSampler'>
INFO 10-17 02:07:05 spec_decode_worker.py:204] [Speculative Decoding] Disabling MQA scorer as the target model is not running in eager mode.
INFO 10-17 02:07:05 model_runner.py:1060] Starting to load model meta-llama/Llama-3.1-8B-Instruct...
INFO 10-17 02:07:05 weight_utils.py:243] Using model weights format ['*.safetensors']

Loading safetensors checkpoint shards:   0% Completed | 0/4 [00:00<?, ?it/s]

Loading safetensors checkpoint shards:  25% Completed | 1/4 [00:00<00:02,  1.44it/s]

Loading safetensors checkpoint shards:  50% Completed | 2/4 [00:01<00:01,  1.23it/s]

Loading safetensors checkpoint shards:  75% Completed | 3/4 [00:02<00:00,  1.20it/s]

Loading safetensors checkpoint shards: 100% Completed | 4/4 [00:02<00:00,  1.60it/s]

Loading safetensors checkpoint shards: 100% Completed | 4/4 [00:02<00:00,  1.45it/s]

INFO 10-17 02:07:08 model_runner.py:1071] Loading model weights took 14.9888 GB
INFO 10-17 02:07:08 model_runner.py:1060] Starting to load model ibm-fms/llama3-8b-accelerator...
INFO 10-17 02:07:09 weight_utils.py:243] Using model weights format ['*.safetensors']

Loading safetensors checkpoint shards:   0% Completed | 0/2 [00:00<?, ?it/s]

Loading safetensors checkpoint shards:  50% Completed | 1/2 [00:00<00:00,  1.28it/s]

Loading safetensors checkpoint shards: 100% Completed | 2/2 [00:01<00:00,  1.81it/s]

Loading safetensors checkpoint shards: 100% Completed | 2/2 [00:01<00:00,  1.70it/s]

