Change flash attention outputs to be SymInt instead of int (#110533)

Fixes #110322

Signed-off-by: Edward Z. Yang <ezyang@meta.com>

Pull Request resolved: #110533
Approved by: https://github.com/albanD

aten/src/ATen/native/native_functions.yaml

@@ -14349,14 +14349,14 @@
   variants: function
   tags: nondeterministic_seeded
 
-- func: _scaled_dot_product_flash_attention(Tensor query, Tensor key, Tensor value, float dropout_p=0.0, bool is_causal=False, bool return_debug_mask=False, *, float? scale=None) -> (Tensor output, Tensor logsumexp, Tensor cum_seq_q, Tensor cum_seq_k, int max_q, int max_k, Tensor philox_seed, Tensor philox_offset, Tensor debug_attn_mask)
+- func: _scaled_dot_product_flash_attention(Tensor query, Tensor key, Tensor value, float dropout_p=0.0, bool is_causal=False, bool return_debug_mask=False, *, float? scale=None) -> (Tensor output, Tensor logsumexp, Tensor cum_seq_q, Tensor cum_seq_k, SymInt max_q, SymInt max_k, Tensor philox_seed, Tensor philox_offset, Tensor debug_attn_mask)
   dispatch:
     CPU: _scaled_dot_product_flash_attention_cpu
     CUDA: _scaled_dot_product_flash_attention_cuda
     NestedTensorCUDA: _scaled_dot_product_flash_attention_nestedtensor_cuda
   tags: nondeterministic_seeded
 
-- func: _scaled_dot_product_flash_attention_backward(Tensor grad_out, Tensor query, Tensor key, Tensor value, Tensor out, Tensor logsumexp, Tensor cum_seq_q, Tensor cum_seq_k, int max_q, int max_k, float dropout_p, bool is_causal, Tensor philox_seed, Tensor philox_offset, *, float? scale=None) -> (Tensor grad_query, Tensor grad_key, Tensor grad_value)
+- func: _scaled_dot_product_flash_attention_backward(Tensor grad_out, Tensor query, Tensor key, Tensor value, Tensor out, Tensor logsumexp, Tensor cum_seq_q, Tensor cum_seq_k, SymInt max_q, SymInt max_k, float dropout_p, bool is_causal, Tensor philox_seed, Tensor philox_offset, *, float? scale=None) -> (Tensor grad_query, Tensor grad_key, Tensor grad_value)
   device_check: NoCheck
   variants: function
   dispatch:
@@ -14375,13 +14375,13 @@
     CUDA: _scaled_dot_product_efficient_attention_backward_cuda
   tags: nondeterministic_seeded
 
-- func: _flash_attention_forward(Tensor query, Tensor key, Tensor value, Tensor? cum_seq_q, Tensor? cum_seq_k, int? max_q, int? max_k, float dropout_p, bool is_causal, bool return_debug_mask, *, float? scale=None) -> (Tensor output, Tensor softmax_logsumexp, Tensor philox_seed, Tensor philox_offset, Tensor debug_attn_mask)
+- func: _flash_attention_forward(Tensor query, Tensor key, Tensor value, Tensor? cum_seq_q, Tensor? cum_seq_k, SymInt? max_q, SymInt? max_k, float dropout_p, bool is_causal, bool return_debug_mask, *, float? scale=None) -> (Tensor output, Tensor softmax_logsumexp, Tensor philox_seed, Tensor philox_offset, Tensor debug_attn_mask)
   variants: function
   dispatch:
     CUDA: _flash_attention_forward
   tags: nondeterministic_seeded
 
-- func: _flash_attention_backward(Tensor grad_out, Tensor query, Tensor key, Tensor value, Tensor out, Tensor logsumexp, Tensor cum_seq_q, Tensor cum_seq_k, int max_q, int max_k, float dropout_p, bool is_causal, Tensor philox_seed, Tensor philox_offset, *, float? scale=None) -> (Tensor, Tensor, Tensor)
+- func: _flash_attention_backward(Tensor grad_out, Tensor query, Tensor key, Tensor value, Tensor out, Tensor logsumexp, Tensor cum_seq_q, Tensor cum_seq_k, SymInt max_q, SymInt max_k, float dropout_p, bool is_causal, Tensor philox_seed, Tensor philox_offset, *, float? scale=None) -> (Tensor, Tensor, Tensor)
   device_check: NoCheck
   variants: function
   dispatch:

