Upload VQVAE

_utils.py

@@ -0,0 +1,86 @@
+# Shifts src_tf dim to dest dim
+# i.e. shift_dim(x, 1, -1) would be (b, c, t, h, w) -> (b, t, h, w, c)
+def shift_dim(x, src_dim=-1, dest_dim=-1, make_contiguous=True):
+    n_dims = len(x.shape)
+    if src_dim < 0:
+        src_dim = n_dims + src_dim
+    if dest_dim < 0:
+        dest_dim = n_dims + dest_dim
+
+    assert 0 <= src_dim < n_dims and 0 <= dest_dim < n_dims
+
+    dims = list(range(n_dims))
+    del dims[src_dim]
+
+    permutation = []
+    ctr = 0
+    for i in range(n_dims):
+        if i == dest_dim:
+            permutation.append(src_dim)
+        else:
+            permutation.append(dims[ctr])
+            ctr += 1
+    x = x.permute(permutation)
+    if make_contiguous:
+        x = x.contiguous()
+    return x
+
+# reshapes tensor start from dim i (inclusive)
+# to dim j (exclusive) to the desired shape
+# e.g. if x.shape = (b, thw, c) then
+# view_range(x, 1, 2, (t, h, w)) returns
+# x of shape (b, t, h, w, c)
+def view_range(x, i, j, shape):
+    shape = tuple(shape)
+
+    n_dims = len(x.shape)
+    if i < 0:
+        i = n_dims + i
+
+    if j is None:
+        j = n_dims
+    elif j < 0:
+        j = n_dims + j
+
+    assert 0 <= i < j <= n_dims
+
+    x_shape = x.shape
+    target_shape = x_shape[:i] + shape + x_shape[j:]
+    return x.view(target_shape)
+
+    
+def tensor_slice(x, begin, size):
+    assert all([b >= 0 for b in begin])
+    size = [l - b if s == -1 else s
+            for s, b, l in zip(size, begin, x.shape)]
+    assert all([s >= 0 for s in size])
+
+    slices = [slice(b, b + s) for b, s in zip(begin, size)]
+    return x[slices]
+
+
+import math
+import numpy as np
+import skvideo.io
+def save_video_grid(video, fname, nrow=None):
+    b, c, t, h, w = video.shape
+    video = video.permute(0, 2, 3, 4, 1)
+    video = (video.cpu().numpy() * 255).astype('uint8')
+
+    if nrow is None:
+        nrow = math.ceil(math.sqrt(b))
+    ncol = math.ceil(b / nrow)
+    padding = 1
+    video_grid = np.zeros((t, (padding + h) * nrow + padding,
+                           (padding + w) * ncol + padding, c), dtype='uint8')
+    for i in range(b):
+        r = i // ncol
+        c = i % ncol
+
+        start_r = (padding + h) * r
+        start_c = (padding + w) * c
+        video_grid[:, start_r:start_r + h, start_c:start_c + w] = video[i]
+
+    skvideo.io.vwrite(fname, video_grid, inputdict={'-r': '5'})
+    print('saved videos to', fname)
+

attention.py

@@ -0,0 +1,567 @@
+import numpy as np
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch.utils.checkpoint import checkpoint
+
+from ._utils import shift_dim, view_range, tensor_slice
+
+
+class AttentionStack(nn.Module):
+    def __init__(
+        self, shape, embd_dim, n_head, n_layer, dropout,
+        attn_type, attn_dropout, class_cond_dim, frame_cond_shape,
+    ):
+        super().__init__()
+        self.shape = shape
+        self.embd_dim = embd_dim
+        self.use_frame_cond = frame_cond_shape is not None
+
+        self.right_shift = RightShift(embd_dim)
+        self.pos_embd = AddBroadcastPosEmbed(
+            shape=shape, embd_dim=embd_dim
+        )
+
+        self.attn_nets = nn.ModuleList(
+            [
+                AttentionBlock(
+                    shape=shape,
+                    embd_dim=embd_dim,
+                    n_head=n_head,
+                    n_layer=n_layer,
+                    dropout=dropout,
+                    attn_type=attn_type,
+                    attn_dropout=attn_dropout,
+                    class_cond_dim=class_cond_dim,
+                    frame_cond_shape=frame_cond_shape
+                )
+                for i in range(n_layer)
+            ]
+        )
+
+    def forward(self, x, cond, decode_step, decode_idx):
+        """
+        Args
+        ------
+            x: (b, d1, d2, ..., dn, embd_dim)
+            cond: a dictionary of conditioning tensors
+
+            (below is used only when sampling for fast decoding)
+            decode: the enumerated rasterscan order of the current idx being sampled
+            decode_step: a tuple representing the current idx being sampled
+        """
+        x = self.right_shift(x, decode_step)
+        x = self.pos_embd(x, decode_step, decode_idx)
+        for net in self.attn_nets:
+            x = net(x, cond, decode_step, decode_idx)
+
+        return x
+
+
+class AttentionBlock(nn.Module):
+    def __init__(self, shape, embd_dim, n_head, n_layer, dropout,
+                 attn_type, attn_dropout, class_cond_dim, frame_cond_shape):
+        super().__init__()
+        self.use_frame_cond = frame_cond_shape is not None
+
+        self.pre_attn_norm = LayerNorm(embd_dim, class_cond_dim)
+        self.post_attn_dp = nn.Dropout(dropout)
+        self.attn = MultiHeadAttention(shape, embd_dim, embd_dim, n_head,
