Fine-tune fastchat-t5-3b-v1.0 with Lora, learning_rate is always 0

Nobody has claimed this yet.

Hi，when I use the script to fine-tune fastchat-t5-3b-v1.0 with Lora:
```
CUDA_VISIBLE_DEVICES=3 python fastchat/train/train_lora_t5.py \
--model_name_or_path /fastchat-t5-3b-v1.0 \
--lora_r 8 \
--lora_alpha 16 \
--lora_dropout 0.05 \
--data_path /home/sunjian/FastChat/data/dummy_conversation.json \
--output_dir ./checkpoints \
--num_train_epochs 150 \
--fp16 True \
--per_device_train_batch_size 2 \
--per_device_eval_batch_size 2 \
--gradient_accumulation_steps 1 \
--evaluation_strategy "steps" \
--eval_steps 100 \
--save_strategy "steps" \
--save_steps 200 \
--save_total_limit 2 \
--learning_rate 2e-5 \
--weight_decay 0. \
--warmup_ratio 0.03 \
--lr_scheduler_type "cosine" \
--logging_strategy "steps" \
--logging_steps 1 \
--tf32 True \
--model_max_length 256 \
--q_lora False \
--gradient_checkpointing True
```
we get the log as follows：
```
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.0}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.0}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.01}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.01}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.01}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.01}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.02}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.02}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.02}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.02}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.02}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.03}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.03}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.03}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.03}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.04}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.04}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.04}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.04}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.04}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.05}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.05}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.05}
{'loss': 0.0, 'learning_rate': 0.0, 'epoch': 0.05}
```
What is wrong with the setting? Why is the learning rate always 0?

Start with fastchat/train/train_lora_t5.py and trace how the shown command-line arguments reach the trainer and scheduler. Reproduce the command with the dummy_conversation.json data, then verify that the reported learning rate reflects --learning_rate 2e-5 instead of remaining 0.

Written by the indexing model from the issue text.