Add support for a rope extension method (#6553)

vllm/config.py

@@ -151,6 +151,15 @@ def __init__(
         self.hf_text_config = get_hf_text_config(self.hf_config)
         self.dtype = _get_and_verify_dtype(self.hf_text_config, dtype)
 
+        if (getattr(self.hf_config, "max_position_embeddings", 0) == 131072
+                and getattr(self.hf_config, "rope_scaling", None) is None):
+            # Note(simon): this is a special case for a model that doesn't
+            # supply rope_scaling. We should remove this once the model is
+            # updated.
+            self.hf_config.update({"rope_scaling": {
+                "type": "extended",
+            }})
+
         if (not self.disable_sliding_window
                 and self.hf_text_config.model_type == "gemma2"
                 and self.hf_text_config.sliding_window is not None):
@@ -1442,8 +1451,9 @@ def _get_and_verify_max_len(
     rope_scaling = getattr(hf_config, "rope_scaling", None)
     # The correct one should be "longrope", kept "su" here
     # to be backward compatible
-    if rope_scaling is not None and rope_scaling["type"] != "su" \
-        and rope_scaling["type"] != "longrope":
+    if rope_scaling is not None and rope_scaling["type"] not in {
+            "su", "longrope", "extended"
+    }:
         if disable_sliding_window:
             # TODO(robertgshaw): Find a model that supports rope_scaling
             # with sliding window to see if this case should be allowed.

vllm/model_executor/layers/rotary_embedding.py

@@ -733,6 +733,36 @@ def _compute_inv_freq(self, base: Union[int, float]) -> torch.Tensor:
         return inv_freq
 
 
+class ExtendedRotaryEmbedding(RotaryEmbedding):
+
+    def _compute_inv_freq(self, base: Union[int, float]) -> torch.Tensor:
+        inv_freqs = super()._compute_inv_freq(base)
+        return self.apply_scaling(inv_freqs)
+
+    def apply_scaling(self, freqs: torch.Tensor):
+        scale_factor = 8
+        low_freq_factor = 1
+        high_freq_factor = 4
+        old_context_len = 8192
+
+        low_freq_wavelen = old_context_len / low_freq_factor
+        high_freq_wavelen = old_context_len / high_freq_factor
+        new_freqs = []
+        for freq in freqs:
+            wavelen = 2 * math.pi / freq
+            if wavelen < high_freq_wavelen:
+                new_freqs.append(freq)
+            elif wavelen > low_freq_wavelen:
+                new_freqs.append(freq / scale_factor)
+            else:
+                assert low_freq_wavelen != high_freq_wavelen
+                smooth = (old_context_len / wavelen - low_freq_factor) / (
+                    high_freq_factor - low_freq_factor)
+                new_freqs.append((1 - smooth) * freq / scale_factor +
+                                 smooth * freq)
+        return torch.tensor(new_freqs, dtype=freqs.dtype, device=freqs.device)
+
+
 _ROPE_DICT: Dict[Tuple, RotaryEmbedding] = {}
 
 
@@ -767,9 +797,13 @@ def get_rope(
         scaling_type = rope_scaling["type"]
         # The correct one should be "longrope" but keep "su" here
         # for backward compatible
-        if scaling_type != "su" and scaling_type != "longrope":
+        if scaling_type not in {"su", "longrope", "extended"}:
             scaling_factor = rope_scaling["factor"]
-        if scaling_type == "linear":
+        if scaling_type == "extended":
+            rotary_emb = ExtendedRotaryEmbedding(head_size, rotary_dim,
+                                                 max_position, base,
+                                                 is_neox_style, dtype)
+        elif scaling_type == "linear":
             rotary_emb = LinearScalingRotaryEmbedding(head_size, rotary_dim,
                                                       max_position, base,
                                                       is_neox_style,