Sync configuration_auristream.py from TuKoResearch/AuriStream100M_0Pred_librilight_200k

configuration_auristream.py

@@ -0,0 +1,26 @@
+from transformers import PretrainedConfig
+
+class AuriStreamConfig(PretrainedConfig):
+    model_type = "AuriStream.AuriStream"
+
+    def __init__(
+        self,
+        n_layer=48,
+        n_head=16,
+        n_embd=1280,
+        vocab_size=8192,
+        dropout=0.0,
+        bias=False,
+        use_rope=False,
+        seq_len=4096,
+        **kwargs
+    ):
+        self.n_layer = n_layer
+        self.n_head = n_head
+        self.n_embd = n_embd
+        self.vocab_size = vocab_size
+        self.dropout = dropout
+        self.bias = bias
+        self.use_rope = use_rope
+        self.seq_len = seq_len
+        super().__init__(**kwargs)