-
Baber Abbasi authored
* max_length - 1 (generation always >= 1) * vllm: fix rolling prefix_token * nit: add comment * fixup! max_length should be handled for logliklihoods
b31f92e8
* max_length - 1 (generation always >= 1) * vllm: fix rolling prefix_token * nit: add comment * fixup! max_length should be handled for logliklihoods