TaH2
Collection
Improving Test-Time Scaling with Adaptive Looped Transformers • 4 items • Updated
TaH2-1.7B-Standard
Standard single-pass Qwen3-1.7B model. Fine-tuned on the 1.7B subset of TaH2 AMTeam Tool.
Code: thu-nics/TaH · Paper: arXiv:2609.35748.
Load with Transformers:
from transformers import AutoModelForCausalLM, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("nics-efc/TaH2-1.7B-Standard")
model = AutoModelForCausalLM.from_pretrained(
"nics-efc/TaH2-1.7B-Standard", dtype="bfloat16", device_map="auto")
- Downloads last month
- -
Model tree for nics-efc/TaH2-1.7B-Standard
Base model
Qwen/Qwen3-1.7B-Base