ggerganov/llama.cpp

Pull Requests Commits
ggml : fix bug in ggml_compute_forward_dup_f32()
ggerganov committed 2 years ago
Verified 9ff334f3
Add Q4_3 support to cuBLAS (#1086)
slaren committed 2 years ago
Verified 2005469e
ggml : do not break cuBLAS build (Q4_3 is not yet implemented)
ggerganov committed 2 years ago
Verified 8a1756ab
ggml : fix Q4_3 quantization
ggerganov committed 2 years ago
Verified 66aab460
llama : multi-threaded quantization (#1075)
ikawrakow committed 2 years ago
Verified 38de86a7
ggml : add Q4_3 quantization (#1082)
ggerganov committed 2 years ago
Verified e0305ead
ci : remove the LLAMA_ACCELERATE matrix dimension from Ubuntu builds in the CI (#1074)
dfyz committed 2 years ago
Verified 6a9661ea
fix: LLAMA_CUBLAS=1 undefined reference 'shm_open' (#1080)
fumiama committed 2 years ago
Verified 5addcb12
AVX2 optimization for vec_dot_q4_2_q8_0 (#1068)
sw committed 2 years ago
Verified c8c2c524
Improve cuBLAS performance by dequantizing on the GPU (#1065)
slaren committed 2 years ago
Verified 02d69881
Minor: Readme fixed grammar, spelling, and misc updates (#1071)
CRD716 committed 2 years ago
Verified 834695fe
Q4_2 quantization with rmse-optimized scale and quants (#1062)
ikawrakow committed 2 years ago
Verified f7d05095
ggml : use 8-bit precision for Q4_1 intermediate results (#1047)
ggerganov committed 2 years ago
Verified 884e7d7a
readme : add warning about Q4_2 and Q4_3
ggerganov committed 2 years ago
Verified 7cd5c4a3
ggml : Q4 cleanup - remove 4-bit dot product code (#1061)
sw committed 2 years ago
Verified f3d4edf5
Add NVIDIA cuBLAS support (#1044)
slaren committed 2 years ago
Verified 8944a132
Multi-threaded ggml_cpy (#1035)
slaren committed 2 years ago
Verified 66674012
ggml : add new Q4_2 quantization (ARM only) (#1046)
ggerganov committed 2 years ago
Verified 77a73403
ggml : scratch that - vmlaq_n_f32 is always better
ggerganov committed 2 years ago
Verified 50a8a2af
gitignore : vdot
ggerganov committed 2 years ago
Verified 4caebf6d
ggml : optimize ggml_vec_dot_q4_0_q8_0() using vectorized accumulators
ggerganov committed 2 years ago
Verified dcdd65e2
Adding a simple program to measure speed of dot products (#1041)
ikawrakow committed 2 years ago
Verified 5ecff351
readme : update hot topics about new LoRA functionality
ggerganov committed 2 years ago
Verified 7faa7460
ci : do not run on drafts
ggerganov committed 2 years ago
Verified 5af8e322
Do not close file after mmap (Windows version) (#1034)
dfyz committed 2 years ago
Verified 42747220
readme : add Ruby bindings (#1029)
yoshoku committed 2 years ago
Verified e9298af3
add 4_0 to default outfile namestr dict (#1031)
cammytown committed 2 years ago
Verified 4ad73137
Add LoRA support (#820)
slaren committed 2 years ago
Verified 315a95a4
llama : well-defined static initialization of complex objects (#927)
arikpoz committed 2 years ago
Verified efd05648
quantize-stats : fix bug in --type argument
ggerganov committed 2 years ago
Verified eb17a026
Terms Of Use Privacy Policy Refund Policy Impressum