能正常推理的情况下，使用官方命令尝试微调时，下载数据集时显示获取数据集文件列表失败，无论使用modelscope还是huggingface的数据集都存在这个问题

(myenv_swift) root@3e1a8642baf5:~/.cache/modelscope/hub/datasets# CUDA_VISIBLE_DEVICES=0
swift sft
--model /workspace/Qwen2.5-VL-7B-Instruct
--train_type lora
--dataset AI-ModelScope/alpaca-gpt4-data-zh#500
AI-ModelScope/alpaca-gpt4-data-en#500
--torch_dtype bfloat16
--num_train_epochs 1
--per_device_train_batch_size 1
--per_device_eval_batch_size 1
--learning_rate 1e-4
--lora_rank 8
--lora_alpha 32
--target_modules all-linear
--gradient_accumulation_steps 16
--eval_steps 50
--save_steps 50
--save_total_limit 2
--logging_steps 5
--max_length 2048
--output_dir output
--system 'You are a helpful assistant.'
--warmup_ratio 0.05
--dataloader_num_workers 4
--model_author swift
--model_name swift-robot
run sh: /workspace/myenv_swift/bin/python /workspace/ms-swift/swift/cli/sft.py --model /workspace/Qwen2.5-VL-7B-Instruct --train_type lora --dataset AI-ModelScope/alpaca-gpt4-data-zh#500 AI-ModelScope/alpaca-gpt4-data-en#500 --torch_dtype bfloat16 --num_train_epochs 1 --per_device_train_batch_size 1 --per_device_eval_batch_size 1 --learning_rate 1e-4 --lora_rank 8 --lora_alpha 32 --target_modules all-linear --gradient_accumulation_steps 16 --eval_steps 50 --save_steps 50 --save_total_limit 2 --logging_steps 5 --max_length 2048 --output_dir output --system You are a helpful assistant. --warmup_ratio 0.05 --dataloader_num_workers 4 --model_author swift --model_name swift-robot
[INFO:swift] Successfully registered /workspace/ms-swift/swift/llm/dataset/data/dataset_info.json.
[INFO:swift] rank: -1, local_rank: -1, world_size: 1, local_world_size: 1
torch_dtype is deprecated! Use dtype instead!
[INFO:swift] Setting args.lazy_tokenize: True
[INFO:swift] output_dir: /root/.cache/modelscope/hub/datasets/output/v1-20251225-050116
[INFO:swift] Global seed set to 42
[INFO:swift] args: TrainArguments(
_n_gpu=-1,
acc_strategy=token,
accelerator_config={'dispatch_batches': False},
adafactor=False,
adalora_beta1=0.85,
adalora_beta2=0.85,
adalora_deltaT=1,
adalora_init_r=12,
adalora_orth_reg_weight=0.5,
adalora_target_r=8,
adalora_tfinal=0,
adalora_tinit=0,
adam_beta1=0.9,
adam_beta2=0.95,
adam_epsilon=1e-08,
adapter_act=gelu,
adapter_length=128,
adapters=[],
add_non_thinking_prefix=True,
add_version=True,
agent_template=None,
aligner_lr=None,
attn_impl=None,
auto_find_batch_size=False,
average_tokens_across_devices=True,
batch_eval_metrics=False,
bf16=True,
bf16_full_eval=False,
bnb_4bit_compute_dtype=torch.bfloat16,
bnb_4bit_quant_storage=None,
bnb_4bit_quant_type=nf4,
bnb_4bit_use_double_quant=True,
boft_block_num=0,
boft_block_size=4,
boft_dropout=0.0,
boft_n_butterfly_factor=1,
cached_dataset=[],
cached_val_dataset=[],
check_model=True,
ckpt_dir=None,
columns={},
create_checkpoint_symlink=False,
custom_dataset_info=[],
custom_register_path=[],
data_seed=42,
dataloader_drop_last=False,
dataloader_num_workers=4,
dataloader_persistent_workers=False,
dataloader_pin_memory=True,
dataloader_prefetch_factor=None,
dataset=['AI-ModelScope/alpaca-gpt4-data-zh#500', 'AI-ModelScope/alpaca-gpt4-data-en#500'],
dataset_num_proc=1,
dataset_shuffle=True,
ddp_backend=None,
ddp_broadcast_buffers=None,
ddp_bucket_cap_mb=None,
ddp_find_unused_parameters=None,
ddp_timeout=18000000,
debug=None,
deepspeed=None,
