微调时出现TypeError: string indices must be integers错误

[20191864218]

Traceback (most recent call last): File "/root/VisualGLM-6B/finetune_XrayGLM.py", line 194, in training_main(args, model_cls=model, forward_step_function=forward_step, create_dataset_function=create_dataset_function, collate_fn=data_collator) File "/root/miniconda3/lib/python3.10/site-packages/sat/training/deepspeed_training.py", line 67, in training_main train_data, val_data, test_data = make_loaders(args, hooks['create_dataset_function'], collate_fn=collate_fn) File "/root/miniconda3/lib/python3.10/site-packages/sat/data_utils/configure_data.py", line 198, in make_loaders train = make_dataset(**data_set_args, args=args, dataset_weights=args.train_data_weights, is_train_data=True) File "/root/miniconda3/lib/python3.10/site-packages/sat/data_utils/configure_data.py", line 124, in make_dataset_full d = create_dataset_function(p, args) File "/root/VisualGLM-6B/finetune_XrayGLM.py", line 160, in create_dataset_function dataset = FewShotDataset(path, image_processor, tokenizer, args) File "/root/VisualGLM-6B/finetune_XrayGLM.py", line 117, in init image = processor(Image.open(item['img']).convert('RGB')) TypeError: string indices must be integers [2023-11-08 20:22:18,687] [INFO] [launch.py:315:sigkill_handler] Killing subprocess 4613 [2023-11-08 20:22:18,687] [ERROR] [launch.py:321:sigkill_handler] ['/root/miniconda3/bin/python', '-u', 'finetune_XrayGLM.py', '--local_rank=0', '--experiment-name', 'finetune-XrayGLM', '--model-parallel-size', '1', '--mode', 'finetune', '--train-iters', '300', '--resume-dataloader', '--max_source_length', '64', '--max_target_length', '256', '--lora_rank', '10', '--pre_seq_len', '4', '--train-data', '/root/autodl-tmp/Xray/openi-zh.json', '--valid-data', '/root/autodl-tmp/Xray/openi-zh.json', '--distributed-backend', 'nccl', '--lr-decay-style', 'cosine', '--warmup', '.02', '--checkpoint-activations', '--save-interval', '300', '--eval-interval', '10000', '--save', './checkpoints', '--split', '1', '--eval-iters', '10', '--eval-batch-size', '8', '--zero-stage', '1', '--lr', '0.0001', '--batch-size', '8', '--skip-init', '--fp16', '--use_lora'] exits with return code = 1

[Jayson236]

Has the issue been resolved?

[GG6Bond]

请问怎么解决的？

[xlsssss]

和visual_GLM 的dataset.json对比可知，openi-zh.json 还不是最终的可训练的 json版本。执行./data/build_ch_prompt.py，把finetune_XrayGLM.sh 里面的 json路径改成刚刚生成的路径。