ggml : implement GEGLU_ERF and GEGLU_QUICK ops (#14445)
This commit is contained in:
@@ -530,6 +530,8 @@ enum ggml_metal_kernel_type {
|
||||
GGML_METAL_KERNEL_TYPE_REGLU,
|
||||
GGML_METAL_KERNEL_TYPE_GEGLU,
|
||||
GGML_METAL_KERNEL_TYPE_SWIGLU,
|
||||
GGML_METAL_KERNEL_TYPE_GEGLU_ERF,
|
||||
GGML_METAL_KERNEL_TYPE_GEGLU_QUICK,
|
||||
GGML_METAL_KERNEL_TYPE_SUM_ROWS,
|
||||
GGML_METAL_KERNEL_TYPE_MEAN,
|
||||
GGML_METAL_KERNEL_TYPE_POOL_2D_AVG_F32,
|
||||
@@ -1510,6 +1512,8 @@ static struct ggml_backend_metal_context * ggml_metal_init(ggml_backend_dev_t de
|
||||
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_REGLU, reglu, true);
|
||||
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_GEGLU, geglu, true);
|
||||
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_SWIGLU, swiglu, true);
|
||||
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_GEGLU_ERF, geglu_erf, true);
|
||||
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_GEGLU_QUICK, geglu_quick, true);
|
||||
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_SUM_ROWS, sum_rows, true);
|
||||
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_MEAN, mean, true);
|
||||
GGML_METAL_ADD_KERNEL(GGML_METAL_KERNEL_TYPE_ARGMAX, argmax, true);
|
||||
@@ -1693,6 +1697,8 @@ static bool ggml_metal_supports_op(const struct ggml_backend_metal_device_contex
|
||||
case GGML_GLU_OP_REGLU:
|
||||
case GGML_GLU_OP_GEGLU:
|
||||
case GGML_GLU_OP_SWIGLU:
|
||||
case GGML_GLU_OP_GEGLU_ERF:
|
||||
case GGML_GLU_OP_GEGLU_QUICK:
|
||||
return ggml_is_contiguous_1(op->src[0]) && op->src[0]->type == GGML_TYPE_F32;
|
||||
default:
|
||||
return false;
|
||||
@@ -2456,6 +2462,12 @@ static bool ggml_metal_encode_node(
|
||||
case GGML_GLU_OP_SWIGLU:
|
||||
pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_SWIGLU].pipeline;
|
||||
break;
|
||||
case GGML_GLU_OP_GEGLU_ERF:
|
||||
pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_GEGLU_ERF].pipeline;
|
||||
break;
|
||||
case GGML_GLU_OP_GEGLU_QUICK:
|
||||
pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_GEGLU_QUICK].pipeline;
|
||||
break;
|
||||
default:
|
||||
GGML_ABORT("fatal error");
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user