aphrodite-engine
aphrodite-engine copied to clipboard
PygmalionAI
[Bug]: WSL Cuda out of Memory when Trying to Load GGUF Model
Your current environment
PyTorch version: 2.2.0+cu121
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: Could not collect
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-5.15.133.1-microsoft-standard-WSL2-x86_64-with-glibc2.35
Is CUDA available: False
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration:
GPU 0: NVIDIA GeForce RTX 2080 Ti
GPU 1: NVIDIA GeForce RTX 2080 Ti
GPU 2: NVIDIA GeForce RTX 2080 Ti
GPU 3: NVIDIA GeForce RTX 2080 Ti
GPU 4: NVIDIA GeForce RTX 2080 Ti
Nvidia driver version: 551.52
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: 48 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 32
On-line CPU(s) list: 0-31
Vendor ID: AuthenticAMD
Model name: AMD EPYC 7F52 16-Core Processor
CPU family: 23
Model: 49
Thread(s) per core: 2
Core(s) per socket: 16
Socket(s): 1
Stepping: 0
BogoMIPS: 6986.90
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl tsc_reliable nonstop_tsc cpuid extd_apicid pni pclmulqdq ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm cmp_legacy svm cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw topoext perfctr_core ssbd ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 clzero xsaveerptr arat npt nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold v_vmsave_vmload umip rdpid
Virtualization: AMD-V
Hypervisor vendor: Microsoft
Virtualization type: full
L1d cache: 512 KiB (16 instances)
L1i cache: 512 KiB (16 instances)
L2 cache: 8 MiB (16 instances)
L3 cache: 16 MiB (1 instance)
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: Mitigation; untrained return thunk; SMT enabled with STIBP protection
Vulnerability Spec rstack overflow: Mitigation; safe RET
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; Retpolines, IBPB conditional, STIBP always-on, RSB filling, PBRSB-eIBRS Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected
Versions of relevant libraries:
[pip3] numpy==1.26.4
[pip3] torch==2.2.0
[pip3] triton==2.2.0
[conda] Could not collect ROCM Version: Could not collect
Aphrodite Version: 0.5.1
Aphrodite Build Flags:
CUDA Archs: Not Set; ROCm: Disabled
Please note that the system is WSL on windows 11, therefore the env.py is not able to gather the correct version of CUDA due to a known bug. The CUDA version installed is 12.1
🐛 Describe the bug
Trying to run a gguf model that has been converted to safetensors results in a Cuda out of memory error. This occurs after some of the ray workers have finished loading the model. 4 RTX2080ti 22GB were used which should have 88GB of VRAM available.
Launch Parameters
python3 -m aphrodite.endpoints.openai.api_server --model miquliz-120b-v2.0.i1-Q4_K_M/ -tp 4 --api-keys 123456 -q gguf --dtype float16 -gmu 0.95
Error Log
INFO: Model weights loaded. Memory usage: 17.15 GiB x 4 = 68.60 GiB
(RayWorkerAphrodite pid=232346) INFO: Model weights loaded. Memory usage: 17.15 GiB x 4 = 68.60 GiB
(RayWorkerAphrodite pid=232501) WARNING: Custom allreduce is disabled because your platform lacks GPU P2P capability. To silence this warning, specify [repeated 2x across cluster] (Ray deduplicates logs by default. Set RAY_DEDUP_LOGS=0 to disable log deduplication, or see https://docs.ray.io/en/master/ray-observability/ray-logging.html#log-deduplication for more options.)
(RayWorkerAphrodite pid=232501) disable_custom_all_reduce=True explicitly. [repeated 2x across cluster]
Traceback (most recent call last):
File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/usr/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/endpoints/openai/api_server.py", line 563, in <module>
engine = AsyncAphrodite.from_engine_args(engine_args)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/async_aphrodite.py", line 676, in from_engine_args
engine = cls(parallel_config.worker_use_ray,
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/async_aphrodite.py", line 341, in __init__
self.engine = self._init_engine(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/async_aphrodite.py", line 410, in _init_engine
return engine_class(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/aphrodite_engine.py", line 118, in __init__
self._init_cache()
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/aphrodite_engine.py", line 321, in _init_cache
num_blocks = self._run_workers(
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/aphrodite_engine.py", line 1028, in _run_workers
driver_worker_output = getattr(self.driver_worker,
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/task_handler/worker.py", line 136, in profile_num_available_blocks
self.model_runner.profile_run()
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/task_handler/model_runner.py", line 758, in profile_run
self.execute_model(seqs, kv_caches)
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/task_handler/model_runner.py", line 692, in execute_model
hidden_states = model_executable(
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/modeling/models/llama.py", line 413, in forward
hidden_states = self.model(input_ids, positions, kv_caches,
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/modeling/models/llama.py", line 340, in forward
hidden_states, residual = layer(
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/modeling/models/llama.py", line 298, in forward
hidden_states = self.mlp(hidden_states)
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/modeling/models/llama.py", line 114, in forward
x, _ = self.down_proj(x)
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/modeling/layers/linear.py", line 617, in forward
output_parallel = self.linear_method.apply_weights(
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/modeling/layers/quantization/gguf.py", line 137, in apply_weights
out = reshaped_x @ weight.T
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 512.00 MiB. GPU 0 has a total capacity of 22.00 GiB of which 0 bytes is free. Process 232346 has 17179869184.00 GiB memory in use. Including non-PyTorch memory, this process has 17179869184.00 GiB memory in use. Of the allocated memory 20.96 GiB is allocated by PyTorch, and 137.73 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)
(RayWorkerAphrodite pid=232430) INFO: Model weights loaded. Memory usage: 17.15 GiB x 4 = 68.60 GiB [repeated 2x across cluster]
The same error occurs. It appears that one of the ray workers is trying to allocate way too much memory.
