search

PygmalionAI / aphrodite-engine

Large-scale LLM inference engine
https://aphrodite.pygmalion.chat
GNU Affero General Public License v3.0
921 stars 100 forks source link

[Bug]: LORA not working after commit e3f2ea4 "make punica kernels work with rocm" on rc_054 branch #564

Closed Nero10578 closed 1 week ago

Nero10578 commented 1 month ago

Your current environment

Collecting environment information...
PyTorch version: 2.3.0
Is debug build: False
CUDA used to build PyTorch: 12.1
ROCM used to build PyTorch: N/A
OS: Ubuntu 22.04.4 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect 
CMake version: version 3.30.1
Libc version: glibc-2.35
Python version: 3.11.9 (main, Apr 19 2024, 16:48:06) [GCC 11.2.0] (64-bit runtime)
Python platform: Linux-6.5.0-44-generic-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: 12.1.105
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: 
GPU 0: NVIDIA GeForce RTX 3090
GPU 1: NVIDIA GeForce RTX 3090

Nvidia driver version: 550.90.07
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):                             16
On-line CPU(s) list:                0-15
Vendor ID:                          GenuineIntel
Model name:                         Intel(R) Core(TM) i7-6900K CPU @ 3.20GHz
CPU family:                         6
Model:                              79
Thread(s) per core:                 2
Core(s) per socket:                 8
Socket(s):                          1
Stepping:                           1
CPU max MHz:                        4200,0000
CPU min MHz:                        1200,0000
BogoMIPS:                           6411.24
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 arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx 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 pti intel_ppin ssbd ibrs ibpb stibp tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm rdt_a rdseed adx smap intel_pt xsaveopt cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts vnmi md_clear flush_l1d
Virtualization:                     VT-x
L1d cache:                          256 KiB (8 instances)
L1i cache:                          256 KiB (8 instances)
L2 cache:                           2 MiB (8 instances)
L3 cache:                           20 MiB (1 instance)
NUMA node(s):                       1
NUMA node0 CPU(s):                  0-15
Vulnerability Gather data sampling: Not affected
Vulnerability Itlb multihit:        KVM: Mitigation: VMX disabled
Vulnerability L1tf:                 Mitigation; PTE Inversion; VMX conditional cache flushes, SMT vulnerable
Vulnerability Mds:                  Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Meltdown:             Mitigation; PTI
Vulnerability Mmio stale data:      Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Retbleed:             Not affected
Vulnerability Spec rstack overflow: 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; Retpolines; IBPB conditional; IBRS_FW; STIBP conditional; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds:                Not affected
Vulnerability Tsx async abort:      Mitigation; Clear CPU buffers; SMT vulnerable
Versions of relevant libraries:
[pip3] numpy==1.26.4
[pip3] torch==2.3.0
[pip3] torchaudio==2.3.0
[pip3] torchvision==0.18.0
[pip3] triton==2.3.0
[conda] blas                      1.0                         mkl  
[conda] ffmpeg                    4.3                  hf484d3e_0    pytorch
[conda] libjpeg-turbo             2.0.0                h9bf148f_0    pytorch
[conda] mkl                       2023.1.0         h213fc3f_46344  
[conda] mkl-service               2.4.0           py311h5eee18b_1  
[conda] mkl_fft                   1.3.8           py311h5eee18b_0  
[conda] mkl_random                1.2.4           py311hdb19cb5_0  
[conda] numpy                     1.26.4          py311h08b1b3b_0  
[conda] numpy-base                1.26.4          py311hf175353_0  
[conda] pytorch                   2.3.0           py3.11_cuda12.1_cudnn8.9.2_0    pytorch
[conda] pytorch-cuda              12.1                 ha16c6d3_5    pytorch
[conda] pytorch-mutex             1.0                        cuda    pytorch
[conda] torch                     2.3.0                    pypi_0    pypi
[conda] torchaudio                2.3.0               py311_cu121    pytorch
[conda] torchtriton               2.3.0                     py311    pytorch
[conda] torchvision               0.18.0              py311_cu121    pytorch
[conda] triton                    2.3.0                    pypi_0    pypiROCM Version: Could not collect 
Aphrodite Version: 0.5.3
Aphrodite Build Flags:
CUDA Archs: Not Set; ROCm: Disabled

🐛 Describe the bug

On rc_054, after commit e3f2ea4 Using LORA is not possible even after building with environment variable set.

export APHRODITE_INSTALL_PUNICA_KERNELS=1
APHRODITE_INSTALL_PUNICA_KERNELS=1 pip install -e .

