Corpus Dataset or source name used to identify this text. Tokenizer Tokenizer used to match text against training transcripts. cl100k_baseo200k_baseo200k_harmonyp50k_basep50k_editr50k_basegpt2llama3deepseek_v3qwen2mistral_v3 Minimum window Smallest token window to index, from 1 to 65536. Maximum window Largest token window to index, at least the minimum and at most 65536.