ggml : support bcast ggml_soft_max_ext, ggml_flash_attn_ext (ggml-org#14435)

ggml-ci

Author: ggerganov

ggml/include/ggml.h

@@ -557,8 +557,6 @@ extern "C" {
         GGML_GLU_OP_REGLU,
         GGML_GLU_OP_GEGLU,
         GGML_GLU_OP_SWIGLU,
-        GGML_GLU_OP_GEGLU_ERF,
-        GGML_GLU_OP_GEGLU_QUICK,
 
         GGML_GLU_OP_COUNT,
     };
@@ -648,9 +646,6 @@ extern "C" {
 
     // misc
 
-    GGML_API const char * ggml_version(void);
-    GGML_API const char * ggml_commit(void);
-
     GGML_API void    ggml_time_init(void); // call this once at the beginning of the program
     GGML_API int64_t ggml_time_ms(void);
     GGML_API int64_t ggml_time_us(void);
@@ -1149,22 +1144,6 @@ extern "C" {
             struct ggml_context * ctx,
             struct ggml_tensor  * a);
 
-    GGML_API struct ggml_tensor * ggml_geglu_erf(
-            struct ggml_context * ctx,
-            struct ggml_tensor  * a);
-
-    GGML_API struct ggml_tensor * ggml_geglu_erf_swapped(
-            struct ggml_context * ctx,
-            struct ggml_tensor  * a);
-
-    GGML_API struct ggml_tensor * ggml_geglu_quick(
-            struct ggml_context * ctx,
-            struct ggml_tensor  * a);
-
-    GGML_API struct ggml_tensor * ggml_geglu_quick_swapped(
-            struct ggml_context * ctx,
-            struct ggml_tensor  * a);
-
     // A: n columns, r rows,
     // B: n columns, r rows,
     GGML_API struct ggml_tensor * ggml_glu_split(
@@ -1188,16 +1167,6 @@ extern "C" {
             struct ggml_tensor  * a,
             struct ggml_tensor  * b);
 
-    GGML_API struct ggml_tensor * ggml_geglu_erf_split(
-            struct ggml_context * ctx,
-            struct ggml_tensor  * a,
-            struct ggml_tensor  * b);
-
-    GGML_API struct ggml_tensor * ggml_geglu_quick_split(
-            struct ggml_context * ctx,
-            struct ggml_tensor  * a,
-            struct ggml_tensor  * b);
-
     // normalize along rows
     GGML_API struct ggml_tensor * ggml_norm(
             struct ggml_context * ctx,
@@ -2011,16 +1980,15 @@ extern "C" {
 
 #define GGML_KQ_MASK_PAD 64
 
-    // q:    [n_embd_k, n_batch,     n_head,    ne3 ]
-    // k:    [n_embd_k, n_kv,        n_head_kv, ne3 ]
-    // v:    [n_embd_v, n_kv,        n_head_kv, ne3 ] !! not transposed !!
-    // mask: [n_kv,     n_batch_pad, ne32,      ne33] !! n_batch_pad = GGML_PAD(n_batch, GGML_KQ_MASK_PAD) !!
-    // res:  [n_embd_v, n_head,      n_batch,   ne3 ] !! permuted !!
+    // q:    [n_embd_k, n_batch,     n_head,    ne3]
+    // k:    [n_embd_k, n_kv,        n_head_kv, ne3]
+    // v:    [n_embd_v, n_kv,        n_head_kv, ne3] !! not transposed !!
+    // mask: [n_kv,     n_batch_pad, ne32,      1] !! n_batch_pad = GGML_PAD(n_batch, GGML_KQ_MASK_PAD) !!
+    // res:  [n_embd_v, n_head,      n_batch,   ne3] !! permuted !!
     //
     // broadcast:
     //   n_head % n_head_kv == 0
-    //   n_head % ne32      == 0
-    //   ne3    % ne33      == 0
+    //   ne3    % ne32      == 0
     //
     GGML_API struct ggml_tensor * ggml_flash_attn_ext(
             struct ggml_context * ctx,
@@ -2060,8 +2028,7 @@ extern "C" {
             struct ggml_tensor  * dt,
             struct ggml_tensor  * A,
             struct ggml_tensor  * B,
-            struct ggml_tensor  * C,
-            struct ggml_tensor  * ids);
+            struct ggml_tensor  * C);
 
     // partition into non-overlapping windows with padding if needed
     // example: