An important note at 11:30 and about fine-tuning (later in the video). Here is some information that I've read on the Crowdin website: you can only fine-tune models if you have your own API key. You will not be able to train a custom model to your specifications if you use the "model managed by Crowdin" option.
That's a very good point! I had to mention it during the session indeed. The Fine-Tuning is available when you use own API key from AI provider + it uses only approved translations to train the model. Also, you should have at least 10 assets (files) in the project to train a model. I've seen quite large projects with just 1 file in Crowdin project many times, so the solution here is to add at least 9-10 dummy files with at least 1 string inside that will contain approved translation. It's ok if content is the same across dummy files. Thanks again for your comment ❤
An important note at 11:30 and about fine-tuning (later in the video). Here is some information that I've read on the Crowdin website: you can only fine-tune models if you have your own API key. You will not be able to train a custom model to your specifications if you use the "model managed by Crowdin" option.
That's a very good point! I had to mention it during the session indeed. The Fine-Tuning is available when you use own API key from AI provider + it uses only approved translations to train the model. Also, you should have at least 10 assets (files) in the project to train a model. I've seen quite large projects with just 1 file in Crowdin project many times, so the solution here is to add at least 9-10 dummy files with at least 1 string inside that will contain approved translation. It's ok if content is the same across dummy files.
Thanks again for your comment ❤