 Georgi Gerganov
|
6ff13987ad
common : normalize naming style (#7462)
|
1 ano atrás |
 jukofyork
|
48b2f9c1fc
Fixed save_imatrix to match old behaviour for MoE (#7099)
|
1 ano atrás |
 Pierrick Hymbert
|
0c4d489e29
quantize: add imatrix and dataset metadata in GGUF (#6658)
|
1 ano atrás |
 slaren
|
0d56246f4b
ggml : group all experts in a single ggml_mul_mat_id (#6505)
|
1 ano atrás |
|
Georgi Gerganov
|
ef21ce4ccb
imatrix : remove invalid assert (#6632)
|
1 ano atrás |
|
Pierrick Hymbert
|
b804b1ef77
eval-callback: Example how to use eval callback for debugging (#6576)
|
1 ano atrás |
 Jared Van Bortel
|
1b67731e18
BERT tokenizer fixes (#6498)
|
1 ano atrás |
|
slaren
|
08a0c02060
ggml : mul_mat_id use the same tensor for all the experts (#6387)
|
1 ano atrás |
 compilade
|
557410b8f0
llama : greatly reduce output buffer memory usage (#6122)
|
1 ano atrás |
|
Georgi Gerganov
|
a0e584defd
imatrix : fix wname for mul_mat_id ops (#6271)
|
1 ano atrás |
|
slaren
|
2bf8d0f7c4
backend : offload large batches to GPU (#6083)
|
1 ano atrás |
 bmwl
|
f486f6e1e5
ggml : add numa options (#5377)
|
1 ano atrás |
 Kawrakow
|
5ed26e1fc9
Adding some imatrix tools (#5302)
|
1 ano atrás |
|
Kawrakow
|
15bceec2d7
imatrix : keep intermediate imatrix results (#5077)
|
2 anos atrás |
|
Kawrakow
|
726c0fa9a2
Slightly faster imatrix (#5050)
|
2 anos atrás |
|
Georgi Gerganov
|
2d5419d08a
imatrix : fix assert for src0 non-cont check
|
2 anos atrás |
|
Georgi Gerganov
|
ba69bbc84c
imatrix : offload to GPU support (#4957)
|
2 anos atrás |
|
Kawrakow
|
326b418b59
Importance Matrix calculation (#4861)
|
2 anos atrás |
