fixup falcon with new ggml

mudler · mudler · commit 3fec197a1dc4 · 2023-06-30T22:42:11.000+02:00
diff --git a/implementations/falcon.cpp b/implementations/falcon.cpp
@@ -461,8 +461,8 @@ bool falcon_eval(
                 fused_qkv_row_nb, (n_embd + head_dim) * sizeof(float));
 
             // using mode = 2 for neox mode
-            Qcur = ggml_rope_inplace(ctx0, Qcur, n_past, head_dim, 2);
-            Kcur = ggml_rope_inplace(ctx0, Kcur, n_past, head_dim, 2);
+            Qcur = ggml_rope_inplace(ctx0, Qcur, n_past, head_dim, 2, 0);
+            Kcur = ggml_rope_inplace(ctx0, Kcur, n_past, head_dim, 2, 0 );
 
             // store key and value to memory
             {

Original file line number	Diff line number	Diff line change
`@@ -461,8 +461,8 @@ bool falcon_eval(`
`461`	`461`	`fused_qkv_row_nb, (n_embd + head_dim) * sizeof(float));`
`462`	`462`
`463`	`463`	`// using mode = 2 for neox mode`
`464`		`- Qcur = ggml_rope_inplace(ctx0, Qcur, n_past, head_dim, 2);`
`465`		`- Kcur = ggml_rope_inplace(ctx0, Kcur, n_past, head_dim, 2);`
	`464`	`+ Qcur = ggml_rope_inplace(ctx0, Qcur, n_past, head_dim, 2, 0);`
	`465`	`+ Kcur = ggml_rope_inplace(ctx0, Kcur, n_past, head_dim, 2, 0 );`
`466`	`466`
`467`	`467`	`// store key and value to memory`
`468`	`468`	`{`