| /aosp_15_r20/external/executorch/backends/vulkan/test/op_tests/ |
| H A D | sdpa_test.cpp | 217 const int num_heads, in test_reference_sdpa()
286 const int num_heads, in test_vulkan_sdpa()
444 const int num_heads = 6; in TEST() local
464 const int num_heads = 6; in TEST() local
483 const int num_heads = 32; in TEST() local
502 const int num_heads = 32; in TEST() local
|
| /aosp_15_r20/external/pytorch/aten/src/ATen/native/nested/cuda/ |
| H A D | NestedTensorTransformerUtils.cpp | 94 const int64_t num_heads = tensor -> opt_size(2).value(); in is_safe_to_get_storage_as_tensor() local
173 const int64_t num_heads, in view_as_dense()
405 const int64_t num_heads = query.size(1); in sdpa_nested_preprocessing() local
482 const int64_t num_heads = query.size(1); in sdpa_nested_preprocessing_backward() local
|
| /aosp_15_r20/external/pytorch/aten/src/ATen/native/transformers/cuda/flash_attn/ |
| H A D | flash_api.cpp | 298 const int num_heads, const int head_size, const int max_seqlen_k, const int max_seqlen_q, in set_params_splitkv()
329 …Flash_fwd_params ¶ms, std::optional<at::Tensor> &alibi_slopes_, int batch_size, int num_heads){ in set_params_alibi()
390 int num_heads = sizes[2]; in mha_fwd() local
604 int num_heads = sizes[1]; in mha_varlen_fwd() local
866 const int num_heads = sizes[2]; in mha_bwd() local
1088 const int num_heads = sizes[1]; in mha_varlen_bwd() local
1310 int num_heads = sizes[2]; in mha_fwd_kvcache() local
|
| /aosp_15_r20/external/pytorch/torch/csrc/api/src/nn/options/ |
| H A D | activation.cpp | 26 int64_t num_heads) in MultiheadAttentionOptions()
42 int64_t num_heads, in MultiheadAttentionForwardFuncOptions()
|
| /aosp_15_r20/external/pytorch/benchmarks/transformer/ |
| H A D | sdp.py | 102 def __init__(self, num_heads, in_proj_weight, in_proj_bias, out_proj): argument
277 batch_sizes, num_heads, max_seq_lens, embed_dims, dtypes, pad_percentages argument
|
| H A D | attention_bias_benchmarks.py | 81 def __init__(self, num_heads, embed_dim, device=None, dtype=None): argument
|
| /aosp_15_r20/external/executorch/examples/qualcomm/oss_scripts/llama2/runner/ |
| H A D | runner.cpp | 542 size_t num_heads = (method_meta_->num_inputs() - 3) / 2; in compute_total_nbytes() local
573 size_t num_heads = (method_meta_->num_inputs() - 3) / 2; in init_tensors() local
|
| /aosp_15_r20/external/executorch/exir/tests/ |
| H A D | transformer.py | 15 def __init__(self, embed_dim, num_heads=2): argument
|
| /aosp_15_r20/external/executorch/examples/models/llama3_2_vision/vision_encoder/ |
| H A D | model.py | 23 num_heads: int = 16 variable in VisionEncoderConfig
|
| /aosp_15_r20/external/pytorch/torch/csrc/api/include/torch/nn/modules/ |
| H A D | activation.h | 826 MultiheadAttentionImpl(int64_t embed_dim, int64_t num_heads) in MultiheadAttentionImpl()
|
| /aosp_15_r20/external/pytorch/aten/src/ATen/native/transformers/ |
| H A D | transformer.cpp | 78 const int64_t num_heads, in transformer_encoder_layer_forward()
|
| /aosp_15_r20/external/pytorch/test/ |
| H A D | test_native_mha.py | 160 def __init__(self, embed_dim, num_heads, qkv, proj): argument
|
| H A D | test_transformers.py | 3326 def rand_nt(sequence_list, num_heads, head_dim): argument
|
| H A D | test_jit.py | 14912 num_heads, # type: int argument
14992 def __init__(self, embed_dim, num_heads): argument
|
| /aosp_15_r20/external/executorch/examples/mediatek/aot_utils/llm_utils/ |
| H A D | utils.py | 244 num_heads, argument
|
| /aosp_15_r20/external/pytorch/aten/src/ATen/native/transformers/cuda/ |
| H A D | attention.cu | 752 const int64_t num_heads = query.size(1); in _scaled_dot_product_cudnn_attention_cuda() local
1024 int64_t num_heads = query.size(-2); in _efficient_attention_forward() local
|
| H A D | attention_backward.cu | 195 const int64_t num_heads = query.size(1); in _scaled_dot_product_cudnn_attention_backward_cuda() local
|
| /aosp_15_r20/external/executorch/backends/vulkan/runtime/graph/ops/impl/ |
| H A D | SDPA.cpp | 259 const int64_t num_heads = graph.size_at<int64_t>(2, q_projected); in sdpa_with_kv_cache_impl() local
|
| /aosp_15_r20/external/pytorch/test/cpp/api/ |
| H A D | sequential.cpp | 501 int64_t num_heads = 4; in TEST_F() local
|
| /aosp_15_r20/external/pytorch/torch/nn/modules/ |
| H A D | activation.py | 1043 num_heads, argument
|
| /aosp_15_r20/external/pytorch/test/cpp_extensions/ |
| H A D | open_registration_extension.cpp | 455 const int64_t num_heads = query.size(1); in custom_scaled_dot_product_fused_attention_overrideable() local
|
| /aosp_15_r20/external/pytorch/aten/src/ATen/native/transformers/cuda/mem_eff_attention/ |
| H A D | kernel_forward.h | 186 int32_t num_heads = 0; member
|
| H A D | kernel_backward.h | 666 int32_t num_heads = -1; member
|
| /aosp_15_r20/external/pytorch/test/inductor/ |
| H A D | test_cpu_repro.py | 296 num_heads, argument
3161 def __init__(self, hidden_size, num_heads): argument
|
| /aosp_15_r20/external/bcc/libbpf-tools/arm64/ |
| H A D | vmlinux_608.h | 27774 long unsigned int num_heads; member
|