Upload model.py
Browse files
model.py
ADDED
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import json
|
2 |
+
import pandas as pd
|
3 |
+
import numpy as np
|
4 |
+
import torch
|
5 |
+
from pathlib import Path
|
6 |
+
import lightning as pl
|
7 |
+
from lightning.pytorch.callbacks import ModelCheckpoint
|
8 |
+
from lightning.pytorch.loggers import TensorBoardLogger
|
9 |
+
from torch.utils.data import Dataset, DataLoader
|
10 |
+
import textwrap
|
11 |
+
from transformers import (
|
12 |
+
AdamW,
|
13 |
+
T5ForConditionalGeneration,
|
14 |
+
T5TokenizerFast as T5Tokenizer
|
15 |
+
)
|
16 |
+
from tqdm.auto import tqdm
|
17 |
+
|
18 |
+
class NewsSummaryModel(pl.LightningModule):
|
19 |
+
def __init__(self):
|
20 |
+
super().__init__()
|
21 |
+
self.model= T5ForConditionalGeneration.from_pretrained("t5-base", return_dict=True)
|
22 |
+
def forward(self,input_ids, attention_mask, decoder_attention_mask, labels=None):
|
23 |
+
output = self.model(
|
24 |
+
input_ids,
|
25 |
+
attention_mask=attention_mask,
|
26 |
+
labels=labels,
|
27 |
+
decoder_attention_mask=decoder_attention_mask
|
28 |
+
)
|
29 |
+
return output.loss, output.logits
|
30 |
+
def training_step(self, batch, batch_idx):
|
31 |
+
input_ids=batch["text_input_ids"]
|
32 |
+
attention_mask=batch["text_attention_mask"]
|
33 |
+
labels=batch["labels"]
|
34 |
+
labels_attention_mask=batch["labels_attention_mask"]
|
35 |
+
|
36 |
+
|
37 |
+
loss, outputs = self(
|
38 |
+
input_ids=input_ids,
|
39 |
+
attention_mask=attention_mask,
|
40 |
+
decoder_attention_mask=labels_attention_mask,
|
41 |
+
labels=labels
|
42 |
+
)
|
43 |
+
self.log("train_loss", loss, prog_bar=True, logger=True)
|
44 |
+
return loss
|
45 |
+
def validation_step(self, batch, batch_idx):
|
46 |
+
input_ids=batch["text_input_ids"]
|
47 |
+
attention_mask=batch["text_attention_mask"]
|
48 |
+
labels=batch["labels"]
|
49 |
+
labels_attention_mask=batch["labels_attention_mask"]
|
50 |
+
|
51 |
+
|
52 |
+
loss, outputs = self(
|
53 |
+
input_ids=input_ids,
|
54 |
+
attention_mask=attention_mask,
|
55 |
+
decoder_attention_mask=labels_attention_mask,
|
56 |
+
labels=labels
|
57 |
+
)
|
58 |
+
self.log("val_loss", loss, prog_bar=True, logger=True)
|
59 |
+
return loss
|
60 |
+
def test_step(self, batch, batch_idx):
|
61 |
+
input_ids=batch["text_input_ids"]
|
62 |
+
attention_mask=batch["text_attention_mask"]
|
63 |
+
labels=batch["labels"]
|
64 |
+
labels_attention_mask=batch["labels_attention_mask"]
|
65 |
+
|
66 |
+
|
67 |
+
loss, outputs = self(
|
68 |
+
input_ids=input_ids,
|
69 |
+
attention_mask=attention_mask,
|
70 |
+
decoder_attention_mask=labels_attention_mask,
|
71 |
+
labels=labels
|
72 |
+
)
|
73 |
+
self.log("test_loss", loss, prog_bar=True, logger=True)
|
74 |
+
return loss
|
75 |
+
def configure_optimizers(self):
|
76 |
+
return AdamW(self.parameters(), lr=0.0001)
|
77 |
+
|
78 |
+
|