Add recurrent gated delta rule custom op for Qwen3.5 attention

examples/models/llama/attention.py

@@ -1,3 +1,4 @@
+import logging
 from abc import ABC, abstractmethod
 from enum import Enum
 from typing import Any, Dict, Optional, Tuple, Type, TypedDict
@@ -52,6 +53,8 @@ def forward(
 
 
 ATTENTION_REGISTRY: Dict[str, Type[Attention]] = {}
+_RECURRENT_GATED_DELTA_RULE_OP = None
+_TRIED_LOADING_RECURRENT_GATED_DELTA_RULE_OP = False
 
 
 def register_attention(name: str):
@@ -64,6 +67,38 @@ def decorator(cls: Type[Attention]):
     return decorator
 
 
+def _get_recurrent_gated_delta_rule_op():
+    global _RECURRENT_GATED_DELTA_RULE_OP
+    global _TRIED_LOADING_RECURRENT_GATED_DELTA_RULE_OP
+
+    if _TRIED_LOADING_RECURRENT_GATED_DELTA_RULE_OP:
+        return _RECURRENT_GATED_DELTA_RULE_OP
+
+    _TRIED_LOADING_RECURRENT_GATED_DELTA_RULE_OP = True
+    try:
+        _RECURRENT_GATED_DELTA_RULE_OP = (
+            torch.ops.llama.recurrent_gated_delta_rule.default
+        )
+        return _RECURRENT_GATED_DELTA_RULE_OP
+    except (AttributeError, RuntimeError):
+        pass
+
+    try:
+        from executorch.extension.llm.custom_ops import custom_ops  # noqa: F401
+    except (ImportError, OSError, RuntimeError):
+        logging.debug("Failed to import custom ops library", exc_info=True)
+        return None
+
+    try:
+        _RECURRENT_GATED_DELTA_RULE_OP = (
+            torch.ops.llama.recurrent_gated_delta_rule.default
+        )
+    except (AttributeError, RuntimeError):
+        _RECURRENT_GATED_DELTA_RULE_OP = None
+
+    return _RECURRENT_GATED_DELTA_RULE_OP
+
+
 class KVCache(nn.Module):
     def __init__(
         self,
@@ -725,28 +760,43 @@ def _apply_causal_conv(self, mixed_qkv: torch.Tensor) -> torch.Tensor:
         out = F.silu(out[:, :, -seq_len:]).to(mixed_qkv.dtype)
         return out.transpose(1, 2).contiguous()
 
-    def _recurrent_gated_delta_rule(
+    def _gated_delta_rule_op(
         self,
         query: torch.Tensor,
         key: torch.Tensor,
         value: torch.Tensor,
         g: torch.Tensor,
         beta: torch.Tensor,
     ) -> torch.Tensor:
-        # query/key/value: (batch, seq_len, num_heads, head_dim)
-        # g/beta: (batch, seq_len, num_heads)
-        initial_dtype = query.dtype
-        query = _l2norm(query, dim=-1, eps=1e-6)
-        key = _l2norm(key, dim=-1, eps=1e-6)
-        query, key, value, beta, g = [
-            x.transpose(1, 2).contiguous().to(torch.float32)
-            for x in (query, key, value, beta, g)
-        ]
+        batch_size = query.shape[0]
+        recurrent_gated_delta_rule_op = _get_recurrent_gated_delta_rule_op()
+        if recurrent_gated_delta_rule_op is not None:
+            return recurrent_gated_delta_rule_op(
+                query,
+                key,
+                value,
+                g,
+                beta,
+                self.recurrent_state[:batch_size],
+            )
+        return self._naive_gated_delta_rule_op(
+            query,
+            key,
+            value,
+            g,
+            beta,
+        )
 
-        batch_size, num_heads, sequence_length, k_head_dim = key.shape
+    def _naive_gated_delta_rule_op(
+        self,
+        query: torch.Tensor,
+        key: torch.Tensor,
+        value: torch.Tensor,
+        g: torch.Tensor,
+        beta: torch.Tensor,
+    ) -> torch.Tensor:
+        batch_size, num_heads, sequence_length, _ = key.shape
         v_head_dim = value.shape[-1]
-        scale = 1.0 / (query.shape[-1] ** 0.5)
-        query = query * scale
 
         core_attn_out = torch.zeros(
             batch_size,
@@ -780,6 +830,36 @@ def _recurrent_gated_delta_rule(
                 last_recurrent_state.to(self.recurrent_state.dtype)
             )
 
