metadata
base_model: mistralai/Mistral-7B-Instruct-v0.1
datasets:
- theoldmandthesea/17k_business_book
library_name: peft
license: apache-2.0
tags:
- trl
- sft
- generated_from_trainer
model-index:
- name: business_mistral-v1.0
results: []
language:
- en
Mistral 7B Fine-Tuned for Business
This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.1 on the theoldmandthesea/17k_business_book dataset. It achieves the following results on the evaluation set:
- Loss: 0.5729
Model description
- Base Model: mistralai/Mistral-7B-Instruct-v0.1
- Quantization: 4 Bit
- Tokenizer: mistralai/Mistral-7B-Instruct-v0.1
Intended uses & limitations
Mainly intended to be used at the backend of a chatbot application meant to assist children and teenager with knowledge of business and finances.
Training and evaluation data
Total number of samples: 17480
Number of Training Samples: 13984
Number of Validation / Test Samples: 3496
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0002
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: constant
- training_steps: 100
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.6197 | 0.1869 | 20 | 0.6099 |
| 0.5927 | 0.3738 | 40 | 0.5905 |
| 0.5664 | 0.5607 | 60 | 0.5824 |
| 0.572 | 0.7477 | 80 | 0.5767 |
| 0.5817 | 0.9346 | 100 | 0.5729 |
Framework versions
- PEFT 0.11.1
- Transformers 4.42.3
- Pytorch 2.1.2
- Datasets 2.20.0
- Tokenizers 0.19.1