Fixing Transformers config

The Yarn scaling is not applied when mscale_all_dim is set to 1. This means that rope values were incorrect leading to worse performanc especially for long context. This should fix the issue.

config.json

@@ -41,7 +41,7 @@
       "factor": 64.0,
       "llama_4_scaling_beta": 0,
       "mscale": 1.0,
-      "mscale_all_dim": 1.0,
+      "mscale_all_dim": 0.0,
       "original_max_position_embeddings": 4096,
       "rope_theta": 1000000.0,
       "rope_type": "yarn",