deepspeed_autotp_size=None,
device_groups=None,
device_map=None,
disable_tqdm=None,
do_eval=False,
do_predict=False,
do_train=False,
download_mode=reuse_dataset_if_exists,
ds3_gather_for_generation=True,
early_stop_interval=None,
enable_channel_loss=False,
enable_dft_loss=False,
enable_thinking=None,
eval_accumulation_steps=None,
eval_dataset=[],
eval_dataset_args=None,
eval_delay=0,
eval_do_concat_batches=True,
eval_generation_config=None,
eval_limit=None,
eval_on_start=False,
eval_steps=50.0,
eval_strategy=no,
eval_use_evalscope=False,
eval_use_gather_object=False,
external_plugins=[],
extra_eval_args=None,
fourier_n_frequency=2000,
fourier_scaling=300.0,
fp16=False,
fp16_backend=auto,
fp16_full_eval=False,
fp16_opt_level=O1,
freeze_aligner=True,
freeze_llm=False,
freeze_parameters=[],
freeze_parameters_ratio=0.0,
freeze_parameters_regex=None,
freeze_vit=True,
fsdp=None,
fsdp_config=None,
fsdp_min_num_params=0,
fsdp_transformer_layer_cls_to_wrap=None,
full_determinism=False,
galore_cos_threshold=0.4,
galore_gamma_proj=2,
galore_optim_per_parameter=False,
galore_proj_bits=4,
galore_proj_group_size=256,
galore_proj_quant=False,
galore_proj_type=std,
galore_quantization=False,
galore_queue_size=5,
galore_rank=128,
galore_scale=1.0,
galore_target_modules=None,
galore_update_proj_gap=50,
galore_with_embedding=False,
generation_config=None,
generation_max_length=None,
generation_num_beams=None,
gradient_accumulation_steps=16,
gradient_checkpointing=True,
gradient_checkpointing_kwargs=None,
greater_is_better=False,
group_by_length=False,
half_precision_backend=auto,
hqq_axis=None,
hub_always_push=False,
hub_model_id=None,
hub_private_repo=None,
hub_revision=None,
hub_strategy=every_save,
hub_token=<HUB_TOKEN>,
ignore_args_error=False,
ignore_data_skip=False,
include_for_metrics=[],
include_inputs_for_metrics=False,
include_num_input_tokens_seen=False,
include_tokens_per_second=False,
init_strategy=None,
init_weights=True,
interleave_prob=None,
jit_mode_eval=False,
label_names=None,
label_smoothing_factor=0.0,
lazy_tokenize=True,
learning_rate=0.0001,
length_column_name=length,
liger_kernel_config=None,
lisa_activated_layers=0,
lisa_step_interval=20,
llamapro_num_groups=None,
llamapro_num_new_blocks=4,
load_args=False,
load_best_model_at_end=False,
load_data_args=False,
load_from_cache_file=False,
local_rank=-1,
local_repo_path=None,
log_level=passive,
log_level_replica=warning,
log_on_each_node=True,
logging_dir=/root/.cache/modelscope/hub/datasets/output/v1-20251225-050116/runs,
logging_first_step=True,
logging_nan_inf_filter=True,
logging_steps=5,
logging_strategy=steps,
logprobs=False,
lora_alpha=32,
lora_bias=none,
lora_dropout=0.05,
lora_dtype=None,
lora_ga_batch_size=2,
lora_ga_direction=ArB2r,
lora_ga_iters=2,
lora_ga_max_length=1024,
lora_ga_scale=stable,
lora_ga_stable_gamma=16,
lora_modules=[],
lora_rank=8,
lorap_lr_ratio=None,
loss_scale=default,
loss_type=None,
lr_scheduler_kwargs=None,
lr_scheduler_type=cosine,
max_epochs=None,
max_grad_norm=1.0,
max_length=2048,
max_memory={},
max_model_len=None,
max_new_tokens=64,
max_pixels=None,
max_steps=-1,
metric=None,
metric_for_best_model=loss,
model=/workspace/Qwen2.5-VL-7B-Instruct,
model_author=['swift'],
model_kwargs={},
model_name=['swift-robot'],
model_revision=None,
model_type=qwen2_5_vl,
modules_to_save=[],
