Text Classifier
Learn how to train a text classification model using Clarifai Python SDK
A text classifier is a machine learning model designed to automatically categorize or classify text documents into predefined categories or labels. These categories could be anything from sentiment analysis (positive, negative, neutral) to topic classification (sports, politics, technology). You can learn more about Text Classifier here.
App Creation
The first part of model training includes the creation of an app under which the training process takes place. Here we are creating an app with the app id as “demo_train” and the base workflow is set as “Universal”. You can change the base workflows to Empty, Universal, Language Understanding, and General according to your use case.
- Python
from clarifai.client.user import User
#replace your "user_id"
client = User(user_id="user_id")
app = client.create_app(app_id="demo_train", base_workflow="Universal")
Dataset Upload
The next step involves dataset upload. You can upload the dataset to your app so that the model accepts the data directly from the platform. The data used for training in this tutorial is available in the examples repository you have cloned.
- Python
#importing load_module_dataloader for calling the dataloader object in dataset.py in the local data folder
from clarifai.datasets.upload.utils import load_module_dataloader
# Construct the path to the dataset folder
CSV_PATH = os.path.join(os.getcwd().split('/models/model_train')[0],'datasets/upload/data/imdb.csv')
# Create a Clarifai dataset with the specified dataset_id
dataset = app.create_dataset(dataset_id="text_dataset")
# Upload the dataset using the provided dataloader and get the upload status
dataset.upload_from_csv(csv_path=CSV_PATH,input_type='text',csv_type='raw', labels=True)
If you have followed the steps correctly you should receive an output that looks like this,
Output
Choose The Model Type
First let's list the all available trainable model types in the platform,
- Python
print(app.list_trainable_model_types())
Output
['visual-classifier',
'visual-detector',
'visual-segmenter',
'visual-embedder',
'clusterer',
'text-classifier',
'embedding-classifier',
'text-to-text']
Click here to know more about Clarifai Model Types.
Model Creation
From the above list of model types we are going to choose text-classifier as it is similar to our use case. Now let's create a model with the above model type.
- Python
MODEL_ID = "model_text_classifier"
MODEL_TYPE_ID = "text-classifier"
# Create a model by passing the model name and model type as parameter
model = app.create_model(model_id=MODEL_ID, model_type_id=MODEL_TYPE_ID)
Output
Template Selection
Inside the Clarifiai platform there is a template feature. Templates give you the control to choose the specific architecture used by your neural network, as well as define a set of hyperparameters you can use to fine-tune the way your model learns. We are going to choose the 'HuggingFace_AdvancedConfig' template for training our model.
- Python
print(model.list_training_templates())
Output
['HF_GPTNeo_125m_lora',
'HF_GPTNeo_2p7b_lora',
'HF_Llama_2_13b_chat_GPTQ_lora',
'HF_Llama_2_7b_chat_GPTQ_lora',
'HF_Mistral_7b_instruct_GPTQ_lora',
'HuggingFace_AdvancedConfig']
Setup Model Parameters
You can update the model params to your need before initiating training.
- Python
# get the model parameters
model_params = model.get_params(template='HuggingFace_AdvancedConfig')
concepts = [concept.id for concept in app.list_concepts()]
# update the concept field in model parameters
model.update_params(dataset_id = 'text_dataset',concepts = ["id-pos","id-neg"])
Output
{'dataset_id': 'text_dataset',
'dataset_version_id': '',
'concepts': ['id-pos', 'id-neg'],
'train_params': {'invalid_data_tolerance_percent': 5.0,
'template': 'HuggingFace_AdvancedConfig',
'model_config': {'problem_type': 'multi_label_classification',
'pretrained_model_name_or_path': 'bert-base-cased',
'torch_dtype': 'torch.float32'},
'tokenizer_config': {},
'trainer_config': {'num_train_epochs': 1.0,
'auto_find_batch_size': True,
'output_dir': 'checkpoint'}},
'inference_params': {'select_concepts': []}}
Initiate Model Training
We can initiate the model training by calling the model.train() method. The Clarifai Python SDK also offers features like showing training status and saving training logs in a local file.
If the status code is 'MODEL-TRAINED', then the user can know the Model is Trained and ready to use.
