| /third_party/mindspore/mindspore-src/source/mindspore/python/mindspore/ |
| D | amp.py | 87 def _grad_scale_map(scale_value, inputs): argument 92 def _grad_unscale_map(scale_value, inputs): argument 265 def __init__(self, scale_value): argument 343 def __init__(self, scale_value, scale_factor, scale_window): argument
|
| /third_party/mindspore/mindspore-src/source/mindspore/ccsrc/plugin/device/cpu/kernel/nnacl/fp32/ |
| D | dropout_fp32_simd.h.in | 29 SIMD_F32 scale_value = SIMD_MOV_F32(scale); variable
|
| /third_party/mindspore/mindspore-src/source/tests/ut/cpp/ops/ |
| D | test_ops_paged_attention.cc | 43 ValuePtr scale_value; member 60 auto scale_value = param.scale_value->ToAbstract(); in TEST_P() local
|
| D | test_ops_paged_attention_mask.cc | 45 ValuePtr scale_value; member 63 auto scale_value = param.scale_value->ToAbstract(); in TEST_P() local
|
| /third_party/mindspore/mindspore-src/source/mindspore/ccsrc/plugin/device/cpu/kernel/nnacl/include/nnacl/sse/ |
| D | dropout_fp32_sse.h | 33 SIMD_F32 scale_value = SIMD_MOV_F32(scale); in DropoutFp32SSE() local
|
| /third_party/mindspore/mindspore-src/source/mindspore/ccsrc/plugin/device/cpu/kernel/nnacl/include/nnacl/avx512/ |
| D | dropout_fp32_avx512.h | 33 SIMD_F32 scale_value = SIMD_MOV_F32(scale); in DropoutFp32AVX512() local
|
| /third_party/mindspore/mindspore-src/source/mindspore/ccsrc/plugin/device/cpu/kernel/nnacl/include/nnacl/neon/ |
| D | dropout_fp32_neon.h | 32 SIMD_F32 scale_value = SIMD_MOV_F32(scale); in DropoutFp32NEON() local
|
| /third_party/mindspore/mindspore-src/source/mindspore/ccsrc/plugin/device/cpu/kernel/nnacl/include/nnacl/avx/ |
| D | dropout_fp32_avx.h | 33 SIMD_F32 scale_value = SIMD_MOV_F32(scale); in DropoutFp32AVX() local
|
| /third_party/mindspore/mindspore-src/source/tests/st/ops/ascend/ |
| D | test_fused_infer_attention_score.py | 31 … deq_scale1, quant_scale1, deq_scale2, quant_scale2, quant_offset2, num_heads, scale_value=1.0, argument 43 def __init__(self, num_heads, input_layout, scale_value, num_key_value_heads): argument 62 def __init__(self, num_heads, input_layout='BSH', scale_value=1.0, num_key_value_heads=0, argument
|
| D | test_prompt_flash_attention.py | 30 … deq_scale1, quant_scale1, deq_scale2, quant_scale2, quant_offset2, num_heads, scale_value=1.0, argument
|
| /third_party/mindspore/mindspore-src/source/mindspore/ccsrc/plugin/device/ascend/kernel/pyboost/customize/ |
| D | flash_attention_score.cc | 49 …const Int64ImmPtr head_num, const FP32ImmPtr keep_prob, const FP32ImmPtr scale_value, const Int64I… in FlashAttentionScoreAscendCall() 104 const FP32ImmPtr scale_value, const Int64ImmPtr pre_tokens, const Int64ImmPtr next_tokens, in FlashAttentionScoreAscendCustomize()
|
| D | flash_attention_score_grad.cc | 38 const FP32ImmPtr scale_value, const Int64ImmPtr pre_tokens, const Int64ImmPtr next_tokens, in FlashAttentionScoreGradAscendCall() 92 const FP32ImmPtr scale_value, const Int64ImmPtr pre_tokens, const Int64ImmPtr next_tokens, in FlashAttentionScoreGradAscendCustomize()
|
| /third_party/mindspore/mindspore-src/source/mindspore/lite/src/litert/kernel/cpu/base/ |
| D | quant_dtype_cast.cc | 291 float scale_value = scale[i]; in DoDequanInt8ToFp32ChannelRow() local 311 float scale_value = scale[j]; in DoDequanInt8ToFp32ChannelCol() local
|
| /third_party/mindspore/mindspore-src/source/mindspore/lite/tools/converter/parser/onnx/ |
| D | onnx_prompt_flash_attention_parser.cc | 46 auto scale_value = onnx_node_attr.f(); in Parse() local