mp_parameters=,
neftune_noise_alpha=None,
new_special_tokens=[],
no_cuda=False,
norm_bbox=None,
num_beams=1,
num_labels=None,
num_train_epochs=1.0,
optim=adamw_torch_fused,
optim_args=None,
optim_target_modules=None,
optimizer=None,
output_dir=/root/.cache/modelscope/hub/datasets/output/v1-20251225-050116,
overwrite_output_dir=False,
packing=False,
packing_length=None,
packing_num_proc=1,
padding_free=False,
padding_side=right,
parallelism_config=None,
past_index=-1,
per_device_eval_batch_size=1,
per_device_train_batch_size=1,
predict_with_generate=False,
prediction_loss_only=False,
problem_type=None,
project=huggingface,
push_to_hub=False,
push_to_hub_model_id=None,
push_to_hub_organization=None,
push_to_hub_token=<PUSH_TO_HUB_TOKEN>,
quant_bits=None,
quant_method=None,
ray_exp_name=None,
ray_scope=last,
reft_args=None,
reft_intervention_type=LoreftIntervention,
reft_layer_key=None,
reft_layers=None,
reft_rank=4,
remove_unused_columns=True,
repetition_penalty=None,
report_to=['tensorboard'],
response_prefix=None,
restore_callback_states_from_checkpoint=False,
resume_from_checkpoint=None,
resume_only_model=False,
rope_scaling=None,
router_aux_loss_coef=0.0,
run_name=/root/.cache/modelscope/hub/datasets/output/v1-20251225-050116,
save_on_each_node=False,
save_only_model=False,
save_safetensors=True,
save_steps=50.0,
save_strategy=steps,
save_total_limit=2,
seed=42,
sequence_parallel_size=1,
shuffle_buffer_size=1000,
skip_memory_metrics=True,
sortish_sampler=False,
split_dataset_ratio=0.0,
stop_words=[],
stopping_strategy=first_exhausted,
stream=False,
streaming=False,
strict=False,
swanlab_exp_name=None,
swanlab_lark_secret=None,
swanlab_lark_webhook_url=None,
swanlab_mode=cloud,
swanlab_project=None,
swanlab_token=<SWANLAB_TOKEN>,
swanlab_workspace=None,
system=You are a helpful assistant.,
target_modules=['all-linear'],
target_parameters=None,
target_regex=None,
task_type=causal_lm,
temperature=0.0,
template=qwen2_5_vl,
template_backend=swift,
tf32=None,
top_k=None,
top_logprobs=None,
top_p=None,
torch_compile=False,
torch_compile_backend=None,
torch_compile_mode=None,
torch_dtype=torch.bfloat16,
torch_empty_cache_steps=None,
torchdynamo=None,
tpu_metrics_debug=False,
tpu_num_cores=None,
trackio_space_id=trackio,
train_dataloader_shuffle=True,
train_type=lora,
trainable_parameters=[],
trainable_parameters_regex=None,
truncation_strategy=delete,
tuner_backend=peft,
use_chat_template=True,
use_cpu=False,
use_dora=False,
use_flash_ckpt=False,
use_galore=False,
use_hf=False,
use_legacy_prediction_loop=False,
use_liger_kernel=False,
use_logits_to_keep=None,
use_mps_device=False,
use_ray=False,
use_rslora=False,
use_swift_lora=False,
val_dataset=[],
val_dataset_shuffle=False,
vera_d_initial=0.1,
vera_dropout=0.0,
vera_projection_prng_key=0,
vera_rank=256,
vit_gradient_checkpointing=None,
vit_lr=None,
warmup_ratio=0.05,
warmup_steps=0,
weight_decay=0.1,
zero_hpz_partition_size=None,
)
The image processor of type Qwen2VLImageProcessor is now loaded as a fast processor by default, even if the model checkpoint was saved with a slow processor. This is a breaking change and may produce slightly different outputs. To continue using the slow processor, instantiate this class with use_fast=False. Note that this behavior will be extended to all models in a future release.
