peft-lora-starcoder15B-v2-personal-copilot-A100-40GB-colab
This model is a fine-tuned version of bigcode/starcoderbase-1b on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 7.5615
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: 0.0005
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- distributed_type: tpu
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 30
- training_steps: 2000
Training results
Training Loss | Epoch | Step | Validation Loss |
---|---|---|---|
6.2182 | 0.05 | 100 | 8.4091 |
4.5691 | 0.1 | 200 | 9.6057 |
6.4441 | 0.15 | 300 | 7.8901 |
3.1214 | 0.2 | 400 | 9.3116 |
6.8473 | 0.25 | 500 | 7.8618 |
6.7939 | 0.3 | 600 | 7.5668 |
5.145 | 0.35 | 700 | 7.8765 |
4.678 | 0.4 | 800 | 7.4816 |
3.0478 | 0.45 | 900 | 8.1724 |
5.1993 | 0.5 | 1000 | 7.5839 |
6.2314 | 0.55 | 1100 | 7.6065 |
4.3643 | 0.6 | 1200 | 9.1999 |
2.8909 | 0.65 | 1300 | 7.7913 |
6.477 | 0.7 | 1400 | 7.4454 |
6.2981 | 0.75 | 1500 | 8.6974 |
5.7769 | 0.8 | 1600 | 8.3724 |
5.6318 | 0.85 | 1700 | 7.5465 |
4.9988 | 0.9 | 1800 | 7.5232 |
4.7996 | 0.95 | 1900 | 7.5701 |
3.7886 | 1.0 | 2000 | 7.5615 |
Framework versions
- Transformers 4.35.0
- Pytorch 2.0.0+cu118
- Datasets 2.11.0
- Tokenizers 0.14.1
Model tree for erishabh/peft-lora-starcoder15B-v2-personal-copilot-A100-40GB-colab
Base model
bigcode/starcoderbase-1b