Functions — github.com/GCYZSL/MoLA

hub / github.com/GCYZSL/MoLA / functions

Functions226 in github.com/GCYZSL/MoLA

⨍ Functions 226◇ Types & classes 26 ↳ Endpoints 1

↓ 1 callersMethodgenerate_prompt

( self, instruction: str, input: Union[None, str] = None, labe

evaluation_scienceqa.py:48

↓ 1 callersFunctionget_answer

(problem, options)

preparation_scienceqa_data.py:30

↓ 1 callersFunctionget_choice_text

(probelm, options)

preparation_scienceqa_data.py:21

↓ 1 callersFunctionget_context_text

(problem, use_caption)

preparation_scienceqa_data.py:12

↓ 1 callersMethodget_eval_dataloader

Returns the evaluation [`~torch.utils.data.DataLoader`]. Subclass and override this method if you want to inject some custom behavio

src/mola_trainer_hacked.py:920

↓ 1 callersFunctionget_lecture_text

(problem)

preparation_scienceqa_data.py:34

↓ 1 callersMethodget_optimizer_cls_and_kwargs

Returns the optimizer class and optimizer parameters based on the training arguments. Args: args (`transformers.training

src/mola_trainer_hacked.py:1093

↓ 1 callersFunctionget_peft_model_state_dict_moe

Get the state dict of the Peft model. Args: model ([`PeftModel`]): The Peft model. When using torch.nn.DistributedDataParallel, Deep

src/mola_peft_model_hacked.py:64

↓ 1 callersMethodget_prompt_embedding_to_save

Returns the prompt embedding to save when saving the model. Only applicable when `peft_config.peft_type != PeftType.LORA`.

src/mola_peft_model_hacked.py:404

↓ 1 callersMethodget_response

(self, output: str)

evaluation_scienceqa.py:70

↓ 1 callersFunctionget_solution_text

(problem)

preparation_scienceqa_data.py:40

↓ 1 callersMethodget_test_dataloader

Returns the test [`~torch.utils.data.DataLoader`]. Subclass and override this method if you want to inject some custom behavior.

src/mola_trainer_hacked.py:970

↓ 1 callersMethodget_train_dataloader

Returns the training [`~torch.utils.data.DataLoader`]. Will use no sampler if `train_dataset` does not implement `__len__`, a random

src/mola_trainer_hacked.py:841

↓ 1 callersMethodipex_optimize_model

(self, model, training=False, dtype=torch.float32)

src/mola_trainer_hacked.py:1338

↓ 1 callersMethodload_adapter

(self, model_id: str, adapter_name: str, is_trainable: bool = False, **kwargs: Any)

src/mola_peft_model_hacked.py:552

↓ 1 callersFunctionload_balancing_loss_func

r""" Computes auxiliary load balancing loss as in Switch Transformer - implemented in Pytorch. See Switch Transformer (https://arxiv.org/abs/

src/mola_modeling_llama_hacked.py:22

↓ 1 callersFunctionmain

()

evaluation_scienceqa.py:74

↓ 1 callersMethodprint_trainable_parameters

Prints the number of trainable parameters in the model.

src/mola_peft_model_hacked.py:461

↓ 1 callersMethodreset_lora_parameters_MoE

(self, adapter_name)

src/mola_lora_hacked.py:740

↓ 1 callersMethodtorch_jit_model_eval

(self, model, dataloader, training=False)

src/mola_trainer_hacked.py:1300

↓ 1 callersMethodupdate_layer_conv2d

(self, adapter_name, r, lora_alpha, lora_dropout, init_lora_weights)

src/mola_lora_hacked.py:584

Method__call__

(self, input_ids, scores)

utils/callbacks.py:19

Method__enter__

(self)

utils/callbacks.py:71

Method__exit__

(self, exc_type, exc_val, exc_tb)

utils/callbacks.py:74

Method__getattr__

Forward missing attributes to the wrapped module.

src/mola_peft_model_hacked.py:480

Method__getattr__

Forward missing attributes to the wrapped module.

