Support index_first_axis with more than 2 dimensions

flash_attn/bert_padding.py

@@ -1,5 +1,7 @@
 # Adapted from https://github.com/mlcommons/training_results_v1.1/blob/main/NVIDIA/benchmarks/bert/implementations/pytorch/padding.py
 
+import numpy as np
+
 import torch
 import torch.nn.functional as F
 
@@ -11,21 +13,26 @@ class IndexFirstAxis(torch.autograd.Function):
     @staticmethod
     def forward(ctx, input, indices):
         ctx.save_for_backward(indices)
-        ctx.first_axis_dim = input.shape[0]
-        assert input.ndim == 2
+        assert input.ndim >= 2
+        ctx.first_axis_dim, other_shape = input.shape[0], input.shape[1:]
+        second_dim = np.prod(other_shape)
         # TD [2022-03-04] For some reason torch.gather is a bit faster than indexing.
         # return input[indices]
-        return torch.gather(input, 0, repeat(indices, 'z -> z d', d=input.shape[1]))
+        return torch.gather(rearrange(input, 'b ... -> b (...)'), 0,
+                            repeat(indices, 'z -> z d', d=second_dim)).reshape(-1, *other_shape)
 
     @staticmethod
     def backward(ctx, grad_output):
         indices, = ctx.saved_tensors
-        grad_input = torch.zeros([ctx.first_axis_dim, *grad_output.shape[1:]],
-                                 device=grad_output.device, dtype=grad_output.dtype)
+        assert grad_output.ndim >= 2
+        other_shape = grad_output.shape[1:]
+        grad_output = rearrange(grad_output, 'b ... -> b (...)')
+        grad_input = torch.zeros([ctx.first_axis_dim, grad_output.shape[1]],
+                                  device=grad_output.device, dtype=grad_output.dtype)
         # TD [2022-03-04] For some reason torch.scatter is a bit faster than indexing.
         # grad_input[indices] = grad_output
         grad_input.scatter_(0, repeat(indices, 'z -> z d', d=grad_output.shape[1]), grad_output)
-        return grad_input, None
+        return grad_input.reshape(ctx.first_axis_dim, *other_shape), None
 
 
 index_first_axis = IndexFirstAxis.apply
@@ -37,8 +44,8 @@ class IndexPutFirstAxis(torch.autograd.Function):
     def forward(ctx, values, indices, first_axis_dim):
         ctx.save_for_backward(indices)
         assert indices.ndim == 1
-        assert values.ndim == 2
-        output = torch.zeros(first_axis_dim, values.shape[1], device=values.device,
+        assert values.ndim >= 2
+        output = torch.zeros(first_axis_dim, *values.shape[1:], device=values.device,
                              dtype=values.dtype)
         # TD [2022-03-04] For some reason torch.scatter is a bit faster than indexing.
         output[indices] = values
@@ -57,13 +64,45 @@ class IndexPutFirstAxis(torch.autograd.Function):
 index_put_first_axis = IndexPutFirstAxis.apply
 
 
+class IndexFirstAxisResidual(torch.autograd.Function):
+
+    @staticmethod
+    def forward(ctx, input, indices):
+        ctx.save_for_backward(indices)
+        assert input.ndim >= 2
+        ctx.first_axis_dim, other_shape = input.shape[0], input.shape[1:]
+        second_dim = np.prod(other_shape)
+        # TD [2022-03-04] For some reason torch.gather is a bit faster than indexing.
+        output = input[indices]
+        # We don't want to reshape input (b ... -> b (...)) since it could change the channel_last
+        # memory format to channel_first. In other words, input might not be contiguous.
+        # If we don't detach, Pytorch complains about output being a view and is being modified inplace
+        return output, input.detach()
+
+    @staticmethod
+    def backward(ctx, grad_output, grad_residual):
+        indices, = ctx.saved_tensors
+        assert grad_output.ndim >= 2
+        other_shape = grad_output.shape[1:]
+        assert grad_residual.shape[1:] == other_shape
+        grad_input = grad_residual
+        # grad_input[indices] += grad_output
+        indices = indices.reshape(indices.shape[0], *((1,) * (grad_output.ndim - 1)))
+        indices = indices.expand_as(grad_output)
+        grad_input.scatter_add_(0, indices, grad_output)
+        return grad_input.reshape(ctx.first_axis_dim, *other_shape), None
+
+
+index_first_axis_residual = IndexFirstAxisResidual.apply
+
+
 def unpad_input(hidden_states, attention_mask):
     """
     Arguments:
-        hidden_states: (batch, seqlen, dim)
+        hidden_states: (batch, seqlen, ...)
         attention_mask: (batch, seqlen), bool / int, 1 means valid and 0 means not valid.
     Return:
-        hidden_states: (total_nnz, dim), where total_nnz = number of tokens in selected in attention_mask.
+        hidden_states: (total_nnz, ...), where total_nnz = number of tokens in selected in attention_mask.
         cu_seqlens: (batch + 1), the cumulative sequence lengths, used to index into hidden_states.
         max_seqlen_in_batch: int
     """
@@ -76,20 +115,20 @@ def unpad_input(hidden_states, attention_mask):
     # times larger than it needs to be, wasting memory. It's faster and more memory-efficient to
     # index with integer indices. Moreover, torch's index is a bit slower than it needs to be,
     # so we write custom forward and backward to make it a bit faster.
-    return (index_first_axis(rearrange(hidden_states, 'b s d -> (b s) d'), indices), indices,
+    return (index_first_axis(rearrange(hidden_states, 'b s ... -> (b s) ...'), indices), indices,
             cu_seqlens, max_seqlen_in_batch)
 
 
 def pad_input(hidden_states, indices, batch, seqlen):
     """
     Arguments:
-        hidden_states: (total_nnz, dim), where total_nnz = number of tokens in selected in attention_mask.
+        hidden_states: (total_nnz, ...), where total_nnz = number of tokens in selected in attention_mask.
         indices: (total_nnz)
     Return:
-        hidden_states: (batch, seqlen, dim)
+        hidden_states: (batch, seqlen, ...)
     """
     dim = hidden_states.shape[-1]
     # output = torch.zeros((batch * seqlen), dim, device=hidden_states.device, dtype=hidden_states.dtype)
     # output[indices] = hidden_states
     output = index_put_first_axis(hidden_states, indices, batch * seqlen)
-    return rearrange(output, '(b s) d -> b s d', b=batch)
+    return rearrange(output, '(b s) ... -> b s ...', b=batch)