llama.cpp
llama : refactor llama_context, llama_kv_cache, llm_build_context (v2)
#12181
Merged
Merged
llama : refactor llama_context, llama_kv_cache, llm_build_context (v2) #12181
ggerganov
merged
16 commits into
master
from
gg/llama-kv-cache-v2
slaren
approved these changes
on 2025-03-11
llama : refactor llama_context, llama_kv_cache, llm_build_context
55909257
graph : don't mutate the KV cache during defrag
75624a20
context : reduce virtuals + remove test function
5aa3518d
context : move interface implementation to source file + factory
0a6648ca
graph : move KV cache build functions to llama_context impl
cc9fa25a
graph : remove model reference from build_pooling
29c9ef56
graph : remove llama_model reference
bc825604
kv_cache : provide rope factors
ff95ffdf
graph : rework inputs to use only unique_ptr, remove attn input abstr…
562a4787
context : remove llama_context_i abstraction
d0cb3196
context : clean-up
a4fc4e8e
graph : clean-up
af9f6b8e
llama : remove redundant keywords (struct, enum)
226ff010
model : adapt gemma3
5fc6dbd9
graph : restore same attention ops as on master
70ef6530
llama : remove TODO + fix indent
31b8eab5
Login to write a write a comment.
Login via GitHub