How to load a fine-tuned model for predictions

[marcopeix]

Once fine-tuning is done, I know the model is saved according to the path specified here, in the YAML file.

However, I don't see any checkpoints there. I only see .log files and a Hydra yaml.

Did I miss something? I used the default YAML config with no modifications.

In my case, the model should be saved in outputs/finetune/moirai_1.0_R_small/store_finetune/store_sales_finetune.

[chenghaoliu89]

Hi @marcopeix , it is supposed to be stored in outputs/finetune/moirai_1.0_R_small/store_finetune/store_sales_finetune/.../checkpoints folder. If you could not find it, please check if there is any error report from the fine-tuning process

[marcopeix]

HI @chenghaoliu89 , there are no errors during fine-tuning and the process runs fine. When I run !ls -la outputs/finetune/moirai_1.0_R_small/store_finetune/store_sales_finetune/, I get this output:

total 20
drwxr-xr-x 4 root root 4096 Feb  7 20:52 .
drwxr-xr-x 3 root root 4096 Feb  7 20:52 ..
drwxr-xr-x 2 root root 4096 Feb  7 20:52 .hydra
drwxr-xr-x 3 root root 4096 Feb  7 20:52 logs
-rw-r--r-- 1 root root  232 Feb  7 20:52 train.log

What am I missing?

Thanks for your help!

[chenghaoliu89]

Hi @marcopeix could you try the fine-tuning example case in readme.md first and see if you can get the model checkpoint file.

[liyujia011025]

您好，您能否先尝试一下 readme.md 中的微调示例案例，看看是否可以获取模型检查点文件。

Hello expert:
May I ask if you have made any minor adjustments using the process and code in the official README file? It's strange that after fine-tuning with my own data according to this operation, the effect is even worse than not fine-tuning at all. Did you make any specific changes to hyperparameter settings or other content during the fine-tuning process? Also, may I ask if you only fine tuned certain layers or all layers during the fine-tuning process?
Looking forward to your reply, thank you very much!