clip/jclip/mha.py

from typing import Optional, Tuple, List
import warnings
import math

import jittor as jt
from jittor import Var
from jittor.nn import Module, Linear, softmax, pad, linear, dropout
from jittor.init import xavier_uniform_, xavier_gauss_, constant_


def _canonical_mask(
    mask: Optional[Var],
    mask_name: str,
    other_type,
    other_name: str,
    target_type,
    check_other: bool = True,
) -> Optional[Var]:

    if mask is not None:
        _mask_dtype = mask.dtype
        _mask_is_float = mask.dtype == jt.float16 or mask.dtype == jt.float32 or mask.dtype == jt.float64
        if _mask_dtype != jt.bool and not _mask_is_float:
            raise AssertionError(
                f"only bool and floating types of {mask_name} are supported")
        if check_other and other_type is not None:
            if _mask_dtype != other_type:
                warnings.warn(
                    f"Support for mismatched {mask_name} and {other_name} "
                    "is deprecated. Use same type for both instead.")
        if not _mask_is_float:
            # WARNING(514flowey): Check Here
            new_mask = jt.zeros_like(mask, dtype=target_type)
            new_mask[mask] = float("-inf")
            mask = new_mask
    return mask


def _none_or_dtype(input: Optional[Var]):
    if input is None:
        return None
    elif isinstance(input, jt.Var):
        return input.dtype


def baddbmm(input_var: jt.Var,
            batch1: jt.Var,
            batch2: jt.Var,
            beta=1,
            alpha=1) -> jt.Var:
    # WARNING(514flowey): Check here
    return beta * input_var + alpha * (batch1 @ batch2)


def scaled_dot_product_attention(query,
                                 key,
                                 value,
                                 attn_mask=None,
                                 dropout_p=0.0,
                                 is_causal=False,
                                 scale=None,
                                 training=True) -> jt.Var:
    # Efficient implementation equivalent to the following:
    L, S = query.size(-2), key.size(-2)
    scale_factor = 1 / math.sqrt(query.size(-1)) if scale is None else scale
    attn_bias = jt.zeros(L, S, dtype=query.dtype)
    if is_causal:
        assert attn_mask is None
        temp_mask = jt.ones(L, S, dtype=jt.bool).tril(diagonal=0)
        attn_bias[jt.logical_not(temp_mask)] = float("-inf")
        # attn_bias.to(query.dtype)
        attn_bias = jt.array(attn_bias, query.dtype)

    if attn_mask is not None:
        if attn_mask.dtype == jt.bool:
            attn_bias[jt.logical_not(temp_mask)] = float("-inf")
        else:
            attn_bias += attn_mask
    attn_weight = query @ key.transpose(-2, -1) * scale_factor
    attn_weight += attn_bias
    attn_weight = softmax(attn_weight, dim=-1)
    attn_weight = dropout(attn_weight, dropout_p, is_train=training)
    return attn_weight @ value


def _mha_shape_check(query: Var, key: Var, value: Var,
                     key_padding_mask: Optional[Var], attn_mask: Optional[Var],
                     num_heads: int):
    if query.dim() == 3:
        is_batched = True
        assert key.dim() == 3 and value.dim() == 3, \
            ("For batched (3-D) `query`, expected `key` and `value` to be 3-D"
             f" but found {key.dim()}-D and {value.dim()}-D Vars respectively")
        if key_padding_mask is not None:
            assert key_padding_mask.dim() == 2, \
                ("For batched (3-D) `query`, expected `key_padding_mask` to be `None` or 2-D"
                 f" but found {key_padding_mask.dim()}-D Var instead")
        if attn_mask is not None:
            assert attn_mask.dim() in (2, 3), \
                ("For batched (3-D) `query`, expected `attn_mask` to be `None`, 2-D or 3-D"
                 f" but found {attn_mask.dim()}-D Var instead")
    elif query.dim() == 2:
        is_batched = False
        assert key.dim() == 2 and value.dim() == 2, \
            ("For unbatched (2-D) `query`, expected `key` and `value` to be 2-D"
             f" but found {key.dim()}-D and {value.dim()}-D Vars respectively")

        if key_padding_mask is not None:
            assert key_padding_mask.dim() == 1, \
                ("For unbatched (2-D) `query`, expected `key_padding_mask` to be `None` or 1-D"
                 f" but found {key_padding_mask.dim()}-D Var instead")

