Model Summary
Maika Buddy Model, fine-tuned from Microsoft Phi3 with our specialized children's conversational dataset, drives BuddyOS to enrich children's communication skills through interactive dialogue. Designed for AI-enhanced toys, Maika encourages children to express themselves, ask questions, and practice active listening in a non-judgmental environment. It not only fosters confidence in communication but also promotes a balanced approach by encouraging activities like outdoor play, hands-on creativity, and sharing experiences with family and friends, ensuring comprehensive developmental support for young users.
Project Website: BuddyOS
Requirements
The code of Maika Buddy has been in the latest Hugging face transformers and we advise you to install transformers>=4.37.0.
Quickstart
Here provides a code snippet to show you how to load the tokenizer and model and how to generate contents.
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
torch.set_default_device("cuda")
model_path = "olliai/Maika-Buddy-3B8-Phi3"
model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype="auto", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
prompt = """A chat between a curious kid and an artificial intelligence assistant. The assistant gives helpful, easy-to-understand, detailed, and polite answers to the questions of the kid.
Kid: I have 10 apples. I gave Bob 2 pencil. How many apples do I have now?
AI: """
inputs = tokenizer(prompt, return_tensors="pt", return_attention_mask=False)
outputs = model.generate(**inputs, max_length=200,
do_sample=True,
repetition_penalty=1.1,
temperature=0.3,
top_k=50,
top_p=0.95,
eos_token_id=tokenizer.encode("<|endoftext|>"))
text = tokenizer.batch_decode(outputs)[0]
print(text)
License
Maika Buddy is distributed under the terms of the MIT. No restrict on the research, but should comply with the Phi3 License.
Contact Us
If you have any questions, please raise an issue or contact us at viet.hoang@olli-ai.com or hai@olli-ai.com.