如何在trainer中调用wandb

```python
train_dataset = SmileDataset(args, smiles, whole_string, max_len, prop=prop, aug_prob=0, scaffold=scaffold, scaffold_maxlen= scaffold_max_len)
valid_dataset = SmileDataset(args, vsmiles, whole_string, max_len, prop=vprop, aug_prob=0, scaffold=vscaffold, scaffold_maxlen= scaffold_max_len)

mconf = GPTConfig(train_dataset.vocab_size, train_dataset.max_len, num_props=num_props,  # args.num_props,
                    n_layer=args.n_layer, n_head=args.n_head, n_embd=args.n_embd, scaffold=args.scaffold, scaffold_maxlen=scaffold_max_len,
                    lstm=args.lstm, lstm_layers=args.lstm_layers)
model = GPT(mconf)

tconf = TrainerConfig(max_epochs=args.max_epochs, batch_size=args.batch_size, learning_rate=args.learning_rate,
                        lr_decay=True, warmup_tokens=0.1*len(train_data)*max_len, final_tokens=args.max_epochs*len(train_data)*max_len,
                        num_workers=10, ckpt_path=f'D:/Jupyter/MolGPT/train-copy1/cond_gpt/weights/{args.run_name}.pt', block_size=train_dataset.max_len, generate=False)
trainer = Trainer(model, train_dataset, valid_dataset,
                    tconf, train_dataset.stoi, train_dataset.itos)
df = trainer.train(wandb)

df.to_csv(f'{args.run_name}.csv', index=False)

```如图调用wandb后会在最后报错’NoneType’ object has no attribute ‘to_csv’，进行检查发现是前面在trainer调用wandb后出现空集导致’to_csv’失败，望指点