Everytime I try to load with LORA module it outputs this:

python -m aphrodite.endpoints.openai.api_server \
--model /home/user/models/Meta-Llama-3-8B-Instruct \
--gpu-memory-utilization 0.95 --max-model-len 8192 --port 8000 \
--max-num-seqs 200 --served-model-name Meta-Llama-3-8B-Instruct --enforce-eager true --tensor-parallel 2 \
--enable-lora --max-lora-rank 64 --lora-modules \
Indo-Formax=/home/user/loras/L3-8B-Indo-Formax
INFO:     ----------------------------------------------------------------------------
INFO:     Initializing the Aphrodite Engine (v0.5.3) with the following config:
INFO:     Model = '/home/user/models/Meta-Llama-3-8B-Instruct'
INFO:     Speculative Config = None
INFO:     DataType = torch.bfloat16
INFO:     Model Load Format = LoadFormat.AUTO
INFO:     Number of GPUs = 2
INFO:     Disable Custom All-Reduce = False
INFO:     Quantization Format = None
INFO:     Context Length = 8192
INFO:     Enforce Eager Mode = True
INFO:     Prefix Caching = False
INFO:     KV Cache DataType = auto
INFO:     Device = cuda
INFO:     Rope Scaling = None
INFO:     Guided Decoding Backend = DecodingConfig(guided_decoding_backend='outlines')
INFO:     ----------------------------------------------------------------------------
(AphroditeWorkerProcess pid=72452) INFO:     Worker ready; awaiting tasks
(AphroditeWorkerProcess pid=72452) INFO:     Found nccl from library /lib/x86_64-linux-gnu/libnccl.so.2
INFO:     Found nccl from library /lib/x86_64-linux-gnu/libnccl.so.2
(AphroditeWorkerProcess pid=72452) INFO:     Aphrodite is using nccl==2.18.3
INFO:     Aphrodite is using nccl==2.18.3
INFO:     reading GPU P2P access cache from /home/user/.config/aphrodite/gpu_p2p_access_cache_for_0,1.json
(AphroditeWorkerProcess pid=72452) INFO:     reading GPU P2P access cache from /home/user/.config/aphrodite/gpu_p2p_access_cache_for_0,1.json
(AphroditeWorkerProcess pid=72452) WARNING:  Custom allreduce is disabled because your platform lacks GPU P2P capability or P2P test failed. To silence this 
(AphroditeWorkerProcess pid=72452) warning, specify disable_custom_all_reduce=True explicitly.
WARNING:  Custom allreduce is disabled because your platform lacks GPU P2P capability or P2P test failed. To silence this 
warning, specify disable_custom_all_reduce=True explicitly.
INFO:     Rank 0: Model weights loaded in 2.09 secs.
INFO:     Memory usage: 7.48 GiB x 2 = 14.97 GiB
(AphroditeWorkerProcess pid=72452) INFO:     Rank 1: Model weights loaded in 2.12 secs.
[rank0]: Traceback (most recent call last):
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/lora/punica.py", line 45, in bgmv
[rank0]:     import aphrodite._punica_C as punica_kernels
[rank0]: ImportError: /home/user/aphrodite-engine/aphrodite/_punica_C.cpython-311-x86_64-linux-gnu.so: undefined symbol: _Z23dispatch_bgmv_low_levelN2at6TensorES0_S0_lflll

[rank0]: The above exception was the direct cause of the following exception:

