imdb_tuned-bloom1b1-sentiment-classifier

A Finetuned Bloom 1b1 Model for Sequence Classification

<!-- Provide a quick summary of what the model is/does. -->

The model was developed as a personal learning experience to fine tune a ready language model for Text Classification and to use it on real life data from the internet to perform sentiment analysis.

It has been generated using this raw template.

Model Details

The model achieves the following scores on the evaluation set during the fine tuning:

Here is the train/ eval/ test split:

DatasetDict({
    train: Dataset({
        features: ['review', 'sentiment'],
        num_rows: 36000
    })
    test: Dataset({
        features: ['review', 'sentiment'],
        num_rows: 5000
    })
    eval: Dataset({
        features: ['review', 'sentiment'],
        num_rows: 9000
    })
})

Model Description

<!-- Provide a longer summary of what this model is. -->

• Developed by: Snoop088
• Model type: Text Classification / Sequence Classification
• Language(s) (NLP): English
• License: Apache 2.0
• **Finetuned from model: bigscience/bloom-1b1

Model Sources [optional]

<!-- Provide the basic links for the model. -->

• Repository: https://huggingface.co/snoop088/imdb_tuned-bloom1b1-sentiment-classifier/tree/main
• Paper [optional]: [More Information Needed]
• Demo [optional]: [More Information Needed]

Uses

The model is intended to be used for Text Classification.

Direct Use

Example script to use the model. Please note that this is peft adapter on the Bloom 1b model:

DEVICE = "cuda:0" if torch.cuda.is_available() else "cpu"
model_name = 'snoop088/imdb_tuned-bloom1b1-sentiment-classifier'
loaded_model = AutoModelForSequenceClassification.from_pretrained(model_name, 
                                                                  trust_remote_code=True, 
                                                                  num_labels=2,
                                                                  device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
tokenizer.pad_token = tokenizer.eos_token

my_set = pd.read_csv("./data/df_manual.csv")

inputs = tokenizer(list(my_set["review"]), truncation=True, padding="max_length", max_length=256,  return_tensors="pt").to(DEVICE)
outputs = loaded_model(**inputs)
outcome = np.argmax(torch.Tensor.cpu(outputs.logits), axis=-1)

[More Information Needed]

Downstream Use [optional]

The purpose of this model is to be used to perform sentiment analysis on a dataset similar to the one by IMDB. It should work well on product reviews, too in my opinion.

[More Information Needed]

Out-of-Scope Use

<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->

[More Information Needed]

Bias, Risks, and Limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->

[More Information Needed]

Recommendations

<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

How to Get Started with the Model

Use the code below to get started with the model.

[More Information Needed]

Training Details

Training Data

<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->

Training is done on the IMDB dataset available on the Hub:

imdb

[More Information Needed]

Training Procedure

training_arguments = TrainingArguments(
    output_dir="your_tuned_model_name",
    save_strategy="epoch",
    per_device_train_batch_size=4,
    per_device_eval_batch_size=4,
    gradient_accumulation_steps=4,
    optim="adamw_torch",
    evaluation_strategy="steps",
    logging_steps=5,
    learning_rate=1e-5,
    max_grad_norm = 0.3,
    eval_steps=0.2,
    num_train_epochs=2,
    warmup_ratio= 0.1,
    # group_by_length=True,
    fp16=False,
    weight_decay=0.001,
    lr_scheduler_type="constant",
)

peft_model = get_peft_model(model, LoraConfig(
                            task_type="SEQ_CLS",
                            r=16,
                            lora_alpha=16,
                            target_modules=[
                                'query_key_value',
                                'dense'
                            ],
                            bias="none",
                            lora_dropout=0.05, # Conventional
                        ))

LORA results in: trainable params: 3,542,016 || all params: 1,068,859,392 || trainable%: 0.3313827830405592

<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->

Preprocessing [optional]

Simple preprocessing with DataCollator:

def process_data(example):
    item = tokenizer(example["review"], truncation=True, max_length=320) # see if this is OK for dyn padding
    item["labels"] = [ 1 if sent == 'positive' else 0 for sent in example["sentiment"]]
    return item

tokenised_data = tokenised_data.remove_columns(["review", "sentiment"])
data_collator = DataCollatorWithPadding(tokenizer=tokenizer)

Training Hyperparameters

• Training regime: [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->

Speeds, Sizes, Times [optional]

<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->

[More Information Needed]

Evaluation

<!-- This section describes the evaluation protocols and provides the results. -->

Evaluation function:

import evaluate

def compute_metrics(eval_pred):
    # All metrics are already predefined in the HF `evaluate` package
    precision_metric = evaluate.load("precision")
    recall_metric = evaluate.load("recall")
    f1_metric= evaluate.load("f1")
    accuracy_metric = evaluate.load("accuracy")

    logits, labels = eval_pred # eval_pred is the tuple of predictions and labels returned by the model
    predictions = np.argmax(logits, axis=-1)
    precision = precision_metric.compute(predictions=predictions, references=labels)["precision"]
    recall = recall_metric.compute(predictions=predictions, references=labels)["recall"]
    f1 = f1_metric.compute(predictions=predictions, references=labels)["f1"]
    accuracy = accuracy_metric.compute(predictions=predictions, references=labels)["accuracy"]
    # The trainer is expecting a dictionary where the keys are the metrics names and the values are the scores. 
    return {"precision": precision, "recall": recall, "f1-score": f1, 'accuracy': accuracy}

Testing Data, Factors & Metrics

Testing Data

<!-- This should link to a Dataset Card if possible. -->

[More Information Needed]

Factors

<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->

[More Information Needed]

Metrics

<!-- These are the evaluation metrics being used, ideally with a description of why. -->

[More Information Needed]

Results

[More Information Needed]

Summary

Model Examination [optional]

<!-- Relevant interpretability work for the model goes here -->

[More Information Needed]

Environmental Impact

<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

• Hardware Type: [More Information Needed]
• Hours used: [More Information Needed]
• Cloud Provider: [More Information Needed]
• Compute Region: [More Information Needed]
• Carbon Emitted: [More Information Needed]

Technical Specifications [optional]

Model Architecture and Objective

[More Information Needed]

Compute Infrastructure

[More Information Needed]

Hardware

• Model: 6.183.1 "13th Gen Intel(R) Core(TM) i9-13900K"
• GPU: Nvidia RTX 4900/ 24 GB
• Memory: 64 GB

Software

• python 3.11.6
• transformers 4.36.2
• torch 2.1.2
• peft 0.7.1
• numpy 1.26.2
• datasets 2.16.0

Citation [optional]

<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->

BibTeX:

[More Information Needed]

APA:

[More Information Needed]

Glossary [optional]

<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->

[More Information Needed]

More Information [optional]

[More Information Needed]

Model Card Authors [optional]

[More Information Needed]

Model Card Contact

[More Information Needed]