微调chatglm3-6b报错Could not locate the tokenization_chatglm.py inside THUDM/chatglm3-6b.

[fan2goa1]

根据xtuner提供的chatglm3-alpaca-qlora的config文件进行了相关更改，采用加载本地模型和自创数据集（格式没有问题）。按照同样的方法微调internlm、mistral、qwen都没有问题，但是微调chatglm3-6b时报错：

Could not locate the tokenization_chatglm.py inside THUDM/chatglm3-6b.
Traceback (most recent call last):
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/urllib3/connection.py", line 198, in _new_conn
    sock = connection.create_connection(
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/urllib3/util/connection.py", line 85, in create_connection
    raise err
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/urllib3/util/connection.py", line 73, in create_connection
    sock.connect(sa)
OSError: [Errno 101] Network is unreachable。`

按照网上的修改，将tokenizer_config.json中的

"AutoTokenizer": [
  "THUDM/chatglm3-6b--tokenization_chatglm.ChatGLMTokenizer",
  null
]

改为

"AutoTokenizer": [
  "tokenization_chatglm.ChatGLMTokenizer",
  null
]

之后仍会报其他错:

Traceback (most recent call last):
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/xtuner/tools/train.py", line 307, in <module>
    main()
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/xtuner/tools/train.py", line 303, in main
    runner.train()
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/runner/_flexible_runner.py", line 1160, in train
    self._train_loop = self.build_train_loop(
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/runner/_flexible_runner.py", line 958, in build_train_loop
    loop = LOOPS.build(
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/registry/registry.py", line 570, in build
    return self.build_func(cfg, *args, **kwargs, registry=self)
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/registry/build_functions.py", line 121, in build_from_cfg
    obj = obj_cls(**args)  # type: ignore
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/xtuner/engine/runner/loops.py", line 32, in __init__
    dataloader = runner.build_dataloader(
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/runner/_flexible_runner.py", line 824, in build_dataloader
    dataset = DATASETS.build(dataset_cfg)
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/registry/registry.py", line 570, in build
    return self.build_func(cfg, *args, **kwargs, registry=self)
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/registry/build_functions.py", line 121, in build_from_cfg
    obj = obj_cls(**args)  # type: ignore
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/xtuner/dataset/huggingface.py", line 235, in process_hf_dataset
    dataset = process(*args, **kwargs)
TypeError: process() got an unexpected keyword argument 'ataset'
Exception ignored in atexit callback: <function matmul_ext_update_autotune_table at 0x7f6911ca11b0>

[LZHgrla]

@fan2goa1

config 中，是否给 process_hf_dataset 传入了一些不应传入的参数？比如报错中提到的 ataset

[fan2goa1]

@fan2goa1

config 中，是否给 process_hf_dataset 传入了一些不应传入的参数？比如报错中提到的 ataset

@LZHgrla

我的config文件是从模板中copy出来并将模型和数据集改成自己的，其他没有动，包括process_hf_dataset.

# Copyright (c) OpenMMLab. All rights reserved.
import torch
from datasets import load_dataset
from mmengine.dataset import DefaultSampler
from mmengine.hooks import (CheckpointHook, DistSamplerSeedHook, IterTimerHook,
                            LoggerHook, ParamSchedulerHook)
from mmengine.optim import AmpOptimWrapper, CosineAnnealingLR, LinearLR
from peft import LoraConfig
from torch.optim import AdamW
from transformers import (AutoModelForCausalLM, AutoTokenizer,
                          BitsAndBytesConfig)

from xtuner.dataset import process_hf_dataset
from xtuner.dataset.collate_fns import default_collate_fn
from xtuner.dataset.map_fns import alpaca_map_fn, template_map_fn_factory
from xtuner.engine.hooks import (DatasetInfoHook, EvaluateChatHook,
                                 VarlenAttnArgsToMessageHubHook)
from xtuner.engine.runner import TrainLoop
from xtuner.model import SupervisedFinetune
from xtuner.utils import PROMPT_TEMPLATE, SYSTEM_TEMPLATE

#######################################################################
#                          PART 1  Settings                           #
#######################################################################
# Model
pretrained_model_name_or_path = '/mnt/hysc/models/chatglm3-6b'
use_varlen_attn = False

# Data
data_path = 'xxxx'
prompt_template = PROMPT_TEMPLATE.chatglm3
max_length = 2048
pack_to_max_length = True

# Scheduler & Optimizer
batch_size = 1  # per_device
accumulative_counts = 16
dataloader_num_workers = 0
max_epochs = 3
optim_type = AdamW
lr = 2e-4
betas = (0.9, 0.999)
weight_decay = 0
max_norm = 1  # grad clip
warmup_ratio = 0.03