        if attn_mask is not None:
            assert attn_mask.dim() in (2, 3), \
                ("For unbatched (2-D) `query`, expected `attn_mask` to be `None`, 2-D or 3-D"
                 f" but found {attn_mask.dim()}-D Var instead")
            if attn_mask.dim() == 3:
                expected_shape = (num_heads, query.shape[0], key.shape[0])
                assert attn_mask.shape == expected_shape, \
                    (f"Expected `attn_mask` shape to be {expected_shape} but got {attn_mask.shape}")
    else:
        raise AssertionError(
            f"query should be unbatched 2D or batched 3D Var but received {query.dim()}-D query Var"
        )

    return is_batched


def _in_projection_packed(
    q: Var,
    k: Var,
    v: Var,
    w: Var,
    b: Optional[Var] = None,
) -> List[Var]:
    E = q.size(-1)
    if k is v:
        if q is k:
            # self-attention
            proj = linear(q, w, b)
            # reshape to 3, E and not E, 3 is deliberate for better memory coalescing and keeping same order as chunk()
            # proj = proj.unflatten(-1, (3, E)).unsqueeze(0).transpose(0, -2).squeeze(-2).contiguous()
            nshape = proj.shape[:-1] + (3, E)
            proj = proj.reshape(nshape).unsqueeze(0).transpose(0,
                                                               -2).squeeze(-2)
            return proj[0], proj[1], proj[2]
        else:
            # encoder-decoder attention
            w_q, w_kv = w.split([E, E * 2])
            if b is None:
                b_q = b_kv = None
            else:
                b_q, b_kv = b.split([E, E * 2])
            q_proj = linear(q, w_q, b_q)
            kv_proj = linear(k, w_kv, b_kv)
            # reshape to 2, E and not E, 2 is deliberate for better memory coalescing and keeping same order as chunk()
            # kv_proj = kv_proj.unflatten(-1, (2, E)).unsqueeze(0).transpose(0, -2).squeeze(-2).contiguous()
            nshape = kv_proj.shape[:-1] + (2, E)
            kv_proj = kv_proj.reshape(nshape).unsqueeze(0).transpose(
                0, -2).squeeze(-2)
            return (q_proj, kv_proj[0], kv_proj[1])
    else:
        w_q, w_k, w_v = w.chunk(3)
        if b is None:
            b_q = b_k = b_v = None
        else:
            b_q, b_k, b_v = b.chunk(3)
        return linear(q, w_q, b_q), linear(k, w_k, b_k), linear(v, w_v, b_v)


def _in_projection(
    q: Var,
    k: Var,
    v: Var,
    w_q: Var,
    w_k: Var,
    w_v: Var,
    b_q: Optional[Var] = None,
    b_k: Optional[Var] = None,
    b_v: Optional[Var] = None,
) -> Tuple[Var, Var, Var]:
    Eq, Ek, Ev = q.size(-1), k.size(-1), v.size(-1)
    assert w_q.shape == (
        Eq, Eq
    ), f"expecting query weights shape of {(Eq, Eq)}, but got {w_q.shape}"
    assert w_k.shape == (
        Eq,
        Ek), f"expecting key weights shape of {(Eq, Ek)}, but got {w_k.shape}"
    assert w_v.shape == (
        Eq, Ev
    ), f"expecting value weights shape of {(Eq, Ev)}, but got {w_v.shape}"
    assert b_q is None or b_q.shape == (
        Eq, ), f"expecting query bias shape of {(Eq,)}, but got {b_q.shape}"
    assert b_k is None or b_k.shape == (
        Eq, ), f"expecting key bias shape of {(Eq,)}, but got {b_k.shape}"
    assert b_v is None or b_v.shape == (
        Eq, ), f"expecting value bias shape of {(Eq,)}, but got {b_v.shape}"
    return linear(q, w_q, b_q), linear(k, w_k, b_k), linear(v, w_v, b_v)


