ISHA Call Center QA Model

This model was trained on a finetuned version(StableBeluga2) of Llama2-13B from stabilityai : (StableBeluga2 tops the LLM leaderboard currently)

Dataset Used : https://huggingface.co/datasets/nateshmbhat/isha-qa-text

Train Params used :

• Base model : stabilityai/StableBeluga-13B
• Quantization Used : 4 bit
• Learning rate : 2e-4
• Batch Size : 2
• Epochs : 3
• Trainer : sft
• Max token length : 2048 (capable of higher token length)

!autotrain llm --train --project_name project-isha-qa --model stabilityai/StableBeluga-13B --data_path nateshmbhat/isha-qa-text --use_peft --use_int4 --learning_rate 2e-4 --train_batch_size 2 --num_train_epochs 3 --trainer sft --model_max_length 2048 --push_to_hub --repo_id nateshmbhat/model-isha-qa