search

vllm-project / vllm

A high-throughput and memory-efficient inference and serving engine for LLMs
https://docs.vllm.ai
Apache License 2.0
23.36k stars 3.33k forks source link

[Bug]: Batch expansion doesn't work with lora #6160

Open Adhyyan1252 opened 2 weeks ago

Adhyyan1252 commented 2 weeks ago

Your current environment


GPU 0: NVIDIA H100 80GB HBM3
GPU 1: NVIDIA H100 80GB HBM3
GPU 2: NVIDIA H100 80GB HBM3
GPU 3: NVIDIA H100 80GB HBM3
GPU 4: NVIDIA H100 80GB HBM3
GPU 5: NVIDIA H100 80GB HBM3
GPU 6: NVIDIA H100 80GB HBM3
GPU 7: NVIDIA H100 80GB HBM3

Nvidia driver version: 535.183.01
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):                             192
On-line CPU(s) list:                0-191
Vendor ID:                          GenuineIntel
Model name:                         Intel(R) Xeon(R) Platinum 8468
CPU family:                         6
Model:                              143
Thread(s) per core:                 2
Core(s) per socket:                 48
Socket(s):                          2
Stepping:                           8
CPU max MHz:                        3800.0000
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 hwp hwp_act_window hwp_epp hwp_pkg_req 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:                          4.5 MiB (96 instances)
L1i cache:                          3 MiB (96 instances)
L2 cache:                           192 MiB (96 instances)
L3 cache:                           210 MiB (2 instances)
NUMA node(s):                       2
NUMA node0 CPU(s):                  0-47,96-143
NUMA node1 CPU(s):                  48-95,144-191
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
Vul
[pip3] torchvision==0.18.0
[pip3] transformers==4.42.3
[pip3] triton==2.3.0
[conda] Could not collect
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.5.0.post1
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
GPU0    GPU1    GPU2    GPU3    GPU4    GPU5    GPU6    GPU7    NIC0    NIC1    NIC2    NIC3    NIC4    NIC5    NIC6    NIC7    NIC8    NIC9    NIC10   NIC11   CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X      NV18    NV18    NV18    NV18    NV18    NV18    NV18    PXB     PXB     NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS     0-47,96-143     0               N/A
GPU1    NV18     X      NV18    NV18    NV18    NV18    NV18    NV18    PXB     PXB     NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS     0-47,96-143     0               N/A
GPU2    NV18    NV18     X      NV18    NV18    NV18    NV18    NV18    NODE    NODE    NODE    NODE    PXB     PXB     SYS     SYS     SYS     SYS     SYS     SYS     0-47,96-143     0               N/A
GPU3    NV18    NV18    NV18     X      NV18    NV18    NV18    NV18    NODE    NODE    NODE    NODE    PXB     PXB     SYS     SYS     SYS     SYS     SYS     SYS     0-47,96-143     0               N/A
GPU4    NV18    NV18    NV18    NV18     X      NV18    NV18    NV18    SYS     SYS     SYS     SYS     SYS     SYS     PXB     PXB     NODE    NODE    NODE    NODE    48-95,144-191   1               N/A
GPU5    NV18    NV18    NV18    NV18    NV18     X      NV18    NV18    SYS     SYS     SYS     SYS     SYS     SYS     PXB     PXB     NODE    NODE    NODE    NODE    48-95,144-191   1               N/A
GPU6    NV18    NV18    NV18    NV18    NV18    NV18     X      NV18    SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    PXB     PXB     48-95,144-191   1               N/A
GPU7    NV18    NV18    NV18    NV18    NV18    NV18    NV18     X      SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    PXB     PXB     48-95,144-191   1               N/A
NIC0    PXB     PXB     NODE    NODE    SYS     SYS     SYS     SYS      X      PXB     NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS
NIC1    PXB     PXB     NODE    NODE    SYS     SYS     SYS     SYS     PXB      X      NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS
NIC2    NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     NODE    NODE     X      PIX     NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS
NIC3    NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     NODE    NODE    PIX      X      NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS
NIC4    NODE    NODE    PXB     PXB     SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE     X      PXB     SYS     SYS     SYS     SYS     SYS     SYS
NIC5    NODE    NODE    PXB     PXB     SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    PXB      X      SYS     SYS     SYS     SYS     SYS     SYS
NIC6    SYS     SYS     SYS     SYS     PXB     PXB     NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS      X      PXB     NODE    NODE    NODE    NODE
NIC7    SYS     SYS     SYS     SYS     PXB     PXB     NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS     PXB      X      NODE    NODE    NODE    NODE
NIC8    SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE     X      PIX     NODE    NODE
NIC9    SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE    PIX      X      NODE    NODE
NIC10   SYS     SYS     SYS     SYS     NODE    NODE    PXB     PXB     SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE     X      PXB
NIC11   SYS     SYS     SYS     SYS     NODE    NODE    PXB     PXB     SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    PXB      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
  NIC1: mlx5_1
  NIC2: mlx5_2
  NIC3: mlx5_3
  NIC4: mlx5_4
  NIC5: mlx5_5
  NIC6: mlx5_6
  NIC7: mlx5_7
  NIC8: mlx5_8
  NIC9: mlx5_9
  NIC10: mlx5_10
  NIC11: mlx5_11

