Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-27 22:59:24 +01:00
Code Issues Packages Projects Releases Wiki Activity

fix q4_0_8_8 format for corrupted tokens issue (#10198)

Browse Source 
Co-authored-by: EC2 Default User <ec2-user@ip-172-31-62-167.us-west-2.compute.internal>
This commit is contained in:
snadampal 2024-11-07 02:02:08 -06:00 committed by GitHub
parent 3bcd40b3c5
commit 2319126a70
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-cpu.c
View File

@ -409,6 +409,8 @@ static const struct ggml_type_traits_cpu type_traits_cpu[GGML_TYPE_COUNT] = {
.gemm = ggml_gemm_q4_0_4x8_q8_0, .gemm = ggml_gemm_q4_0_4x8_q8_0,
}, },
[GGML_TYPE_Q4_0_8_8] = { [GGML_TYPE_Q4_0_8_8] = {
.vec_dot = NULL,
.vec_dot_type = GGML_TYPE_Q8_0,
.nrows = 1, .nrows = 1,
.ncols = 8, .ncols = 8,
.gemv = ggml_gemv_q4_0_8x8_q8_0, .gemv = ggml_gemv_q4_0_8x8_q8_0,