Fine Tune
Train or Fine Tune a model via supervised, unsupervised or rlhf fine tuning
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model_name |
str
|
The name of the model - which is referenced in the model_ids table alongside the model_id. |
required |
model_type |
str
|
Model that is used to do the fine tuning - could be "FT-GPT", "Llama3" for supervised, "MLM" for unsupervised. |
FTGPT
|
fine_tuning_type |
str
|
The type of fine tuning - could be "supervised", "unsupervised" or "rlhf" |
required |
x_train_csv |
DataFrame
|
training data for fine tuning (string is csv of file path)- for each row could contain question or context entries |
required |
y_train_csv |
DataFrame
|
training labels for fine tuning (string is csv of file path) - for each row could contain answer entries |
required |
initial_model_id |
str
|
if rlhf fine tuning type, can add initial unsupervised model id from pretraining step for transfer learning |
required |
document_files |
list[str]
|
A list of file paths to documents for document-based tasks for training. Example: ['path/to/file1.pdf', 'path/to/file2.pdf']. |
required |
Returns:
| Name | Type | Description |
|---|---|---|
response |
dict
|
A JSON response from the API, including the |