[INFO:swift] model_kwargs: {'device_map': 'cuda:0', 'dtype': torch.bfloat16}
Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5/5 [00:05<00:00, 1.10s/it]
[INFO:swift] Setting image_factor: 28. You can adjust this hyperparameter through the environment variable: IMAGE_FACTOR.
[INFO:swift] Setting min_pixels: 3136. You can adjust this hyperparameter through the environment variable: MIN_PIXELS.
[INFO:swift] Setting max_pixels: 12845056. You can adjust this hyperparameter through the environment variable: MAX_PIXELS.
[INFO:swift] Setting video_min_pixels: 100352. You can adjust this hyperparameter through the environment variable: VIDEO_MIN_PIXELS.
[INFO:swift] Setting video_max_pixels: 602112. You can adjust this hyperparameter through the environment variable: VIDEO_MAX_PIXELS.
[INFO:swift] Setting video_total_pixels: 19267584. You can adjust this hyperparameter through the environment variable: VIDEO_TOTAL_PIXELS.
[INFO:swift] Setting max_ratio: 200. You can adjust this hyperparameter through the environment variable: MAX_RATIO.
[INFO:swift] Setting frame_factor: 2. You can adjust this hyperparameter through the environment variable: FRAME_FACTOR.
[INFO:swift] Setting fps: 2.0. You can adjust this hyperparameter through the environment variable: FPS.
[INFO:swift] Setting fps_min_frames: 4. You can adjust this hyperparameter through the environment variable: FPS_MIN_FRAMES.
[INFO:swift] Setting fps_max_frames: 768. You can adjust this hyperparameter through the environment variable: FPS_MAX_FRAMES.
[INFO:swift] model.hf_device_map: {'': device(type='cuda', index=0)}
[INFO:swift] model_info: ModelInfo(model_type='qwen2_5_vl', model_dir='/workspace/Qwen2.5-VL-7B-Instruct', torch_dtype=torch.bfloat16, max_model_len=128000, quant_method=None, quant_bits=None, rope_scaling={'type': 'default', 'mrope_section': [16, 24, 24], 'rope_type': 'default'}, is_moe_model=False, is_multimodal=True, config=Qwen2_5_VLConfig {
"architectures": [
"Qwen2_5_VLForConditionalGeneration"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"dtype": "bfloat16",
"eos_token_id": 151645,
"hidden_act": "silu",
"hidden_size": 3584,
"image_token_id": 151655,
"initializer_range": 0.02,
"intermediate_size": 18944,
"max_position_embeddings": 128000,
"max_window_layers": 28,
"model_type": "qwen2_5_vl",
"num_attention_heads": 28,
"num_hidden_layers": 28,
"num_key_value_heads": 4,
"rms_norm_eps": 1e-06,
"rope_scaling": {
"mrope_section": [
16,
24,
24
],
"rope_type": "default",
"type": "default"
},
"rope_theta": 1000000.0,
"sliding_window": 32768,
"text_config": {
"_name_or_path": "/workspace/Qwen2.5-VL-7B-Instruct",
"architectures": [
"Qwen2_5_VLForConditionalGeneration"
],
"attention_dropout": 0.0,
"bos_token_id": 151643,
"dtype": "bfloat16",
"eos_token_id": 151645,
"hidden_act": "silu",
"hidden_size": 3584,
"initializer_range": 0.02,
"intermediate_size": 18944,
"layer_types": [
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention",
"full_attention"
],
"max_position_embeddings": 128000,
"max_window_layers": 28,
"model_type": "qwen2_5_vl_text",
"num_attention_heads": 28,
"num_hidden_layers": 28,
"num_key_value_heads": 4,
"pad_token_id": 151643,
"rms_norm_eps": 1e-06,
"rope_scaling": {
"mrope_section": [
16,
24,
24
],
"rope_type": "default",
"type": "default"
},
"rope_theta": 1000000.0,
"sliding_window": null,
"use_cache": true,
"use_sliding_window": false,
"vision_token_id": 151654,
"vocab_size": 152064
},
"tie_word_embeddings": false,
"transformers_version": "4.57.3",
"use_cache": true,
"use_sliding_window": false,
"video_token_id": 151656,
"vision_config": {
"depth": 32,
"dtype": "bfloat16",
"fullatt_block_indexes": [
7,
15,
23,
31
],
"hidden_act": "silu",
"hidden_size": 1280,
"in_channels": 3,
"in_chans": 3,
"initializer_range": 0.02,
"intermediate_size": 3420,
"model_type": "qwen2_5_vl",
"num_heads": 16,
"out_hidden_size": 3584,
"pad_token_id": 151643,
"patch_size": 14,
"spatial_merge_size": 2,
"spatial_patch_size": 14,
"temporal_patch_size": 2,
"tokens_per_second": 2,
"window_size": 112
},
"vision_end_token_id": 151653,
"vision_start_token_id": 151652,
"vision_token_id": 151654,
"vocab_size": 152064
}
, task_type='causal_lm', num_labels=None)
[INFO:swift] model.generation_config: GenerationConfig {
"bos_token_id": 151643,
"eos_token_id": [
151645,
151643
],
"max_new_tokens": 64,
"pad_token_id": 151643,
"repetition_penalty": 1.05
}

[INFO:swift] default_system: 'You are a helpful assistant.'
