llama.cpp

musa: enable building fat binaries, enable unified memory, and disable Flash Attention on QY1 (MTT S80) (#9526)

#221:Commit c35e586ea5 pushed by root

b3804

2024-09-22 21:16:21 +00:00

0s

Fix merge error in #9454 (#9589)

#220:Commit 912c331d3d pushed by root

b3803

2024-09-22 21:16:21 +00:00

0s

musa: enable building fat binaries, enable unified memory, and disable Flash Attention on QY1 (MTT S80) (#9526)

#219:Commit c35e586ea5 pushed by root

master

2024-09-22 21:16:20 +00:00

0s

perplexity : remove extra new lines after chunks

#216:Commit 3b6470f228 pushed by root

gg/perplexity-nl

2024-09-22 21:16:20 +00:00

0s

metal : use F32 prec for K*Q in vec FA

#215:Commit 5d888c48a3 pushed by root

gg/metal-fa-f32-qk

2024-09-22 21:16:20 +00:00

0s

CUDA: enable Gemma FA for HIP/Pascal (#9581)

#214:Commit a5b57b08ce pushed by root

b3802

2024-09-22 13:06:20 +00:00

0s

CUDA: enable Gemma FA for HIP/Pascal (#9581)

#213:Commit a5b57b08ce pushed by root

master

2024-09-22 21:16:20 +00:00

0s

llama: remove redundant loop when constructing ubatch (#9574)

#209:Commit ecd5d6b65b pushed by root

b3801

2024-09-22 04:56:21 +00:00

0s

RWKV v6: RWKV_WKV op CUDA implementation (#9454)

#208:Commit 2a63caaa69 pushed by root

b3800

2024-09-22 04:56:21 +00:00

0s

flake.lock: Update

#206:Commit db660f5a40 pushed by root

update_flake_lock_action

2024-09-22 04:56:20 +00:00

0s

llama: remove redundant loop when constructing ubatch (#9574)

#204:Commit ecd5d6b65b pushed by root

master

2024-09-22 13:06:19 +00:00

0s

ggml-alloc : fix list of allocated tensors with GGML_ALLOCATOR_DEBUG (#9573)

#199:Commit d09770cae7 pushed by root

b3799

2024-09-22 21:36:17 +00:00

0s

ggml-alloc : fix list of allocated tensors with GGML_ALLOCATOR_DEBUG (#9573)

#198:Commit d09770cae7 pushed by root

master

2024-09-22 04:56:20 +00:00

0s

Update CUDA graph on scale change plus clear nodes/params (#9550)

#194:Commit 41f477879f pushed by root

b3798

2024-09-22 09:36:17 +00:00

0s

CI: Provide prebuilt windows binary for hip (#9467)

#193:Commit e948a7da7a pushed by root

b3797

2024-09-22 09:36:17 +00:00

0s

ggml-alloc : fix list of allocated tensors with GGML_ALLOCATOR_DEBUG

#192:Commit d9ce02ae82 pushed by root

sl/fix-debug-alloc

2024-09-22 09:36:17 +00:00

0s

Update CUDA graph on scale change plus clear nodes/params (#9550)

#191:Commit 41f477879f pushed by root

master

2024-09-21 12:36:19 +00:00

0s

ggml : fix builds (#0)

#187:Commit d13edb17ed pushed by root

b3795

2024-09-21 21:36:17 +00:00

0s

CUDA: fix sum.cu compilation for CUDA < 11.7 (#9562)

#186:Commit 5cb12f6839 pushed by root

b3790

2024-09-21 21:36:17 +00:00

0s

quantize : improve type name parsing (#9570)

#185:Commit 63351143b2 pushed by root

master

2024-09-21 04:26:20 +00:00

0s

examples : flush log upon ctrl+c (#9559)

#181:Commit d39e26741f pushed by root

b3789

2024-09-21 15:36:17 +00:00

0s

perplexity : do not escape input data by default (#9548)

#180:Commit 722ec1eb51 pushed by root

b3788

2024-09-21 15:36:17 +00:00

0s

examples : flush log upon ctrl+c (#9559)

#179:Commit d39e26741f pushed by root

master

2024-09-20 20:16:21 +00:00

0s

llama : make llm_tokenizer more private

#176:Commit 6e873e561a pushed by root

gg/tokenizer-cleanup

2024-09-21 15:36:17 +00:00

0s

server : add rerank endpoint

#174:Commit 5f95dccea8 pushed by root

gg/rerank

2024-09-20 21:36:17 +00:00

0s

server : clean-up completed tasks from waiting list (#9531)

#172:Commit 6026da52d6 pushed by root

b3787

2024-09-20 15:36:17 +00:00

0s

imatrix : disable prompt escape by default (#9543)

#171:Commit eca0fab44e pushed by root

b3786

2024-09-20 15:36:17 +00:00

0s

server : clean-up completed tasks from waiting list (#9531)

#170:Commit 6026da52d6 pushed by root

master

2024-09-20 12:06:20 +00:00

0s

llama-bench : add time-to-first-byte stat

#167:Commit ff231de553 pushed by root

gg/ttfb

2024-09-20 15:36:17 +00:00

0s

llama : add "rank" pooling type

#166:Commit f03bcd84e7 pushed by root

gg/rerank

2024-09-19 19:46:21 +00:00

0s