falcon_spider_v1
This model is a fine-tuned version of ybelkada/falcon-7b-sharded-bf16 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.5862 The repo contains the adapters for the model, as well as the tokenizer files.
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: 0.0002
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: constant
- lr_scheduler_warmup_ratio: 0.03
- training_steps: 100
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 1.1754 | 0.03 | 10 | 0.6316 |
| 0.5011 | 0.05 | 20 | 0.4848 |
| 0.4355 | 0.08 | 30 | 0.4586 |
| 0.3921 | 0.1 | 40 | 0.4555 |
| 0.3327 | 0.13 | 50 | 0.4674 |
| 0.2662 | 0.15 | 60 | 0.4507 |
| 0.2258 | 0.18 | 70 | 0.4813 |
| 0.1569 | 0.2 | 80 | 0.5407 |
| 0.1313 | 0.23 | 90 | 0.5924 |
| 0.1215 | 0.25 | 100 | 0.5862 |
Framework versions
- Transformers 4.30.2
- Pytorch 2.0.1+cu117
- Datasets 2.13.1
- Tokenizers 0.13.3