Views
No views yet
1from transformers import AutoModelForCausalLM, AutoTokenizer
2
3repo_id = "atrost/climbmix-baseline-large-353m-1p2b-h100"
4tokenizer = AutoTokenizer.from_pretrained(repo_id)
5model = AutoModelForCausalLM.from_pretrained(
6 repo_id,
7 trust_remote_code=True,
8 torch_dtype="auto",
9)trust_remote_code=True is required for the custom StairFormer/asymmetric checkpoints and harmless for the dense Llama baseline.atrost/climbmix-llama-288m-2p8b-h100 at commit 1cb4607.