[INFO:swift] max_length: 2048
[INFO:swift] response_prefix: ''
[INFO:swift] agent_template: hermes
[INFO:swift] norm_bbox: none
[INFO:swift] Setting ROOT_IMAGE_DIR: None. You can adjust this hyperparameter through the environment variable: ROOT_IMAGE_DIR.
[INFO:swift] Setting QWENVL_BBOX_FORMAT: legacy. You can adjust this hyperparameter through the environment variable: QWENVL_BBOX_FORMAT.
[INFO:swift] Start time of running main: 2025-12-25 05:01:23.326869
[INFO:swift] swift.version: 3.12.0.dev0
[INFO:swift] SelfCognitionPreprocessor has been successfully configured with name: ('swift-robot', 'swift-robot'), author: ('swift', 'swift').
[INFO:swift] Downloading the dataset from ModelScope, dataset_id: AI-ModelScope/alpaca-gpt4-data-zh
[ERROR:modelscope] Get dataset: AI-ModelScope/alpaca-gpt4-data-zh file list failed, message: Url = https://www.modelscope.cn/api/v1/datasets/4061/repo/tree, Request id=da935da9-ed26-4753-935d-4dce8ce2b415 Code = 10020101029 Message = 获取数据集文件列表失败: 获取数据集文件列表失败: %s, Please specify correct dataset_name and namespace.
[ERROR:modelscope] Get dataset: AI-ModelScope/alpaca-gpt4-data-zh file list failed, message: Url = https://www.modelscope.cn/api/v1/datasets/4061/repo/tree, Request id=cc2e60db-fab7-405c-ac5f-d9135a5d6b4f Code = 10020101029 Message = 获取数据集文件列表失败: 获取数据集文件列表失败: %s, Please specify correct dataset_name and namespace.
Map: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 48818/48818 [00:00<00:00, 63193.57 examples/s]
[INFO:swift] Downloading the dataset from ModelScope, dataset_id: AI-ModelScope/alpaca-gpt4-data-en
[ERROR:modelscope] Get dataset: AI-ModelScope/alpaca-gpt4-data-en file list failed, message: Url = https://www.modelscope.cn/api/v1/datasets/4266/repo/tree, Request id=c02b54d7-6363-4f28-a069-b1bcce9d3a0e Code = 10020101029 Message = 获取数据集文件列表失败: 获取数据集文件列表失败: %s, Please specify correct dataset_name and namespace.
[ERROR:modelscope] Get dataset: AI-ModelScope/alpaca-gpt4-data-en file list failed, message: Url = https://www.modelscope.cn/api/v1/datasets/4266/repo/tree, Request id=41652219-3f98-4af5-8b72-a0fa19ca0577 Code = 10020101029 Message = 获取数据集文件列表失败: 获取数据集文件列表失败: %s, Please specify correct dataset_name and namespace.
Map: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 52002/52002 [00:01<00:00, 44139.59 examples/s]

The reported entry point is swift/cli/sft.py, and dataset metadata is registered from swift/llm/dataset/data/dataset_info.json. Start by reproducing the shown swift sft command with either dataset source and capture the omitted dataset-list failure. Done means both ModelScope and Hugging Face dataset references can fetch their file lists during fine-tuning.