Views
No views yet
1from transformers import AutoModelForCausalLM, AutoTokenizer
2
3REPO = "omar81939/Ouro-1.4B-Thinking-depth-SFT"
4DEPTH = 16
5
6tokenizer = AutoTokenizer.from_pretrained(REPO)
7model = AutoModelForCausalLM.from_pretrained(
8 REPO,
9 trust_remote_code=True,
10 dtype="bfloat16",
11 total_ut_steps=DEPTH,
12)hf_overrides={"total_ut_steps": DEPTH} when creating the engine.