peft/examples/sequence_classification/Prefix-Tuning.ipynb

In [1]:

import argparse
import os

import torch
from torch.optim import AdamW
from torch.utils.data import DataLoader
from pet import get_pet_config,get_pet_model, get_pet_model_state_dict, set_pet_model_state_dict, LoRAConfig, PETType, \
PrefixTuningConfig, PromptEncoderConfig

import evaluate
from datasets import load_dataset
from transformers import AutoModelForSequenceClassification, AutoTokenizer, get_linear_schedule_with_warmup, set_seed
from tqdm import tqdm

In [2]:

batch_size = 32
model_name_or_path = "roberta-large"
task = "mrpc"
pet_type = PETType.PREFIX_TUNING
device = "cuda"
num_epochs = 20

In [3]:

pet_config = PrefixTuningConfig(
    task_type="SEQ_CLS",
    num_virtual_tokens=20
)
lr = 1e-2

In [4]:

if any(k in model_name_or_path for k in ("gpt", "opt", "bloom")):
    padding_side = "left"
else:
    padding_side = "right"
    
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, padding_side=padding_side)
if getattr(tokenizer, "pad_token_id") is None:
    tokenizer.pad_token_id = tokenizer.eos_token_id
    
datasets = load_dataset("glue", task)
metric = evaluate.load("glue", task)

def tokenize_function(examples):
    # max_length=None => use the model max length (it's actually the default)
    outputs = tokenizer(examples["sentence1"], examples["sentence2"], truncation=True, max_length=None)
    return outputs

tokenized_datasets = datasets.map(
    tokenize_function,
    batched=True,
    remove_columns=["idx", "sentence1", "sentence2"],
)

# We also rename the 'label' column to 'labels' which is the expected name for labels by the models of the
# transformers library
tokenized_datasets = tokenized_datasets.rename_column("label", "labels")

def collate_fn(examples):
    return tokenizer.pad(examples, padding="longest", return_tensors="pt")

# Instantiate dataloaders.
train_dataloader = DataLoader(
    tokenized_datasets["train"], shuffle=True, collate_fn=collate_fn, batch_size=batch_size
)
eval_dataloader = DataLoader(
    tokenized_datasets["validation"], shuffle=False, collate_fn=collate_fn, batch_size=batch_size
)

Found cached dataset glue (/home/sourab/.cache/huggingface/datasets/glue/mrpc/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad)

  0%|          | 0/3 [00:00<?, ?it/s]

Loading cached processed dataset at /home/sourab/.cache/huggingface/datasets/glue/mrpc/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad/cache-121b991f592093a4.arrow
Loading cached processed dataset at /home/sourab/.cache/huggingface/datasets/glue/mrpc/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad/cache-6c55b9fb8fbb12c7.arrow
Loading cached processed dataset at /home/sourab/.cache/huggingface/datasets/glue/mrpc/1.0.0/dacbe3125aa31d7f70367a07a8a9e72a5a0bfeb5fc42e75c9db75b96da6053ad/cache-b9740d82185f93e5.arrow

In [5]:

model = AutoModelForSequenceClassification.from_pretrained(model_name_or_path, return_dict=True)
model = get_pet_model(model, pet_config)
model.print_trainable_parameters()
model

Out [5]:

Some weights of the model checkpoint at roberta-large were not used when initializing RobertaForSequenceClassification: ['lm_head.decoder.weight', 'lm_head.layer_norm.bias', 'lm_head.bias', 'roberta.pooler.dense.weight', 'lm_head.layer_norm.weight', 'lm_head.dense.bias', 'roberta.pooler.dense.bias', 'lm_head.dense.weight']
- This IS expected if you are initializing RobertaForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing RobertaForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
Some weights of RobertaForSequenceClassification were not initialized from the model checkpoint at roberta-large and are newly initialized: ['classifier.out_proj.bias', 'classifier.dense.weight', 'classifier.out_proj.weight', 'classifier.dense.bias']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.

trainable params: 2034690 || all params: 356344834 || trainable%: 0.5709890549444586

PETModelForSequenceClassification(
  (base_model): RobertaForSequenceClassification(
    (roberta): RobertaModel(
      (embeddings): RobertaEmbeddings(
        (word_embeddings): Embedding(50265, 1024, padding_idx=1)
        (position_embeddings): Embedding(514, 1024, padding_idx=1)
        (token_type_embeddings): Embedding(1, 1024)
        (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
        (dropout): Dropout(p=0.1, inplace=False)
      )
      (encoder): RobertaEncoder(
        (layer): ModuleList(
          (0): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (1): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (2): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (3): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (4): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (5): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (6): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (7): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (8): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (9): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (10): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (11): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (12): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (13): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (14): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (15): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (16): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (17): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (18): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (19): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (20): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (21): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (22): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (23): RobertaLayer(
            (attention): RobertaAttention(
              (self): RobertaSelfAttention(
                (query): Linear(in_features=1024, out_features=1024, bias=True)
                (key): Linear(in_features=1024, out_features=1024, bias=True)
                (value): Linear(in_features=1024, out_features=1024, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): RobertaSelfOutput(
                (dense): Linear(in_features=1024, out_features=1024, bias=True)
                (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): RobertaIntermediate(
              (dense): Linear(in_features=1024, out_features=4096, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): RobertaOutput(
              (dense): Linear(in_features=4096, out_features=1024, bias=True)
              (LayerNorm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
        )
      )
    )
    (classifier): RobertaClassificationHead(
      (dense): Linear(in_features=1024, out_features=1024, bias=True)
      (dropout): Dropout(p=0.1, inplace=False)
      (out_proj): Linear(in_features=1024, out_features=2, bias=True)
    )
  )
  (word_embeddings): Embedding(50265, 1024, padding_idx=1)
  (prompt_encoder): PrefixEncoder(
    (embedding): Embedding(20, 49152)
  )
)

