Add yarn option for qwen2

llms/mlx_lm/models/qwen2.py

@@ -7,6 +7,7 @@
 import mlx.nn as nn
 
 from .base import BaseModelArgs, create_attention_mask, scaled_dot_product_attention
+from .rope_utils import initialize_rope
 
 
 @dataclass
@@ -18,24 +19,13 @@ class ModelArgs(BaseModelArgs):
     num_attention_heads: int
     rms_norm_eps: float
     vocab_size: int
-    num_key_value_heads: Optional[int] = None
+    num_key_value_heads: int
+    max_position_embeddings: int = 32768
     rope_theta: float = 1000000
     rope_traditional: bool = False
     rope_scaling: Optional[Dict[str, Union[float, str]]] = None
     tie_word_embeddings: bool = True
 
-    def __post_init__(self):
-        if self.num_key_value_heads is None:
-            self.num_key_value_heads = self.num_attention_heads
-
-        if self.rope_scaling:
-            required_keys = {"factor", "type"}
-            if not all(key in self.rope_scaling for key in required_keys):
-                raise ValueError(f"rope_scaling must contain keys {required_keys}")
-
-            if self.rope_scaling["type"] != "linear":
-                raise ValueError("rope_scaling 'type' currently only supports 'linear'")
-
 
 class Attention(nn.Module):
     def __init__(self, args: ModelArgs):
@@ -54,16 +44,12 @@ def __init__(self, args: ModelArgs):
         self.v_proj = nn.Linear(dim, n_kv_heads * head_dim, bias=True)
         self.o_proj = nn.Linear(n_heads * head_dim, dim, bias=False)
 
-        rope_scale = (
-            1 / args.rope_scaling["factor"]
-            if args.rope_scaling is not None and args.rope_scaling["type"] == "linear"
-            else 1
-        )
-        self.rope = nn.RoPE(
+        self.rope = initialize_rope(
             head_dim,
-            traditional=args.rope_traditional,
             base=args.rope_theta,
-            scale=rope_scale,
+            traditional=args.rope_traditional,
+            scaling_config=args.rope_scaling,
+            max_position_embeddings=args.max_position_embeddings,
         )
 
     def __call__(

llms/mlx_lm/models/rope_utils.py

@@ -1,5 +1,6 @@
 # Copyright © 2023-2024 Apple Inc.
 
+import math
 from typing import Optional
 
 import mlx.core as mx
@@ -61,6 +62,78 @@ def __call__(self, x, offset: int = 0):
         )
 
 
+class YarnRoPE(nn.Module):
+    def __init__(
+        self,
+        dims,
+        traditional=False,
+        max_position_embeddings=2048,
+        base=10000,
+        scaling_factor=1.0,
+        original_max_position_embeddings=4096,
+        beta_fast=32,
+        beta_slow=1,
+        mscale=1,
+        mscale_all_dim=0,
+    ):
+        super().__init__()
+
+        def yarn_find_correction_dim(num_rotations):
+            return (
+                dims
+                * math.log(
+                    original_max_position_embeddings / (num_rotations * 2 * math.pi)
+                )
+            ) / (2 * math.log(base))
+
+        def yarn_find_correction_range():
+            low = math.floor(yarn_find_correction_dim(beta_fast))
+            high = math.ceil(yarn_find_correction_dim(beta_slow))
+            return max(low, 0), min(high, dims - 1)
+
+        def yarn_get_mscale(scale=1, mscale=1):
+            if scale <= 1:
+                return 1.0
+            return 0.1 * mscale * math.log(scale) + 1.0
+
+        def yarn_linear_ramp_mask(min_val, max_val, dim):
+            if min_val == max_val:
+                max_val += 0.001  # Prevent singularity
+
+            linear_func = (mx.arange(dim, dtype=mx.float32) - min_val) / (
+                max_val - min_val
+            )
+            return mx.clip(linear_func, 0, 1)
+
+        self.mscale = yarn_get_mscale(scaling_factor, mscale) / yarn_get_mscale(
+            scaling_factor, mscale_all_dim
+        )
+        freq_extra = base ** (mx.arange(0, dims, 2, dtype=mx.float32) / dims)
+        freq_inter = scaling_factor * base ** (
+            mx.arange(0, dims, 2, dtype=mx.float32) / dims
+        )
+        low, high = yarn_find_correction_range()
+        freq_mask = 1.0 - yarn_linear_ramp_mask(low, high, dims // 2)
+        self._freqs = (freq_inter * freq_extra) / (
+            freq_inter * freq_mask + freq_extra * (1 - freq_mask)
+        )
+        self.dims = dims
+        self.traditional = traditional
+
+    def __call__(self, x, offset=0):
+        if self.mscale != 1.0:
+            x[..., : self.dims] = self.mscale * x[..., : self.dims]
+        return mx.fast.rope(
+            x,
+            self.dims,
+            traditional=self.traditional,
+            base=None,
+            scale=1.0,
+            offset=offset,
+            freqs=self._freqs,
+        )
+
+
 def initialize_rope(
     dims,
     base,
@@ -87,5 +160,25 @@ def initialize_rope(
             base=base,
             scaling_config=scaling_config,
         )
+    elif rope_type == "yarn":
+        scaling_factor = scaling_config["factor"]
+        rope_kwargs = {
+            key: scaling_config[key]
+            for key in [
+                "original_max_position_embeddings",
+                "beta_fast",
+                "beta_slow",
+                "mscale",
+                "mscale_all_dim",
+            ]
+            if key in scaling_config
+        }
+        return YarnRoPE(
+            dims=dims,
+            max_position_embeddings=max_position_embeddings,
+            traditional=traditional,
+            base=base,
+            **rope_kwargs,
+        )
     else:
         raise ValueError(f"Unsupported RoPE type {rope_type}")

llms/tests/test_models.py

@@ -336,6 +336,7 @@ def test_qwen2(self):
             num_hidden_layers=4,
             intermediate_size=2048,
             num_attention_heads=4,
+            num_key_value_heads=4,
             rms_norm_eps=1e-5,
             vocab_size=10_000,
         )