ggml-org/llama.cpp

Pull Requests Commits
ggml : alternative Q4_3 implementation using modified Q8_0 (#1109)
ggerganov committed 3 years ago
Verified 955ef9a5
ggml : AVX2 optimization for vec_dot_q4_3_q8_0 and refactoring (#1099)
sw committed 3 years ago
Verified c5aa5e57
examples : Improve Alpaca Default Repeat Penalty: Better Match Alpaca.cpp Experience (#1107)
HanClinto committed 3 years ago
Verified e9a9cb0c
llama : add api for getting/setting the complete state: rng, logits, embedding and kv_cache (#1105)
xaedes committed 3 years ago
Verified b6e7f9b0
Improve cuBLAS performance by using a memory pool (#1094)
slaren committed 3 years ago
Verified 50cb666b
llama : fixed rlimit error message (#888)
apaz-cli committed 3 years ago
Verified 25d7abbd
cmake : link threads publicly to ggml (#1042)
fumiama committed 3 years ago
Verified 018f2279
main : evaluate tokens in batches after swapping context (#1014)
grencez committed 3 years ago
Verified 94112882
llama : remember and restore kv cache data pointers (#1104)
xaedes committed 3 years ago
Verified 8687c1f2
ggml : a faster version for Q4_1 x Q8_0 dot products (#1083)
ikawrakow committed 3 years ago
Verified 1bfc153e
Show perplexity ETA in hours and minutes (#1096)
slaren committed 3 years ago
Verified 3d59769c
llama : fix comment for "output.weight" tensor
ggerganov committed 3 years ago
Verified d40fded9
Add ggml-model-*.bin checksums for 7B, 13B, 30B, 65B (#1088)
sw committed 3 years ago
Verified 2510c183
ggml : sync ggml (add GPT-NeoX RoPE implementation)
ggerganov committed 3 years ago
Verified 12b5900d
ggml : fix bug in ggml_compute_forward_dup_f32()
ggerganov committed 3 years ago
Verified 9ff334f3
Add Q4_3 support to cuBLAS (#1086)
slaren committed 3 years ago
Verified 2005469e
ggml : do not break cuBLAS build (Q4_3 is not yet implemented)
ggerganov committed 3 years ago
Verified 8a1756ab
ggml : fix Q4_3 quantization
ggerganov committed 3 years ago
Verified 66aab460
llama : multi-threaded quantization (#1075)
ikawrakow committed 3 years ago
Verified 38de86a7
ggml : add Q4_3 quantization (#1082)
ggerganov committed 3 years ago
Verified e0305ead
ci : remove the LLAMA_ACCELERATE matrix dimension from Ubuntu builds in the CI (#1074)
dfyz committed 3 years ago
Verified 6a9661ea
fix: LLAMA_CUBLAS=1 undefined reference 'shm_open' (#1080)
fumiama committed 3 years ago
Verified 5addcb12
AVX2 optimization for vec_dot_q4_2_q8_0 (#1068)
sw committed 3 years ago
Verified c8c2c524
Improve cuBLAS performance by dequantizing on the GPU (#1065)
slaren committed 3 years ago
Verified 02d69881
Minor: Readme fixed grammar, spelling, and misc updates (#1071)
CRD716 committed 3 years ago
Verified 834695fe
Q4_2 quantization with rmse-optimized scale and quants (#1062)
ikawrakow committed 3 years ago
Verified f7d05095
ggml : use 8-bit precision for Q4_1 intermediate results (#1047)
ggerganov committed 3 years ago
Verified 884e7d7a
readme : add warning about Q4_2 and Q4_3
ggerganov committed 3 years ago
Verified 7cd5c4a3
ggml : Q4 cleanup - remove 4-bit dot product code (#1061)
sw committed 3 years ago
Verified f3d4edf5
Add NVIDIA cuBLAS support (#1044)
slaren committed 3 years ago
Verified 8944a132
Terms Of Use Privacy Policy Refund Policy Impressum