SwanLabCallback isn't initialized correctly when transformers==4.50.0

[Luffy-ZY-Wang]

Reminder

• I have read the above rules and searched the existing issues.

System Info

[2025-04-03 06:02:20,328] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
INFO 04-03 06:02:21 [init.py:256] Automatically detected platform cuda.

• llamafactory version: 0.9.3.dev0
• Platform: Linux-5.4.0-162-generic-x86_64-with-glibc2.31
• Python version: 3.12.9
• PyTorch version: 2.6.0+cu118 (GPU)
• Transformers version: 4.50.0
• Datasets version: 3.4.1
• Accelerate version: 1.5.2
• PEFT version: 0.15.0
• TRL version: 0.9.6
• GPU type: NVIDIA GeForce RTX 4090
• GPU number: 8
• GPU memory: 23.65GB
• DeepSpeed version: 0.16.4
• vLLM version: 0.8.1

Reproduction

When I used transformers==4.50.0, and specified swanlab_run_name and swanlab_project like:

swanlab_project: llamafactory
swanlab_run_name: "{{BASE_MODEL_NAME}}_train_lora_{{TASK}}_{{CURRENT_DATE}}" # would be replaced before running

I couldn't get correct project and runname on swanlab cloud. The project name and runname weren't correct. But it did receive those 2 arguments in finetuning_args.

But when I just downgraded transformers==4.49.0, it could init project and runname correctly. Like this:

I guess this is a bug. But I'm wondering who should be the main fault of this issue, swanlab or llamafactory?

Others

No response

[hiyouga]

cc @Zeyi-Lin

[Zeyi-Lin]

Due to transformers==4.50.0 internally integrating swanlab, some conflicts have arisen between the two —— I will soon submit a PR to resolve this issue.