+                                       n_layer, causal=True, attn_type=attn_type,
+                                       attn_kwargs=dict(attn_dropout=attn_dropout))
+
+        if frame_cond_shape is not None:
+            enc_len = np.prod(frame_cond_shape[:-1])
+            self.pre_enc_norm = LayerNorm(embd_dim, class_cond_dim)
+            self.post_enc_dp = nn.Dropout(dropout)
+            self.enc_attn = MultiHeadAttention(shape, embd_dim, frame_cond_shape[-1],
+                                               n_head, n_layer, attn_type='full',
+                                               attn_kwargs=dict(attn_dropout=0.), causal=False)
+
+        self.pre_fc_norm = LayerNorm(embd_dim, class_cond_dim)
+        self.post_fc_dp = nn.Dropout(dropout)
+        self.fc_block = nn.Sequential(
+            nn.Linear(in_features=embd_dim, out_features=embd_dim * 4),
+            GeLU2(),
+            nn.Linear(in_features=embd_dim * 4, out_features=embd_dim),
+        )
+
+    def forward(self, x, cond, decode_step, decode_idx):
+        h = self.pre_attn_norm(x, cond)
+        if self.training:
+            h = checkpoint(self.attn, h, h, h, decode_step, decode_idx)
+        else:
+            h = self.attn(h, h, h, decode_step, decode_idx)
+        h = self.post_attn_dp(h)
+        x = x + h
+
+        if self.use_frame_cond:
+            h = self.pre_enc_norm(x, cond)
+            if self.training:
+                h = checkpoint(self.enc_attn, h, cond['frame_cond'], cond['frame_cond'],
+                               decode_step, decode_idx)
+            else:
+                h = self.enc_attn(h, cond['frame_cond'], cond['frame_cond'],
+                                  decode_step, decode_idx)
+            h = self.post_enc_dp(h)
+            x = x + h
+
+        h = self.pre_fc_norm(x, cond)
+        if self.training:
+            h = checkpoint(self.fc_block, h)
+        else:
+            h = self.fc_block(h)
+        h = self.post_fc_dp(h)
+        x = x + h
+
+        return x
+
+
+class MultiHeadAttention(nn.Module):
+    def __init__(self, shape, dim_q, dim_kv, n_head, n_layer,
+                 causal, attn_type, attn_kwargs):
+        super().__init__()
+        self.causal = causal
+        self.shape = shape
+
+        self.d_k = dim_q // n_head
+        self.d_v = dim_kv // n_head
+        self.n_head = n_head
+
+        self.w_qs = nn.Linear(dim_q, n_head * self.d_k, bias=False) # q
+        self.w_qs.weight.data.normal_(std=1.0 / np.sqrt(dim_q))
+
+        self.w_ks = nn.Linear(dim_kv, n_head * self.d_k, bias=False) # k
+        self.w_ks.weight.data.normal_(std=1.0 / np.sqrt(dim_kv))
+
+        self.w_vs = nn.Linear(dim_kv, n_head * self.d_v, bias=False) # v
+        self.w_vs.weight.data.normal_(std=1.0 / np.sqrt(dim_kv))
+
+        self.fc = nn.Linear(n_head * self.d_v, dim_q, bias=True) # c
+        self.fc.weight.data.normal_(std=1.0 / np.sqrt(dim_q * n_layer))
+
+        if attn_type == 'full':
+            self.attn = FullAttention(shape, causal, **attn_kwargs)
+        elif attn_type == 'axial':
+            assert not causal, 'causal axial attention is not supported'
+            self.attn = AxialAttention(len(shape), **attn_kwargs)
+        elif attn_type == 'sparse':
+            self.attn = SparseAttention(shape, n_head, causal, **attn_kwargs)
+
+        self.cache = None
+
+    def forward(self, q, k, v, decode_step=None, decode_idx=None):
+        """ Compute multi-head attention
+        Args
+            q, k, v: a [b, d1, ..., dn, c] tensor or
+                     a [b, 1, ..., 1, c] tensor if decode_step is not None
+
+        Returns
+            The output after performing attention
+        """
+
+        # compute k, q, v
+        d_k, d_v, n_head = self.d_k, self.d_v, self.n_head
+        q = view_range(self.w_qs(q), -1, None, (n_head, d_k))
+        k = view_range(self.w_ks(k), -1, None, (n_head, d_k))
+        v = view_range(self.w_vs(v), -1, None, (n_head, d_v))
+
+        # b x n_head x seq_len x d
+        # (b, *d_shape, n_head, d) -> (b, n_head, *d_shape, d)
+        q = shift_dim(q, -2, 1)
+        k = shift_dim(k, -2, 1)
+        v = shift_dim(v, -2, 1)
+
+        # fast decoding
+        if decode_step is not None:
+            if decode_step == 0:
+                if self.causal:
+                    k_shape = (q.shape[0], n_head, *self.shape, self.d_k)
+                    v_shape = (q.shape[0], n_head, *self.shape, self.d_v)
+                    self.cache = dict(k=torch.zeros(k_shape, dtype=k.dtype, device=q.device),
+                                    v=torch.zeros(v_shape, dtype=v.dtype, device=q.device))
+                else:
+                    # cache only once in the non-causal case
+                    self.cache = dict(k=k.clone(), v=v.clone())
