# a fundamental problem, our current implementation does not support # this case. common_prefix_len = min( common_prefix_len, self.input_batch.num_computed_tokens_cpu[:num_reqs].min()) # common_prefix_len should be a multiple of the block size. common_prefix_len = (common_prefix_len ...
# a fundamental problem, our current implementation does not support # this case. common_prefix_len = min( common_prefix_len, self.input_batch.num_computed_tokens_cpu[:num_reqs].min()) # common_prefix_len should be a multiple of the block size. common_prefix_len = (common_prefix_...