Modelpublic

nabin2004/AOS-qwen25-coder-7b-manim-merged

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face

0likes24downloads

Model Card

AOS Qwen2.5-Coder-7B Manim SFT (Merged)

Full merged bf16 weights (LoRA baked into base) for Manim animation generation via multi-turn tool calling.

Model URL: https://huggingface.co/nabin2004/AOS-qwen25-coder-7b-manim-merged

Related repos

Artifact	Repo
LoRA adapter	nabin2004/AOS-qwen25-coder-7b-manim-sft
GGUF (Ollama / llama.cpp)	nabin2004/AOS-qwen25-coder-7b-manim-gguf

Base model

Merged from Qwen/Qwen2.5-Coder-7B-Instruct and the AOS Manim SFT LoRA adapter.

Training data

Fine-tuned on nabin2004/AOS-Trajectories using the AOS Phase 1 SFT trainer (`apps/sft`).

Usage

Load with Transformers

python

import os
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

token = os.environ.get("HF_TOKEN")
model_id = "nabin2004/AOS-qwen25-coder-7b-manim-merged"

tokenizer = AutoTokenizer.from_pretrained(model_id, token=token)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    token=token,
)