|
| /third_party/mindspore/mindspore-src/source/tests/st/ops/ascend/test_acl_ops/ |
| D | test_incre_flash_attention.py | 27 def __init__(self, num_heads, input_layout, scale_value, num_key_value_heads): argument
|
| /third_party/mindspore/mindspore-src/source/mindspore/lite/tools/optimizer/fusion/ |
| D | flash_attention_fusion.cc | 96 bool PFACheckShape(float scale_value, const std::vector<int64_t> &q_shape, const std::vector<int64_… in PFACheckShape() 1302 const AnfNodePtr &atten_mask, int64_t num_heads, int64_t next_token, float scale_value, in CreatePromptFlashAttentionCnodeForBNSD() 1357 …dePtr &atten_mask, const AnfNodePtr &pse, int64_t num_heads, int64_t next_token, float scale_value, in CreatePromptFlashAttentionCnodeForBNSDWithPse() 1422 const AnfNodePtr &atten_mask, int64_t num_heads, int64_t next_token, float scale_value, in CreatePromptFlashAttentionCnodeForBSH() 1510 float scale_value = 1 / (pow(input_tensor_q_shape[kNumIndex3], kNumPowerHalf)); in CreateFAForBNSDWithAttenMask() local 1582 float scale_value = 1 / (pow(input_tensor_q_shape[kNumIndex3], kNumPowerHalf)); in CreateGQACNodeForBNSD() local 1599 …const AnfNodePtr &atten_mask, int64_t num_heads, float scale_value, int64_t num_key_value_heads) c… in CreateIncreFlashAttentionCnodeForBNSD() 1640 float scale_value, in CreateFAForSD15() 1671 int64_t next_token, float scale_value, in CreateFAWithPadAndPse() 1786 float scale_value = 1 / (pow(input_tensor_q_shape[kNumIndex3], kNumPowerHalf)); in CreateFlashAttentionNodeForMsSDXL() local [all …]
|
| /third_party/mindspore/mindspore-src/source/mindspore/ccsrc/plugin/device/ascend/kernel/opapi/aclnn/ |
| D | flash_attention_score_grad_aclnn_kernel.h | 78 auto scale_value = inputs[kIndex17]; in DEFINE_GET_WORKSPACE_FOR_RESIZE() local
|
| D | flash_attention_score_aclnn_kernel.cc | 49 auto scale_value = inputs[kIndex12]; in GetWorkSpaceInfo() local
|
| D | flash_attention_score_grad_aclnn_kernel.cc | 49 auto scale_value = inputs[kIndex17]; in GetWorkSpaceInfo() local
|
| D | flash_attention_score_aclnn_kernel.h | 78 auto scale_value = inputs[kIndex12]; in DEFINE_GET_WORKSPACE_FOR_RESIZE() local
|
| /third_party/mindspore/mindspore-src/source/mindspore/ccsrc/plugin/device/ascend/optimizer/ir_fusion/ |
| D | flash_attention_fusion.cc | 110 …AnfNodePtr &atten_mask, const int64_t num_heads, const int64_t next_token, const float scale_value, in CreatePromptFlashAttentionCnodeForBNSD() 250 const float scale_value = 1 / (pow(input_tensor_q_shape[kNumIndex3], 0.5)); in CreateFlashAttentionNode() local 364 const float scale_value = 1 / (pow(input_tensor_q_shape[kNumIndex3], 0.5)); in CreateFlashAttentionNode() local
|
| /third_party/mindspore/mindspore-src/source/mindspore/ccsrc/plugin/device/ascend/optimizer/ir_fission/ |
| D | scale_grad_fission.cc | 78 float scale_value = 0.0; in Process() local
|
| /third_party/mindspore/mindspore-src/source/tests/st/auto_parallel/ |
| D | test_prompt_flash_attention_info.py | 122 …def __init__(self, num_heads, scale_value=1.0, pre_tokens=2147483547, next_tokens=0, input_layout=… argument
|
| /third_party/mindspore/mindspore-src/source/tests/ut/python/parallel/ |
| D | test_incre_flash_attention.py | 88 scale_value=1.0, argument
|
| /third_party/mindspore/mindspore-src/source/mindspore/core/ops/ |
| D | scale_and_translate.cc | 101 std::vector<float> scale_value = GetArrayValue<float>(scale_abs).value().ToVector(); in ScaleAndTranslateInferShape() local
|