- Python
import time
#Starting the training
model_version_id = model.train()
#Checking the status of training
while True:
status = model.training_status(version_id=model_version_id,training_logs=False)
if status.code == 21106: #MODEL_TRAINING_FAILED
print(status)
break
elif status.code == 21100: #MODEL_TRAINED
print(status)
break
else:
print("Current Status:",status)
print("Waiting---")
time.sleep(120)
Output
Model Prediction
Since the model is trained and ready let’s run some predictions to view the model performance,
- Python
# Getting the predictions
TEXT = b"This is a great place to work"
model_prediction = model.predict_by_bytes(TEXT, input_type="text")
# Get the output
print('Input: ',TEXT)
for concept in model_prediction.outputs[0].data.concepts:
print(concept.id,':',round(concept.value,2))
Output
Input: b'This is a great place to work'
id-neg : 0.56
id-pos : 0.39
Model Evaluation
Now let's evaluate the model using train and test datasets. First let's see the evaluation metrics for the training dataset,
- Python
# Evaluate the model using the specified dataset ID 'text_dataset' and evaluation ID 'one'.
model.evaluate(dataset_id='text_dataset', eval_id='one')
# Retrieve the evaluation result for the evaluation ID 'one'.
result = model.get_eval_by_id(eval_id="one")
# Print the summary of the evaluation result.
print(result.summary)
Output
macro_avg_roc_auc: 0.6499999761581421
macro_std_roc_auc: 0.07468751072883606
macro_avg_f1_score: 0.75
macro_avg_precision: 0.6000000238418579
macro_avg_recall: 0.5
Before evaluating with a test dataset, we have to first upload the dataset using the data loader and then perform model evaluation,
- Python
#importing load_module_dataloader for calling the dataloader object in dataset.py in the local data folder
from clarifai.datasets.upload.utils import load_module_dataloader
# Construct the path to the dataset folder
CSV_PATH = os.path.join(os.getcwd().split('/models/model_train')[0],'datasets/upload/data/test_imdb.csv')
# Create a Clarifai dataset with the specified dataset_id
test_dataset = app.create_dataset(dataset_id="test_text_dataset")
# Upload the dataset using the provided dataloader and get the upload status
test_dataset.upload_from_csv(csv_path=CSV_PATH,input_type='text',csv_type='raw', labels=True)
# Evaluate the model using the specified test text dataset identified as 'test_text_dataset'
# and the evaluation identifier 'two'.
model.evaluate(dataset_id='test_text_dataset', eval_id='two')
# Retrieve the evaluation result with the identifier 'two'.
result = model.get_eval_by_id("two")
# Print the summary of the evaluation result.
print(result.summary)
Output
macro_avg_roc_auc: 0.6161290407180786
macro_std_roc_auc: 0.1225806474685669
macro_avg_f1_score: 0.7207207679748535
macro_avg_precision: 0.5633803009986877
macro_avg_recall: 0.5
Finally let's compare the results from multiple datasets using EvalResultCompare feature from Clarifai Python SDK to get a better understanding of the model's performance.
- Python
from clarifai.utils.evaluation import EvalResultCompare
# Creating an instance of EvalResultCompare class with specified models and datasets
eval_result = EvalResultCompare(models=[model], datasets=[dataset, test_dataset])
# Printing a detailed summary of the evaluation result
print(eval_result.detailed_summary())
Output
( Concept Accuracy (ROC AUC) Total Labeled Total Predicted True Positives \
0 id-pos 0.725 80 0 0
0 id-neg 0.575 120 200 120
0 id-pos 0.739 31 0 0
0 id-neg 0.494 40 71 40
False Negatives False Positives Recall Precision F1 \
0 80 0 0.0 1.0000 0.000000
0 0 80 1.0 0.6000 0.750000
0 31 0 0.0 1.0000 0.000000
0 0 31 1.0 0.5634 0.720737
Dataset
0 text_dataset2
0 text_dataset2
0 test_text_dataset3
0 test_text_dataset3 ,
Total Concept Accuracy (ROC AUC) Total Labeled \
0 Dataset:text_dataset2 0.650000 200
0 Dataset:test_text_dataset3 0.616129 71
Total Predicted True Positives False Negatives False Positives Recall \
0 200 120 80 80 0.60000
0 71 40 31 31 0.56338
Precision F1
0 0.760000 0.670588
0 0.754028 0.644909 )