Skip to content
GitLab
Explore
Sign in
Tags
Tags give the ability to mark specific points in history as being important
master-65bdd52
65bdd52a
·
tests : sync test-grad0 from ggml
·
Jun 24, 2023
master-f2c754e
f2c754e1
·
ggml : improve ggml_graph_dump_dot, add ggml_format_name (#1978)
·
Jun 24, 2023
master-b061ba9
b061ba9e
·
llama : fix top-p sampling to match the canonical definition (#1953)
·
Jun 24, 2023
master-527b6fb
527b6fba
·
llama : make model stateless and context stateful (llama_state) (#1797)
·
Jun 24, 2023
master-7487137
74871372
·
rework convert.py to read hyper-parameters from config.json (#1958)
·
Jun 22, 2023
master-bbca06e
bbca06e2
·
cmake: revert CUDA arch default to 52, 61 if f16 (#1959)
·
Jun 21, 2023
master-aacdbd4
aacdbd40
·
llama : fix params struct slignment (#1936)
·
Jun 20, 2023
master-20568fe
20568fe6
·
[Fix] Reenable server embedding endpoint (#1937)
·
Jun 20, 2023
master-18b3562
18b35625
·
ggml : fix bug in LBFGS optimizer (found by ggml tests)
·
Jun 19, 2023
master-ba4e85a
ba4e85a8
·
llama : use aligned memory during ggml_init call from loading saved sessions (#1934)
·
Jun 19, 2023
master-23fc5c2
23fc5c21
·
cmake : fix trailing whitespaces
·
Jun 19, 2023
master-cb40dfc
cb40dfca
·
llama : only use Q6_K for output weights if tensor size is multiple of 256 (#1932)
·
Jun 19, 2023
master-ca7c3f4
ca7c3f4d
·
cuda : faster k-quants on older GPUs (#1930)
·
Jun 19, 2023
master-b97ca43
b97ca431
·
ggml : sync latest ggml repo (#1924)
·
Jun 19, 2023
master-1e3abfc
1e3abfce
·
cmake : fix build shared ggml when CUDA is enabled (#1929)
·
Jun 19, 2023
master-16b9cd1
16b9cd19
·
Convert vector to f16 for dequantize mul mat vec (#1913)
·
Jun 19, 2023
master-b24c304
b24c3049
·
Added tokens per second to info prints (#1928)
·
Jun 18, 2023
master-0ede372
0ede372a
·
Fixed incorrectly applying RMS norm twice (#1925)
·
Jun 18, 2023
master-8596af4
8596af42
·
ggml : fix bug in ggml_compute_forward_add_q_f32 (#1918)
·
Jun 18, 2023
master-8ab8ba6
8ab8ba62
·
llama : prevent usage of k-quants when tensor size is not a multiple of 256 (#1921)
·
Jun 18, 2023
Prev
1
…
73
74
75
76
77
78
79
80
81
…
99
Next