+            if self.causal:
+                idx = (slice(None, None), slice(None, None), *[slice(i, i+ 1) for i in decode_idx])
+                self.cache['k'][idx] = k
+                self.cache['v'][idx] = v
+            k, v = self.cache['k'], self.cache['v']
+
+        a = self.attn(q, k, v, decode_step, decode_idx)
+
+        # (b, *d_shape, n_head, d) -> (b, *d_shape, n_head * d)
+        a = shift_dim(a, 1, -2).flatten(start_dim=-2)
+        a = self.fc(a) # (b x seq_len x embd_dim)
+
+        return a
+
+############## Attention #######################
+class FullAttention(nn.Module):
+    def __init__(self, shape, causal, attn_dropout):
+        super().__init__()
+        self.causal = causal
+        self.attn_dropout = attn_dropout
+
+        seq_len = np.prod(shape)
+        if self.causal:
+            self.register_buffer('mask', torch.tril(torch.ones(seq_len, seq_len)))
+
+    def forward(self, q, k, v, decode_step, decode_idx):
+        mask = self.mask if self.causal else None
+        if decode_step is not None and mask is not None:
+            mask = mask[[decode_step]]
+
+        old_shape = q.shape[2:-1]
+        q = q.flatten(start_dim=2, end_dim=-2)
+        k = k.flatten(start_dim=2, end_dim=-2)
+        v = v.flatten(start_dim=2, end_dim=-2)
+
+        out = scaled_dot_product_attention(q, k, v, mask=mask,
+                                           attn_dropout=self.attn_dropout,
+                                           training=self.training)
+
+        return view_range(out, 2, 3, old_shape)
+
+class AxialAttention(nn.Module):
+    def __init__(self, n_dim, axial_dim):
+        super().__init__()
+        if axial_dim < 0:
+            axial_dim = 2 + n_dim + 1 + axial_dim
+        else:
+            axial_dim += 2 # account for batch, head, dim
+        self.axial_dim = axial_dim
+
+    def forward(self, q, k, v, decode_step, decode_idx):
+        q = shift_dim(q, self.axial_dim, -2).flatten(end_dim=-3)
+        k = shift_dim(k, self.axial_dim, -2).flatten(end_dim=-3)
+        v = shift_dim(v, self.axial_dim, -2)
+        old_shape = list(v.shape)
+        v = v.flatten(end_dim=-3)
+
+        out = scaled_dot_product_attention(q, k, v, training=self.training)
+        out = out.view(*old_shape)
+        out = shift_dim(out, -2, self.axial_dim)
+        return out
+
+
+class SparseAttention(nn.Module):
+    ops = dict()
+    attn_mask = dict()
+    block_layout = dict()
+
+    def __init__(self, shape, n_head, causal, num_local_blocks=4, block=32,
+                 attn_dropout=0.): # does not use attn_dropout
+        super().__init__()
+        self.causal = causal
+        self.shape = shape
+
+        self.sparsity_config = StridedSparsityConfig(shape=shape, n_head=n_head,
+                                                     causal=causal, block=block,
+                                                     num_local_blocks=num_local_blocks)
+
+        if self.shape not in SparseAttention.block_layout:
+            SparseAttention.block_layout[self.shape] = self.sparsity_config.make_layout()
+        if causal and self.shape not in SparseAttention.attn_mask:
+            SparseAttention.attn_mask[self.shape] = self.sparsity_config.make_sparse_attn_mask()
+
+    def get_ops(self):
+        try:
+            from deepspeed.ops.sparse_attention import MatMul, Softmax
+        except:
+            raise Exception('Error importing deepspeed. Please install using `DS_BUILD_SPARSE_ATTN=1 pip install deepspeed`')
+        if self.shape not in SparseAttention.ops:
+            sparsity_layout = self.sparsity_config.make_layout()
+            sparse_dot_sdd_nt = MatMul(sparsity_layout,
+                                       self.sparsity_config.block,
+                                       'sdd',
+                                       trans_a=False,
+                                       trans_b=True)
+
+            sparse_dot_dsd_nn = MatMul(sparsity_layout,
+                                       self.sparsity_config.block,
+                                       'dsd',
+                                       trans_a=False,
+                                       trans_b=False)
+
+            sparse_softmax = Softmax(sparsity_layout, self.sparsity_config.block)
+
+            SparseAttention.ops[self.shape] = (sparse_dot_sdd_nt,
+                                               sparse_dot_dsd_nn,
+                                               sparse_softmax)
+        return SparseAttention.ops[self.shape]
+
+    def forward(self, q, k, v, decode_step, decode_idx):
+        if self.training and self.shape not in SparseAttention.ops:
+            self.get_ops()
+
+        SparseAttention.block_layout[self.shape] = SparseAttention.block_layout[self.shape].to(q)
+        if self.causal:
