wip

test/test_pallas_spmd.py

@@ -22,8 +22,14 @@
 
 class PallasTest(unittest.TestCase):
 
-  def _attention(self, q, k, v):
+  def _attention(self, q, k, v, *, attn_mask=None, ab=None):
     attn_weight = q @ k.transpose(-2, -1)
+    if attn_mask is not None:
+      # Masked out the unrelevant parts.
+      attn_weight = attn_weight.masked_fill(attn_mask,
+                                            torch.finfo(attn_weight.dtype).min)
+    if ab is not None:
+      attn_weight = attn_weight + ab
     attn_weight = nn.functional.softmax(attn_weight, dim=-1)
     attn_output = attn_weight @ v
     return attn_output
@@ -98,6 +104,41 @@ def test_flash_attention_backward_spmd_data_parallel(self):
       self.assertTrue(torch.allclose(i[0].grad.cpu(), i[1].cpu(), atol=1e-05))
     jax.config.update('jax_default_matmul_precision', "default")
 
+  @unittest.skipIf(xr.device_type() != 'TPU' or tpu.version() < 3,
+                   "This test only works on TPUv3+.")
+  def test_flash_attention_spmd_data_parallel_with_segment_ids(self):
+    jax.config.update('jax_default_matmul_precision', "highest")
+    n_devices = xr.global_runtime_device_count()
+    xs.set_global_mesh(xs.Mesh(range(n_devices), (n_devices, 1, 1, 1)))
+
+    q = torch.randn(4, 2, 128, 4).to("xla")
+    k = torch.randn(4, 2, 128, 4).to("xla")
+    v = torch.randn(4, 2, 128, 4).to("xla")
+    q_segment_ids = torch.ones(4, 128, device=q.device, dtype=torch.float32).to("xla")
+    kv_segment_ids = torch.rand(4, 128).to("xla")
+
+    o = flash_attention(q, k, v, q_segment_ids, kv_segment_ids, partition_spec=range(4))
+    self.assertEqual(
+        torch_xla._XLAC._get_xla_sharding_spec(o),
+        f"{{devices=[{n_devices},1,1,1]0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15}}")
+
+    attention_mask = kv_segment_ids.repeat_interleave(32, dim=0)
+    attention_mask = attention_mask.view(4, 32, -1, 128)
+    # head_size = self.heads
+    # current_length: int = attention_mask.shape[-1]
+    # if current_length != target_length:
+    #   attention_mask = F.pad(attention_mask, (0, target_length), value=0.0)
+
+    # if attention_mask.shape[0] < 4 * head_size:
+    #   attention_mask = attention_mask.repeat_interleave(head_size, dim=0)
+    #
+    # attention_mask = attention_mask.view(
+    #   batch_size, attn.heads, -1, attention_mask.shape[-1]
+    # )
+
+    expected_o = self._attention(q, k, v, attn_mask=attention_mask)
+    self.assertTrue(torch.allclose(o.cpu(), expected_o.cpu(), atol=1e-05))
+    jax.config.update('jax_default_matmul_precision', "default")
 
 if __name__ == '__main__':
   logging.getLogger().setLevel(logging.INFO)