CB: support different number of K and V heads per layer (#1610) #20

Job	Run time
cpp-beam_search_causal_lm-notus-7b-v1	0s
cpp-beam_search_causal_lm-Phi-2	0s
cpp-beam_search_causal_lm-Qwen1_5-7B-Chat	0s
cpp-beam_search_causal_lm-ubuntu (./build/samples/cpp/text_generation/beam_search_causal_lm)	18m 35s
cpp-greedy_causal_lm-Qwen-7B-Chat	0s
cpp-beam_search_causal_lm-ubuntu (python ./samples/python/text_generation/beam_search_causal_lm.py)	19m 59s
cpp-greedy_causal_lm-windows	35m 12s
cpp-multinomial-greedy_causal_lm-ubuntu	0s
cpp-speculative_decoding_lm-ubuntu	0s
cpp-prompt_lookup_decoding_lm-ubuntu	0s
cpp-chat_sample-ubuntu	13m 38s
cpp-continuous-batching-macos	20m 39s
cpp-continuous-batching-ubuntu	0s
cpp-continuous-batching-windows	26m 53s
cpp-greedy_causal_lm-redpajama-3b-chat	0s
visual_language_chat_sample-ubuntu-internvl2	0s
cpp-Phi-1_5	0s
visual_language_chat_sample-ubuntu-minicpm_v2_6	0s
visual_language_chat_sample-ubuntu-qwen2vl	0s
visual_language_chat_sample-ubuntu-llava_1_5 / visual_language_chat_sample-ubuntu-llava	0s
visual_language_chat_sample-ubuntu-llava_next / visual_language_chat_sample-ubuntu-llava	0s
	2h 14m 56s

Provide feedback