[rank0]: Traceback (most recent call last):
[rank0]:   File "<frozen runpy>", line 198, in _run_module_as_main
[rank0]:   File "<frozen runpy>", line 88, in _run_code
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/endpoints/openai/api_server.py", line 589, in <module>
[rank0]:     run_server(args)
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/endpoints/openai/api_server.py", line 542, in run_server
[rank0]:     engine = AsyncAphrodite.from_engine_args(engine_args)
[rank0]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/engine/async_aphrodite.py", line 386, in from_engine_args
[rank0]:     engine = cls(
[rank0]:              ^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/engine/async_aphrodite.py", line 341, in __init__
[rank0]:     self.engine = self._init_engine(*args, **kwargs)
[rank0]:                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/engine/async_aphrodite.py", line 461, in _init_engine
[rank0]:     return engine_class(*args, **kwargs)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/engine/aphrodite_engine.py", line 160, in __init__
[rank0]:     self._initialize_kv_caches()
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/engine/aphrodite_engine.py", line 202, in _initialize_kv_caches
[rank0]:     self.model_executor.determine_num_available_blocks())
[rank0]:     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/executor/distributed_gpu_executor.py", line 35, in determine_num_available_blocks
[rank0]:     num_blocks = self._run_workers("determine_num_available_blocks", )
[rank0]:                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/executor/multiproc_gpu_executor.py", line 113, in _run_workers
[rank0]:     driver_worker_output = driver_worker_method(*args, **kwargs)
[rank0]:                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/miniconda3/envs/aphro/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
[rank0]:     return func(*args, **kwargs)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/task_handler/worker.py", line 157, in determine_num_available_blocks
[rank0]:     self.model_runner.profile_run()
[rank0]:   File "/home/user/miniconda3/envs/aphro/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
[rank0]:     return func(*args, **kwargs)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/task_handler/model_runner.py", line 825, in profile_run
[rank0]:     self.execute_model(seqs, kv_caches)
[rank0]:   File "/home/user/miniconda3/envs/aphro/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
[rank0]:     return func(*args, **kwargs)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/task_handler/model_runner.py", line 743, in execute_model
[rank0]:     hidden_states = model_executable(**execute_model_kwargs)
[rank0]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/miniconda3/envs/aphro/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
[rank0]:     return self._call_impl(*args, **kwargs)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/miniconda3/envs/aphro/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
[rank0]:     return forward_call(*args, **kwargs)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/modeling/models/llama.py", line 366, in forward
[rank0]:     hidden_states = self.model(input_ids, positions, kv_caches,
[rank0]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/miniconda3/envs/aphro/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
[rank0]:     return self._call_impl(*args, **kwargs)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/miniconda3/envs/aphro/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
[rank0]:     return forward_call(*args, **kwargs)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/modeling/models/llama.py", line 287, in forward
[rank0]:     hidden_states = self.get_input_embeddings(input_ids)
[rank0]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/modeling/models/llama.py", line 274, in get_input_embeddings
[rank0]:     return self.embed_tokens(input_ids)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/miniconda3/envs/aphro/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
[rank0]:     return self._call_impl(*args, **kwargs)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/miniconda3/envs/aphro/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
[rank0]:     return forward_call(*args, **kwargs)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/lora/layers.py", line 339, in forward
[rank0]:     bgmv(full_output, full_lora_a_embeddings, self.lora_b_stacked,
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/lora/punica.py", line 47, in bgmv
[rank0]:     _raise_import_error(e)
[rank0]:   File "/home/user/aphrodite-engine/aphrodite/lora/punica.py", line 13, in _raise_import_error
[rank0]:     raise ImportError(
[rank0]: ImportError: punica LoRA kernels could not be imported. If you built Aphrodite from source, make sure APHRODITE_INSTALL_PUNICA_KERNELS=1 env var was set.
AlpinDale commented 1 month ago

Thanks for reporting. Does this only happen when you attempt launching a lora?

Nero10578 commented 1 month ago

Thanks for reporting. Does this only happen when you attempt launching a lora?

Yeap running without lora works just fine.

Nero10578 commented 1 week ago

Fixed in latest rc_054