mistral-QA-ViMMRC-Squad-v1.1

This model is a fine-tuned version of unsloth/mistral-7b-instruct-v0.2-bnb-4bit on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.0484

Model description

More information needed

Intended uses & limitations

• Prompt 1: Given the following reference, create a question and a corresponding answer to the question: + [context]
• Prompt 2: Given the following reference, create a multiple-choice question and its corresponding answer: + [context]

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 16
• eval_batch_size: 8
• seed: 3407
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 5
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
1.0039	0.2307	320	1.4915
0.8101	0.4614	640	1.5005
0.6909	0.6921	960	1.5480
0.5851	0.9229	1280	1.5734
0.4374	1.1536	1600	1.6432
0.3462	1.3843	1920	1.6886
0.2845	1.6150	2240	1.7347
0.2236	1.8457	2560	1.8193
0.158	2.0764	2880	1.9148
0.1124	2.3071	3200	1.9873
0.0981	2.5379	3520	2.0051
0.0892	2.7686	3840	2.0392
0.0856	2.9993	4160	2.0484

Framework versions

• PEFT 0.10.0
• Transformers 4.40.2
• Pytorch 2.3.0
• Datasets 2.19.1
• Tokenizers 0.19.1