+        return core_attn_out
+
+    def _recurrent_gated_delta_rule(
+        self,
+        query: torch.Tensor,
+        key: torch.Tensor,
+        value: torch.Tensor,
+        g: torch.Tensor,
+        beta: torch.Tensor,
+    ) -> torch.Tensor:
+        # query/key/value: (batch, seq_len, num_heads, head_dim)
+        # g/beta: (batch, seq_len, num_heads)
+        initial_dtype = query.dtype
+        query = _l2norm(query, dim=-1, eps=1e-6)
+        key = _l2norm(key, dim=-1, eps=1e-6)
+        query, key, value, beta, g = [
+            x.transpose(1, 2).contiguous().to(torch.float32)
+            for x in (query, key, value, beta, g)
+        ]
+
+        scale = 1.0 / (query.shape[-1] ** 0.5)
+        query = query * scale
+
+        core_attn_out = self._gated_delta_rule_op(
+            query,
+            key,
+            value,
+            g,
+            beta,
+        )
         return core_attn_out.transpose(1, 2).contiguous().to(initial_dtype)
 
     def forward(

examples/models/llama/tests/test_export_llama_lib.py

@@ -5,7 +5,10 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
+import json
+import tempfile
 import unittest
+from pathlib import Path
 
 from executorch.devtools.backend_debug import get_delegation_info
 
@@ -25,6 +28,7 @@
 
 from executorch.examples.models.llama.export_llama_lib import (
     _export_llama,
+    _prepare_for_llama_export,
     build_args_parser,
     get_quantizer_and_quant_params,
 )
@@ -37,6 +41,39 @@
 
 
 class ExportLlamaLibTest(unittest.TestCase):
+    def _make_tiny_qwen35_params(self) -> dict:
+        return {
+            "dim": 64,
+            "hidden_dim": 128,
+            "n_heads": 4,
+            "head_dim": 16,
+            "n_kv_heads": 2,
+            "n_layers": 4,
+            "norm_eps": 1e-6,
+            "rope_theta": 10000000.0,
+            "use_scaled_rope": False,
+            "vocab_size": 256,
+            "use_hf_rope": True,
+            "partial_rotary_factor": 0.25,
+            "attention_qkv_bias": False,
+            "use_qk_norm": True,
+            "qk_norm_before_rope": True,
+            "attention_type": "mha",
+            "use_q_gate": True,
+            "rms_norm_add_unit_offset": True,
+            "linear_conv_kernel_dim": 4,
+            "linear_key_head_dim": 8,
+            "linear_value_head_dim": 8,
+            "linear_num_key_heads": 4,
+            "linear_num_value_heads": 4,
+            "layer_types": [
+                "linear_attention",
+                "full_attention",
+                "linear_attention",
+                "full_attention",
+            ],
+        }
+
     def test_has_expected_ops_and_op_counts(self):
         """
         Checks the presence of unwanted expensive ops.
@@ -66,6 +103,41 @@ def test_has_expected_ops_and_op_counts(self):
         for op, _op_info in delegation_info.delegation_by_operator.items():
             self.assertTrue(op not in UNWANTED_OPS)
 
+    def test_tiny_qwen35_export_uses_recurrent_gated_delta_rule(self):
+        with tempfile.TemporaryDirectory() as temp_dir:
+            params_path = Path(temp_dir) / "tiny_qwen35.json"
+            params_path.write_text(json.dumps(self._make_tiny_qwen35_params()))
+
+            parser = build_args_parser()
+            args = parser.parse_args(
+                [
+                    "--model",
+                    "qwen3_5_0_8b",
+                    "--params",
+                    str(params_path),
+                    "--use_kv_cache",
+                    "--disable_dynamic_shape",
+                    "--max_seq_length",
+                    "8",
+                    "--max_context_length",
+                    "8",
+                ]
+            )
+
+            llm_config = LlmConfig.from_args(args)
+            builder = _prepare_for_llama_export(llm_config).export()
+            assert builder.pre_autograd_graph_module is not None
+
+            recurrent_nodes = [
+                node
+                for node in builder.pre_autograd_graph_module.graph.nodes
+                if "auto_functionalized_v2" in str(node.target)
+                and node.args
+                and "llama.recurrent_gated_delta_rule" in str(node.args[0])
+            ]
+
+            self.assertEqual(len(recurrent_nodes), 2)
+
     @unittest.skipUnless(HAS_ARM_BACKEND, "ARM backend not available")
     def test_get_quantizer_and_quant_params_returns_tosa_quantizer(self):
         llm_config = LlmConfig()

examples/models/llama/tests/test_qwen3_5_attention.py

@@ -6,7 +6,9 @@
 
 import unittest
 
+import executorch.examples.models.llama.attention as attention_module
 import torch
+
 from executorch.examples.models.llama.attention import ATTENTION_REGISTRY
 from executorch.examples.models.llama.model_args import ModelArgs
 from executorch.examples.models.llama.norm import RMSNorm
@@ -123,6 +125,109 @@ def test_gated_deltanet_no_input_pos_does_not_leak_state(self):
             torch.allclose(state_after_first, state_after_second, atol=1e-5)
         )
 
