update attn_impl for 70m and 1b models

1b-model/model_config.yml

@@ -8,7 +8,8 @@ n_heads: 16
 use_glu: false
 n_layers: 24
 attn_config:
-  attn_impl: triton
+  attn_impl: flash
+  use_attn_mask: false
   attn_type: grouped_query_attention
   kv_nheads: 16
   attn_pdrop: 0

70m-model/model_config.yml

@@ -8,7 +8,8 @@ n_heads: 8
 use_glu: false
 n_layers: 12
 attn_config:
-  attn_impl: triton
+  attn_impl: flash
+  use_attn_mask: false
   attn_type: grouped_query_attention
   kv_nheads: 8
   attn_pdrop: 0