def multi_head_attention_forward(
    query: Var,
    key: Var,
    value: Var,
    embed_dim_to_check: int,
    num_heads: int,
    in_proj_weight: Optional[Var],
    in_proj_bias: Optional[Var],
    bias_k: Optional[Var],
    bias_v: Optional[Var],
    add_zero_attn: bool,
    dropout_p: float,
    out_proj_weight: Var,
    out_proj_bias: Optional[Var],
    training: bool = True,
    key_padding_mask: Optional[Var] = None,
    need_weights: bool = True,
    attn_mask: Optional[Var] = None,
    use_separate_proj_weight: bool = False,
    q_proj_weight: Optional[Var] = None,
    k_proj_weight: Optional[Var] = None,
    v_proj_weight: Optional[Var] = None,
    static_k: Optional[Var] = None,
    static_v: Optional[Var] = None,
    average_attn_weights: bool = True,
    is_causal: bool = False,
) -> Tuple[Var, Optional[Var]]:

    is_batched = _mha_shape_check(query, key, value, key_padding_mask,
                                  attn_mask, num_heads)

    # For unbatched input, we unsqueeze at the expected batch-dim to pretend that the input
    # is batched, run the computation and before returning squeeze the
    # batch dimension so that the output doesn't carry this temporary batch dimension.
    if not is_batched:
        # unsqueeze if the input is unbatched
        query = query.unsqueeze(1)
        key = key.unsqueeze(1)
        value = value.unsqueeze(1)
        if key_padding_mask is not None:
            key_padding_mask = key_padding_mask.unsqueeze(0)

    # set up shape vars
    tgt_len, bsz, embed_dim = query.shape
    src_len, _, _ = key.shape

    key_padding_mask = _canonical_mask(mask=key_padding_mask,
                                       mask_name="key_padding_mask",
                                       other_type=_none_or_dtype(attn_mask),
                                       other_name="attn_mask",
                                       target_type=query.dtype)

    if is_causal and attn_mask is None:
        raise RuntimeError(
            "Need attn_mask if specifying the is_causal hint. "
            "You may use the Transformer module method "
            "`generate_square_subsequent_mask` to create this mask.")

    if is_causal and key_padding_mask is None and not need_weights:
        # when we have a kpm or need weights, we need attn_mask
        # Otherwise, we use the is_causal hint go as is_causal
        # indicator to SDPA.
        attn_mask = None
    else:
        attn_mask = _canonical_mask(
            mask=attn_mask,
            mask_name="attn_mask",
            other_type=None,
            other_name="",
            target_type=query.dtype,
            check_other=False,
        )

        if key_padding_mask is not None:
            # We have the attn_mask, and use that to merge kpm into it.
            # Turn off use of is_causal hint, as the merged mask is no
            # longer causal.
            is_causal = False

    assert embed_dim == embed_dim_to_check, \
        f"was expecting embedding dimension of {embed_dim_to_check}, but got {embed_dim}"
    if isinstance(embed_dim, jt.Var):
        # embed_dim can be a Var when JIT tracing
        head_dim = embed_dim.div(num_heads, rounding_mode='trunc')
    else:
        head_dim = embed_dim // num_heads
    assert head_dim * num_heads == embed_dim, f"embed_dim {embed_dim} not divisible by num_heads {num_heads}"
    if use_separate_proj_weight:
        # allow MHA to have different embedding dimensions when separate projection weights are used
        assert key.shape[:2] == value.shape[:2], \
            f"key's sequence and batch dims {key.shape[:2]} do not match value's {value.shape[:2]}"
    else:
        assert key.shape == value.shape, f"key shape {key.shape} does not match value shape {value.shape}"

    #
    # compute in-projection
    #
    if not use_separate_proj_weight:
        assert in_proj_weight is not None, "use_separate_proj_weight is False but in_proj_weight is None"
        q, k, v = _in_projection_packed(query, key, value, in_proj_weight,
                                        in_proj_bias)
    else:
        assert q_proj_weight is not None, "use_separate_proj_weight is True but q_proj_weight is None"
        assert k_proj_weight is not None, "use_separate_proj_weight is True but k_proj_weight is None"
        assert v_proj_weight is not None, "use_separate_proj_weight is True but v_proj_weight is None"
        if in_proj_bias is None:
            b_q = b_k = b_v = None
        else:
            b_q, b_k, b_v = in_proj_bias.chunk(3)
        q, k, v = _in_projection(query, key, value, q_proj_weight,
                                 k_proj_weight, v_proj_weight, b_q, b_k, b_v)

