LLM Parameters
These parameters are used only in case of Text Classification category.
To set parameters related to LoRa, one needs to enable it first using the following command:
trainingObject.enable_lora(True)
Now, the following are the parameters which can be set for LoRa-enabled experiments:
- lora_r: Specifies the rank for the LoRA layer. Must be a positive integer. The default value is 256.
- lora_aplha: Defines the scaling factor alpha for LoRA. Must be a positive integer. The default value is 512.
- lora_dropout: Sets the dropout rate for LoRA layers. Must be a float between 0 and 1. The default value is 0.05.
- q_lora: Determines whether to enable or disable Q LoRA. Must be a boolean value (True or False). The default value is False.
Use the following command to set all or any of the parameters; the default values for each parameter are shown in the example:
trainingObject.set_lora_parameters(q_lora=False,lora_alpha=512,lora_dropout=0.05,lora_r=256)
Note: These parameters are supported only for PyTorch.