torch.cuda.OutOfMemoryError: CUDA out of memory.
Tried to allocate 512.00 MiB. GPU 0 has a total capacity of 22.00 GiB of which 0 bytes is free.
Process 232346 has 17179869184.00 GiB memory in use. Including non-PyTorch memory, this process has 17179869184.00 GiB memory in use.
Of the allocated memory 20.96 GiB is allocated by PyTorch, and 137.73 MiB is reserved by PyTorch but unallocated.
If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)
You may need to lower your context length by specifying --max-model-len 4096
Different error after lowering context length.
*** SIGSEGV received at time=1711454288 on cpu 22 ***
@ 0x7f9340ccc520 1479940200 (unknown)
@ 0x2000 (unknown) (unknown)
@ 0x2 1528629200 (unknown)
@ 0x7f933f8d2e30 (unknown) (unknown)
@ 0xec8348fb89485355 (unknown) (unknown)
[2024-03-26 19:58:08,630 E 235769 235769] logging.cc:361: *** SIGSEGV received at time=1711454288 on cpu 22 ***
[2024-03-26 19:58:08,632 E 235769 235769] logging.cc:361: @ 0x7f9340ccc520 1479940200 (unknown)
[2024-03-26 19:58:08,634 E 235769 235769] logging.cc:361: @ 0x2000 (unknown) (unknown)
[2024-03-26 19:58:08,637 E 235769 235769] logging.cc:361: @ 0x2 1528629200 (unknown)
[2024-03-26 19:58:08,638 E 235769 235769] logging.cc:361: @ 0x7f933f8d2e30 (unknown) (unknown)
[2024-03-26 19:58:08,640 E 235769 235769] logging.cc:361: @ 0xec8348fb89485355 (unknown) (unknown)
Fatal Python error: Segmentation fault
Stack (most recent call first):
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/modeling/layers/quantization/gguf.py", line 135 in apply_weights
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/modeling/layers/vocab_parallel_embedding.py", line 170 in forward
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520 in _call_impl
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511 in _wrapped_call_impl
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/modeling/models/llama.py", line 422 in sample
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/task_handler/model_runner.py", line 700 in execute_model
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115 in decorate_context
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/task_handler/model_runner.py", line 758 in profile_run
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115 in decorate_context
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/task_handler/worker.py", line 136 in profile_num_available_blocks
File "/home/lirikana/.local/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115 in decorate_context
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/aphrodite_engine.py", line 1028 in _run_workers
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/aphrodite_engine.py", line 321 in _init_cache
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/aphrodite_engine.py", line 118 in __init__
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/async_aphrodite.py", line 410 in _init_engine
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/async_aphrodite.py", line 341 in __init__
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/engine/async_aphrodite.py", line 676 in from_engine_args
File "/home/lirikana/.local/lib/python3.10/site-packages/aphrodite/endpoints/openai/api_server.py", line 563 in <module>
File "/usr/lib/python3.10/runpy.py", line 86 in _run_code
File "/usr/lib/python3.10/runpy.py", line 196 in _run_module_as_main
Extension modules: