### Mistral tokenizer

|       |    num_examples |   num_words |   dataset_gb |   num_tokens |   avg_word |   avg_tokens |
|:------|----------------:|------------:|-------------:|-------------:|-----------:|-------------:|
| train |     5.89863e+07 | 4.01475e+10 |      242.115 |  6.51262e+10 |    680.623 |      1104.09 |
| test  | 10000           | 6.7228e+06  |        0.04  |  1.09506e+07 |    672.28  |      1095.06 |

### Custom tokenizer

|       |    num_examples |   num_words |   dataset_gb |   num_tokens |   avg_word |   avg_tokens |   ratio |
|:------|----------------:|------------:|-------------:|-------------:|-----------:|-------------:|--------:|
| train |     5.89863e+07 | 4.01475e+10 |      242.115 |  6.38042e+10 |    680.623 |      1081.68 | 1.99966 |
| test  | 10000           | 6.7228e+06  |        0.04  |  1.07121e+07 |    672.28  |      1071.21 | 0.00034 |