 Georgi Gerganov
|
3e1d29348b
kv-cache : simplify + fix warning for recurrent models (#12756)
|
9 months ago |
 Diego Devesa
|
e0e912f49b
llama : add option to override model tensor buffers (#11397)
|
9 months ago |
|
Georgi Gerganov
|
a10b36c91a
llama : refactor kv cache guard (#12695)
|
9 months ago |
 Xuan-Son Nguyen
|
af6ae1efb2
llama : fix non-causal mask for gemma 3 (#12615)
|
9 months ago |
|
Georgi Gerganov
|
b4ae50810e
metal : improve FA + improve MoE (#12612)
|
9 months ago |
|
Georgi Gerganov
|
2d77d88e70
context : fix worst-case reserve outputs (#12545)
|
10 months ago |
 fairydreaming
|
568013d0cd
context : clear sets containing encoder output sequence ids before storing new values (#12470)
|
10 months ago |
|
Georgi Gerganov
|
75422e8bc4
graph : normalize Q, K, V shapes + sync cross attention (#12449)
|
10 months ago |
|
Georgi Gerganov
|
8551c44d84
context : always use non-causal attention for encoder graphs (#12447)
|
10 months ago |
|
Georgi Gerganov
|
dc079cfdff
context : fix init of n_outputs (#12397)
|
10 months ago |
|
fairydreaming
|
8fcb563613
Load all MoE experts during warmup (#11571)
|
10 months ago |
|
Georgi Gerganov
|
081bee8c64
hparams : add SWA rope parameters (#12374)
|
10 months ago |
|
Georgi Gerganov
|
84d5475541
llama : fix Gemma3 SWA KV cache shift (#12373)
|
10 months ago |
|
Georgi Gerganov
|
e0dbec0bc6
llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181)
|
10 months ago |
|
Georgi Gerganov
|
afa8a9ec9b
llama : add `llama_vocab`, functions -> methods, naming (#11110)
|
1 year ago |
|
Georgi Gerganov
|
f66f582927
llama : refactor `src/llama.cpp` (#10902)
|
1 year ago |
