ggerganov/llama.cpp

Pull Requests Commits
make : fix CPPFLAGS (#3035)
cebtenzzre committed 2 years ago
Verified 5ffab089
llama-bench : use two tokens in the warmup run for prompt evals (#3059)
slaren committed 2 years ago
Verified 15b67a66
metal : parallel RoPE on Metal (#3024)
ikawrakow committed 2 years ago
Verified be8c9c24
metal : correct fix of kernel_norm (#3060)
ikawrakow committed 2 years ago
Verified be6beeb8
metal : fix kernel_norm (fixes Falcon on Metal) (#3057)
ggerganov committed 2 years ago
Verified c4f49664
ggml : posixify madvise and pagesize (#3037)
przemoc committed 2 years ago
Verified fec2fb19
k-quants : fix zero-weight guard in Q6_K (ref #3040)
ggerganov committed 2 years ago
Verified 178b1850
convert-llama-ggml-to-gguf: Try to handle files older than GGJTv3 (#3023)
KerfuffleV2 committed 2 years ago
Verified ea2c85d5
build : add LLAMA_METAL_NDEBUG flag (#3033)
cebtenzzre committed 2 years ago
Verified 9912b9ef
make : use new flag variables for recent changes (#3019)
cebtenzzre committed 2 years ago
Verified 9e202315
examples : replace fprintf to stdout with printf (#3017)
cebtenzzre committed 2 years ago
Verified de2fe892
convert: fix convert.py not working with int filename_stem (#3028)
Green-Sky committed 2 years ago
Verified c9c3220c
Guard against all weights in a super-block being zero (#3010)
ikawrakow committed 2 years ago
Verified d59bd970
llama : update logic for number of threads when using BLAS
ggerganov committed 2 years ago
Verified 35938ee3
speculative : add grammar support (#2991)
ggerganov committed 2 years ago
Verified 92177210
py : minor
ggerganov committed 2 years ago
Verified 2ba85c86
build : on Mac OS enable Metal by default (#2901)
ggerganov committed 2 years ago
Verified e36ecdcc
ggml-opencl : store GPU buffer in ggml_tensor::extra (#2994)
slaren committed 2 years ago
Verified bd33e5ab
llama-bench : make cpp file non-executable (#2999)
cebtenzzre committed 2 years ago
Verified 31035681
make : add speculative example (#3003)
leng-yue committed 2 years ago
Verified 5b8530d8
server : add a subtle loading animation to the edit box (#2466)
akx committed 2 years ago
Verified e4386f41
2x faster (rms) norm cuda kernels (3.7% e2e improvement) (#2985)
li-plus committed 2 years ago
Verified 35195689
ggml-alloc : use virtual memory for measurement (#2973)
slaren committed 2 years ago
Verified cf9b0848
speculative : PoC for speeding-up inference via speculative sampling (#2926)
ggerganov committed 2 years ago
Verified 47068e51
perplexity : fix ETA by warming up the model with an empty run
ggerganov committed 2 years ago
Verified 8f429fa5
gguf(python): Fix special vocab handling when id < 0 (#2984)
KerfuffleV2 committed 2 years ago
Verified 6519e9c9
metal : restore 363f0bf and fix reduce in F16_F32 kernels (#2986)
ggerganov committed 2 years ago
Verified b7f2aa9e
cov : disable comment in PRs (#2989)
alonfaraj committed 2 years ago
Verified 73a12a63
llama : fix bpe tokenize from byte (#2889)
opparco committed 2 years ago
Verified 37301347
metal : revert 6af0bab until we fix it
ggerganov committed 2 years ago
Verified d9151e6f
Terms Of Use Privacy Policy Refund Policy Impressum