Skip to content

Support dup & cont ops on CUDA #2242

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jul 17, 2023
Merged

Support dup & cont ops on CUDA #2242

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 18 additions & 1 deletion ggml-cuda.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3537,6 +3537,11 @@ void ggml_cuda_cpy(const ggml_tensor * src0, const ggml_tensor * src1, ggml_tens
(void) dst;
}

void ggml_cuda_dup(const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
ggml_cuda_cpy(src0, dst, nullptr);
(void) src1;
}

void ggml_cuda_diag_mask_inf(const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
GGML_ASSERT(src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32);
ggml_cuda_op(src0, src1, dst, ggml_cuda_op_diag_mask_inf, true, true);
Expand Down Expand Up @@ -3670,7 +3675,7 @@ void ggml_cuda_assign_buffers_impl(struct ggml_tensor * tensor, bool scratch, bo
// recursively assign CUDA buffers until a compute tensor is found
if (tensor->src[0] != nullptr && tensor->src[0]->backend == GGML_BACKEND_CPU) {
const ggml_op src0_op = tensor->src[0]->op;
if (src0_op == GGML_OP_RESHAPE || src0_op == GGML_OP_TRANSPOSE || src0_op == GGML_OP_VIEW) {
if (src0_op == GGML_OP_RESHAPE || src0_op == GGML_OP_TRANSPOSE || src0_op == GGML_OP_VIEW || src0_op == GGML_OP_PERMUTE) {
ggml_cuda_assign_buffers_impl(tensor->src[0], scratch, force_inplace);
}
}
Expand Down Expand Up @@ -3776,6 +3781,12 @@ bool ggml_cuda_compute_forward(struct ggml_compute_params * params, struct ggml_
|| (tensor->src[1] != nullptr && tensor->src[1]->backend == GGML_BACKEND_GPU);

switch (tensor->op) {
case GGML_OP_DUP:
if (!any_on_device) {
return false;
}
func = ggml_cuda_dup;
break;
case GGML_OP_ADD:
if (!any_on_device) {
return false;
Expand Down Expand Up @@ -3830,6 +3841,12 @@ bool ggml_cuda_compute_forward(struct ggml_compute_params * params, struct ggml_
}
func = ggml_cuda_cpy;
break;
case GGML_OP_CONT:
if (!any_on_device) {
return false;
}
func = ggml_cuda_dup;
break;
case GGML_OP_RESHAPE:
case GGML_OP_VIEW:
case GGML_OP_PERMUTE:
Expand Down