neurai.finetune package

Subpackages

• neurai.finetune.tuning package
  • Subpackages
    • neurai.finetune.tuning.lorax package
      • Submodules
  • Submodules

Submodules

class neurai.finetune.finetune_config.LoRAConfig(base_model_name_or_path=None, revision=None, tuning_type=None, inference_mode=False, r=8, target_modules=None, lora_alpha=8)

Bases: TuningConfig

This is the configuration class to store the configuration of a [LoraModel].

Parameters:

• r (int) – Lora attention dimension.

• lora_alpha (int) – The alpha parameter for Lora scaling.

• base_model_name_or_path (Optional[str]) –

• revision (Optional[str]) –

• tuning_type (Union[str, TuningType, None]) –

• inference_mode (bool) –

• target_modules (Union[List[str], str, None]) –

property is_prompt_learning: bool

Utility method to check if the configuration is for prompt learning.

class neurai.finetune.finetune_config.PrefixTuningConfig(base_model_name_or_path=None, revision=None, tuning_type=None, inference_mode=False, r=8, target_modules=None, num_virtual_tokens=None, token_dim=None, num_transformer_submodules=None, num_attention_heads=None, num_layers=None, encoder_hidden_size=None, prefix_projection=False)

Bases: PromptLearningConfig

This is the configuration class to store the configuration of a [PrefixEncoder].

Parameters:

• encoder_hidden_size (int) – The hidden size of the prompt encoder.

• prefix_projection (bool) – Whether to project the prefix embeddings.

• base_model_name_or_path (Optional[str]) –

• revision (Optional[str]) –

• tuning_type (Union[str, TuningType, None]) –

• inference_mode (bool) –

• r (int) –

• target_modules (Union[List[str], str, None]) –

• num_virtual_tokens (Optional[int]) –

• token_dim (Optional[int]) –

• num_transformer_submodules (Optional[int]) –

• num_attention_heads (Optional[int]) –

• num_layers (Optional[int]) –

class neurai.finetune.finetune_config.PromptLearningConfig(base_model_name_or_path=None, revision=None, tuning_type=None, inference_mode=False, r=8, target_modules=None, num_virtual_tokens=None, token_dim=None, num_transformer_submodules=None, num_attention_heads=None, num_layers=None)

Bases: TuningConfig

This is the base configuration class to store the configuration of [PrefixTuning], [PromptEncoder], or [PromptTuning].

Parameters:

• num_virtual_tokens (int) – The number of virtual tokens to use.

• token_dim (int) – The hidden embedding dimension of the base transformer model.

• num_transformer_submodules (int) – The number of transformer submodules in the base transformer model.

• num_attention_heads (int) – The number of attention heads in the base transformer model.

• num_layers (int) – The number of layers in the base transformer model.

• base_model_name_or_path (Optional[str]) –

• revision (Optional[str]) –

• tuning_type (Union[str, TuningType, None]) –

• inference_mode (bool) –

• r (int) –

• target_modules (Union[List[str], str, None]) –

property is_prompt_learning: bool

Utility method to check if the configuration is for prompt learning.

class neurai.finetune.finetune_config.TuningConfig(base_model_name_or_path=None, revision=None, tuning_type=None, inference_mode=False, r=8, target_modules=None)

Bases: object

This is the base configuration class to store the configuration of a [PeftModel].

Parameters:

• base_model_name_or_path (str) – The name of the base model to use.

• revision (str) – The specific model version to use.

• tuning_type (Union[str, TuningType]) – Peft type.

• inference_mode (bool) – Whether to use inference mode.

• r (int) – Lora attention dimension.

• target_modules (Optional[Union[List[str], str]]) – List of module names or regex expression of the module names to replace with Lora. For example, [‘q’, ‘v’] or ‘.*decoder.*(SelfAttention|EncDecAttention).*(q|v)$’

class neurai.finetune.finetune_config.TuningType(value)

Bases: str, Enum

An enumeration of finetuning type.

Module contents