merge

flexflow · Oct 29, 2024 · 48570a3 · 48570a3
1 parent 2b99168
commit 48570a3
Showing 1 changed file with 2 additions and 87 deletions.
diff --git a/include/flexflow/model.h b/include/flexflow/model.h
@@ -709,51 +709,11 @@ class FFModel {
                              DataType data_type = DT_NONE,
                              Initializer *kernel_initializer = NULL,
                              char const *name = NULL);
-<<<<<<< HEAD
-  Tensor inc_multihead_self_attention(Tensor const input,
-                                      int embed_dim,
-                                      int num_heads,
-                                      int num_hidden_layers,
-                                      int kdim = 0,
-                                      int vdim = 0,
-                                      float dropout = 0.0f,
-                                      bool bias = false,
-                                      bool add_bias_kv = false,
-                                      bool add_zero_attn = false,
-                                      DataType data_type = DT_NONE,
-                                      Initializer *kernel_initializer = NULL,
-                                      bool apply_rotary_embedding = false,
-                                      bool scaling_query = false,
-                                      float scaling_factor = 1.0f,
-                                      bool qk_prod_scaling = true,
-                                      bool position_bias = false,
-                                      bool streaming_cache = false,
-                                      char const *name = NULL);
-  Tensor
-      spec_inc_multihead_self_attention(Tensor const input,
-                                        int embed_dim,
-                                        int num_heads,
-                                        int num_hidden_layers,
-                                        int kdim = 0,
-                                        int vdim = 0,
-                                        float dropout = 0.0f,
-                                        bool bias = false,
-                                        bool add_bias_kv = false,
-                                        bool add_zero_attn = false,
-                                        DataType data_type = DT_NONE,
-                                        Initializer *kernel_initializer = NULL,
-                                        bool apply_rotary_embedding = false,
-                                        bool scaling_query = false,
-                                        float scaling_factor = 1.0f,
-                                        bool qk_prod_scaling = true,
-                                        bool position_bias = false,
-                                        bool streaming_cache = false,
-                                        char const *name = NULL);
-=======
   Tensor inc_multihead_self_attention(
       const Tensor input,
       int embed_dim,
       int num_heads,
+                                      int num_hidden_layers,
       int kdim = 0,
       int vdim = 0,
       float dropout = 0.0f,
@@ -773,6 +733,7 @@ class FFModel {
       const Tensor input,
       int embed_dim,
       int num_heads,
+                                        int num_hidden_layers,
       int kdim = 0,
       int vdim = 0,
       float dropout = 0.0f,
@@ -788,7 +749,6 @@ class FFModel {
       bool position_bias = false,
       bool streaming_cache = false,
       char const *name = NULL);
->>>>>>> origin/specscheduler
   Tensor inc_multihead_self_attention_verify(
       Tensor const input,
       int embed_dim,
@@ -808,12 +768,10 @@ class FFModel {
       bool qk_prod_scaling = true,
       bool position_bias = false,
       char const *name = NULL);
-<<<<<<< HEAD
   Tensor groupquery_self_attention(Tensor const input,
                                    int embed_dim,
                                    int num_q_heads,
                                    int num_kv_heads,
-                                   int num_hidden_layers,
                                    int kdim = 0,
                                    int vdim = 0,
                                    float dropout = 0.0f,
@@ -834,7 +792,6 @@ class FFModel {
                                          int embed_dim,
                                          int num_q_heads,
                                          int num_kv_heads,
-                                         int num_hidden_layers,
                                          int kdim = 0,
                                          int vdim = 0,
                                          float dropout = 0.0f,
@@ -850,48 +807,6 @@ class FFModel {
                                          bool position_bias = false,
                                          bool streaming_cache = false,
                                          char const *name = NULL);
-=======
-  Tensor groupquery_self_attention(
-      Tensor const input,
-      int embed_dim,
-      int num_q_heads,
-      int num_kv_heads,
-      int kdim = 0,
-      int vdim = 0,
-      float dropout = 0.0f,
-      bool bias = false,
-      bool add_bias_kv = false,
-      bool add_zero_attn = false,
-      DataType data_type = DT_NONE,
-      Initializer *kernel_initializer = NULL,
-      RotaryEmbeddingMeta rotary_embedding_meta = RotaryEmbeddingMeta(),
-      bool scaling_query = false,
-      float scaling_factor = 1.0f,
-      bool qk_prod_scaling = true,
-      bool position_bias = false,
-      bool streaming_cache = false,
-      char const *name = NULL);
-  Tensor spec_inc_multiquery_self_attention(
-      Tensor const input,
-      int embed_dim,
-      int num_q_heads,
-      int num_kv_heads,
-      int kdim = 0,
-      int vdim = 0,
-      float dropout = 0.0f,
-      bool bias = false,
-      bool add_bias_kv = false,
-      bool add_zero_attn = false,
-      DataType data_type = DT_NONE,
-      Initializer *kernel_initializer = NULL,
-      RotaryEmbeddingMeta rotary_embedding_meta = RotaryEmbeddingMeta(),
-      bool scaling_query = false,
-      float scaling_factor = 1.0f,
-      bool qk_prod_scaling = true,
-      bool position_bias = false,
-      bool streaming_cache = false,
-      char const *name = NULL);
->>>>>>> origin/specscheduler
   Tensor inc_multiquery_self_attention_verify(
       Tensor const input,
       int embed_dim,