Quantization made by Richard Erkhov.
mpt_1000_STEPS_1e6_SFT_SFT - bnb 4bits
- Model creator: https://huggingface.co/tsavage68/
- Original model: https://huggingface.co/tsavage68/mpt_1000_STEPS_1e6_SFT_SFT/
Original model description:
license: apache-2.0 base_model: mosaicml/mpt-7b-instruct tags: - trl - sft - generated_from_trainer model-index: - name: mpt_1000_STEPS_1e6_SFT_SFT results: []
mpt_1000_STEPS_1e6_SFT_SFT
This model is a fine-tuned version of mosaicml/mpt-7b-instruct on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.4128
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-06
- train_batch_size: 2
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- training_steps: 1000
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 1.568 | 0.05 | 50 | 1.4778 |
| 0.4601 | 0.1 | 100 | 0.4773 |
| 0.4655 | 0.15 | 150 | 0.4432 |
| 0.4866 | 0.2 | 200 | 0.4338 |
| 0.4309 | 0.24 | 250 | 0.4279 |
| 0.4481 | 0.29 | 300 | 0.4238 |
| 0.4239 | 0.34 | 350 | 0.4206 |
| 0.4025 | 0.39 | 400 | 0.4184 |
| 0.4377 | 0.44 | 450 | 0.4169 |
| 0.4192 | 0.49 | 500 | 0.4154 |
| 0.407 | 0.54 | 550 | 0.4145 |
| 0.4291 | 0.59 | 600 | 0.4136 |
| 0.4048 | 0.64 | 650 | 0.4133 |
| 0.397 | 0.68 | 700 | 0.4131 |
| 0.4016 | 0.73 | 750 | 0.4128 |
| 0.4108 | 0.78 | 800 | 0.4128 |
| 0.4427 | 0.83 | 850 | 0.4128 |
| 0.3882 | 0.88 | 900 | 0.4127 |
| 0.3929 | 0.93 | 950 | 0.4127 |
| 0.4125 | 0.98 | 1000 | 0.4128 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.0.0+cu117
- Datasets 2.18.0
- Tokenizers 0.15.2