src/mola_lora_hacked.py:353

Method__init__

(self, template_name: str = "", verbose: bool = False)

evaluation_scienceqa.py:34

Method__init__

(self, callback_func=None)

utils/callbacks.py:16

Method__init__

(self, func, kwargs={}, callback=None)

utils/callbacks.py:32

Method__init__

(self, template_name: str = "", verbose: bool = False)

utils/prompter.py:13

Method__init__

( self, model: Union[PreTrainedModel, nn.Module] = None, args: TrainingArguments = Non

src/mola_trainer_hacked.py:296

Method__init__

(self, model, peft_config: PeftConfig, adapter_name="default")

src/mola_peft_model_hacked.py:763

Method__init__

(self, model, peft_config: PeftConfig, adapter_name="default", number_experts= [8] * 32, top_k=[2] * 32)

src/mola_peft_model_hacked.py:948

Method__init__

(self, model, peft_config: PeftConfig, adapter_name="default")

src/mola_peft_model_hacked.py:1116

Method__init__

(self, model, peft_config: PeftConfig = None, adapter_name="default")

src/mola_peft_model_hacked.py:1350

Method__init__

(self, model, peft_config: PeftConfig = None, adapter_name="default")

src/mola_peft_model_hacked.py:1520

Method__init__

(self, in_features: int, out_features: int, **kwargs)

src/mola_lora_hacked.py:544

Method__init__

(self, in_features: int, out_features: int, **kwargs)

src/mola_lora_hacked.py:643

Method__init__

( self, adapter_name: str, in_features: int, out_features: int, r: int

src/mola_lora_hacked.py:765

Method__init__

LlamaRMSNorm is equivalent to T5LayerNorm

src/mola_modeling_llama_hacked.py:165

Method__init__

(self, dim, max_position_embeddings=2048, base=10000, device=None)

src/mola_modeling_llama_hacked.py:185

Method__init__

( self, hidden_size: int, num_heads: int, config: LlamaConfig, ##modified

src/mola_modeling_llama_hacked.py:280

Method__init__

(self, config: LlamaConfig)

src/mola_modeling_llama_hacked.py:456

Method__init__

(self, config: LlamaConfig)

src/mola_modeling_llama_hacked.py:664

Method__init__

(self, config)

src/mola_modeling_llama_hacked.py:958

Method__init__

(self, config)

src/mola_modeling_llama_hacked.py:1146

Method__init__

(self, config)

src/mola_modeling_llama_hacked.py:1390

Method__iter__

(self)

utils/callbacks.py:61

Method__next__

(self)

utils/callbacks.py:64

Method__post_init__

(self)

src/mola_lora_hacked.py:94

Method_callback

(val)

utils/callbacks.py:40

Method_init_weights

(self, module)

src/mola_modeling_llama_hacked.py:576

Method_inner_training_loop

( self, batch_size=None, args=None, resume_from_checkpoint=None, trial=None, ignore_keys_for_eval=None

src/mola_trainer_hacked.py:1645

Method_reorder_cache

(past_key_values, beam_idx)

src/mola_modeling_llama_hacked.py:1137

Method_reorder_cache

(past_key_values, beam_idx)

src/mola_modeling_llama_hacked.py:1366

Method_set_gradient_checkpointing

(self, module, value=False)

src/mola_modeling_llama_hacked.py:587

Method_shape

(self, tensor: torch.Tensor, seq_len: int, bsz: int)

src/mola_modeling_llama_hacked.py:320

Methodactive_peft_config

(self)

src/mola_peft_model_hacked.py:679

Methodadd_weighted_adapter

(self, adapters, weights, adapter_name)

src/mola_lora_hacked.py:459

Methodauto_wrapper_callable

(m, *args, **kwargs)

src/mola_trainer_hacked.py:1502

Methodcreate_custom_forward

(module)

src/mola_modeling_llama_hacked.py:898

Methodcustom_forward

(*inputs)

src/mola_modeling_llama_hacked.py:899

Methoddisable_adapter

Disables the adapter module.

