Views
No views yet
<strategy>-<sequence-length>, e.g. baseline-causal-1024
(original run) and baseline-causal-2048 (continued pre-training at 2048).1repo_id = "mmcarpi/flexqwen-stil"
2config_name = "baseline-causal-2048/best" # strategy-seqlen/checkpoint, e.g. "hybrid-shift-1024/step_000500"
3
4# Load tokenizer from the root
5tokenizer = AutoTokenizer.from_pretrained(repo_id, trust_remote_code=True)
6
7# Load model from the specific configuration subfolder
8model = AutoModelForCausalLM.from_pretrained(
9 repo_id,
10 subfolder=config_name,
11 trust_remote_code=True,
12)