INFO 10-17 02:07:10 model_runner.py:1071] Loading model weights took 5.9512 GB
INFO 10-17 02:07:10 spec_decode_worker.py:310] [Speculative Decoding] Use batch expansion for scoring proposals.
INFO 10-17 02:07:12 gpu_executor.py:122] # GPU blocks: 7630, # CPU blocks: 2048
INFO 10-17 02:07:12 gpu_executor.py:126] Maximum concurrency for 1984 tokens per request: 61.53x
INFO 10-17 02:07:13 model_runner.py:1402] Capturing the model for CUDA graphs. This may lead to unexpected consequences if the model is not static. To run the model in eager mode, set 'enforce_eager=True' or use '--enforce-eager' in the CLI.
INFO 10-17 02:07:13 model_runner.py:1406] CUDA graphs can take additional 1~3 GiB memory per GPU. If you are running out of memory, consider decreasing `gpu_memory_utilization` or enforcing eager mode. You can also reduce the `max_num_seqs` as needed to decrease memory usage.
INFO 10-17 02:07:24 model_runner.py:1530] Graph capturing finished in 10 secs.
INFO 10-17 02:07:24 engine.py:292] Added request da9a4028-09d7-4141-a2bc-6b275cd7fa56.
INFO 10-17 02:07:24 engine.py:292] Added request a6c6f527-eef8-43f9-a2df-93f629f44e50.
INFO 10-17 02:07:24 engine.py:292] Added request 61b089c8-9a85-4d01-bc5a-6ab56815f144.
INFO 10-17 02:07:24 engine.py:292] Added request f4f092be-6e22-4507-b3ca-3bd81d3843d6.
INFO 10-17 02:07:24 engine.py:292] Added request 6f4c0aae-f17c-46ff-9af7-ec9f863d318e.
INFO 10-17 02:07:24 engine.py:292] Added request 22dc3f94-57eb-494a-b2eb-4e810b78b93f.
INFO 10-17 02:07:24 engine.py:292] Added request 57808f41-90f3-48b4-9b1d-f6db08a816c1.
INFO 10-17 02:07:24 engine.py:292] Added request b3ac5c88-c9f7-46bc-a7f5-8350921a0fa7.
INFO 10-17 02:07:24 engine.py:292] Added request 76db0527-be1f-47ed-b552-be868b5f850f.
INFO 10-17 02:07:24 engine.py:292] Added request 8837bd8f-23e9-467c-af9e-872ab8a51c82.
INFO 10-17 02:07:24 spec_decode_worker.py:871] SpecDecodeWorker stage times: average_time_per_proposal_tok_ms=1.73 scoring_time_ms=24.96 verification_time_ms=4.70
INFO 10-17 02:07:29 metrics.py:345] Avg prompt throughput: 57.7 tokens/s, Avg generation throughput: 274.8 tokens/s, Running: 10 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 1.7%, CPU KV cache usage: 0.0%.
INFO 10-17 02:07:29 metrics.py:367] Speculative metrics: Draft acceptance rate: 0.500, System efficiency: 0.400, Number of speculative tokens: 4, Number of accepted tokens: 20, Number of draft tokens: 40, Number of emitted tokens: 20.
INFO 10-17 02:07:29 engine.py:292] Added request e24eba59-2b8f-4279-995d-151390473e93.
INFO 10-17 02:07:29 engine.py:292] Added request bb77ed15-9746-4022-a6b0-ac8f32bd3493.
INFO 10-17 02:07:29 engine.py:292] Added request 618e83fa-cc89-4fc7-85d3-8640869ea5b0.
INFO 10-17 02:07:29 engine.py:292] Added request eb3977b1-f247-4d97-a000-ef977c28b857.
INFO 10-17 02:07:29 engine.py:292] Added request e9475f87-cf59-4324-bf76-bd56ab76f305.
INFO 10-17 02:07:29 engine.py:292] Added request 33d40f6d-64b6-4a8f-be23-80155a29401d.
INFO 10-17 02:07:29 engine.py:292] Added request f84f3716-e594-499e-bb48-a714963101dc.
INFO 10-17 02:07:29 engine.py:292] Added request f1dad6b1-5443-4787-8ddf-9dd5b6cc65be.
INFO 10-17 02:07:29 engine.py:292] Added request b6c86031-f57e-4fc7-9fa6-c6bb8f7365e4.
INFO 10-17 02:07:29 engine.py:292] Added request 7dadb741-c8d2-4d18-ab9d-8b785f3bf1b5.
INFO 10-17 02:07:29 spec_decode_worker.py:871] SpecDecodeWorker stage times: average_time_per_proposal_tok_ms=1.64 scoring_time_ms=24.71 verification_time_ms=1.45
INFO 10-17 02:07:34 engine.py:292] Added request a29b2c3d-ab1b-45bd-b1d7-ee8556faced3.
INFO 10-17 02:07:34 engine.py:292] Added request aa537b10-1e01-4faf-b52d-94d0329d2a0a.
INFO 10-17 02:07:34 engine.py:292] Added request ca454e6d-baac-4a50-be9d-ae39b14fc424.
INFO 10-17 02:07:34 engine.py:292] Added request 7481adc3-3b0a-4517-aa27-05d39b6e79f0.
INFO 10-17 02:07:34 engine.py:292] Added request 5a1be1bc-5172-4481-a900-24883c472140.
INFO 10-17 02:07:34 engine.py:292] Added request f7c511bb-1eb4-41c5-833d-0327f2619018.
