convert : bailingmoe : set yarn metadata if present (#13312)

diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py
index 838999531e58041976c8e612bc97b382874bf227..34bed7a08a1b28383cc2231ece8c03660d4b8229 100755 (executable)
--- a/convert_hf_to_gguf.py
+++ b/convert_hf_to_gguf.py
@@ -5680,7 +5680,12 @@ class BailingMoeModel(TextModel):
         rope_dim = hparams.get("head_dim") or hparams["hidden_size"] // hparams["num_attention_heads"]
 
         self.gguf_writer.add_rope_dimension_count(rope_dim)
-        self.gguf_writer.add_rope_scaling_type(gguf.RopeScalingType.NONE)
+        if (self.hparams.get("rope_scaling") or {}).get("type") == "yarn" and "factor" in self.hparams["rope_scaling"]:
+            self.gguf_writer.add_rope_scaling_type(gguf.RopeScalingType.YARN)
+            self.gguf_writer.add_rope_scaling_factor(self.hparams["rope_scaling"]["factor"])
+            self.gguf_writer.add_rope_scaling_orig_ctx_len(self.hparams["rope_scaling"]["original_max_position_embeddings"])
+        else:
+            self.gguf_writer.add_rope_scaling_type(gguf.RopeScalingType.NONE)
         self.gguf_writer.add_leading_dense_block_count(hparams["first_k_dense_replace"])
         self.gguf_writer.add_vocab_size(hparams["vocab_size"])
         self.gguf_writer.add_expert_feed_forward_length(hparams["moe_intermediate_size"])