Quantization made by Richard Erkhov.
mpt_1000_STEPS_1e7_SFT_SFT - bnb 8bits
- Model creator: https://huggingface.co/tsavage68/
- Original model: https://huggingface.co/tsavage68/mpt_1000_STEPS_1e7_SFT_SFT/
Original model description:
license: apache-2.0 base_model: mosaicml/mpt-7b-instruct tags: - trl - sft - generated_from_trainer model-index: - name: v1_1000_STEPS_1e7_SFT_SFT results: []
v1_1000_STEPS_1e7_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: 1.4957
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-07
- 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 |
|---|---|---|---|
| 1.8395 | 0.05 | 50 | 1.8528 |
| 1.8225 | 0.1 | 100 | 1.8184 |
| 1.7682 | 0.15 | 150 | 1.7635 |
| 1.7241 | 0.2 | 200 | 1.7138 |
| 1.6752 | 0.24 | 250 | 1.6696 |
| 1.6395 | 0.29 | 300 | 1.6321 |
| 1.6 | 0.34 | 350 | 1.5992 |
| 1.5692 | 0.39 | 400 | 1.5718 |
| 1.5362 | 0.44 | 450 | 1.5487 |
| 1.5344 | 0.49 | 500 | 1.5312 |
| 1.508 | 0.54 | 550 | 1.5173 |
| 1.5041 | 0.59 | 600 | 1.5080 |
| 1.5054 | 0.64 | 650 | 1.5018 |
| 1.488 | 0.68 | 700 | 1.4984 |
| 1.5004 | 0.73 | 750 | 1.4964 |
| 1.4982 | 0.78 | 800 | 1.4959 |
| 1.5145 | 0.83 | 850 | 1.4958 |
| 1.4867 | 0.88 | 900 | 1.4959 |
| 1.4961 | 0.93 | 950 | 1.4957 |
| 1.5039 | 0.98 | 1000 | 1.4957 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.0.0+cu117
- Datasets 2.18.0
- Tokenizers 0.15.2