numpy.core._multiarray_umath, numpy.core._multiarray_tests, numpy.linalg._umath_linalg, numpy.fft._pocketfft_internal, numpy.random._common, numpy.random.bit_generator, numpy.random._bounded_integers, numpy.random._mt19937, numpy.random.mtrand, numpy.random._philox, numpy.random._pcg64, numpy.random._sfc64, numpy.random._generator, torch._C, torch._C._fft, torch._C._linalg, torch._C._nested, torch._C._nn, torch._C._sparse, torch._C._special, charset_normalizer.md, yaml._yaml, sentencepiece._sentencepiece, psutil._psutil_linux, psutil._psutil_posix, msgpack._cmsgpack, google._upb._message, setproctitle, ray._raylet, cupy_backends.cuda.api._runtime_enum, cupy_backends.cuda.api.runtime, cupy_backends.cuda.stream, cupy_backends.cuda.libs.cublas, cupy_backends.cuda.libs.cusolver, cupy_backends.cuda._softlink, cupy_backends.cuda.libs.cusparse, cupy._util, cupy.cuda.device, fastrlock.rlock, cupy.cuda.memory_hook, cupy.cuda.graph, cupy.cuda.stream, cupy_backends.cuda.api._driver_enum, cupy_backends.cuda.api.driver, cupy.cuda.memory, cupy._core.internal, cupy._core._carray, cupy.cuda.texture, cupy.cuda.function, cupy_backends.cuda.libs.nvrtc, cupy.cuda.jitify, cupy.cuda.pinned_memory, cupy_backends.cuda.libs.curand, cupy_backends.cuda.libs.profiler, cupy.cuda.common, cupy.cuda.cub, cupy_backends.cuda.libs.nvtx, cupy.cuda.thrust, cupy._core._dtype, cupy._core._scalar, cupy._core._accelerator, cupy._core._memory_range, cupy._core._fusion_thread_local, cupy._core._kernel, cupy._core._routines_manipulation, cupy._core._routines_binary, cupy._core._optimize_config, cupy._core._cub_reduction, cupy._core._reduction, cupy._core._routines_math, cupy._core._routines_indexing, cupy._core._routines_linalg, cupy._core._routines_logic, cupy._core._routines_sorting, cupy._core._routines_statistics, cupy._core.dlpack, cupy._core.flags, cupy._core.core, cupy._core._fusion_variable, cupy._core._fusion_trace, cupy._core._fusion_kernel, cupy._core.new_fusion, cupy._core.fusion, cupy._core.raw, cupyx.cusolver, scipy._lib._ccallback_c, scipy.sparse._sparsetools, _csparsetools, scipy.sparse._csparsetools, scipy.linalg._fblas, scipy.linalg._flapack, scipy.linalg.cython_lapack, scipy.linalg._cythonized_array_utils, scipy.linalg._solve_toeplitz, scipy.linalg._flinalg, scipy.linalg._decomp_lu_cython, scipy.linalg._matfuncs_sqrtm_triu, scipy.linalg.cython_blas, scipy.linalg._matfuncs_expm, scipy.linalg._decomp_update, scipy.sparse.linalg._dsolve._superlu, scipy.sparse.linalg._eigen.arpack._arpack, scipy.sparse.csgraph._tools, scipy.sparse.csgraph._shortest_path, scipy.sparse.csgraph._traversal, scipy.sparse.csgraph._min_spanning_tree, scipy.sparse.csgraph._flow, scipy.sparse.csgraph._matching, scipy.sparse.csgraph._reordering, cupy.cuda.cufft, cupy.fft._cache, cupy.fft._callback, cupy.random._generator_api, cupy.random._bit_generator, scipy._lib._uarray._uarray, scipy.special._ufuncs_cxx, scipy.special._ufuncs, scipy.special._specfun, scipy.special._comb, scipy.special._ellip_harm_2, cupy.lib._polynomial, cupy_backends.cuda.libs.nccl, regex._regex, numba.core.typeconv._typeconv, numba._helperlib, numba._dynfunc, numba._dispatcher, numba.core.runtime._nrt_python, numba.np.ufunc._internal, numba.experimental.jitclass._box, markupsafe._speedups, scipy.optimize._minpack2, scipy.optimize._group_columns, scipy._lib.messagestream, scipy.optimize._trlib._trlib, scipy.optimize._lbfgsb, _moduleTNC, scipy.optimize._moduleTNC, scipy.optimize._cobyla, scipy.optimize._slsqp, scipy.optimize._minpack, scipy.optimize._lsq.givens_elimination, scipy.optimize._zeros, scipy.optimize._highs.cython.src._highs_wrapper, scipy.optimize._highs._highs_wrapper, scipy.optimize._highs.cython.src._highs_constants, scipy.optimize._highs._highs_constants, scipy.linalg._interpolative, scipy.optimize._bglu_dense, scipy.optimize._lsap, scipy.spatial._ckdtree, scipy.spatial._qhull, scipy.spatial._voronoi, scipy.spatial._distance_wrap, scipy.spatial._hausdorff, scipy.spatial.transform._rotation, scipy.optimize._direct (total: 157)
Segmentation fault
Error appears to be happening on line 563 in api_server.py
engine = AsyncAphrodite.from_engine_args(engine_args)
Not sure where exactly it fails since efforts at putting in debug code in the format of
logger.debug()
doesn't yield any messages in the console.
Sorry I've been away for a while. Have you tried the docker image? This is probably a WSL issue. GPU docker on windows uses WSL too, but who knows...
Not sure how to run a docker image on windows. Hopefully official support for windows will be added in the future.