# Save
save_steps = 500
save_total_limit = 2  # Maximum checkpoints to keep (-1 means unlimited)

# Evaluate the generation performance during the training
evaluation_freq = 500
SYSTEM = ""
evaluation_inputs = [
]

#######################################################################
#                      PART 2  Model & Tokenizer                      #
#######################################################################
tokenizer = dict(
    type=AutoTokenizer.from_pretrained,
    pretrained_model_name_or_path=pretrained_model_name_or_path,
    trust_remote_code=True,
    encode_special_tokens=True,
    padding_side='left')

model = dict(
    type=SupervisedFinetune,
    use_varlen_attn=use_varlen_attn,
    llm=dict(
        type=AutoModelForCausalLM.from_pretrained,
        pretrained_model_name_or_path=pretrained_model_name_or_path,
        trust_remote_code=True,
        torch_dtype=torch.float16,
        quantization_config=dict(
            type=BitsAndBytesConfig,
            load_in_4bit=True,
            load_in_8bit=False,
            llm_int8_threshold=6.0,
            llm_int8_has_fp16_weight=False,
            bnb_4bit_compute_dtype=torch.float16,
            bnb_4bit_use_double_quant=True,
            bnb_4bit_quant_type='nf4')),
    lora=dict(
        type=LoraConfig,
        r=64,
        lora_alpha=16,
        lora_dropout=0.1,
        bias='none',
        task_type='CAUSAL_LM'))

#######################################################################
#                      PART 3  Dataset & Dataloader                   #
#######################################################################
train_dataset = dict(
    type=process_hf_dataset,
    ataset=dict(
        type=load_dataset, path='json', data_files=dict(train=data_path)),
    tokenizer=tokenizer,
    max_length=max_length,
    dataset_map_fn=None,
    template_map_fn=dict(
        type=template_map_fn_factory, template=prompt_template),
    remove_unused_columns=True,
    shuffle_before_pack=True,
    pack_to_max_length=pack_to_max_length,
    use_varlen_attn=use_varlen_attn)

train_dataloader = dict(
    batch_size=batch_size,
    num_workers=dataloader_num_workers,
    dataset=train_dataset,
    sampler=dict(type=DefaultSampler, shuffle=True),
    collate_fn=dict(type=default_collate_fn, use_varlen_attn=use_varlen_attn))

#######################################################################
#                    PART 4  Scheduler & Optimizer                    #
#######################################################################
# optimizer
optim_wrapper = dict(
    type=AmpOptimWrapper,
    optimizer=dict(
        type=optim_type, lr=lr, betas=betas, weight_decay=weight_decay),
    clip_grad=dict(max_norm=max_norm, error_if_nonfinite=False),
    accumulative_counts=accumulative_counts,
    loss_scale='dynamic',
    dtype='float16')

# learning policy
# More information: https://github.com/open-mmlab/mmengine/blob/main/docs/en/tutorials/param_scheduler.md  # noqa: E501
param_scheduler = [
    dict(
        type=LinearLR,
        start_factor=1e-5,
        by_epoch=True,
        begin=0,
        end=warmup_ratio * max_epochs,
        convert_to_iter_based=True),
    dict(
        type=CosineAnnealingLR,
        eta_min=0.0,
        by_epoch=True,
        begin=warmup_ratio * max_epochs,
        end=max_epochs,
        convert_to_iter_based=True)
]

# train, val, test setting
train_cfg = dict(type=TrainLoop, max_epochs=max_epochs)

#######################################################################
#                           PART 5  Runtime                           #
#######################################################################
# Log the dialogue periodically during the training process, optional
custom_hooks = [
    dict(type=DatasetInfoHook, tokenizer=tokenizer),
    dict(
        type=EvaluateChatHook,
        tokenizer=tokenizer,
        every_n_iters=evaluation_freq,
        evaluation_inputs=evaluation_inputs,
        system=SYSTEM,
        prompt_template=prompt_template)
]

if use_varlen_attn:
    custom_hooks += [dict(type=VarlenAttnArgsToMessageHubHook)]

# configure default hooks
default_hooks = dict(
    # record the time of every iteration.
    timer=dict(type=IterTimerHook),
    # print log every 10 iterations.
    logger=dict(type=LoggerHook, log_metric_by_epoch=False, interval=10),
    # enable the parameter scheduler.
    param_scheduler=dict(type=ParamSchedulerHook),
    # save checkpoint per `save_steps`.
    checkpoint=dict(
        type=CheckpointHook,
        by_epoch=False,
        interval=save_steps,
        max_keep_ckpts=save_total_limit),
    # set sampler seed in distributed evrionment.
    sampler_seed=dict(type=DistSamplerSeedHook),
)