+            SparseAttention.attn_mask[self.shape] = SparseAttention.attn_mask[self.shape].to(q).type_as(q)
+        attn_mask = SparseAttention.attn_mask[self.shape] if self.causal else None
+
+        old_shape = q.shape[2:-1]
+        q = q.flatten(start_dim=2, end_dim=-2)
+        k = k.flatten(start_dim=2, end_dim=-2)
+        v = v.flatten(start_dim=2, end_dim=-2)
+
+        if decode_step is not None:
+            mask = self.sparsity_config.get_non_block_layout_row(SparseAttention.block_layout[self.shape], decode_step)
+            out = scaled_dot_product_attention(q, k, v, mask=mask, training=self.training)
+        else:
+            if q.shape != k.shape or k.shape != v.shape:
+                raise Exception('SparseAttention only support self-attention')
+            sparse_dot_sdd_nt, sparse_dot_dsd_nn, sparse_softmax = self.get_ops()
+            scaling = float(q.shape[-1]) ** -0.5
+
+            attn_output_weights = sparse_dot_sdd_nt(q, k)
+            if attn_mask is not None:
+                attn_output_weights = attn_output_weights.masked_fill(attn_mask == 0,
+                                                                      float('-inf'))
+            attn_output_weights = sparse_softmax(
+                attn_output_weights,
+                scale=scaling
+            )
+
+            out = sparse_dot_dsd_nn(attn_output_weights, v)
+
+        return view_range(out, 2, 3, old_shape)
+
+
+class StridedSparsityConfig(object):
+    """
+    Strided Sparse configuration specified in https://arxiv.org/abs/1904.10509 that
+    generalizes to arbitrary dimensions
+    """
+    def __init__(self, shape, n_head, causal, block, num_local_blocks):
+        self.n_head = n_head
+        self.shape = shape
+        self.causal = causal
+        self.block = block
+        self.num_local_blocks = num_local_blocks
+
+        assert self.num_local_blocks >= 1, 'Must have at least 1 local block'
+        assert self.seq_len % self.block == 0, 'seq len must be divisible by block size'
+
+        self._block_shape = self._compute_block_shape()
+        self._block_shape_cum = self._block_shape_cum_sizes()
+
+    @property
+    def seq_len(self):
+        return np.prod(self.shape)
+
+    @property
+    def num_blocks(self):
+        return self.seq_len // self.block
+
+    def set_local_layout(self, layout):
+        num_blocks = self.num_blocks
+        for row in range(0, num_blocks):
+            end = min(row + self.num_local_blocks, num_blocks)
+            for col in range(
+                    max(0, row - self.num_local_blocks),
+                    (row + 1 if self.causal else end)):
+                layout[:, row, col] = 1
+        return layout
+
+    def set_global_layout(self, layout):
+        num_blocks = self.num_blocks
+        n_dim = len(self._block_shape)
+        for row in range(num_blocks):
+            assert self._to_flattened_idx(self._to_unflattened_idx(row)) == row
+            cur_idx = self._to_unflattened_idx(row)
+            # no strided attention over last dim
+            for d in range(n_dim - 1):
+                end = self._block_shape[d]
+                for i in range(0, (cur_idx[d] + 1 if self.causal else end)):
+                    new_idx = list(cur_idx)
+                    new_idx[d] = i
+                    new_idx = tuple(new_idx)
+
+                    col = self._to_flattened_idx(new_idx)
+                    layout[:, row, col] = 1
+
+        return layout
+
+    def make_layout(self):
+        layout = torch.zeros((self.n_head, self.num_blocks, self.num_blocks), dtype=torch.int64)
+        layout = self.set_local_layout(layout)
+        layout = self.set_global_layout(layout)
+        return layout
+
+    def make_sparse_attn_mask(self):
+        block_layout = self.make_layout()
+        assert block_layout.shape[1] == block_layout.shape[2] == self.num_blocks
+
+        num_dense_blocks = block_layout.sum().item()
+        attn_mask = torch.ones(num_dense_blocks, self.block, self.block)
+        counter = 0
+        for h in range(self.n_head):
+            for i in range(self.num_blocks):
+                for j in range(self.num_blocks):
+                    elem = block_layout[h, i, j].item()
+                    if elem == 1:
+                        assert i >= j
+                        if i == j: # need to mask within block on diagonals
+                            attn_mask[counter] = torch.tril(attn_mask[counter])
+                        counter += 1
+        assert counter == num_dense_blocks
+
+        return attn_mask.unsqueeze(0)
+
+    def get_non_block_layout_row(self, block_layout, row):
+        block_row = row // self.block
+        block_row = block_layout[:, [block_row]] # n_head x 1 x n_blocks
+        block_row = block_row.repeat_interleave(self.block, dim=-1)
+        block_row[:, :, row + 1:] = 0.