src/mola_peft_model_hacked.py:494

Methodevaluation_loop

Prediction/evaluation loop, shared by `Trainer.evaluate()` and `Trainer.predict()`. Works both with or without labels.

src/mola_trainer_hacked.py:3042

Methodforward

Forward pass of the model.

src/mola_peft_model_hacked.py:487

Methodforward

( self, input_ids=None, attention_mask=None, inputs_embeds=None, label

src/mola_peft_model_hacked.py:778

Methodforward

( self, input_ids=None, attention_mask=None, inputs_embeds=None, label

src/mola_peft_model_hacked.py:952

Methodforward

( self, input_ids=None, attention_mask=None, inputs_embeds=None, decod

src/mola_peft_model_hacked.py:1123

Methodforward

( self, input_ids=None, attention_mask=None, inputs_embeds=None, label

src/mola_peft_model_hacked.py:1365

Methodforward

( self, input_ids=None, attention_mask=None, token_type_ids=None, posi

src/mola_peft_model_hacked.py:1535

Methodforward

(self, x: torch.Tensor)

src/mola_lora_hacked.py:831

Methodforward

(self, hidden_states)

src/mola_modeling_llama_hacked.py:173

Methodforward

(self, x, seq_len=None)

src/mola_modeling_llama_hacked.py:199

Methodforward

(self, x)

src/mola_modeling_llama_hacked.py:246

Methodforward

Input shape: Batch x Time x Channel

src/mola_modeling_llama_hacked.py:323

Methodforward

Args: hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)` attention_mask (

src/mola_modeling_llama_hacked.py:478

Methodforward

r""" Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence toke

src/mola_modeling_llama_hacked.py:711

Methodforward

r""" Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence toke

src/mola_modeling_llama_hacked.py:986

Methodforward

r""" Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence toke

src/mola_modeling_llama_hacked.py:1184

Methodforward

r""" labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): Labels for computing the sequence classification/regression

src/mola_modeling_llama_hacked.py:1406

Methodgenerate

(self, **kwargs)

src/mola_peft_model_hacked.py:1234

Functiongenerate_and_tokenize_prompt

(data_point)

mola_training.py:173

Functiongenerate_and_tokenize_prompt_meta

(data_point)

mola_training_instruction.py:170

Methodgentask

()

utils/callbacks.py:45

Methodget_decoder

(self)

src/mola_modeling_llama_hacked.py:982

Methodget_decoder

(self)

src/mola_modeling_llama_hacked.py:1179

Methodget_input_embeddings

(self)

src/mola_modeling_llama_hacked.py:967

Methodget_input_embeddings

(self)

src/mola_modeling_llama_hacked.py:1164

Methodget_input_embeddings

(self)

src/mola_modeling_llama_hacked.py:1399

Methodget_output_embeddings

(self)

src/mola_modeling_llama_hacked.py:973

Methodget_output_embeddings

(self)

src/mola_modeling_llama_hacked.py:1170

Functionget_peft_config

Returns a Peft config object from a dictionary. Args: config_dict (`Dict[str, Any]`): Dictionary containing the configuration parame

src/mola_mapping_hacked.py:61

Methodget_peft_config_as_dict

(self, inference: bool = False)

src/mola_lora_hacked.py:360

Methodget_response

(self, output: str)

utils/prompter.py:73

Methodhyperparameter_search

Launch an hyperparameter search using `optuna` or `Ray Tune` or `SigOpt`. The optimized quantity is determined by `compute_objective`

src/mola_trainer_hacked.py:2461

Methodmake_inputs_require_grad

(module, input, output)

src/mola_peft_model_hacked.py:398

Functionmark_only_lora_B_as_trainable

(model: nn.Module, bias: str = "none")

src/mola_lora_hacked.py:525

Methodmerge_adapter

(self)

src/mola_lora_hacked.py:388

Methodmerge_and_unload

r""" This method merges the LoRa layers into the base model. This is needed if someone wants to use the base model as a standalone mod

src/mola_lora_hacked.py:406

Methodopt_load_hook

(mod, opt)

src/mola_trainer_hacked.py:2433

← previous next →101–200 of 226, ranked by callers