# configure environment
env_cfg = dict(
    # whether to enable cudnn benchmark
    cudnn_benchmark=False,
    # set multi process parameters
    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
    # set distributed parameters
    dist_cfg=dict(backend='nccl'),
)

# set visualizer
visualizer = None

# set log level
log_level = 'INFO'

# load from which checkpoint
load_from = None

# whether to resume training from the loaded checkpoint
resume = False

# Defaults to use random seed and disable `deterministic`
randomness = dict(seed=None, deterministic=False)

# set log processor
log_processor = dict(by_epoch=False)

[LZHgrla]

@fan2goa1

train_dataset = dict(
    type=process_hf_dataset,
-   ataset=dict(
+   dataset=dict(
        type=load_dataset, path='json', data_files=dict(train=data_path)),
    tokenizer=tokenizer,
    max_length=max_length,
    dataset_map_fn=None,
    template_map_fn=dict(
        type=template_map_fn_factory, template=prompt_template),
    remove_unused_columns=True,
    shuffle_before_pack=True,
    pack_to_max_length=pack_to_max_length,
    use_varlen_attn=use_varlen_attn)

[fan2goa1]

@fan2goa1

train_dataset = dict(
    type=process_hf_dataset,
-   ataset=dict(
+   dataset=dict(
        type=load_dataset, path='json', data_files=dict(train=data_path)),
    tokenizer=tokenizer,
    max_length=max_length,
    dataset_map_fn=None,
    template_map_fn=dict(
        type=template_map_fn_factory, template=prompt_template),
    remove_unused_columns=True,
    shuffle_before_pack=True,
    pack_to_max_length=pack_to_max_length,
    use_varlen_attn=use_varlen_attn)

@LZHgrla
修改完后更燃会有其他报错：

Traceback (most recent call last):
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/xtuner/tools/train.py", line 307, in <module>
    main()
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/xtuner/tools/train.py", line 296, in main
    runner = Runner.from_cfg(cfg)
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/runner/runner.py", line 462, in from_cfg
    runner = cls(
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/runner/runner.py", line 429, in __init__
    self.model = self.build_model(model)
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/runner/runner.py", line 836, in build_model
    model = MODELS.build(model)
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/registry/registry.py", line 570, in build
    return self.build_func(cfg, *args, **kwargs, registry=self)
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/registry/build_functions.py", line 232, in build_model_from_cfg
    return build_from_cfg(cfg, registry, default_args)
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/registry/build_functions.py", line 121, in build_from_cfg
    obj = obj_cls(**args)  # type: ignore
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/xtuner/model/sft.py", line 27, in __init__
    self.llm = self._build_from_cfg_or_module(llm)
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/xtuner/model/sft.py", line 91, in _build_from_cfg_or_module
    return BUILDER.build(cfg_or_mod)
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/registry/registry.py", line 570, in build
    return self.build_func(cfg, *args, **kwargs, registry=self)
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/mmengine/registry/build_functions.py", line 121, in build_from_cfg
    obj = obj_cls(**args)  # type: ignore
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/transformers/models/auto/auto_factory.py", line 556, in from_pretrained
    return model_class.from_pretrained(
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/transformers/modeling_utils.py", line 3502, in from_pretrained
    ) = cls._load_pretrained_model(
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/transformers/modeling_utils.py", line 3926, in _load_pretrained_model
    new_error_msgs, offload_index, state_dict_index = _load_state_dict_into_meta_model(
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/transformers/modeling_utils.py", line 802, in _load_state_dict_into_meta_model
    or (not hf_quantizer.check_quantized_param(model, param, param_name, state_dict))
  File "/root/anaconda3/envs/xt-new/lib/python3.10/site-packages/transformers/quantizers/quantizer_bnb_4bit.py", line 124, in check_quantized_param
    if isinstance(module._parameters[tensor_name], bnb.nn.Params4bit):
KeyError: 'inv_freq'

[LZHgrla]

@fan2goa1
huggingface/transformers#29329 修复了这一问题，可以从源码安装 transformers 解决。或者降低 transformers 版本至 4.37.2

[fan2goa1]

@fan2goa1
huggingface/transformers#29329 修复了这一问题，可以从源码安装 transformers 解决。或者降低 transformers 版本至 4.37.2

感谢，降低到4.37.2后问题解决。