🐛 Describe the bug

Spec decode doesn't work with lora

Running an OpenAI server with Mixtral8x22B with r=2 lora

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
python3 -u -m vllm.entrypoints.openai.api_server \
--host 0.0.0.0 \
--model /scratch/8x22_converted \
--api-key 'HcswvB_5lHt0RSckCtIa7u2yWwrIt5ky-koolncR5O4' \
--tensor-parallel-size 8 \
--served-model-name "base" \
--tokenizer /scratch/8x22_converted/ \
--max-model-len 32000 \
--max-num-batched-tokens 32000 \
--enforce-eager \
--enable-lora \
--max-loras 4 \
--lora-modules summary=/scratch/lora_modules/summary_0703 \
--disable-log-requests \
--speculative-model [ngram] \
--num-speculative-tokens 128 \
--ngram-prompt-lookup-max 8 \
--ngram-prompt-lookup-min 8 \
--use-v2-block-manager

ERROR 07-05 13:40:56 async_llm_engine.py:53] Traceback (most recent call last):
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/home/adhyyan/vllm/vllm/engine/async_llm_engine.py", line 43, in _log_task_completion
ERROR 07-05 13:40:56 async_llm_engine.py:53]     return_value = task.result()
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/home/adhyyan/vllm/vllm/engine/async_llm_engine.py", line 588, in run_engine_loop
ERROR 07-05 13:40:56 async_llm_engine.py:53]     result = task.result()
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/home/adhyyan/vllm/vllm/engine/async_llm_engine.py", line 533, in engine_step
ERROR 07-05 13:40:56 async_llm_engine.py:53]     request_outputs = await self.engine.step_async(virtual_engine)
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/home/adhyyan/vllm/vllm/engine/async_llm_engine.py", line 239, in step_async
ERROR 07-05 13:40:56 async_llm_engine.py:53]     output = await self.model_executor.execute_model_async(
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/home/adhyyan/vllm/vllm/executor/distributed_gpu_executor.py", line 173, in execute_model_async
ERROR 07-05 13:40:56 async_llm_engine.py:53]     return await self._driver_execute_model_async(execute_model_req)
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/home/adhyyan/vllm/vllm/executor/multiproc_gpu_executor.py", line 157, in _driver_execute_model_async
ERROR 07-05 13:40:56 async_llm_engine.py:53]     return await self.driver_exec_model(execute_model_req)
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/usr/lib/python3.10/concurrent/futures/thread.py", line 58, in run
ERROR 07-05 13:40:56 async_llm_engine.py:53]     result = self.fn(*self.args, **self.kwargs)
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/home/adhyyan/vllm/.vllm_venv/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
ERROR 07-05 13:40:56 async_llm_engine.py:53]     return func(*args, **kwargs)
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/home/adhyyan/vllm/vllm/spec_decode/spec_decode_worker.py", line 341, in execute_model
ERROR 07-05 13:40:56 async_llm_engine.py:53]     return self._run_speculative_decoding_step(execute_model_req,
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/usr/lib/python3.10/contextlib.py", line 79, in inner
ERROR 07-05 13:40:56 async_llm_engine.py:53]     return func(*args, **kwds)
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/home/adhyyan/vllm/vllm/spec_decode/spec_decode_worker.py", line 453, in _run_speculative_decoding_step
ERROR 07-05 13:40:56 async_llm_engine.py:53]     proposal_scores = self.scorer.score_proposals(
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/usr/lib/python3.10/contextlib.py", line 79, in inner
ERROR 07-05 13:40:56 async_llm_engine.py:53]     return func(*args, **kwds)
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/home/adhyyan/vllm/vllm/spec_decode/batch_expansion.py", line 86, in score_proposals
ERROR 07-05 13:40:56 async_llm_engine.py:53]     all_tokens, all_probs, spec_logprobs = self._contract_batch(
ERROR 07-05 13:40:56 async_llm_engine.py:53]   File "/home/adhyyan/vllm/vllm/spec_decode/batch_expansion.py", line 182, in _contract_batch
ERROR 07-05 13:40:56 async_llm_engine.py:53]     all_probs[non_spec_indices, :1, :] = non_spec_target_probs
ERROR 07-05 13:40:56 async_llm_engine.py:53] RuntimeError: shape mismatch: value tensor of shape [4, 1, 33024] cannot be broadcast to indexing result of shape [4, 1, 32768]
Exception in callback functools.partial(<function _log_task_completion at 0x7f22220cf2e0>, error_callback=<bound method AsyncLLMEngine._error_callback of <vllm.engine.async_llm_engine.AsyncLLMEngine object at 0x7f22056eeef0>>)
handle: <Handle functools.partial(<function _log_task_completion at 0x7f22220cf2e0>, error_callback=<bound method AsyncLLMEngine._error_callback of <vllm.engine.async_llm_engine.AsyncLLMEngine object at 0x7f22056eeef0>>)>```

The vocab size is 32769 so I am not sure where 33024 comes from. Seems like an issue with the batch_expansion tensor logic. Spec decode and Lora work fine independently but when mixed it breaks.
Adhyyan1252 commented 2 weeks ago

@LiuXiaoxuanPKU Seems like an issue with batch expansion

DhruvaBansal00 commented 1 week ago

I believe a similar issue is also leading to this - https://github.com/vllm-project/vllm/issues/4872