INFO 10-17 02:07:34 engine.py:292] Added request 7471c619-841c-4a23-9da7-ff987c24b7d7.
INFO 10-17 02:07:34 engine.py:292] Added request a9f74718-3a97-439b-84c1-583cf987c2e3.
INFO 10-17 02:07:34 engine.py:292] Added request 139e87dd-f3da-4723-94fc-3bc17ae58e99.
INFO 10-17 02:07:34 engine.py:292] Added request 32f792fc-f031-4d05-8ca5-bb77a700e7c3.
INFO 10-17 02:07:34 metrics.py:345] Avg prompt throughput: 67.6 tokens/s, Avg generation throughput: 258.3 tokens/s, Running: 10 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 0.3%, CPU KV cache usage: 0.0%.
INFO 10-17 02:07:34 metrics.py:367] Speculative metrics: Draft acceptance rate: 0.181, System efficiency: 0.258, Number of speculative tokens: 4, Number of accepted tokens: 1070, Number of draft tokens: 5920, Number of emitted tokens: 1910.
INFO 10-17 02:07:34 spec_decode_worker.py:871] SpecDecodeWorker stage times: average_time_per_proposal_tok_ms=1.63 scoring_time_ms=24.73 verification_time_ms=1.42
>>>  Running only the repeated prompt
da9a4028-09d7-4141-a2bc-6b275cd7fa56 Has 19 occurrences
a6c6f527-eef8-43f9-a2df-93f629f44e50 Has 19 occurrences
61b089c8-9a85-4d01-bc5a-6ab56815f144 Has 19 occurrences
f4f092be-6e22-4507-b3ca-3bd81d3843d6 Has 19 occurrences
6f4c0aae-f17c-46ff-9af7-ec9f863d318e Has 19 occurrences
22dc3f94-57eb-494a-b2eb-4e810b78b93f Has 19 occurrences
57808f41-90f3-48b4-9b1d-f6db08a816c1 Has 19 occurrences
b3ac5c88-c9f7-46bc-a7f5-8350921a0fa7 Has 19 occurrences
76db0527-be1f-47ed-b552-be868b5f850f Has 19 occurrences
8837bd8f-23e9-467c-af9e-872ab8a51c82 Has 19 occurrences
>>>  Running the repeated prompt with other prompts with same seed
e24eba59-2b8f-4279-995d-151390473e93 Has 19 occurrences
bb77ed15-9746-4022-a6b0-ac8f32bd3493 Has 19 occurrences
618e83fa-cc89-4fc7-85d3-8640869ea5b0 Has 19 occurrences
eb3977b1-f247-4d97-a000-ef977c28b857 Has 19 occurrences
e9475f87-cf59-4324-bf76-bd56ab76f305 Has 19 occurrences
>>>  Running the repeated prompt with other prompts with different seeds
a29b2c3d-ab1b-45bd-b1d7-ee8556faced3 Has 1 occurrences
 I will!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!""!! I!!!"!"!!!"!"!"!"!!!!!"!"!"!"!"!!!!!!!"!"!"!! I!"!"!"!"!"!! I!"!! I!! I""!! I! I""!! I"!! I! I! I"!! I"!! I! I! I"!! I! I"!! I"! I! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"!
aa537b10-1e01-4faf-b52d-94d0329d2a0a Has 1 occurrences
 I will!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!""!! I!!!"!"!!!"!"!"!"!!!!!"!"!"!"!"!!!!!!!"!"!"!! I!"!"!"!"!"!! I!"!! I!! I""!! I! I""!! I"!! I! I! I"!! I"!! I! I! I"!! I! I"!! I"! I! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"!
ca454e6d-baac-4a50-be9d-ae39b14fc424 Has 1 occurrences
 I will!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!""!! I!!!"!"!!!"!"!"!"!!!!!"!"!"!"!"!!!!!!!"!"!"!! I!"!"!"!"!"!! I!"!! I!! I""!! I! I""!! I"!! I! I! I"!! I"!! I! I! I"!! I! I"!! I"! I! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"!
7481adc3-3b0a-4517-aa27-05d39b6e79f0 Has 1 occurrences
 I will!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!""!! I!!!"!"!!!"!"!"!"!!!!!"!"!"!"!"!!!!!!!"!"!"!! I!"!"!"!"!"!! I!"!! I!! I""!! I! I""!! I"!! I! I! I"!! I"!! I! I! I"!! I! I"!! I"! I! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"!
5a1be1bc-5172-4481-a900-24883c472140 Has 1 occurrences
 I will!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!""!! I!!!"!"!!!"!"!"!"!!!!!"!"!"!"!"!!!!!!!"!"!"!! I!"!"!"!"!"!! I!"!! I!! I""!! I! I""!! I"!! I! I! I"!! I"!! I! I! I"!! I! I"!! I"! I! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"! I"!
Done
/usr/lib64/python3.12/multiprocessing/resource_tracker.py:254: UserWarning: resource_tracker: There appear to be 2 leaked semaphore objects to clean up at shutdown
  warnings.warn('resource_tracker: There appear to be %d '

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[venkatakesav]

Hi @wallashss, I'd like to work on this issue. Could you please assign it to me? Thanks!

[wallashss]

Hey @venkatakesav any updates on that?