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

ggml : fix YARN + add tests + add asserts (#7617)

Browse source 
* tests : add rope tests

ggml-ci

* ggml : fixes (hopefully)

ggml-ci

* tests : add non-cont tests

ggml-ci

* cuda : add asserts for rope/norm + fix DS2

ggml-ci

* ggml : assert contiguousness

* tests : reduce RoPE tests

ggml-ci
This commit is contained in:
Georgi Gerganov 2024-05-29 20:17:31 +03:00 • committed by GitHub
parent cce3dcffc5
commit fb76ec31a9
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
12 changed files with 167 additions and 104 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml-cuda.cu
View file

@ -1870,7 +1870,7 @@ static void ggml_cuda_mul_mat_batched_cublas(ggml_backend_cuda_context & ctx, co
}
}
#else
if (r2 == 1 && r3 == 1 && src0->nb[2]*src0->ne[2] == src0->nb[3] && src1->nb[2]*src1->ne[2] == src1->nb[3]) {
if (r2 == 1 && r3 == 1 && ggml_is_contiguous_2(src0) && ggml_is_contiguous_2(src1)) {
// there is no broadcast and src0, src1 are contiguous across dims 2, 3
// use cublasGemmStridedBatchedEx
CUBLAS_CHECK(
@ -2886,7 +2886,9 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
case GGML_OP_CONT:
case GGML_OP_DIAG_MASK_INF:
case GGML_OP_SOFT_MAX:
return true;
case GGML_OP_ROPE:
return ggml_is_contiguous(op->src[0]);
case GGML_OP_IM2COL:
case GGML_OP_POOL_2D:
case GGML_OP_SUM_ROWS: