examples : fix a hyperparameter value in gpt-neox (#161) (#162)

diff --git a/examples/gpt-neox/convert-h5-to-ggml.py b/examples/gpt-neox/convert-h5-to-ggml.py
index 0019810e28e1ff2a7ca7ad2795e4fb1e2eb41a1a..9d21226f536e65fd4e2acac65a701dd50c05c625 100644 (file)
--- a/examples/gpt-neox/convert-h5-to-ggml.py
+++ b/examples/gpt-neox/convert-h5-to-ggml.py
@@ -58,7 +58,7 @@ fout.write(struct.pack("i", hparams["hidden_size"]))
 fout.write(struct.pack("i", hparams["num_attention_heads"]))
 fout.write(struct.pack("i", hparams["num_hidden_layers"]))
 fout.write(struct.pack("i", int(hparams["rotary_pct"]*(hparams["hidden_size"]//hparams["num_attention_heads"]))))
-fout.write(struct.pack("i", hparams["use_parallel_residual"]))
+fout.write(struct.pack("i", hparams["use_parallel_residual"] if "use_parallel_residual" in hparams else True))
 fout.write(struct.pack("i", ftype))
 
 # TODO: temporary hack to not deal with implementing the tokenizer