vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

llama : fix buffer checks for mamba and rwk (#10111)

Browse source 
* llama : fix buffer checks for mamba and rwk

* llama : fix missing worst case flag during reserve

* cuda : fix supports_op for norm

* disable sched SET_CAUSE
This commit is contained in:
Diego Devesa 2024-10-31 22:54:23 +01:00 • committed by GitHub
parent ab3d71f97f
commit c02e5ab2a6
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 35 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

1
ggml/src/ggml.c
View file

@ -7272,6 +7272,7 @@ struct ggml_tensor * ggml_ssm_conv(
const int64_t n_s = sx->ne[2];
// TODO: maybe support other strides than 1?
// FIXME: this is always true?
GGML_ASSERT(sx->ne[0] == d_conv - 1 + n_t);
GGML_ASSERT(sx->ne[1] == d_inner);
GGML_ASSERT(n_t >= 0);