    # prep attention mask

    if attn_mask is not None:
        # ensure attn_mask's dim is 3
        if attn_mask.dim() == 2:
            correct_2d_size = (tgt_len, src_len)
            if attn_mask.shape != correct_2d_size:
                raise RuntimeError(
                    f"The shape of the 2D attn_mask is {attn_mask.shape}, but should be {correct_2d_size}."
                )
            attn_mask = attn_mask.unsqueeze(0)
        elif attn_mask.dim() == 3:
            correct_3d_size = (bsz * num_heads, tgt_len, src_len)
            if attn_mask.shape != correct_3d_size:
                raise RuntimeError(
                    f"The shape of the 3D attn_mask is {attn_mask.shape}, but should be {correct_3d_size}."
                )
        else:
            raise RuntimeError(
                f"attn_mask's dimension {attn_mask.dim()} is not supported")

    # add bias along batch dimension (currently second)
    if bias_k is not None and bias_v is not None:
        assert static_k is None, "bias cannot be added to static key."
        assert static_v is None, "bias cannot be added to static value."
        k = jt.concat([k, bias_k.repeat(1, bsz, 1)])
        v = jt.concat([v, bias_v.repeat(1, bsz, 1)])
        if attn_mask is not None:
            attn_mask = pad(attn_mask, (0, 1))
        if key_padding_mask is not None:
            key_padding_mask = pad(key_padding_mask, (0, 1))
    else:
        assert bias_k is None
        assert bias_v is None

    #
    # reshape q, k, v for multihead attention and make em batch first
    #
    q = q.view(tgt_len, bsz * num_heads, head_dim).transpose(0, 1)
    if static_k is None:
        k = k.view(k.shape[0], bsz * num_heads, head_dim).transpose(0, 1)
    else:
        # TODO finish disentangling control flow so we don't do in-projections when statics are passed
        assert static_k.size(0) == bsz * num_heads, \
            f"expecting static_k.size(0) of {bsz * num_heads}, but got {static_k.size(0)}"
        assert static_k.size(2) == head_dim, \
            f"expecting static_k.size(2) of {head_dim}, but got {static_k.size(2)}"
        k = static_k
    if static_v is None:
        v = v.view(v.shape[0], bsz * num_heads, head_dim).transpose(0, 1)
    else:
        # TODO finish disentangling control flow so we don't do in-projections when statics are passed
        assert static_v.size(0) == bsz * num_heads, \
            f"expecting static_v.size(0) of {bsz * num_heads}, but got {static_v.size(0)}"
        assert static_v.size(2) == head_dim, \
            f"expecting static_v.size(2) of {head_dim}, but got {static_v.size(2)}"
        v = static_v

    # add zero attention along batch dimension (now first)
    if add_zero_attn:
        zero_attn_shape = (bsz * num_heads, 1, head_dim)
        k = jt.concat([k, jt.zeros(zero_attn_shape, dtype=k.dtype)], dim=1)
        v = jt.concat([v, jt.zeros(zero_attn_shape, dtype=v.dtype)], dim=1)
        if attn_mask is not None:
            attn_mask = pad(attn_mask, (0, 1))
        if key_padding_mask is not None:
            key_padding_mask = pad(key_padding_mask, (0, 1))

    # update source sequence length after adjustments
    src_len = k.size(1)

    # merge key padding and attention masks
    if key_padding_mask is not None:
        assert key_padding_mask.shape == (bsz, src_len), \
            f"expecting key_padding_mask shape of {(bsz, src_len)}, but got {key_padding_mask.shape}"
        key_padding_mask = key_padding_mask.view(bsz, 1, 1, src_len).   \
            expand(-1, num_heads, -1, -1).reshape(bsz * num_heads, 1, src_len)
        if attn_mask is None:
            attn_mask = key_padding_mask
        else:
            attn_mask = attn_mask + key_padding_mask

