ggerganov/llama.cpp

Pull Requests Commits

SYCL: set extras only on GGML_TYPE_Q4_0 (#12366)

qnixsynapse committed 1 year ago

Verified b3c9a656

llama : fix OLMo-2-0325-32B-Instruct K-norm size (#12400)

CISC committed 1 year ago

Verified 8ba95dca

context : fix init of n_outputs (#12397)

ggerganov committed 1 year ago

Verified dc079cfd

ci : add --symlinks to xcframework zip command (#12409)

danbev committed 1 year ago

Verified 7b61bcc8

llama-tts : add '-o' option (#12398)

marcoStocchi committed 1 year ago

Verified f4c3dd5d

SYCL: Delete redundant plus sign and space (#12391)

aubreyli committed 1 year ago

Verified 3d35d87b

SYCL : support non-contiguous tensors in binary ops (add, sub, etc) (#12399)

fairydreaming committed 1 year ago

Verified b19bd064

[CANN]MUL_MAT optimization (#12382)

noemotiovon committed 1 year ago

Verified 92a39132

ngxson committed 1 year ago

624a683c

rename to init_from_text

ngxson committed 1 year ago

116b9a16

Add CLI arg to llama-run to adjust the number of threads used (#12370)

ericcurtin committed 1 year ago

Verified 9f2250ba

llama_batch_ext_ptr::from_text/embd

ngxson committed 1 year ago

eaffba0f

main : add -sysf / --system-prompt-file (#12249) (#12250)

CISC committed 1 year ago

Verified 774973b8

Load all MoE experts during warmup (#11571)

fairydreaming committed 1 year ago

Verified 8fcb5636

ngxson committed 1 year ago

8e7714fa

qwen2vl: use llama_batch_ext_set_pos

ngxson committed 1 year ago

a363251f

server: fix "--grammar-file" parameter (#12285)

dodekapod committed 1 year ago

Verified add2a3aa

fix llama_batch_ext_init_from_embd

ngxson committed 1 year ago

ba793696

fix missing n_past in various places

ngxson committed 1 year ago

07d84fa3

ngxson committed 1 year ago

32940369

fix llama-run n_past

ngxson committed 1 year ago

5e6a6d4e

graph : simplify attn input build for unified KV cache (#12381)

ggerganov committed 1 year ago

Verified c522ce41

hparams : add SWA rope parameters (#12374)

ggerganov committed 1 year ago

Verified 081bee8c

bring back mistakenly deleted llama_batch_init/free

ngxson committed 1 year ago

bfdddbc1

correct comment

ngxson committed 1 year ago

54566ad9

rm redundant llama_batch_ext_set_output_last

ngxson committed 1 year ago

04f86418

fix llama_batch_ext_init_from_text

ngxson committed 1 year ago

c3dd7900

ngxson committed 1 year ago

65f01845

fix common_batch missing seq_id

ngxson committed 1 year ago

9fb2d81e

apply to the rest

ngxson committed 1 year ago

47086fa8

Newer Older