tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-07-19 09:08:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

cuda/cpu: Increase support for fp16 unary operations (ggml/1125)

Browse Source 
* Support fp16 unary operations in the CUDA backend

* cpu: increase fp16 support for unary operators in the CPU backend

* cuda: increase fp16 support for unary operators in the CUDA backend

* Add test cases for fp16 unary operators

* metal: update supports_op for unary operators that don't support fp16, to prevent test-backend-ops from failing

* metal: fix PR comments for unary op support after fp16 unary tests
This commit is contained in:
cmdr2
2025-02-28 12:34:39 +05:30
committed by Georgi Gerganov
parent 6d4c23b81b
commit 87abb7e903
7 changed files with 1266 additions and 175 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
ggml/src/ggml-metal/ggml-metal.m
View File

@ -1200,7 +1200,7 @@ static bool ggml_metal_supports_op(const struct ggml_backend_metal_device_contex
case GGML_UNARY_OP_GELU_QUICK:
case GGML_UNARY_OP_SILU:
case GGML_UNARY_OP_ELU:
return ggml_is_contiguous(op->src[0]);
return ggml_is_contiguous(op->src[0]) && op->src[0]->type == GGML_TYPE_F32;
default:
return false;
}
@ -1210,21 +1210,26 @@ static bool ggml_metal_supports_op(const struct ggml_backend_metal_device_contex
case GGML_OP_TRANSPOSE:
case GGML_OP_PERMUTE:
case GGML_OP_CONCAT:
return true;
case GGML_OP_ADD:
case GGML_OP_SUB:
case GGML_OP_ACC:
case GGML_OP_MUL:
case GGML_OP_DIV:
return op->src[0]->type == GGML_TYPE_F32;
case GGML_OP_ACC:
case GGML_OP_REPEAT:
case GGML_OP_SCALE:
case GGML_OP_CLAMP:
case GGML_OP_CONV_TRANSPOSE_1D:
return true;
case GGML_OP_CLAMP:
return op->src[0]->type == GGML_TYPE_F32;
case GGML_OP_SQR:
case GGML_OP_SQRT:
case GGML_OP_SIN:
case GGML_OP_COS:
return ggml_is_contiguous(op->src[0]);
return ggml_is_contiguous(op->src[0]) && op->src[0]->type == GGML_TYPE_F32;
case GGML_OP_LOG:
return false; // TODO: implement
case GGML_OP_SUM_ROWS:
case GGML_OP_SOFT_MAX:
case GGML_OP_GROUP_NORM:
@ -1254,10 +1259,11 @@ static bool ggml_metal_supports_op(const struct ggml_backend_metal_device_contex
case GGML_OP_UPSCALE:
case GGML_OP_PAD:
case GGML_OP_PAD_REFLECT_1D:
case GGML_OP_ARANGE:
case GGML_OP_TIMESTEP_EMBEDDING:
case GGML_OP_ARGSORT:
case GGML_OP_LEAKY_RELU:
return op->src[0]->type == GGML_TYPE_F32;
case GGML_OP_ARANGE:
return true;
case GGML_OP_FLASH_ATTN_EXT:
if (op->src[1]->type != op->src[2]->type) {