Shuddho Qwen3-8B Vaiyakarana 98K Continued LoRA
A continued LoRA adapter for Bangla grammar, spelling, punctuation, spacing and writing correction.
Base model
Qwen/Qwen3-8B
Continued from
rummanali65/shuddho-qwen3-8b-vaiyakarana-83k-continued-lora
Cumulative training path
- V1: 10K examples
- V2: additional 20K examples
- V3: additional 23K training rows
- V4: additional 30K examples
- V5: additional 15K examples
Total cumulative training exposure: approximately 98K examples.
Important
This repository contains the LoRA adapter only.
For inference, load Qwen/Qwen3-8B first and then attach this adapter using PEFT.
- Downloads last month
- 73
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-8B") model = PeftModel.from_pretrained(base_model, "rummanali65/shuddho-qwen3-8b-vaiyakarana-98k-continued-lora")