ZygAI/ZygAI_OSS_Flash

ZygAI OSS Flash

ZygAI OSS Flash is a personal legacy release by the creator of ZygAI, published at age 23 as an open-source milestone.

This model represents a point in time: a commitment to build in public, share the work openly, and leave something useful behind for the community.
It is called “Flash” because it is meant to be practical, lightweight in spirit, and easy to run across different setups.

Legacy Note

This release is more than a checkpoint.
It is a legacy snapshot of the ZygAI journey so far:

• built with limited resources but long-term vision
• released openly so others can learn, adapt, and continue
• dedicated to the idea that meaningful AI work can come from anywhere

If you use this model, improve it, or fork it, you are part of that legacy.

Repository purpose

This repository contains the Hugging Face-format source model assets for ZygAI_OSS_Flash, including merged weights and training outputs.

Main practical folder:

• merged_fp16/ - primary source for inference and GGUF conversion

Personal Thanks

A huge, heartfelt thank you to Ruby2001, 0daysophie, italian_tech_person, and Julia's Tech Spot.

Your help was not just technical, it was life-changing.
Without you, my life would have been dark.
Thank you for standing with me and helping make ZygAI OSS Flash real.

Quick Transformers example

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_id = "ZygAI/ZygAI_OSS_Flash"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.float16,
    device_map="auto",
)

prompt = "Labas! Parašyk trumpą motyvacinę žinutę."
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=120)
print(tokenizer.decode(out[0], skip_special_tokens=True))