+        return block_row
+
+    ############# Helper functions ##########################
+
+    def _compute_block_shape(self):
+        n_dim = len(self.shape)
+        cum_prod = 1
+        for i in range(n_dim - 1, -1, -1):
+            cum_prod *= self.shape[i]
+            if cum_prod > self.block:
+                break
+        assert cum_prod % self.block == 0
+        new_shape = (*self.shape[:i], cum_prod // self.block)
+
+        assert np.prod(new_shape) == np.prod(self.shape) // self.block
+
+        return new_shape
+
+    def _block_shape_cum_sizes(self):
+        bs = np.flip(np.array(self._block_shape))
+        return tuple(np.flip(np.cumprod(bs)[:-1])) + (1,)
+
+    def _to_flattened_idx(self, idx):
+        assert len(idx) == len(self._block_shape), f"{len(idx)} != {len(self._block_shape)}"
+        flat_idx = 0
+        for i in range(len(self._block_shape)):
+            flat_idx += idx[i] * self._block_shape_cum[i]
+        return flat_idx
+
+    def _to_unflattened_idx(self, flat_idx):
+        assert flat_idx < np.prod(self._block_shape)
+        idx = []
+        for i in range(len(self._block_shape)):
+            idx.append(flat_idx // self._block_shape_cum[i])
+            flat_idx %= self._block_shape_cum[i]
+        return tuple(idx)
+
+
+################ Spatiotemporal broadcasted positional embeddings ###############
+class AddBroadcastPosEmbed(nn.Module):
+    def __init__(self, shape, embd_dim, dim=-1):
+        super().__init__()
+        assert dim in [-1, 1] # only first or last dim supported
+        self.shape = shape
+        self.n_dim = n_dim = len(shape)
+        self.embd_dim = embd_dim
+        self.dim = dim
+
+        assert embd_dim % n_dim == 0, f"{embd_dim} % {n_dim} != 0"
+        self.emb = nn.ParameterDict({
+             f'd_{i}': nn.Parameter(torch.randn(shape[i], embd_dim // n_dim) * 0.01
+                                    if dim == -1 else
+                                    torch.randn(embd_dim // n_dim, shape[i]) * 0.01)
+             for i in range(n_dim)
+        })
+
+    def forward(self, x, decode_step=None, decode_idx=None):
+        embs = []
+        for i in range(self.n_dim):
+            e = self.emb[f'd_{i}']
+            if self.dim == -1:
+                # (1, 1, ..., 1, self.shape[i], 1, ..., -1)
+                e = e.view(1, *((1,) * i), self.shape[i], *((1,) * (self.n_dim - i - 1)), -1)
+                e = e.expand(1, *self.shape, -1)
+            else:
+                e = e.view(1, -1, *((1,) * i), self.shape[i], *((1,) * (self.n_dim - i - 1)))
+                e = e.expand(1, -1, *self.shape)
+            embs.append(e)
+
+        embs = torch.cat(embs, dim=self.dim)
+        if decode_step is not None:
+            embs = tensor_slice(embs, [0, *decode_idx, 0],
+                                [x.shape[0], *(1,) * self.n_dim, x.shape[-1]])
+
+        return x + embs
+
+################# Helper Functions ###################################
+def scaled_dot_product_attention(q, k, v, mask=None, attn_dropout=0., training=True):
+    # Performs scaled dot-product attention over the second to last dimension dn
+
+    # (b, n_head, d1, ..., dn, d)
+    attn = torch.matmul(q, k.transpose(-1, -2))
+    attn = attn / np.sqrt(q.shape[-1])
+    if mask is not None:
+        attn = attn.masked_fill(mask == 0, float('-inf'))
+    attn_float = F.softmax(attn, dim=-1)
+    attn = attn_float.type_as(attn) # b x n_head x d1 x ... x dn x d
+    attn = F.dropout(attn, p=attn_dropout, training=training)
+
+    a = torch.matmul(attn, v) # b x n_head x d1 x ... x dn x d
+
+    return a
+
+
+class RightShift(nn.Module):
+    def __init__(self, embd_dim):
+        super().__init__()
+        self.embd_dim = embd_dim
+        self.sos = nn.Parameter(torch.FloatTensor(embd_dim).normal_(std=0.02), requires_grad=True)
+
+    def forward(self, x, decode_step):
+        if decode_step is not None and decode_step > 0:
+            return x
+
+        x_shape = list(x.shape)
+        x = x.flatten(start_dim=1, end_dim=-2) # (b, seq_len, embd_dim)
+        sos = torch.ones(x_shape[0], 1, self.embd_dim, dtype=torch.float32).to(self.sos) * self.sos
+        sos = sos.type_as(x)
+        x = torch.cat([sos, x[:, :-1, :]], axis=1)
+        x = x.view(*x_shape)
+
+        return x
+
+
+class GeLU2(nn.Module):
+    def forward(self, x):
+        return (1.702 * x).sigmoid() * x
+
+
+class LayerNorm(nn.Module):
+    def __init__(self, embd_dim, class_cond_dim):
+        super().__init__()
+        self.conditional = class_cond_dim is not None
+
+        if self.conditional:
+            self.w = nn.Linear(class_cond_dim, embd_dim, bias=False)
+            nn.init.constant_(self.w.weight.data, 1. / np.sqrt(class_cond_dim))
