|
--- |
|
base_model: |
|
- google/gemma-2-2b |
|
pipeline_tag: text-generation |
|
metrics: |
|
- accuracy |
|
license: mit |
|
language: |
|
- en |
|
tags: |
|
- pytorch |
|
- transformers |
|
- keras |
|
- Spell Checker |
|
--- |
|
|
|
## Description |
|
The RLM-spell-checker is a fine-tuned version of gemma-2b-V3, enhanced using LoRA (Low-Rank Adaptation) to specialize in spelling correction. LoRA fine-tunes models efficiently by adjusting only a few parameters, allowing the RLM-spell-checker to retain the robust language understanding of gemma-2b-V3 while focusing on identifying and correcting spelling errors. This fine-tuning enables the model to provide context-aware suggestions for corrections, making it a powerful tool for real-time applications like automated writing assistance, chatbots, and word processors. By improving spelling accuracy without interrupting the natural flow of text, the RLM-spell-checker enhances text quality and user experience in various tasks. |
|
|
|
### Author: [Rudra Shah](https://www.linkedin.com/in/rudra-shah-b044781b4/) |
|
|
|
## Running Model |
|
``` python |
|
# Load model directly |
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
|
tokenizer = AutoTokenizer.from_pretrained("rudrashah/RLM-spell-checker") |
|
model = AutoModelForCausalLM.from_pretrained("rudrashah/RLM-spell-checker") |
|
|
|
sent = "Whaat iss the mision?" |
|
template = "Sentence:\n{org}\n\nCorrect_Grammar:\n{new}" |
|
input_text = template.format(org=sent, new="") |
|
|
|
input_ids = tokenizer(input_text, return_tensors="pt") |
|
outputs = model.generate(**input_ids, max_new_tokens=128) |
|
print(tokenizer.decode(outputs[0])) |
|
``` |
