Auto Text Classification

In [ ]:

import os
import sys

os.chdir("../../")

import os import sys os.chdir("../../")

In [ ]:

import warnings

warnings.filterwarnings("ignore")

import warnings warnings.filterwarnings("ignore")

In [ ]:

!pip install git+https://github.com/gradsflow/gradsflow@main

!pip install git+https://github.com/gradsflow/gradsflow@main

Installing collected packages: gradsflow
Successfully installed gradsflow-0.0.8

In [ ]:

!pip install "lightning-flash[text]" wandb

!pip install "lightning-flash[text]" wandb

In [ ]:

from flash.core.data.utils import download_data
from flash.text import TextClassificationData
from gradsflow import AutoTextClassifier

import ray

from flash.core.data.utils import download_data from flash.text import TextClassificationData from gradsflow import AutoTextClassifier import ray

In [ ]:

download_data("https://pl-flash-data.s3.amazonaws.com/imdb.zip", "./data/")

datamodule = TextClassificationData.from_csv(
    "review",
    "sentiment",
    train_file="data/imdb/train.csv",
    val_file="data/imdb/valid.csv",
    batch_size=64,
)

download_data("https://pl-flash-data.s3.amazonaws.com/imdb.zip", "./data/") datamodule = TextClassificationData.from_csv( "review", "sentiment", train_file="data/imdb/train.csv", val_file="data/imdb/valid.csv", batch_size=64, )

Using custom data configuration default-9e8b8625e23aa791

Downloading and preparing dataset csv/default to /root/.cache/huggingface/datasets/csv/default-9e8b8625e23aa791/0.0.0/433e0ccc46f9880962cc2b12065189766fbb2bee57a221866138fb9203c83519...

Downloading data files:   0%|          | 0/1 [00:00<?, ?it/s]

Extracting data files:   0%|          | 0/1 [00:00<?, ?it/s]

Dataset csv downloaded and prepared to /root/.cache/huggingface/datasets/csv/default-9e8b8625e23aa791/0.0.0/433e0ccc46f9880962cc2b12065189766fbb2bee57a221866138fb9203c83519. Subsequent calls will reuse this data.

  0%|          | 0/1 [00:00<?, ?it/s]

  0%|          | 0/22500 [00:00<?, ?ex/s]

Using custom data configuration default-6a8d347e65f511f3

Downloading and preparing dataset csv/default to /root/.cache/huggingface/datasets/csv/default-6a8d347e65f511f3/0.0.0/433e0ccc46f9880962cc2b12065189766fbb2bee57a221866138fb9203c83519...

Downloading data files:   0%|          | 0/1 [00:00<?, ?it/s]

Extracting data files:   0%|          | 0/1 [00:00<?, ?it/s]

Dataset csv downloaded and prepared to /root/.cache/huggingface/datasets/csv/default-6a8d347e65f511f3/0.0.0/433e0ccc46f9880962cc2b12065189766fbb2bee57a221866138fb9203c83519. Subsequent calls will reuse this data.

  0%|          | 0/1 [00:00<?, ?it/s]

  0%|          | 0/2500 [00:00<?, ?ex/s]

In [ ]:

suggested_conf = dict(
    optimizer=["adam", "adamw"],
    lr=(5e-4, 1e-3),
)

model = AutoTextClassifier(
    datamodule,
    suggested_backbones=["prajjwal1/bert-tiny"],
    suggested_conf=suggested_conf,
    max_epochs=4,
    optimization_metric="val_accuracy",
    n_trials=1,
)

print("AutoTextClassifier initialised!")
trainer_config = {"accelerator":"auto", "devices":1}
model.hp_tune(gpu=1, trainer_config=trainer_config)

suggested_conf = dict( optimizer=["adam", "adamw"], lr=(5e-4, 1e-3), ) model = AutoTextClassifier( datamodule, suggested_backbones=["prajjwal1/bert-tiny"], suggested_conf=suggested_conf, max_epochs=4, optimization_metric="val_accuracy", n_trials=1, ) print("AutoTextClassifier initialised!") trainer_config = {"accelerator":"auto", "devices":1} model.hp_tune(gpu=1, trainer_config=trainer_config)

In [ ]:

model.analysis.dataframe()

model.analysis.dataframe()

Out[ ]:

	val_accuracy	train_accuracy	time_this_iter_s	should_checkpoint	done	timesteps_total	episodes_total	training_iteration	trial_id	experiment_id	...	hostname	node_ip	time_since_restore	timesteps_since_restore	iterations_since_restore	warmup_time	config/backbone	config/lr	config/optimizer	logdir
0	0.8004	0.984375	30.232423	True	False	NaN	NaN	4	f3e66_00000	7ada6e549a4e4a7fbdfddadc99012877	...	b206c2fd5e9b	172.28.0.2	139.360543	0	4	0.007067	prajjwal1/bert-tiny	0.000919	adam	/root/ray_results/optimization_objective_2022-...

1 rows × 24 columns

In [ ]:

from flash import Trainer
trainer = Trainer(accelerator="auto", devices=1)

from flash import Trainer trainer = Trainer(accelerator="auto", devices=1)

GPU available: True, used: True
TPU available: False, using: 0 TPU cores
IPU available: False, using: 0 IPUs
HPU available: False, using: 0 HPUs

In [ ]:

trainer.validate(model.model, datamodule=datamodule)

trainer.validate(model.model, datamodule=datamodule)

LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]

Validation: 0it [00:00, ?it/s]

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃      Validate metric      ┃       DataLoader 0        ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━┩
│       val_accuracy        │    0.8787053227424622     │
│     val_cross_entropy     │     0.410677433013916     │
└───────────────────────────┴───────────────────────────┘

Out[ ]:

[{'val_accuracy': 0.8787053227424622, 'val_cross_entropy': 0.410677433013916}]

---

Last update: May 18, 2022