+            self.wb = nn.Linear(class_cond_dim, embd_dim, bias=False)
+        else:
+            self.g = nn.Parameter(torch.ones(embd_dim, dtype=torch.float32), requires_grad=True)
+            self.b = nn.Parameter(torch.zeros(embd_dim, dtype=torch.float32), requires_grad=True)
+
+    def forward(self, x, cond):
+        if self.conditional:  # (b, cond_dim)
+            g = 1 + self.w(cond['class_cond']).view(x.shape[0], *(1,)*(len(x.shape)-2), x.shape[-1]) # (b, ..., embd_dim)
+            b = self.wb(cond['class_cond']).view(x.shape[0], *(1,)*(len(x.shape)-2), x.shape[-1])
+        else:
+            g = self.g  # (embd_dim,)
+            b = self.b
+
+        x_float = x.float()
+
+        mu = x_float.mean(dim=-1, keepdims=True)
+        s = (x_float - mu).square().mean(dim=-1, keepdims=True)
+        x_float = (x_float - mu) * (1e-5 + s.rsqrt())  # (b, ..., embd_dim)
+        x_float = x_float * g + b
+
+        x = x_float.type_as(x)
+        return x

config.json

@@ -0,0 +1,21 @@
+{
+  "architectures": [
+    "VQVAE"
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_vqvae.VQVAEConfig",
+    "AutoModel": "modeling_vqvae.VQVAE"
+  },
+  "downsample": [
+    2,
+    4,
+    4
+  ],
+  "embedding_dim": 256,
+  "model_type": "VQVAE",
+  "n_codes": 2048,
+  "n_hiddens": 240,
+  "n_res_layers": 4,
+  "torch_dtype": "float32",
+  "transformers_version": "4.37.2"
+}

configuration_vqvae.py

@@ -0,0 +1,22 @@
+from transformers import PretrainedConfig
+from typing import List
+
+
+class VQVAEConfig(PretrainedConfig):
+    model_type = "VQVAE"
+
+    def __init__(
+        self,
+        embedding_dim: int = 256,
+        n_codes: int = 2048,
+        n_hiddens: int = 240,
+        n_res_layers: int = 4,
+        downsample: List[int] = [2, 4, 4],
+        **kwargs,
+    ):
+        self.embedding_dim = embedding_dim
+        self.n_codes = n_codes
+        self.n_hiddens = n_hiddens
+        self.n_res_layers = n_res_layers
+        self.downsample = downsample
+        super().__init__(**kwargs)

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9fda02bdef17ca1378a9392adc5b1d9692fa194ccaabff3b8352ce7548af0de
+size 88842260

modeling_vqvae.py

@@ -0,0 +1,321 @@
+import os
+import math
+import numpy as np
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.distributed as dist
+import gdown
+
+from .attention import MultiHeadAttention
+from ._utils import shift_dim
+from transformers import PreTrainedModel
+from typing import Tuple
+from .configuration_vqvae import VQVAEConfig
+
+
+_VQVAE = {
+    'bair_stride4x2x2': '1iIAYJ2Qqrx5Q94s5eIXQYJgAydzvT_8L', # trained on 16 frames of 64 x 64 images
+    'ucf101_stride4x4x4': '1uuB_8WzHP_bbBmfuaIV7PK_Itl3DyHY5', # trained on 16 frames of 128 x 128 images
+    'kinetics_stride4x4x4': '1DOvOZnFAIQmux6hG7pN_HkyJZy3lXbCB', # trained on 16 frames of 128 x 128 images
+    'kinetics_stride2x4x4': '1jvtjjtrtE4cy6pl7DK_zWFEPY3RZt2pB' # trained on 16 frames of 128 x 128 images
+}
+
+def download(id, fname, root=None):
+    """
+    Download the VQVAE weights from Google Drive.
+    
+    Args:
+        id (str): the ID of the file to download
+        fname (str): the name of the file to save
+        root (str): the directory to save the file to
+    """
+    if root is None:
+        root = os.path.expanduser('~/.cache/sora')
+    os.makedirs(root, exist_ok=True)
+    destination = os.path.join(root, fname)
+
+    if os.path.exists(destination):
+        return destination
+
+    gdown.download(id=id, output=destination, quiet=False)
+    return destination
+
+
+class VQVAE(PreTrainedModel):
+    config_class = VQVAEConfig
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.embedding_dim = config.embedding_dim
+        self.n_codes = config.n_codes
+
+        self.encoder = Encoder(config.n_hiddens, config.n_res_layers, config.downsample)
+        self.decoder = Decoder(config.n_hiddens, config.n_res_layers, config.downsample)
+
+        self.pre_vq_conv = SamePadConv3d(config.n_hiddens, config.embedding_dim, 1)
+        self.post_vq_conv = SamePadConv3d(config.embedding_dim, config.n_hiddens, 1)
+
+        self.codebook = Codebook(config.n_codes, config.embedding_dim)
+
+    @property
+    def latent_shape(self):
+        input_shape = (self.args.sequence_length, self.args.resolution,
+                       self.args.resolution)
+        return tuple([s // d for s, d in zip(input_shape,
+                                             self.args.downsample)])
+
+    def encode(self, x, include_embeddings=False):
+        h = self.pre_vq_conv(self.encoder(x))
+        vq_output = self.codebook(h)
+        if include_embeddings:
+            return vq_output['encodings'], vq_output['embeddings']
+        else:
