AOS Qwen3 8B Narrated (Continued SFT Merged)
This is the full bf16 Safetensors release of the Supervised Fine-Tuned (SFT) model for mathematical and scientific animation in Manim Community Edition (CE) synchronized with manim-voiceover.
- Base Foundation Model:
Qwen/Qwen3-8B - Trained Adapter:
nabin2004/AOS-qwen3-8b-narrated-adapter - Merged Model:
nabin2004/AOS-qwen3-8b-narrated-sft-merged - Quantized GGUF Models:
nabin2004/AOS-qwen3-8b-narrated-sft-gguf
Model Capabilities
- VoiceoverScene Architecture: Generates complete, executable Manim scripts inheriting from
VoiceoverScene. - Audio Bookmarking: Integrates synchronized audio markers (
tracker.wait_until_bookmark(...)) paired with visual animations. - CE API Compliance: Strict Manim Community Edition syntax, eliminating deprecated legacy APIs.
Quickstart (vLLM / Transformers)
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
model_id = "nabin2004/AOS-qwen3-8b-narrated-sft-merged"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16, device_map="auto")
prompt = "Write a complete Manim VoiceoverScene explaining Bayes Theorem with visual probability trees."
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=2048, temperature=0.2)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))