v2 loader: VOCAB 8192 -> 12288

load_model.py

@@ -14,7 +14,7 @@ import torch.nn as nn
 import torch.nn.functional as F
 from tokenizers import Tokenizer
 
-D, L, H, FFN, VOCAB, SEQ = 512, 16, 8, 2048, 8192, 512
+D, L, H, FFN, VOCAB, SEQ = 512, 16, 8, 2048, 12288, 512
 
 
 class RMSNorm(nn.Module):
@@ -93,4 +93,4 @@ if __name__ == "__main__":
     ids = tok.encode("Once upon a time,")
     ids = torch.tensor([ids], device="cuda")
     out = m.generate(ids, 100, temp=0.8, top_k=40)
-    print(tok.decode(out[0].tolist(), skip_special_tokens=True))
+    print(tok.decode(out[0].tolist(), skip_special_tokens=True))