extern/whisper.cpp
1
0
Fork 1
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2024-10-05 01:22:22 +02:00
Code Issues Packages Projects Releases Wiki Activity

ggml : fix cont with transposed tensors when one dimension is 1 (ggml/934)

Browse Source 
* ggml_cont: fix issue with transposed tensors when one dimension is 1

when using multiple threads, it is not enough
to check for the tensors to be contiguous for
ggml_compute_forward_dup_same_cont to work correctly.
The tensors strides also need to match.

Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>

* Add ggml_cont tests

Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>

* Remove dead code

it isn't possible to reach this code because
all these functions are invoked by ggml_compute_forward_dup
if and only if src0->type != dst->type

Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>

* Make ggml_compute_forward_dup_same_cont work with contiguous tensors

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>

---------

Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Salvatore Mesoraca 2024-08-28 10:23:02 +02:00 committed by Georgi Gerganov
parent c96906d84d
commit 4a4a52bf98
1 changed files with 5 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
ggml/src/ggml.c
View File

@ -8120,8 +8120,7 @@ static void ggml_compute_forward_dup_same_cont(
GGML_ASSERT(ggml_is_contiguous(dst) && ggml_is_contiguous(src0));
GGML_ASSERT(src0->type == dst->type);
const size_t nb00 = src0->nb[0];
const size_t nb0 = dst->nb[0];
const size_t nb0 = ggml_type_size(src0->type);
const int ith = params->ith; // thread index
const int nth = params->nth; // number of threads
@ -8135,8 +8134,8 @@ static void ggml_compute_forward_dup_same_cont(
if (ie0 < ie1) {
memcpy(
((char *) dst->data + ie0*nb0),
((char *) src0->data + ie0*nb00),
(ie1 - ie0) * ggml_type_size(src0->type));
((char *) src0->data + ie0*nb0),
(ie1 - ie0) * nb0);
}
}
@ -8153,11 +8152,6 @@ static void ggml_compute_forward_dup_f16(
const int ith = params->ith; // thread index
const int nth = params->nth; // number of threads
if (ggml_is_contiguous(src0) && ggml_is_contiguous(dst) && src0->type == dst->type) {
ggml_compute_forward_dup_same_cont(params, dst);
return;
}
// parallelize by rows
const int nr = ne01;
// number of rows per thread
@ -8422,11 +8416,6 @@ static void ggml_compute_forward_dup_bf16(
const int ith = params->ith; // thread index
const int nth = params->nth; // number of threads
if (ggml_is_contiguous(src0) && ggml_is_contiguous(dst) && src0->type == dst->type) {
ggml_compute_forward_dup_same_cont(params, dst);
return;
}
// parallelize by rows
const int nr = ne01;
// number of rows per thread
@ -8778,11 +8767,6 @@ static void ggml_compute_forward_dup_f32(
const int ith = params->ith; // thread index
const int nth = params->nth; // number of threads
if (ggml_is_contiguous(src0) && ggml_is_contiguous(dst) && src0->type == dst->type) {
ggml_compute_forward_dup_same_cont(params, dst);
return;
}
// parallelize by rows
const int nr = ne01;
// number of rows per thread
@ -9092,13 +9076,13 @@ static void ggml_compute_forward_dup_bytes(
GGML_ASSERT(ggml_nelements(dst) == ggml_nelements(src0));
GGML_ASSERT(src0->type == dst->type);
GGML_TENSOR_UNARY_OP_LOCALS;
if (ggml_is_contiguous(src0) && ggml_is_contiguous(dst)) {
ggml_compute_forward_dup_same_cont(params, dst);
return;
}
GGML_TENSOR_UNARY_OP_LOCALS;
const size_t type_size = ggml_type_size(src0->type);
const int ith = params->ith; // thread index
const int nth = params->nth; // number of threads