[Bug] 断点二次训练merge lora失败

linssonSUSUSU commented 3 months ago

Checklist

[X] 1. I have searched related issues but cannot get the expected help.
[ ] 2. The bug has not been fixed in the latest version.
[ ] 3. Please note that if the bug-related issue you submitted lacks corresponding environment info and a minimal reproducible demo, it will be challenging for us to reproduce and resolve the issue, reducing the likelihood of receiving feedback.

Describe the bug

merge lora报错，我是在使用lora微调8b模型的脚本训练了一个checkpoint后又将本次checkpoint作为第二次训练的输入模型。在对第二次微调后的模型进行lora merge

Reproduction

merge_lora.py

Environment

sys.platform: linux
Python: 3.12.3 | packaged by Anaconda, Inc. | (main, May  6 2024, 19:46:43) [GCC 11.2.0]
CUDA available: True
MUSA available: False
numpy_random_seed: 2147483648
GPU 0: NVIDIA GeForce RTX 4090
CUDA_HOME: /usr/local/cuda
NVCC: Cuda compilation tools, release 12.1, V12.1.105
GCC: gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
PyTorch: 2.3.0+cu121
PyTorch compiling details: PyTorch built with:
  - GCC 9.3
  - C++ Version: 201703
  - Intel(R) oneAPI Math Kernel Library Version 2022.2-Product Build 20220804 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v3.3.6 (Git Hash 86e6af5974177e513fd3fee58425e1063e7f1361)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - LAPACK is enabled (usually provided by MKL)
  - NNPACK is enabled
  - CPU capability usage: AVX512
  - CUDA Runtime 12.1
  - NVCC architecture flags: -gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86;-gencode;arch=compute_90,code=sm_90
  - CuDNN 8.9.2
  - Magma 2.6.1
  - Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=12.1, CUDNN_VERSION=8.9.2, CXX_COMPILER=/opt/rh/devtoolset-9/root/usr/bin/c++, CXX_FLAGS= -D_GLIBCXX_USE_CXX11_ABI=0 -fabi-version=11 -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOROCTRACER -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-stringop-overflow -Wsuggest-override -Wno-psabi -Wno-error=pedantic -Wno-error=old-style-cast -Wno-missing-braces -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_VERSION=2.3.0, USE_CUDA=ON, USE_CUDNN=ON, USE_CUSPARSELT=1, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_GLOO=ON, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=1, USE_NNPACK=ON, USE_OPENMP=ON, USE_ROCM=OFF, USE_ROCM_KERNEL_ASSERT=OFF, 

TorchVision: 0.18.0+cu121
LMDeploy: 0.5.3+6a230b3
transformers: 4.40.0
gradio: 4.41.0
fastapi: 0.112.0
pydantic: 2.8.2
triton: 2.3.1
NVIDIA Topology: 
    GPU0    NIC0    NIC1    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X  NODE    NODE    32-63,96-127    1       N/A
NIC0    NODE     X  PIX             
NIC1    NODE    PIX  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

Error traceback

- This IS expected if you are initializing InternVLChatModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing InternVLChatModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of InternVLChatModel were not initialized from the model checkpoint at /root/autodl-tmp/mytrain_2/ and are newly initialized: ['language_model.base_model.model.model.layers.0.attention.wo.lora_A.default.weight', 'language_model.base_model.model.model.layers.0.attention.wo.lora_B.default.weight'
......]
Loading tokenizer...
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
Traceback (most recent call last):
  File "/root/InternVL/internvl_chat/tools/merge_lora.py", line 25, in <module>
    model.language_model.merge_and_unload()
  File "/root/miniconda3/lib/python3.12/site-packages/peft/tuners/lora/model.py", line 838, in merge_and_unload
    return self._unload_and_optionally_merge(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/lib/python3.12/site-packages/peft/tuners/lora/model.py", line 457, in _unload_and_optionally_merge
    target.merge(safe_merge=safe_merge, adapter_names=adapter_names)
  File "/root/miniconda3/lib/python3.12/site-packages/peft/tuners/lora/layer.py", line 472, in merge
    base_layer.weight.data = base_layer.weight.data + delta_weight
                             ~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~
  File "/root/miniconda3/lib/python3.12/site-packages/torch/_prims_common/wrappers.py", line 252, in _fn
    result = fn(*args, **kwargs)
             ^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/lib/python3.12/site-packages/torch/_prims_common/wrappers.py", line 137, in _fn
    result = fn(**bound.arguments)
             ^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/lib/python3.12/site-packages/torch/_refs/__init__.py", line 1091, in add
    output = prims.add(a, b)
             ^^^^^^^^^^^^^^^
  File "/root/miniconda3/lib/python3.12/site-packages/torch/_ops.py", line 594, in __call__
    return self_._op(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/lib/python3.12/site-packages/torch/_prims/__init__.py", line 359, in _prim_elementwise_meta
    utils.check_same_device(*args_, allow_cpu_scalar_tensors=True)
  File "/root/miniconda3/lib/python3.12/site-packages/torch/_prims_common/__init__.py", line 740, in check_same_device
    raise RuntimeError(msg)
RuntimeError: Tensor on device cpu is not on the expected device meta!

VJaGG commented 2 months ago

大佬解决了吗？

linssonSUSUSU commented 2 months ago

没有欸，目前还是调整数据后重启训练

一叶之秋 @.***> 于2024年9月28日周六 13:37写道：

大佬解决了吗？

— Reply to this email directly, view it on GitHub https://github.com/OpenGVLab/InternVL/issues/527#issuecomment-2380417249, or unsubscribe https://github.com/notifications/unsubscribe-auth/BETZR3RGU22ZSAAT4DMG3QLZYY6ATAVCNFSM6AAAAABM3YDAT2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGOBQGQYTOMRUHE . You are receiving this because you authored the thread.Message ID: @.***>

czczup commented 1 week ago

这个bug看着好像可以通过调整一下权重的device来解决

OpenGVLab / InternVL

[Bug] 断点二次训练merge lora失败 #527

Checklist

Describe the bug

Reproduction

Environment

Error traceback