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

implement ggml_cont backward pass

Browse source
This commit is contained in:
xaedes 2023-04-28 18:12:25 +02:00
parent 02d3fd0894
commit 3d21f2646e
No known key found for this signature in database
GPG key ID: 30030EDD817EA2B1
1 changed files with 9 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

11
ggml.c
View file

@ -5822,7 +5822,6 @@ struct ggml_tensor * ggml_cont_impl(
bool is_node = false; bool is_node = false;
if (!inplace && a->grad) { if (!inplace && a->grad) {
// TODO: implement backward
is_node = true; is_node = true;
} }
@ -13188,7 +13187,15 @@ static void ggml_compute_backward(struct ggml_context * ctx, struct ggml_tensor
} break; } break;
case GGML_OP_CONT: case GGML_OP_CONT:
{ {
GGML_ASSERT(false); // TODO: not implemented // same as cpy
if (src0->grad) {
GGML_ASSERT(ggml_is_contiguous(src0->grad));
GGML_ASSERT(ggml_is_contiguous(tensor->grad));
src0->grad = ggml_add_impl(ctx, src0->grad, tensor->grad, inplace);
}
if (src1->grad) {
// noop
}
} break; } break;
case GGML_OP_RESHAPE: case GGML_OP_RESHAPE:
{ {