    # adjust dropout probability
    if not training:
        dropout_p = 0.0

    #
    # (deep breath) calculate attention and out projection
    #

    if need_weights:
        B, Nt, E = q.shape
        q_scaled = q / math.sqrt(E)

        assert not (is_causal and attn_mask is None
                    ), "FIXME: is_causal not implemented for need_weights"

        if attn_mask is not None:
            attn_output_weights = baddbmm(attn_mask, q_scaled,
                                          k.transpose(-2, -1))
        else:
            attn_output_weights = jt.bmm(q_scaled, k.transpose(-2, -1))
        attn_output_weights = softmax(attn_output_weights, dim=-1)
        if dropout_p > 0.0:
            attn_output_weights = dropout(attn_output_weights, p=dropout_p)

        attn_output = jt.bmm(attn_output_weights, v)

        attn_output = attn_output.transpose(0, 1).contiguous().view(
            tgt_len * bsz, embed_dim)
        attn_output = linear(attn_output, out_proj_weight, out_proj_bias)
        attn_output = attn_output.view(tgt_len, bsz, attn_output.size(1))

        # optionally average attention weights over heads
        attn_output_weights = attn_output_weights.view(bsz, num_heads, tgt_len,
                                                       src_len)
        if average_attn_weights:
            attn_output_weights = attn_output_weights.mean(dim=1)

        if not is_batched:
            # squeeze the output if input was unbatched
            attn_output = attn_output.squeeze(1)
            attn_output_weights = attn_output_weights.squeeze(0)
        return attn_output, attn_output_weights
    else:
        # attn_mask can be either (L,S) or (N*num_heads, L, S)
        # if attn_mask's shape is (1, L, S) we need to unsqueeze to (1, 1, L, S)
        # in order to match the input for SDPA of (N, num_heads, L, S)
        if attn_mask is not None:
            if attn_mask.size(0) == 1 and attn_mask.dim() == 3:
                attn_mask = attn_mask.unsqueeze(0)
            else:
                attn_mask = attn_mask.view(bsz, num_heads, -1, src_len)

        q = q.view(bsz, num_heads, tgt_len, head_dim)
        k = k.view(bsz, num_heads, src_len, head_dim)
        v = v.view(bsz, num_heads, src_len, head_dim)

        attn_output = scaled_dot_product_attention(q,
                                                   k,
                                                   v,
                                                   attn_mask,
                                                   dropout_p,
                                                   is_causal,
                                                   training=training)
        attn_output = attn_output.permute(2, 0, 1, 3).contiguous().view(
            bsz * tgt_len, embed_dim)

        attn_output = linear(attn_output, out_proj_weight, out_proj_bias)
        attn_output = attn_output.view(tgt_len, bsz, attn_output.size(1))
        if not is_batched:
            # squeeze the output if input was unbatched
            attn_output = attn_output.squeeze(1)
        return attn_output, None


class MultiheadAttention(Module):
    __constants__ = ['batch_first']
    bias_k: Optional[jt.Var]
    bias_v: Optional[jt.Var]

    def __init__(self,
                 embed_dim,
                 num_heads,
                 dropout=0.,
                 bias=True,
                 add_bias_kv=False,
                 add_zero_attn=False,
                 kdim=None,
                 vdim=None,
                 batch_first=False,
                 dtype=jt.float32) -> None:
        if embed_dim <= 0 or num_heads <= 0:
            raise ValueError(
                f"embed_dim and num_heads must be greater than 0,"
                f" got embed_dim={embed_dim} and num_heads={num_heads} instead"
            )
        factory_kwargs = {'dtype': dtype}
        super().__init__()
        self.embed_dim = embed_dim
        self.kdim = kdim if kdim is not None else embed_dim
        self.vdim = vdim if vdim is not None else embed_dim
        self._qkv_same_embed_dim = self.kdim == embed_dim and self.vdim == embed_dim

        self.num_heads = num_heads
        self.dropout = dropout
        self.batch_first = batch_first
        self.head_dim = embed_dim // num_heads
        assert self.head_dim * num_heads == self.embed_dim, "embed_dim must be divisible by num_heads"

        if not self._qkv_same_embed_dim:
            self.q_proj_weight = jt.empty((embed_dim, embed_dim),
                                          **factory_kwargs)
            self.k_proj_weight = jt.empty((embed_dim, self.kdim),
                                          **factory_kwargs)
            self.v_proj_weight = jt.empty((embed_dim, self.vdim),
                                          **factory_kwargs)
            self.in_proj_weight = None
        else:
            self.q_proj_weight = None
            self.k_proj_weight = None
            self.v_proj_weight = None
            self.in_proj_weight = jt.empty((3 * embed_dim, embed_dim),
                                           **factory_kwargs)

