license: apache-2.0
tags:
- generated_from_trainer
model-index:
- name: Hieroglyph-Translator-Using-Gardiner-Codes
results: []
Hieroglyph Gardiner Code Translator Model
This model was created to translate hieroglyphs into english.
Egyptian Hieroglyphs have been grouped into different classes and given a referencing method called Gardiner Codes using Gardiner Classification.
Using the Gardiner Codes we can assign meanings to different combinations of hieroglyphs.
To Translate any sequence of hieroglyphs using this model, provide the following input :-
"Translate hieroglyph unicode sequence to English: {Gardiner Codes of the Hieroglyphs}"
Examples :
"Translate hieroglyph unicode sequence to English: A1 B6 F8"
"Translate hieroglyph unicode sequence to English: G4 H9 P3"
Model description
This model is a fine-tuned version of t5-small on a custom dataset derived from the Dictionary of Middle Egyptian.
The Inference Api on the hugging face model page doesn't work well, load the model in jupyter notebook using the following code snippet:
text = "" # add your hieroglyph gardiner code combination in the string
from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("AnushS/hieroglyph_unicode_translator_t5_small")
inputs = tokenizer(text, return_tensors="pt").input_ids
from transformers import AutoModelForSeq2SeqLM
model = AutoModelForSeq2SeqLM.from_pretrained("AnushS/hieroglyph_unicode_translator_t5_small")
outputs = model.generate(inputs, max_new_tokens=40, do_sample=True, top_k=30, top_p=0.95)
translated_keywords = str(tokenizer.decode(outputs[0], skip_special_tokens=True))
Intended uses & limitations
The Model is intended to be used to translate hieroglyphs. The model does not provide full sentences, it only outputs bits and keywords.
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 2
Training results
| Training Loss | Epoch | Step | Validation Loss | Gen Len |
|---|---|---|---|---|
| 5.0665 | 1.0 | 688 | 4.2034 | 6.946 |
| 4.4621 | 2.0 | 1376 | 4.1388 | 6.946 |
Framework versions
- Transformers 4.27.4
- Pytorch 2.2.0.dev20231113
- Datasets 2.12.0
- Tokenizers 0.13.3