+    def test_gated_deltanet_chunked_prefill_matches_full_sequence(self):
+        torch.manual_seed(0)
+        args = self._make_args(
+            use_kv_cache=True,
+            use_q_gate=True,
+            linear_conv_kernel_dim=4,
+            linear_key_head_dim=4,
+            linear_value_head_dim=4,
+            linear_num_key_heads=2,
+            linear_num_value_heads=4,
+        )
+        rope = Rope(args)
+        attn_full = ATTENTION_REGISTRY["gated_deltanet"](args, 0, rope)
+        attn_chunked = ATTENTION_REGISTRY["gated_deltanet"](args, 0, rope)
+        attn_chunked.load_state_dict(attn_full.state_dict())
+
+        x = torch.randn(1, 5, args.dim)
+        dummy_freq = torch.zeros(1, 1)
+
+        full_output, _ = attn_full(
+            x,
+            dummy_freq,
+            dummy_freq,
+            input_pos=torch.tensor([0], dtype=torch.long),
+        )
+
+        chunk_outputs = []
+        for start, end in ((0, 3), (3, 4), (4, 5)):
+            output, _ = attn_chunked(
+                x[:, start:end],
+                dummy_freq,
+                dummy_freq,
+                input_pos=torch.tensor([start], dtype=torch.long),
+            )
+            chunk_outputs.append(output)
+
+        chunked_output = torch.cat(chunk_outputs, dim=1)
+
+        self.assertTrue(torch.allclose(chunked_output, full_output, atol=1e-5))
+        self.assertTrue(
+            torch.allclose(
+                attn_chunked.recurrent_state, attn_full.recurrent_state, atol=1e-5
+            )
+        )
+        self.assertTrue(
+            torch.allclose(attn_chunked.conv_state, attn_full.conv_state, atol=1e-5)
+        )
+
+    def test_gated_deltanet_custom_op_matches_fallback(self):
+        recurrent_op = attention_module._get_recurrent_gated_delta_rule_op()
+        if recurrent_op is None:
+            self.skipTest("llama::recurrent_gated_delta_rule is not available")
+
+        torch.manual_seed(0)
+        args = self._make_args(
+            use_kv_cache=True,
+            use_q_gate=True,
+            linear_conv_kernel_dim=4,
+            linear_key_head_dim=4,
+            linear_value_head_dim=4,
+            linear_num_key_heads=2,
+            linear_num_value_heads=4,
+        )
+        rope = Rope(args)
+        attn_custom = ATTENTION_REGISTRY["gated_deltanet"](args, 0, rope)
+        attn_fallback = ATTENTION_REGISTRY["gated_deltanet"](args, 0, rope)
+        attn_fallback.load_state_dict(attn_custom.state_dict())
+
+        query = torch.randn(1, 3, attn_custom.num_v_heads, attn_custom.head_k_dim)
+        key = torch.randn(1, 3, attn_custom.num_v_heads, attn_custom.head_k_dim)
+        value = torch.randn(1, 3, attn_custom.num_v_heads, attn_custom.head_v_dim)
+        g = torch.randn(1, 3, attn_custom.num_v_heads)
+        beta = torch.sigmoid(torch.randn(1, 3, attn_custom.num_v_heads))
+
+        original_op = attention_module._RECURRENT_GATED_DELTA_RULE_OP
+        original_tried_loading = (
+            attention_module._TRIED_LOADING_RECURRENT_GATED_DELTA_RULE_OP
+        )
+        try:
+            attention_module._RECURRENT_GATED_DELTA_RULE_OP = recurrent_op
+            attention_module._TRIED_LOADING_RECURRENT_GATED_DELTA_RULE_OP = True
+            custom_output = attn_custom._recurrent_gated_delta_rule(
+                query, key, value, g, beta
+            )
+
+            attention_module._RECURRENT_GATED_DELTA_RULE_OP = None
+            attention_module._TRIED_LOADING_RECURRENT_GATED_DELTA_RULE_OP = True
+            fallback_output = attn_fallback._recurrent_gated_delta_rule(
+                query, key, value, g, beta
+            )
+        finally:
+            attention_module._RECURRENT_GATED_DELTA_RULE_OP = original_op
+            attention_module._TRIED_LOADING_RECURRENT_GATED_DELTA_RULE_OP = (
+                original_tried_loading
+            )
+
+        self.assertTrue(torch.allclose(custom_output, fallback_output, atol=1e-5))
+        self.assertTrue(
+            torch.allclose(
+                attn_custom.recurrent_state, attn_fallback.recurrent_state, atol=1e-5
+            )
+        )
+
 
 if __name__ == "__main__":
     unittest.main()

extension/aten_util/make_aten_functor_from_et_functor.h

@@ -15,7 +15,8 @@
 #pragma once
 #include <type_traits>
 #include <vector>
-#if __cplusplus < 201703L
+#if (defined(_MSC_VER) && (!defined(_MSVC_LANG) || _MSVC_LANG < 201703L)) || \
+    (!defined(_MSC_VER) && __cplusplus < 201703L)
 #error "This header requires C++17"
 #endif
 #include <ATen/native/Resize.h>