hash-map/got_fine_tuned

Mistral-7B-Instruct-v0.3-ORPO

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.3 on the got_qa_pairs dataset added some special tokens for charcters,places for better tokenizing to mistral tokenizer.

Model description

More information needed

Intended uses & limitations

for question answering better to use the script on gpu includes retreivers for better contextual answers

Training and evaluation data

dataset uploaded in hash-map/got_qa_pairs 95% is train and 5% is validation randomly shuffled

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-06
• train_batch_size: 3
• eval_batch_size: 3
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 12
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 1.0

Framework versions

• PEFT 0.13.0
• Transformers 4.45.1

metrics

• bertscore 0.4
• bleu 0.59999
• rouge 0.43