ggerganov/llama.cpp

Pull Requests Commits
ggml : allocate graphs in a context (#2392)
slaren committed 2 years ago
Verified 5488fb78
Add LLAMA_DEFAULT_RMS_EPS so we can change the default (#2384)
ikawrakow committed 2 years ago
Verified eb542d39
ggml : fix ggml_flash_attn to use op_params (#2387)
slaren committed 2 years ago
Verified 07aaa0f6
convert.py : support bpe tokenizer (#2228)
ftgreat committed 2 years ago
Verified fce48caf
ggml : relax contiguous constraints in activation function (#2371)
li-plus committed 2 years ago
Verified 875086bd
ggml : improve graph build time via hash table lookup (#2329)
slaren committed 2 years ago
Verified da188983
build : fix line breaking error in build-info.sh (#2349)
hesenp committed 2 years ago
Verified 82552b7f
main : add `--in-prefix-bos` to prefix BOS to user inputs; keep EOS (#2304)
jxy committed 2 years ago
Verified 0c06204f
ci : add non-AVX scalar build/test (#2356)
netrunnereve committed 2 years ago
Verified 1fed755b
k_quants : add AVX support to dot functions with QK_K as 64 (#2339)
katsu560 committed 2 years ago
Verified be2301bc
metal : concurrently dispatch commands (#2358)
lshzh-ww committed 2 years ago
Verified 1aa18ef9
Another speed gain for Q4_0 and Q4_1 on Metal (#2375)
ikawrakow committed 2 years ago
Verified 9a08eaf3
Fix Q4_K and Q5_K for QK_K = 64 on CUDA (#2359)
ikawrakow committed 2 years ago
Verified 129d844c
server: add rms_norm_eps parameter (#2380)
slaren committed 2 years ago
Verified d5512b78
[Server] Escape HTML in webchat (#2368)
SlyEcho committed 2 years ago
Verified c798308e
make rms_norm_eps a parameter (#2374)
slaren committed 2 years ago
Verified 41c67416
Chat UI extras (#2366)
akx committed 2 years ago
Verified b3f138d0
ggml : sync (unary ops refactor, static-correctness) (#2370)
ggerganov committed 2 years ago
Verified 5b2b2dc6
chat css
SlyEcho committed 2 years ago
Verified ca2467d1
Merge remote-tracking branch 'origin/master' into server-cfg
SlyEcho committed 2 years ago
Verified f77972f9
Fix scalar version of Q5_K when QK_K = 64 (#2362)
ikawrakow committed 2 years ago
Verified 42f70cb2
llama : add grammar-based sampling (#1773)
ejones committed 2 years ago
Verified 84e09a7d
Some more Q4_K and Q5_K speedup on CUDA (#2346)
ikawrakow committed 2 years ago
Verified 2f9cf974
Add gqa parameter support to the server (#2351)
IgnacioFDM committed 2 years ago
Verified 4f06592c
Fix __dp4a documentation (#2348)
JohannesGaessler committed 2 years ago
Verified 70d26ac3
common : n_threads == -1 uses std::thread::hardware_concurrency() (#2347)
Freed-Wu committed 2 years ago
Verified 57921ca6
fix n_tasks (#2342)
slaren committed 2 years ago
Verified 3602ac42
ggml: move op parameters from tensors to ggml_tensor::op_params (#2333)
slaren committed 2 years ago
Verified 95a6c595
llama : grouped-query attention + LLaMAv2 70B support (#2276)
ggerganov committed 2 years ago
Verified e76d630d
llama : print help to stdout (#2338)
maddes8cht committed 2 years ago
Verified 1d0824b2
Terms Of Use Privacy Policy Refund Policy Impressum