In [6]:

optimizer = AdamW(params=model.parameters(), lr=lr)

# Instantiate scheduler
lr_scheduler = get_linear_schedule_with_warmup(
    optimizer=optimizer,
    num_warmup_steps=0.06*(len(train_dataloader) * num_epochs),
    num_training_steps=(len(train_dataloader) * num_epochs),
)

In [7]:

model.to(device)
for epoch in range(num_epochs):
    model.train()
    for step, batch in enumerate(tqdm(train_dataloader)):
        batch.to(device)
        outputs = model(**batch)
        loss = outputs.loss
        loss.backward()
        optimizer.step()
        lr_scheduler.step()
        optimizer.zero_grad()

    model.eval()
    for step, batch in enumerate(tqdm(eval_dataloader)):
        batch.to(device)
        with torch.no_grad():
            outputs = model(**batch)
        predictions = outputs.logits.argmax(dim=-1)
        predictions, references = predictions, batch["labels"]
        metric.add_batch(
            predictions=predictions,
            references=references,
        )

    eval_metric = metric.compute()
    print(f"epoch {epoch}:", eval_metric)

  0%|                                                                       | 0/115 [00:00<?, ?it/s]You're using a RobertaTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
100%|█████████████████████████████████████████████████████████████| 115/115 [01:04<00:00,  1.77it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:03<00:00,  3.73it/s]

epoch 0: {'accuracy': 0.7181372549019608, 'f1': 0.8200312989045383}

100%|█████████████████████████████████████████████████████████████| 115/115 [00:58<00:00,  1.97it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:03<00:00,  3.72it/s]

epoch 1: {'accuracy': 0.7720588235294118, 'f1': 0.8467874794069192}

100%|█████████████████████████████████████████████████████████████| 115/115 [00:58<00:00,  1.98it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:03<00:00,  3.70it/s]

epoch 2: {'accuracy': 0.7990196078431373, 'f1': 0.8637873754152825}

100%|█████████████████████████████████████████████████████████████| 115/115 [00:58<00:00,  1.97it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:03<00:00,  3.68it/s]

epoch 3: {'accuracy': 0.8259803921568627, 'f1': 0.883415435139573}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:27<00:00,  1.32it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:06<00:00,  1.86it/s]

epoch 4: {'accuracy': 0.8676470588235294, 'f1': 0.9}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:56<00:00,  1.01s/it]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:07<00:00,  1.75it/s]

epoch 5: {'accuracy': 0.8578431372549019, 'f1': 0.8921933085501859}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:56<00:00,  1.01s/it]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:06<00:00,  1.87it/s]

epoch 6: {'accuracy': 0.8700980392156863, 'f1': 0.9051878354203935}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:22<00:00,  1.39it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:03<00:00,  3.63it/s]

epoch 7: {'accuracy': 0.8774509803921569, 'f1': 0.9084249084249084}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:44<00:00,  1.10it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:07<00:00,  1.74it/s]

epoch 8: {'accuracy': 0.8799019607843137, 'f1': 0.9135802469135803}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:57<00:00,  1.02s/it]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:07<00:00,  1.73it/s]

epoch 9: {'accuracy': 0.8676470588235294, 'f1': 0.9052631578947367}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:56<00:00,  1.01s/it]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:06<00:00,  1.87it/s]

epoch 10: {'accuracy': 0.8725490196078431, 'f1': 0.9084507042253521}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:26<00:00,  1.33it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:03<00:00,  3.75it/s]

epoch 11: {'accuracy': 0.8799019607843137, 'f1': 0.9113924050632911}

100%|█████████████████████████████████████████████████████████████| 115/115 [00:54<00:00,  2.10it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:01<00:00,  8.25it/s]

epoch 12: {'accuracy': 0.8651960784313726, 'f1': 0.9053356282271946}

100%|█████████████████████████████████████████████████████████████| 115/115 [00:26<00:00,  4.27it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:02<00:00,  6.06it/s]

epoch 13: {'accuracy': 0.8774509803921569, 'f1': 0.912280701754386}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:11<00:00,  1.62it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:05<00:00,  2.52it/s]

epoch 14: {'accuracy': 0.8799019607843137, 'f1': 0.9123434704830053}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:26<00:00,  1.32it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:05<00:00,  2.52it/s]

epoch 15: {'accuracy': 0.8725490196078431, 'f1': 0.907473309608541}

100%|█████████████████████████████████████████████████████████████| 115/115 [00:53<00:00,  2.14it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:01<00:00,  8.04it/s]

epoch 16: {'accuracy': 0.8651960784313726, 'f1': 0.905982905982906}

100%|█████████████████████████████████████████████████████████████| 115/115 [00:51<00:00,  2.23it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:03<00:00,  3.73it/s]

epoch 17: {'accuracy': 0.8799019607843137, 'f1': 0.9129662522202486}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:18<00:00,  1.47it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:05<00:00,  2.51it/s]

epoch 18: {'accuracy': 0.8799019607843137, 'f1': 0.913884007029877}

100%|█████████████████████████████████████████████████████████████| 115/115 [01:26<00:00,  1.32it/s]
100%|███████████████████████████████████████████████████████████████| 13/13 [00:05<00:00,  2.48it/s]

epoch 19: {'accuracy': 0.8774509803921569, 'f1': 0.9125874125874125}

In [ ]:

In [ ]: