Tags

Tags give the ability to mark specific points in history as being important

master-e0305ea

e0305ead · ggml : add Q4_3 quantization (#1082) · Apr 20, 2023
master-6a9661e

6a9661ea · ci : remove the LLAMA_ACCELERATE matrix dimension from Ubuntu builds in the CI (#1074) · Apr 20, 2023
master-5addcb1

5addcb12 · fix: LLAMA_CUBLAS=1 undefined reference 'shm_open' (#1080) · Apr 20, 2023
master-c8c2c52

c8c2c524 · AVX2 optimization for vec_dot_q4_2_q8_0 (#1068) · Apr 20, 2023
master-02d6988

02d69881 · Improve cuBLAS performance by dequantizing on the GPU (#1065) · Apr 20, 2023
master-f7d0509

f7d05095 · Q4_2 quantization with rmse-optimized scale and quants (#1062) · Apr 19, 2023
master-884e7d7

884e7d7a · ggml : use 8-bit precision for Q4_1 intermediate results (#1047) · Apr 19, 2023
master-f3d4edf

f3d4edf5 · ggml : Q4 cleanup - remove 4-bit dot product code (#1061) · Apr 19, 2023
master-8944a13

8944a132 · Add NVIDIA cuBLAS support (#1044) · Apr 19, 2023
master-6667401

66674012 · Multi-threaded ggml_cpy (#1035) · Apr 19, 2023
master-77a7340

77a73403 · ggml : add new Q4_2 quantization (ARM only) (#1046) · Apr 18, 2023
master-50a8a2a

50a8a2af · ggml : scratch that - vmlaq_n_f32 is always better · Apr 18, 2023
master-dcdd65e

dcdd65e2 · ggml : optimize ggml_vec_dot_q4_0_q8_0() using vectorized accumulators · Apr 18, 2023
master-5ecff35

5ecff351 · Adding a simple program to measure speed of dot products (#1041) · Apr 18, 2023
master-5af8e32

5af8e322 · ci : do not run on drafts · Apr 18, 2023
master-4274722

42747220 · Do not close file after mmap (Windows version) (#1034) · Apr 18, 2023
master-315a95a

315a95a4 · Add LoRA support (#820) · Apr 17, 2023
master-efd0564

efd05648 · llama : well-defined static initialization of complex objects (#927) · Apr 17, 2023
master-eb17a02

eb17a026 · quantize-stats : fix bug in --type argument · Apr 17, 2023
master-69b7402

69b74028 · ggml : avoid using ggml_fp16_to_fp32() and ggml_fp32_to_fp16() in ggml.c · Apr 17, 2023

Prev
1
…
86
87
88
89
90
91
92
93
94
…
99
Next