        if bias:
            self.in_proj_bias = jt.empty(3 * embed_dim, **factory_kwargs)
        else:
            self.in_proj_bias = None
        self.out_proj = Linear(embed_dim, embed_dim, bias=bias)

        if add_bias_kv:
            self.bias_k = jt.empty((1, 1, embed_dim), **factory_kwargs)
            self.bias_v = jt.empty((1, 1, embed_dim), **factory_kwargs)
        else:
            self.bias_k = self.bias_v = None

        self.add_zero_attn = add_zero_attn

        self._reset_parameters()

    def _reset_parameters(self):
        if self._qkv_same_embed_dim:
            xavier_uniform_(self.in_proj_weight)
        else:
            xavier_uniform_(self.q_proj_weight)
            xavier_uniform_(self.k_proj_weight)
            xavier_uniform_(self.v_proj_weight)

        if self.in_proj_bias is not None:
            constant_(self.in_proj_bias, 0.)
            constant_(self.out_proj.bias, 0.)
        if self.bias_k is not None:
            xavier_gauss_(self.bias_k)
        if self.bias_v is not None:
            xavier_gauss_(self.bias_v)

    def __setstate__(self, state):
        # Support loading old MultiheadAttention checkpoints generated by v1.1.0
        if '_qkv_same_embed_dim' not in state:
            state['_qkv_same_embed_dim'] = True

        super().__setstate__(state)

    def execute(self,
                query: Var,
                key: Var,
                value: Var,
                key_padding_mask: Optional[Var] = None,
                need_weights: bool = True,
                attn_mask: Optional[Var] = None,
                average_attn_weights: bool = True,
                is_causal: bool = False) -> Tuple[Var, Optional[Var]]:

        #####
        # Fast Path is not Supported.
        #####

        is_batched = query.dim() == 3

        key_padding_mask = _canonical_mask(
            mask=key_padding_mask,
            mask_name="key_padding_mask",
            other_type=_none_or_dtype(attn_mask),
            other_name="attn_mask",
            target_type=query.dtype)

        attn_mask = _canonical_mask(
            mask=attn_mask,
            mask_name="attn_mask",
            other_type=None,
            other_name="",
            target_type=query.dtype,
            check_other=False,
        )

        if self.batch_first and is_batched:
            # make sure that the transpose op does not affect the "is" property
            if key is value:
                if query is key:
                    query = key = value = query.transpose(1, 0)
                else:
                    query, key = (x.transpose(1, 0) for x in (query, key))
                    value = key
            else:
                query, key, value = (x.transpose(1, 0)
                                     for x in (query, key, value))

        if not self._qkv_same_embed_dim:
            attn_output, attn_output_weights = multi_head_attention_forward(
                query,
                key,
                value,
                self.embed_dim,
                self.num_heads,
                self.in_proj_weight,
                self.in_proj_bias,
                self.bias_k,
                self.bias_v,
                self.add_zero_attn,
                self.dropout,
                self.out_proj.weight,
                self.out_proj.bias,
                training=self.is_training(),
                key_padding_mask=key_padding_mask,
                need_weights=need_weights,
                attn_mask=attn_mask,
                use_separate_proj_weight=True,
                q_proj_weight=self.q_proj_weight,
                k_proj_weight=self.k_proj_weight,
                v_proj_weight=self.v_proj_weight,
                average_attn_weights=average_attn_weights,
                is_causal=is_causal)
        else:
            attn_output, attn_output_weights = multi_head_attention_forward(
                query,
                key,
                value,
                self.embed_dim,
                self.num_heads,
                self.in_proj_weight,
                self.in_proj_bias,
                self.bias_k,
                self.bias_v,
                self.add_zero_attn,
                self.dropout,
                self.out_proj.weight,
                self.out_proj.bias,
                training=self.is_training(),
                key_padding_mask=key_padding_mask,
                need_weights=need_weights,
                attn_mask=attn_mask,
                average_attn_weights=average_attn_weights,
                is_causal=is_causal)
        if self.batch_first and is_batched:
            return attn_output.transpose(1, 0), attn_output_weights
        else:
            return attn_output, attn_output_weights