aten/src/ATen/native/nested/cuda/NestedTensorTransformerFunctions.cpp

@@ -220,8 +220,8 @@ std::tuple<
     Tensor,
     Tensor,
     Tensor,
-    int64_t,
-    int64_t,
+    c10::SymInt,
+    c10::SymInt,
     Tensor,
     Tensor,
     Tensor>

aten/src/ATen/native/transformers/attention.cpp

@@ -744,8 +744,8 @@ std::tuple<
     at::Tensor,
     at::Tensor,
     at::Tensor,
-    int64_t,
-    int64_t,
+    c10::SymInt,
+    c10::SymInt,
     at::Tensor,
     at::Tensor,
     at::Tensor>

aten/src/ATen/native/transformers/cuda/attention.cu

@@ -668,7 +668,7 @@ std::tuple<Tensor, Tensor> native_multi_head_attention_cuda(
   }
   return std::make_tuple(std::move(proj), std::move(qkt));
 }
-std::tuple<Tensor, Tensor, Tensor, Tensor, int64_t, int64_t, Tensor, Tensor, Tensor> _scaled_dot_product_flash_attention_cuda(
+std::tuple<Tensor, Tensor, Tensor, Tensor, c10::SymInt, c10::SymInt, Tensor, Tensor, Tensor> _scaled_dot_product_flash_attention_cuda(
     const Tensor& query,
     const Tensor& key,
     const Tensor& value,

test/inductor/test_cuda_repro.py

@@ -5,13 +5,16 @@
 
 import torch
 import torch._dynamo.config as dynamo_config
+import torch.backends.cuda
+import torch.nn.functional as F
 from torch import nn
 from torch._dynamo.debug_utils import same_two_models
 from torch._dynamo.testing import rand_strided
 from torch._dynamo.utils import same
 from torch._inductor import config
 from torch._inductor.compile_fx import compile_fx_inner
 from torch.fx.experimental.proxy_tensor import make_fx
+from torch.testing._internal.common_cuda import PLATFORM_SUPPORTS_FLASH_ATTENTION
 from torch.testing._internal.common_utils import (
     DeterministicGuard,
     freeze_rng_state,
@@ -982,6 +985,51 @@ def fn(x, y, z):
 
         self.assertEqual(ref, res)
 
+    @unittest.skipIf(
+        not PLATFORM_SUPPORTS_FLASH_ATTENTION, "flash attention not supported"
+    )
+    def test_flash_attention_dynamic(self):
+        class Model(nn.Module):
+            def __init__(self, *args, **kwargs) -> None:
+                super().__init__(*args, **kwargs)
+
+                self.q = nn.Linear(1024, 1024)
+                self.k = nn.Linear(1024, 1024)
+                self.v = nn.Linear(1024, 1024)
+
+            def forward(self, x):
+                batch_size, seq_len, _ = x.size()
+
+                queries = self.q(x).view(batch_size, seq_len, 8, 128).transpose(2, 1)
+                keys = self.k(x).view(batch_size, seq_len, 8, 128).transpose(2, 1)
+                values = self.v(x).view(batch_size, seq_len, 8, 128).transpose(2, 1)
+
+                attn = F.scaled_dot_product_attention(
+                    queries,
+                    keys,
+                    values,
+                )
+
+                return attn
+
+        cnts = torch._dynamo.testing.CompileCounterWithBackend("inductor")
+
+        model = Model().cuda().half()
+        model = torch.compile(model, backend=cnts, dynamic=True)
+
+        with torch.backends.cuda.sdp_kernel(
+            enable_flash=True, enable_math=False, enable_mem_efficient=False
+        ):
+            input1 = torch.rand(5, 512, 1024, device="cuda", dtype=torch.float16)
+            input2 = torch.rand(5, 513, 1024, device="cuda", dtype=torch.float16)
+            input3 = torch.rand(5, 514, 1024, device="cuda", dtype=torch.float16)
+
+            out1 = model(input1)
+            out2 = model(input2)
+            out3 = model(input3)
+
+        self.assertEqual(cnts.frame_count, 1)
+
     @config.patch({"triton.cudagraphs": True})
     def test_index_put_no_fallback_cudagraph(self):
         def fn(x, y, z):

tools/autograd/derivatives.yaml

@@ -2764,9 +2764,9 @@
   output_differentiability: [True, False, False, False]
   query, key, value, attn_bias: _scaled_dot_product_efficient_attention_backward(grad, query, key, value, attn_bias, output, log_sumexp, philox_seed, philox_offset, dropout_p, grad_input_mask, is_causal, scale)
 
