diff --git a/src/maxtext/inference/vllm_decode.py b/src/maxtext/inference/vllm_decode.py index 5908ab2e2a..422a2be04f 100644 --- a/src/maxtext/inference/vllm_decode.py +++ b/src/maxtext/inference/vllm_decode.py @@ -93,7 +93,7 @@ def decode_with_vllm(config: Config) -> None: "additional_config": { "maxtext_config": { "model_name": config.model_name, - "weight_dtype": config.weight_dtype, + "weight_dtype": config.weight_dtype.name, "allow_split_physical_axes": True, "debug_sharding": config.debug_sharding, "prefuse_moe_weights": config.prefuse_moe_weights, @@ -250,7 +250,7 @@ def decode_with_tunix( rollout_vllm_additional_config = { "maxtext_config": { "model_name": config.model_name, - "weight_dtype": config.weight_dtype, + "weight_dtype": config.weight_dtype.name, "allow_split_physical_axes": True, "debug_sharding": config.debug_sharding, "prefuse_moe_weights": config.prefuse_moe_weights,