| from torch import nn |
| import torch |
| from typing import Dict |
|
|
| task_config = { |
| "task1590_diplomacy_text_generation": "configs/SuperNI/task1590_diplomacy_text_generation", |
| "task181_outcome_extraction": "configs/SuperNI/task181_outcome_extraction", |
| "task591_sciq_answer_generation": "configs/SuperNI/task591_sciq_answer_generation", |
| "task1729_personachat_generate_next": "configs/SuperNI/task1729_personachat_generate_next", |
| "task1572_samsum_summary": "configs/SuperNI/task1572_samsum_summary", |
| "task1510_evalution_relation_extraction": "configs/SuperNI/task1510_evalution_relation_extraction", |
| "task748_glucose_reverse_cause_event_detection": "configs/SuperNI/task748_glucose_reverse_cause_event_detection", |
| "task002_quoref_answer_generation": "configs/SuperNI/task002_quoref_answer_generation", |
| "task1687_sentiment140_classification": "configs/SuperNI/task1687_sentiment140_classification", |
| "task511_reddit_tifu_long_text_summarization": "configs/SuperNI/task511_reddit_tifu_long_text_summarization", |
| "task875_emotion_classification": "configs/SuperNI/task511_reddit_tifu_long_text_summarization", |
| "task639_multi_woz_user_utterance_generation": "configs/SuperNI/task639_multi_woz_user_utterance_generation", |
| "task1290_xsum_summarization": "configs/SuperNI/task1290_xsum_summarization", |
| "task073_commonsenseqa_answer_generation": "configs/SuperNI/task073_commonsenseqa_answer_generation", |
| "task363_sst2_polarity_classification": "configs/SuperNI/task363_sst2_polarity_classification", |
| "dbpedia": "configs/Long_Sequence/dbpedia", |
| "amazon": "configs/Long_Sequence/amazon", |
| "agnews": "configs/Long_Sequence/agnews", |
| "yahoo": "configs/Long_Sequence/yahoo", |
| "yelp": "configs/Long_Sequence/yelp", |
| "copa": "configs/Long_Sequence/copa", |
| "mnli": "configs/Long_Sequence/mnli", |
| "cb": "configs/Long_Sequence/cb", |
| "imdb": "configs/Long_Sequence/imdb", |
| "multirc": "configs/Long_Sequence/multirc", |
| "sst2": "configs/Long_Sequence/sst2", |
| "boolq": "configs/Long_Sequence/boolq", |
| "rte": "configs/Long_Sequence/rte", |
| "wic": "configs/Long_Sequence/wic", |
| "qqp": "configs/Long_Sequence/qqp", |
| } |
|
|
| def lora_state_dict_A(model: nn.Module, bias: str = 'none', task_name=None) -> Dict[str, torch.Tensor]: |
| my_state_dict = model.state_dict() |
| if bias == 'none': |
| return {k: my_state_dict[k] for k in my_state_dict if 'lora_A' in k} |
| elif bias == 'all': |
| return {k: my_state_dict[k] for k in my_state_dict if 'lora_A' in k or 'bias' in k} |
| elif bias == 'lora_only': |
| to_return = {} |
| for k in my_state_dict: |
| if 'lora_' in k: |
| to_return[k] = my_state_dict[k] |
| bias_name = k.split('lora_A')[0]+'bias' |
| if bias_name in my_state_dict: |
| to_return[bias_name] = my_state_dict[bias_name] |
| return to_return |
| else: |
| raise NotImplementedError |
|
|
| def lora_state_dict_B(model: nn.Module, bias: str = 'none', task_name=None) -> Dict[str, torch.Tensor]: |
| my_state_dict = model.state_dict() |
| if bias == 'none': |
| return {k: my_state_dict[k] for k in my_state_dict if 'lora_B' in k} |
| elif bias == 'all': |
| return {k: my_state_dict[k] for k in my_state_dict if 'lora_B' in k or 'bias' in k} |
| elif bias == 'lora_only': |
| to_return = {} |
| for k in my_state_dict: |
| if 'lora_' in k: |
| to_return[k] = my_state_dict[k] |
| bias_name = k.split('lora_B')[0]+'bias' |
| if bias_name in my_state_dict: |
| to_return[bias_name] = my_state_dict[bias_name] |
| return to_return |
| else: |
| raise NotImplementedError |
| |
| def lora_state_dict_s(model: nn.Module, bias: str = 'none', task_name=None) -> Dict[str, torch.Tensor]: |
| my_state_dict = model.state_dict() |
| if bias == 'none': |
| return {k: my_state_dict[k] for k in my_state_dict if 'lora_s' in k} |
| else: |
| raise NotImplementedError |
