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

tests: add gradient tests for all backends (ggml/932)

Browse source 
* tests: add gradient checking to test-backend-ops

* remove old comment

* reorder includes

* adjust SIN/COS parameters

* add documentation, use supports_op if possible
This commit is contained in:
Johannes Gäßler 2024-09-03 17:21:46 +02:00 • committed by Georgi Gerganov
parent dbbebcab33
commit 202084d31d
10 changed files with 1080 additions and 92 deletions
Download patch file Download diff file Expand all files Collapse all files

12
ggml/src/ggml-cuda.cu
View file

@ -27,6 +27,7 @@
#include "ggml-cuda/rope.cuh"
#include "ggml-cuda/scale.cuh"
#include "ggml-cuda/softmax.cuh"
#include "ggml-cuda/sum.cuh"
#include "ggml-cuda/sumrows.cuh"
#include "ggml-cuda/tsembd.cuh"
#include "ggml-cuda/unary.cuh"
@ -2180,6 +2181,7 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
ggml_cuda_dup(ctx, dst);
break;
case GGML_OP_ADD:
case GGML_OP_ADD1: // TODO: more efficient implementation
ggml_cuda_op_add(ctx, dst);
break;
case GGML_OP_SUB:
@ -2196,6 +2198,9 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
break;
case GGML_OP_UNARY:
switch (ggml_get_unary_op(dst)) {
case GGML_UNARY_OP_NEG:
ggml_cuda_op_neg(ctx, dst);
break;
case GGML_UNARY_OP_GELU:
ggml_cuda_op_gelu(ctx, dst);
break;
@ -2304,6 +2309,9 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
case GGML_OP_POOL_2D:
ggml_cuda_op_pool2d(ctx, dst);
break;
case GGML_OP_SUM:
ggml_cuda_op_sum(ctx, dst);
break;
case GGML_OP_SUM_ROWS:
ggml_cuda_op_sum_rows(ctx, dst);
break;
@ -2748,6 +2756,7 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
switch (op->op) {
case GGML_OP_UNARY:
switch (ggml_get_unary_op(op)) {
case GGML_UNARY_OP_NEG:
case GGML_UNARY_OP_GELU:
case GGML_UNARY_OP_SILU:
case GGML_UNARY_OP_RELU:
@ -2877,6 +2886,7 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
case GGML_OP_TRANSPOSE:
case GGML_OP_NORM:
case GGML_OP_ADD:
case GGML_OP_ADD1:
case GGML_OP_SUB:
case GGML_OP_MUL:
case GGML_OP_DIV:
@ -2896,7 +2906,9 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
case GGML_OP_ROPE:
return ggml_is_contiguous(op->src[0]);
case GGML_OP_IM2COL:
return op->src[0]->type == GGML_TYPE_F16;
case GGML_OP_POOL_2D:
case GGML_OP_SUM:
case GGML_OP_SUM_ROWS:
case GGML_OP_ARGSORT:
case GGML_OP_ACC: