qwenjev / scripts /download_weights.py
tchbcb's picture
QwenJev: multimodal-retrofitted NanoJev for ARC-AGI-3 (initial skeleton)
3e04895 verified
Raw History Blame Contribute Delete
1.28 kB
"""Download the required weights.
* Qwen/Qwen3-0.6B - System-1 backbone (NanoJev-style student)
* Qwen/Qwen3.5-9B - System-2 multimodal controller / teacher
* C-Tianyu/NanoJev - optional reference checkpoint (unified-games-v1)
Tokens: gated repos need HF_TOKEN; the NanoJev and Qwen repos are public.
"""
from __future__ import annotations
import argparse
def main() -> None:
ap = argparse.ArgumentParser()
ap.add_argument("--skip-nanojev", action="store_true")
args = ap.parse_args()
from huggingface_hub import snapshot_download
print(">> Qwen3-0.6B (System-1 backbone)")
snapshot_download("Qwen/Qwen3-0.6B", local_dir="checkpoints/Qwen3-0.6B")
if not args.skip_nanojev:
print(">> NanoJev reference checkpoint (optional, for head surgery)")
snapshot_download(
"C-Tianyu/NanoJev",
revision="unified-games-v1",
local_dir="checkpoints/NanoJev-ref",
allow_patterns=["best.safetensors", "config.json", "tokenizer/*"],
)
print(">> Qwen3.5-9B (System-2 / teacher) - large download")
snapshot_download("Qwen/Qwen3.5-9B", local_dir="checkpoints/Qwen3.5-9B")
print("done -> checkpoints/")
if __name__ == "__main__":
main()