Add bias to gptj (#1363)

Author: jiminha

optimum/habana/transformers/models/gptj/modeling_gptj.py

@@ -73,6 +73,14 @@ def __init__(self, config: GPTJConfig, layer_idx=None):
         super().__init__(config)
         self.config = config
 
+        max_positions = config.max_position_embeddings
+        self.register_buffer(
+            "bias",
+            torch.tril(torch.ones((max_positions, max_positions), dtype=torch.bool)).view(
+                1, 1, max_positions, max_positions
+            ),
+            persistent=False,
+        )
         self.matmul_qk = Matmul()
         self.matmul_av = Matmul()
         self.k_cache = KVCache()