NAME0x0/AVA-v2-merged

AVA v2 (merged weights)

Standalone bf16 weights of AVA v2 — the QLoRA adapter pre-merged into Qwen/Qwen3.5-2B. Load directly with transformers, no PEFT required:

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "NAME0x0/AVA-v2-merged", device_map="auto", dtype="bfloat16"
)
tokenizer = AutoTokenizer.from_pretrained("NAME0x0/AVA-v2-merged")

• Benchmarks, training details, limitations: see the adapter card — 82.0% ARC-Challenge, 92.0% ARC-Easy, 59.2% MMLU at 2B params, trained on a single 4 GB laptop GPU.
• No Python / CPU-only: use the GGUF builds (Ollama, llama.cpp, LM Studio).
• Reproduce everything: github.com/NAME0x0/AVA.