+            return vq_output['encodings']
+
+    def decode(self, encodings):
+        h = F.embedding(encodings, self.codebook.embeddings)
+        h = self.post_vq_conv(shift_dim(h, -1, 1))
+        return self.decoder(h)
+
+    def forward(self, x):
+        z = self.pre_vq_conv(self.encoder(x))
+        vq_output = self.codebook(z)
+        x_recon = self.decoder(self.post_vq_conv(vq_output['embeddings']))
+        recon_loss = F.mse_loss(x_recon, x) / 0.06
+
+        return recon_loss, x_recon, vq_output
+
+
+class AxialBlock(nn.Module):
+    def __init__(self, n_hiddens, n_head):
+        super().__init__()
+        kwargs = dict(shape=(0,) * 3, dim_q=n_hiddens,
+                      dim_kv=n_hiddens, n_head=n_head,
+                      n_layer=1, causal=False, attn_type='axial')
+        self.attn_w = MultiHeadAttention(attn_kwargs=dict(axial_dim=-2),
+                                         **kwargs)
+        self.attn_h = MultiHeadAttention(attn_kwargs=dict(axial_dim=-3),
+                                         **kwargs)
+        self.attn_t = MultiHeadAttention(attn_kwargs=dict(axial_dim=-4),
+                                         **kwargs)
+
+    def forward(self, x):
+        x = shift_dim(x, 1, -1)
+        x = self.attn_w(x, x, x) + self.attn_h(x, x, x) + self.attn_t(x, x, x)
+        x = shift_dim(x, -1, 1)
+        return x
+
+
+class AttentionResidualBlock(nn.Module):
+    def __init__(self, n_hiddens):
+        super().__init__()
+        self.block = nn.Sequential(
+            nn.BatchNorm3d(n_hiddens),
+            nn.ReLU(),
+            SamePadConv3d(n_hiddens, n_hiddens // 2, 3, bias=False),
+            nn.BatchNorm3d(n_hiddens // 2),
+            nn.ReLU(),
+            SamePadConv3d(n_hiddens // 2, n_hiddens, 1, bias=False),
+            nn.BatchNorm3d(n_hiddens),
+            nn.ReLU(),
+            AxialBlock(n_hiddens, 2)
+        )
+
+    def forward(self, x):
+        return x + self.block(x)
+
+class Codebook(nn.Module):
+    def __init__(self, n_codes, embedding_dim):
+        super().__init__()
+        self.register_buffer('embeddings', torch.randn(n_codes, embedding_dim))
+        self.register_buffer('N', torch.zeros(n_codes))
+        self.register_buffer('z_avg', self.embeddings.data.clone())
+
+        self.n_codes = n_codes
+        self.embedding_dim = embedding_dim
+        self._need_init = True
+
+    def _tile(self, x):
+        d, ew = x.shape
+        if d < self.n_codes:
+            n_repeats = (self.n_codes + d - 1) // d
+            std = 0.01 / np.sqrt(ew)
+            x = x.repeat(n_repeats, 1)
+            x = x + torch.randn_like(x) * std
+        return x
+
+    def _init_embeddings(self, z):
+        # z: [b, c, t, h, w]
+        self._need_init = False
+        flat_inputs = shift_dim(z, 1, -1).flatten(end_dim=-2)
+        y = self._tile(flat_inputs)
+
+        d = y.shape[0]
+        _k_rand = y[torch.randperm(y.shape[0])][:self.n_codes]
+        if dist.is_initialized():
+            dist.broadcast(_k_rand, 0)
+        self.embeddings.data.copy_(_k_rand)
+        self.z_avg.data.copy_(_k_rand)
+        self.N.data.copy_(torch.ones(self.n_codes))
+
+    def forward(self, z):
+        # z: [b, c, t, h, w]
+        if self._need_init and self.training:
+            self._init_embeddings(z)
+        flat_inputs = shift_dim(z, 1, -1).flatten(end_dim=-2)
+        distances = (flat_inputs ** 2).sum(dim=1, keepdim=True) \
+                    - 2 * flat_inputs @ self.embeddings.t() \
+                    + (self.embeddings.t() ** 2).sum(dim=0, keepdim=True)
+
+        encoding_indices = torch.argmin(distances, dim=1)
+        encode_onehot = F.one_hot(encoding_indices, self.n_codes).type_as(flat_inputs)
+        encoding_indices = encoding_indices.view(z.shape[0], *z.shape[2:])
+
+        embeddings = F.embedding(encoding_indices, self.embeddings)
+        embeddings = shift_dim(embeddings, -1, 1)
+
+        commitment_loss = 0.25 * F.mse_loss(z, embeddings.detach())
+
+        # EMA codebook update
+        if self.training:
+            n_total = encode_onehot.sum(dim=0)
+            encode_sum = flat_inputs.t() @ encode_onehot
+            if dist.is_initialized():
+                dist.all_reduce(n_total)
+                dist.all_reduce(encode_sum)
+
+            self.N.data.mul_(0.99).add_(n_total, alpha=0.01)
+            self.z_avg.data.mul_(0.99).add_(encode_sum.t(), alpha=0.01)
+
+            n = self.N.sum()
+            weights = (self.N + 1e-7) / (n + self.n_codes * 1e-7) * n
+            encode_normalized = self.z_avg / weights.unsqueeze(1)
+            self.embeddings.data.copy_(encode_normalized)
+
+            y = self._tile(flat_inputs)
+            _k_rand = y[torch.randperm(y.shape[0])][:self.n_codes]
