metadata
license: apache-2.0
base_model: pszemraj/mega-ar-small-4096-NC-simplewiki-v1
tags:
- generated_from_trainer
metrics:
- accuracy
inference:
parameters:
max_new_tokens: 96
do_sample: true
repetition_penalty: 1.1
no_repeat_ngram_size: 5
guidance_scale: 1.02
eta_cutoff: 0.001
datasets:
- JeanKaddour/minipile
mega-ar-small-4096-NC-minipile-v1
This model is a fine-tuned version of pszemraj/mega-ar-small-4096-NC-simplewiki-v1 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 3.7502
- Accuracy: 0.3650
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: 0.0003
- train_batch_size: 1
- eval_batch_size: 1
- seed: 80085
- gradient_accumulation_steps: 64
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-07
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.05
- num_epochs: 1.0
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 5.7062 | 0.04 | 200 | 5.5259 | 0.1829 |
| 5.4056 | 0.07 | 400 | 5.1650 | 0.2073 |
| 5.1319 | 0.11 | 600 | 4.9704 | 0.2261 |
| 4.9674 | 0.14 | 800 | 4.8417 | 0.2370 |
| 4.951 | 0.18 | 1000 | 4.7261 | 0.2467 |
| 4.7753 | 0.21 | 1200 | 4.6288 | 0.2554 |
| 4.6721 | 0.25 | 1400 | 4.5260 | 0.2678 |
| 4.6081 | 0.28 | 1600 | 4.4409 | 0.2768 |
| 4.4929 | 0.32 | 1800 | 4.3566 | 0.2857 |
| 4.4345 | 0.35 | 2000 | 4.2812 | 0.2957 |
| 4.3024 | 0.39 | 2200 | 4.2085 | 0.3052 |
| 4.2505 | 0.42 | 2400 | 4.1424 | 0.3151 |
| 4.2294 | 0.46 | 2600 | 4.0859 | 0.3217 |
| 4.2766 | 0.49 | 2800 | 4.0372 | 0.3298 |
| 4.1229 | 0.53 | 3000 | 3.9901 | 0.3357 |
| 4.2007 | 0.56 | 3200 | 3.9538 | 0.3410 |
| 3.9723 | 0.6 | 3400 | 3.9186 | 0.3453 |
| 4.0599 | 0.63 | 3600 | 3.8881 | 0.3487 |
| 4.0351 | 0.67 | 3800 | 3.8650 | 0.3515 |
| 3.9324 | 0.7 | 4000 | 3.8419 | 0.3546 |
| 3.9408 | 0.74 | 4200 | 3.8234 | 0.3564 |
| 3.9499 | 0.77 | 4400 | 3.8091 | 0.3585 |
| 3.9456 | 0.81 | 4600 | 3.7938 | 0.3602 |
| 3.9035 | 0.84 | 4800 | 3.7809 | 0.3618 |
| 3.8709 | 0.88 | 5000 | 3.7712 | 0.3631 |
| 3.8189 | 0.92 | 5200 | 3.7612 | 0.3639 |
| 3.7973 | 0.95 | 5400 | 3.7549 | 0.3646 |
| 3.8952 | 0.99 | 5600 | 3.7502 | 0.3650 |
Framework versions
- Transformers 4.33.1
- Pytorch 2.2.0.dev20230907+cu118
- Datasets 2.14.5
- Tokenizers 0.13.3