llama : add 128k yarn context for Qwen (ggml-org#10698)

* add 128k yarn context for Qwen

* added property for model tensors

* removing useless line

Author: robbiemu

convert_hf_to_gguf.py

@@ -1992,6 +1992,14 @@ def set_vocab(self):
         except FileNotFoundError:
             self._set_vocab_gpt2()
 
+    def set_gguf_parameters(self):
+        super().set_gguf_parameters()
+        if self.hparams.get("rope_scaling") is not None and "factor" in self.hparams["rope_scaling"]:
+            if self.hparams["rope_scaling"].get("type") == "yarn":
+                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"])
+
 
 @Model.register("Qwen2MoeForCausalLM")
 class Qwen2MoeModel(Model):

gguf-py/gguf/constants.py

@@ -761,6 +761,7 @@ class MODEL_TENSOR(IntEnum):
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,
         MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ROPE_FREQS,
         MODEL_TENSOR.ATTN_NORM,
         MODEL_TENSOR.ATTN_Q,
         MODEL_TENSOR.ATTN_K,