-- name: _scaled_dot_product_flash_attention(Tensor query, Tensor key, Tensor value, float dropout_p=0.0, bool is_causal=False, bool return_debug_mask=False, *, float? scale=None) -> (Tensor output, Tensor logsumexp, Tensor cum_seq_q, Tensor cum_seq_k, int max_q, int max_k, Tensor philox_seed, Tensor philox_offset, Tensor debug_attn_mask)
+- name: _scaled_dot_product_flash_attention(Tensor query, Tensor key, Tensor value, float dropout_p=0.0, bool is_causal=False, bool return_debug_mask=False, *, float? scale=None) -> (Tensor output, Tensor logsumexp, Tensor cum_seq_q, Tensor cum_seq_k, SymInt max_q, SymInt max_k, Tensor philox_seed, Tensor philox_offset, Tensor debug_attn_mask)
   output_differentiability: [True, False, False, False, False, False, False, False, False]
-  query, key, value: _scaled_dot_product_flash_attention_backward(grad, query, key, value, output, logsumexp, cum_seq_q, cum_seq_k, max_q, max_k, dropout_p, is_causal, philox_seed, philox_offset, scale)
+  query, key, value: _scaled_dot_product_flash_attention_backward_symint(grad, query, key, value, output, logsumexp, cum_seq_q, cum_seq_k, max_q, max_k, dropout_p, is_causal, philox_seed, philox_offset, scale)
 
 # - name: _flash_attention_forward(Tensor query, Tensor key, Tensor value, Tensor? cum_seq_q, Tensor? cum_seq_k, int? max_q, int? max_k, float dropout_p, bool is_causal, bool return_debug_mask, *, float? scale=None) -> (Tensor output, Tensor query_padded, Tensor key_padded, Tensor value_padded, Tensor softmax_logsumexp, Tensor philox_seed, Tensor philox_offset, Tensor debug_attn_mask)
 #   output_differentiability: [True, False, False, False, False, False, False, False]

torch/_C/return_types.pyi.in

@@ -16,7 +16,7 @@ from typing import (
     Union,
 )
 
-from torch import contiguous_format, Generator, inf, memory_format, strided, Tensor
+from torch import contiguous_format, Generator, inf, memory_format, strided, Tensor, SymInt
 from torch.types import (
     _bool,
     _device,

torch/_inductor/ir.py

@@ -3984,6 +3984,8 @@ def generate_output(output, indices):
                 )
             elif isinstance(output, int):
                 return output
+            elif isinstance(output, torch.SymInt):
+                return output.node.expr
             else:
                 assert (
                     output is None

torch/csrc/inductor/aoti_torch/shim_common.cpp

@@ -228,8 +228,8 @@ AOTITorchError aoti_torch__scaled_dot_product_flash_attention(
       at::Tensor* ret3_tensor = new at::Tensor(std::move(r3));
       *ret3 = tensor_pointer_to_tensor_handle(ret3_tensor);
     }
-    *ret4 = r4;
-    *ret5 = r5;
+    *ret4 = r4.expect_int();
+    *ret5 = r5.expect_int();
     at::Tensor* ret6_tensor = new at::Tensor(std::move(r6));
     *ret6 = tensor_pointer_to_tensor_handle(ret6_tensor);
     at::Tensor* ret7_tensor = new at::Tensor(std::move(r7));

torchgen/api/python.py

@@ -1129,7 +1129,7 @@ def dispatch_lambda_arg(cpp_arg: Binding) -> DispatchLambdaArgument:
     "::std::tuple<at::Tensor,::std::vector<at::Tensor>>",
     "::std::vector<at::Tensor>",
     # Needed for flash attention forw/backward
-    "::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor,int64_t,int64_t,at::Tensor,at::Tensor,at::Tensor>",
+    "::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor,c10::SymInt,c10::SymInt,at::Tensor,at::Tensor,at::Tensor>",
     "at::Scalar",
     "bool",
     "int64_t",