linux单卡训练正常多卡训练报错

[magicwang1111]

Traceback (most recent call last):
File "/mnt/data/wangxi/lora-scripts/./sd-scripts/train_db.py", line 501, in
train(args)
File "/mnt/data/wangxi/lora-scripts/./sd-scripts/train_db.py", line 321, in train
encoder_hidden_states = train_util.get_hidden_states(
File "/mnt/data/wangxi/lora-scripts/sd-scripts/library/train_util.py", line 4003, in get_hidden_states
encoder_hidden_states = text_encoder.text_model.final_layer_norm(encoder_hidden_states)
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1614, in getattr
raise AttributeError("'{}' object has no attribute '{}'".format(
AttributeError: 'DistributedDataParallel' object has no attribute 'text_model'
steps: 0%| | 0/22080 [00:01<?, ?it/s]
Traceback (most recent call last):
File "/mnt/data/wangxi/lora-scripts/./sd-scripts/train_db.py", line 501, in
train(args)
File "/mnt/data/wangxi/lora-scripts/./sd-scripts/train_db.py", line 321, in train
encoder_hidden_states = train_util.get_hidden_states(
File "/mnt/data/wangxi/lora-scripts/sd-scripts/library/train_util.py", line 4003, in get_hidden_states
encoder_hidden_states = text_encoder.text_model.final_layer_norm(encoder_hidden_states)
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1614, in getattr
raise AttributeError("'{}' object has no attribute '{}'".format(
AttributeError: 'DistributedDataParallel' object has no attribute 'text_model'
^CWARNING:torch.distributed.elastic.agent.server.api:Received 2 death signal, shutting down workers
Traceback (most recent call last):
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/accelerate/commands/launch.py", line 996, in
main()
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/accelerate/commands/launch.py", line 992, in main
launch_command(args)
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/accelerate/commands/launch.py", line 977, in launch_command
multi_gpu_launcher(args)
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/accelerate/commands/launch.py", line 646, in multi_gpu_launcher
distrib_run.run(args)
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/torch/distributed/run.py", line 785, in run
elastic_launch(
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 134, in call
return launch_agent(self._config, self._entrypoint, list(args))
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 241, in launch_agent
result = agent.run()
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/torch/distributed/elastic/metrics/api.py", line 129, in wrapper
result = f(*args, **kwargs)
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/torch/distributed/elastic/agent/server/api.py", line 723, in run
result = self._invoke_run(role)
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/torch/distributed/elastic/agent/server/api.py", line 864, in _invoke_run
time.sleep(monitor_interval)
File "/home/wangxi/miniconda3/envs/lora/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/api.py", line 62, in _terminate_process_handler
raise SignalException(f"Process {os.getpid()} got signal: {sigval}", sigval=sigval)
torch.distributed.elastic.multiprocessing.api.SignalException: Process 35054 got signal: 2
12:51:03-224816 ERROR Training failed / 训练失败
(lora) [wangxi@v100-4 lora-scripts]$ ^C
(lora) [wangxi@v100-4 lora-scripts]$

[askyang]

请教一下，linux，ubuntu 上如何安装成功的

[ik4sumi]

我也遇到了这个问题

[shanshouchen]

单卡训练正常，多卡训练卡在
loading VAE from checkpoint VAE: <All keys matched successfully>不走了，
但是显卡利用了一直保持在99%不变，也不报错，也米有任何进度条，各位有碰到过这类问题吗？

以下是我的参数

model_train_type = "sdxl-finetune"
pretrained_model_name_or_path = "/gemini/pretrain/juggernautXL_version6Rundiffusion.safetensors"
v2 = false
train_data_dir = "/gemini/code/data-3"
resolution = "1024,1024"
enable_bucket = true
min_bucket_reso = 256
max_bucket_reso = 1024
bucket_reso_steps = 64
output_name = "xysSDXL_jugg_dir1375and32106_rep10_e200_bs1_test0001"
output_dir = "./output"
save_model_as = "safetensors"
save_precision = "bf16"
save_every_n_epochs = 40
max_train_epochs = 800
train_batch_size = 4
gradient_checkpointing = true
learning_rate = 0.00001
learning_rate_te1 = 0.0000025
learning_rate_te2 = 0.0000025
lr_scheduler = "cosine_with_restarts"
lr_warmup_steps = 0
lr_scheduler_num_cycles = 0
optimizer_type = "AdamW8bit"
log_with = "tensorboard"
logging_dir = "./logs"
caption_extension = ".txt"
shuffle_caption = true
weighted_captions = false
keep_tokens = 1
max_token_length = 255
seed = 1337
no_token_padding = false
mixed_precision = "bf16"
full_bf16 = true
xformers = true
lowram = false
cache_latents = true
cache_latents_to_disk = true
persistent_data_loader_workers = true
gpu_ids = [ "0", "2" ]
train_text_encoder = true

@Akegarasu