|
--- |
|
library_name: peft |
|
base_model: google/flan-t5-large |
|
--- |
|
|
|
# Model Card for honicky/t5-short-story-character-extractor |
|
|
|
I trained this model as part of a learning project to build a children's story authoring tool for parents of young children. See http://www.storytime.glass/ |
|
This model takes in a short story and outputs a comma separated list of characters in the story. |
|
|
|
I'm not sure yet how useful this fine-tune is: rather it is for me to learn about the nuts and bolts of fine-tuning. |
|
|
|
|
|
## Model Details |
|
|
|
The model is a fine-tune of a sequence-to-sequence model [Flan T5 Large](https://huggingface.co./google/flan-t5-large), so a different architecture from decoder-only models like GPT*. Maybe this allows it to perform this transformation task (transform a story into a list of characters) using a smaller model? |
|
|
|
* Trained using `transformers.Seq2SeqTrainer` plus the corresponding collator, tokenizer etc |
|
|
|
- **Developed by:** RJ Honicky |
|
- **Model type:** Encoder-Decoder Transformer |
|
- **Language(s):** English (fine tune data set) |
|
- **License:** MIT |
|
- **Finetuned from model:** `google/flan-t5-large` |
|
|
|
### Model Sources |
|
|
|
- **Repository:** https://github.com/honicky/character-extraction |
|
- **Weights and Biases**: https://wandb.ai/honicky/t5_target_finetune_for_character_extraction/runs/mx57gh45 |
|
|
|
## Uses |
|
|
|
Primarily for use in https://github.com/honicky/story-time and for learning. |
|
|
