| --- |
| license: mit |
| task_categories: |
| - text-generation |
| language: |
| - en |
| tags: |
| - model_configs |
| - json |
| - hfmc |
| - harley-ml |
| - configs |
| - model_json_configs |
| size_categories: |
| - 10K<n<100K |
| --- |
| |
| # HFMC |
|
|
| HFMC stands for "Hugging Face Model Configs." This dataset has over 7k json model configs from Hugging Face. We used the [Hugging Face API](https://huggingface.co/docs/hub/api) to scrape each one. |
|
|
| ## Stats |
|
|
| | Metric | Value | |
| |--------------------|-----------| |
| | Tokens (GPT2) | 7,117,346 | |
| | Entries | 18,781 | |
| | Tokens/Entry (med) | 318 | |
| | Words | 1,420,637 | |
| | Words/Entry (med) | 62 | |
|
|
| We counted the tokens using [GPT2](https://huggingface.co/openai-community/gpt2)'s tokenizer. |
| ## Cleaning |
|
|
| We deduped, filtered via lang (only English), and length (1024 tokens using an in-domain tokenizer). |
|
|
| ## Use Cases |
|
|
| 1. Pretraining or fine-tuning small models. |
| 2. Including HFMC in a much larger dataset for coverage |