+            if dist.is_initialized():
+                dist.broadcast(_k_rand, 0)
+
+            usage = (self.N.view(self.n_codes, 1) >= 1).float()
+            self.embeddings.data.mul_(usage).add_(_k_rand * (1 - usage))
+
+        embeddings_st = (embeddings - z).detach() + z
+
+        avg_probs = torch.mean(encode_onehot, dim=0)
+        perplexity = torch.exp(-torch.sum(avg_probs * torch.log(avg_probs + 1e-10)))
+
+        return dict(embeddings=embeddings_st, encodings=encoding_indices,
+                    commitment_loss=commitment_loss, perplexity=perplexity)
+
+    def dictionary_lookup(self, encodings):
+        embeddings = F.embedding(encodings, self.embeddings)
+        return embeddings
+
+class Encoder(nn.Module):
+    def __init__(self, n_hiddens, n_res_layers, downsample):
+        super().__init__()
+        n_times_downsample = np.array([int(math.log2(d)) for d in downsample])
+        self.convs = nn.ModuleList()
+        max_ds = n_times_downsample.max()
+        for i in range(max_ds):
+            in_channels = 3 if i == 0 else n_hiddens
+            stride = tuple([2 if d > 0 else 1 for d in n_times_downsample])
+            conv = SamePadConv3d(in_channels, n_hiddens, 4, stride=stride)
+            self.convs.append(conv)
+            n_times_downsample -= 1
+        self.conv_last = SamePadConv3d(in_channels, n_hiddens, kernel_size=3)
+
+        self.res_stack = nn.Sequential(
+            *[AttentionResidualBlock(n_hiddens)
+              for _ in range(n_res_layers)],
+            nn.BatchNorm3d(n_hiddens),
+            nn.ReLU()
+        )
+
+    def forward(self, x):
+        h = x
+        for conv in self.convs:
+            h = F.relu(conv(h))
+        h = self.conv_last(h)
+        h = self.res_stack(h)
+        return h
+
+
+class Decoder(nn.Module):
+    def __init__(self, n_hiddens, n_res_layers, upsample):
+        super().__init__()
+        self.res_stack = nn.Sequential(
+            *[AttentionResidualBlock(n_hiddens)
+              for _ in range(n_res_layers)],
+            nn.BatchNorm3d(n_hiddens),
+            nn.ReLU()
+        )
+
+        n_times_upsample = np.array([int(math.log2(d)) for d in upsample])
+        max_us = n_times_upsample.max()
+        self.convts = nn.ModuleList()
+        for i in range(max_us):
+            out_channels = 3 if i == max_us - 1 else n_hiddens
+            us = tuple([2 if d > 0 else 1 for d in n_times_upsample])
+            convt = SamePadConvTranspose3d(n_hiddens, out_channels, 4,
+                                           stride=us)
+            self.convts.append(convt)
+            n_times_upsample -= 1
+
+    def forward(self, x):
+        h = self.res_stack(x)
+        for i, convt in enumerate(self.convts):
+            h = convt(h)
+            if i < len(self.convts) - 1:
+                h = F.relu(h)
+        return h
+
+
+# Does not support dilation
+class SamePadConv3d(nn.Module):
+    def __init__(self, in_channels, out_channels, kernel_size, stride=1, bias=True):
+        super().__init__()
+        if isinstance(kernel_size, int):
+            kernel_size = (kernel_size,) * 3
+        if isinstance(stride, int):
+            stride = (stride,) * 3
+
+        # assumes that the input shape is divisible by stride
+        total_pad = tuple([k - s for k, s in zip(kernel_size, stride)])
+        pad_input = []
+        for p in total_pad[::-1]: # reverse since F.pad starts from last dim
+            pad_input.append((p // 2 + p % 2, p // 2))
+        pad_input = sum(pad_input, tuple())
+        self.pad_input = pad_input
+
+        self.conv = nn.Conv3d(in_channels, out_channels, kernel_size,
+                              stride=stride, padding=0, bias=bias)
+
+    def forward(self, x):
+        return self.conv(F.pad(x, self.pad_input))
+
+
+class SamePadConvTranspose3d(nn.Module):
+    def __init__(self, in_channels, out_channels, kernel_size, stride=1, bias=True):
+        super().__init__()
+        if isinstance(kernel_size, int):
+            kernel_size = (kernel_size,) * 3
+        if isinstance(stride, int):
+            stride = (stride,) * 3
+
+        total_pad = tuple([k - s for k, s in zip(kernel_size, stride)])
+        pad_input = []
+        for p in total_pad[::-1]: # reverse since F.pad starts from last dim
+            pad_input.append((p // 2 + p % 2, p // 2))
+        pad_input = sum(pad_input, tuple())
+        self.pad_input = pad_input
+
+        self.convt = nn.ConvTranspose3d(in_channels, out_channels, kernel_size,
+                                        stride=stride, bias=bias,
+                                        padding=tuple([k - 1 for k in kernel_size]))
+
+    def forward(self, x):
+        return self.convt(F.pad(x, self.pad_input))