Somali Agriculture GPT-2 Model
This is a GPT-2 language model trained on ~10,000 Somali prompt-response pairs related to agriculture. The goal is to help generate Somali text for agricultural questions and educational content.
Model Details
- Architecture: GPT-2 (4 layers, 4 attention heads)
- Vocabulary: Custom trained Byte-Pair Encoding tokenizer
- Training Data: 10,000 prompt-response pairs in Somali language
- Epochs: 5
- Embedding Size: 256
- Context Length: 512 tokens
Intended Use
This model can be used to:
- Generate Somali answers to agricultural questions
- Create educational materials for Somali farmers
- Build Somali chatbots focused on agriculture
Example Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("tacab/somali-agriculture")
tokenizer = AutoTokenizer.from_pretrained("tacab/somali-agriculture")
prompt = "Maxay tahay faa'iidada bacriminta?"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_length=100)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))