Bump max_position_embeddings to 256K; double yarn rope factor to 64

config.json

@@ -14,7 +14,7 @@
   "num_attention_heads": 64,
   "num_key_value_heads": 8,
   "head_dim": 128,
-  "max_position_embeddings": 131072,
+  "max_position_embeddings": 262144,
   "attention_bias": false,
   "attention_dropout": 0.0,
   "rms_norm_eps": 1e-06,
@@ -38,7 +38,7 @@
     "full_attention": {
       "rope_theta": 500000.0,
       "rope_type": "yarn",
-      "factor": 32.0,
+      "factor": 64.0,
       "original_max_position_embeddings": 4096,
       "beta_slow": 1.0,
       "beta_fast": 64.0,
@@ -516,4 +516,4 @@
     "transform_config": {},
     "version": "0.11.0"
   }
-}
+}