Instructions to use pavanBuduguppa/asr_inverse_text_normalization with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use pavanBuduguppa/asr_inverse_text_normalization with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="pavanBuduguppa/asr_inverse_text_normalization")

# Load model directly
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer = AutoTokenizer.from_pretrained("pavanBuduguppa/asr_inverse_text_normalization")
model = AutoModelForSeq2SeqLM.from_pretrained("pavanBuduguppa/asr_inverse_text_normalization")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use pavanBuduguppa/asr_inverse_text_normalization with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "pavanBuduguppa/asr_inverse_text_normalization"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "pavanBuduguppa/asr_inverse_text_normalization",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/pavanBuduguppa/asr_inverse_text_normalization

SGLang

How to use pavanBuduguppa/asr_inverse_text_normalization with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "pavanBuduguppa/asr_inverse_text_normalization" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "pavanBuduguppa/asr_inverse_text_normalization",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "pavanBuduguppa/asr_inverse_text_normalization" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "pavanBuduguppa/asr_inverse_text_normalization",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use pavanBuduguppa/asr_inverse_text_normalization with Docker Model Runner:
```
docker model run hf.co/pavanBuduguppa/asr_inverse_text_normalization
```

YAML Metadata Warning:The pipeline tag "text2text-generation" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, audio-text-to-text, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-ranking, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, image-text-to-image, image-text-to-video, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, video-text-to-text, keypoint-detection, visual-document-retrieval, any-to-any, video-to-video, other

asr_inverse_text_normalization

Finetuned a facebook/bart-base Pretrained model on the ASR inverse text normalization dataset by treating it as a seq2seq task. Other approaches which may be considered is by considering it as a TokenClassification task and the one mentioned here https://machinelearning.apple.com/research/inverse-text-normal.

Model description

BART (Bidirectional and Auto-Regressive Transformers) is a pre-trained transformer-based neural network model developed by Facebook AI Research (FAIR) for various natural language processing (NLP) tasks

The BART architecture is based on the Transformer model, which is a type of neural network architecture that processes sequential input data, such as text, by applying self-attention mechanisms to capture the relationships between different words in the input sequence. BART includes both auto-regressive and bidirectional encoder-decoder transformer architectures, which enable it to perform both generation and prediction tasks

BART was trained on a diverse range of NLP tasks, including machine translation, summarization, and question answering, and has shown strong performance across multiple benchmarks. Its training process involves corrupting text with different types of noise and training the model to reconstruct the original text, which has been shown to improve the model's ability to generalize to new tasks and outperform other pre-trained language models like GPT and BERT

The model flavour which was chosen is that of "facebook/bart-base" and columns "after" is used as the source while "before" column is used as the targets.

Intended uses & limitations

This model can be used as an out-of-the-box solution to the invesrse text normalization which can convert ASR generated un-normalized text such as "my c v v for my card is five six seven and it expires on november twenty three" -> "my CVV for my card is 567 and it expires on November 23"

The model needs to be explored for various min and max length setting at the time of generation for your specific usecase

How to use


>>> from transformers import pipeline
>>> generator = pipeline(model="pavanBuduguppa/asr_inverse_text_normalization")

>>> generator("my c v v for my card is five six seven and it expires on november twenty three")

Training data

All credits and rights for the training data belongs to Google. The data was merely obtained and processed for this model and the original data can be found here https://www.kaggle.com/competitions/text-normalization-challenge-english-language/data

Downloads last month: 13

Safetensors

Model size

0.1B params

Tensor type

F32

pavanBuduguppa
/

asr_inverse_text_normalization

asr_inverse_text_normalization

Model description

Intended uses & limitations

How to use

Training data

Dataset used to train pavanBuduguppa/asr_inverse_text_normalization