 Sigbjørn Skjæret
|
9008027aa3
hparams : add n_embd_inp() to support extended embed (#16928)
|
2 месяцев назад |
 Georgi Gerganov
|
16bcc1259d
kv-cache : pad the cache size to 256 for performance (#17046)
|
2 месяцев назад |
 Johannes Gäßler
|
aa374175c3
CUDA: fix crash on uneven context without FA (#16988)
|
2 месяцев назад |
|
Georgi Gerganov
|
cd5e3b5754
server : support unified cache across slots (#16736)
|
2 месяцев назад |
 Diego Devesa
|
5a4ff43e7d
llama : disable pipeline parallelism if compute buffer allocation fails (#16748)
|
2 месяцев назад |
 takuya kodama
|
7062dd8460
llama-context: only warn on pooling_type when user specified (#16674)
|
3 месяцев назад |
 Saba Fallah
|
e08db42595
model: EmbeddingGemma Adding Support for SentenceTransformers Dense Modules (#16367)
|
3 месяцев назад |
|
Johannes Gäßler
|
e789095502
llama: print memory breakdown on exit (#15860)
|
3 месяцев назад |
|
Sigbjørn Skjæret
|
b8e09f08b9
model : add grok-2 support (#15539)
|
4 месяцев назад |
 Haiyue Wang
|
f4e664f838
context : remove redundant explicit casting to the same type (#15948)
|
4 месяцев назад |
 Daniel Bevenius
|
86587da03b
llama : check returned fn ptrs from ggml_backend_reg_get_proc_address (#15893)
|
4 месяцев назад |
|
Georgi Gerganov
|
663027fd54
context : fix n_outputs during reserve (#15858)
|
4 месяцев назад |
|
Daniel Bevenius
|
d1e2adba65
llama : set n_outputs to 1 to avoid 0 outputs mean-pooling (#15791)
|
4 месяцев назад |
|
Diego Devesa
|
274966226f
llama : fix fattn reserve call n_seqs parameter (#15699)
|
4 месяцев назад |
|
Diego Devesa
|
9777032dcc
llama : separate compute buffer reserve from fattn check (#15696)
|
4 месяцев назад |
|
Johannes Gäßler
|
e81b8e4b7f
llama: use FA + max. GPU layers by default (#15434)
|
4 месяцев назад |
|
Georgi Gerganov
|
8a4280ce43
kv-cache : remove LLAMA_SET_ROWS checks (#15505)
|
4 месяцев назад |
|
Georgi Gerganov
|
85cc1ae998
context : print graph stats for memory-less contexts (#15586)
|
4 месяцев назад |
|
Georgi Gerganov
|
9ebebef62f
llama : remove KV cache defragmentation logic (#15473)
|
4 месяцев назад |
|
Georgi Gerganov
|
cd36b5e5c7
llama : remove deprecated llama_kv_self API (#15472)
|
5 месяцев назад |
|
Georgi Gerganov
|
715a6db02c
kv-cache : drop the "unified" prefix (#15467)
|
5 месяцев назад |
|
Georgi Gerganov
|
9d262f4bad
server : remove swa_full warning (#15399)
|
5 месяцев назад |
|
Georgi Gerganov
|
d32e03f449
server : add SWA checkpoints (#15293)
|
5 месяцев назад |
 Jonathan Graehl
|
5cdb27e091
finetune: SGD optimizer, more CLI args (#13873)
|
5 месяцев назад |
 compilade
|
ee3a9fcf88
context : fix index overflow on huge outputs (#15080)
|
5 месяцев назад |
|
Georgi Gerganov
|
a4569c41fd
llama : enable LLAMA_SET_ROWS=1 by default (#14959)
|
5 месяцев назад |
|
Georgi Gerganov
|
ba42794c9e
graph : fix equal_seq() check (#14986)
|
5 месяцев назад |
|
Daniel Bevenius
|
ca0ef2dddb
llama : clarify comment about pp and tg graphs [no ci] (#14895)
|
5 месяцев назад |
|
Georgi Gerganov
|
c1dbea752a
context : restore preemptive sched reset when LLAMA_SET_ROWS=0 (#14870)
|
5 месяцев назад |
|
Georgi Gerganov
|
e4868d16d2
context : perform output reorder lazily upon access after sync (#14853)
|
5 месяцев назад |
