 Georgi Gerganov
|
75422e8bc4
graph : normalize Q, K, V shapes + sync cross attention (#12449)
|
преди 10 месеца |
|
Georgi Gerganov
|
8551c44d84
context : always use non-causal attention for encoder graphs (#12447)
|
преди 10 месеца |
|
Georgi Gerganov
|
dc079cfdff
context : fix init of n_outputs (#12397)
|
преди 11 месеца |
 fairydreaming
|
8fcb563613
Load all MoE experts during warmup (#11571)
|
преди 11 месеца |
|
Georgi Gerganov
|
081bee8c64
hparams : add SWA rope parameters (#12374)
|
преди 11 месеца |
|
Georgi Gerganov
|
84d5475541
llama : fix Gemma3 SWA KV cache shift (#12373)
|
преди 11 месеца |
|
Georgi Gerganov
|
e0dbec0bc6
llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181)
|
преди 11 месеца |
|
Georgi Gerganov
|
afa8a9ec9b
llama : add `llama_vocab`, functions -> methods, naming (#11110)
|
преди 1 година |
|
Georgi Gerganov
|
f66f582927
llama : refactor `src/llama.cpp` (#10902)
|
преди 1 година |
