minigemini/train/llama_flash_attn_monkey_patch.py

from typing import Optional, Tuple
import warnings

import torch

import transformers
from transformers.models.llama.modeling_llama import apply_rotary_pos_emb, repeat_kv, rotate_half

try:
    from flash_attn.flash_attn_interface import flash_attn_unpadded_qkvpacked_func
except ImportError:
    from flash_attn.flash_attn_interface import flash_attn_varlen_qkvpacked_func as flash_attn_unpadded_qkvpacked_func
from flash_attn.bert_padding import unpad_input, pad_input
from flash_attn import __version__ as flash_attn_version
from flash_attn.flash_attn_interface import (
    flash_attn_func,
    flash_attn_varlen_kvpacked_func,
)


def forward(
    self,
    hidden_states: torch.Tensor,
    attention_mask: Optional[torch.Tensor] = None,
    position_ids: Optional[torch.Tensor] = None,
    past_key_value: Optional[Tuple[torch.Tensor]] = None,
    output_attentions: bool = False,
    use_cache: bool = False,
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
    if output_attentions:
        warnings.warn(
            "Output attentions is not supported for patched `LlamaAttention`, returning `None` instead."
        )

    bsz, q_len, _ = hidden_states.size()

    query_states = (
        self.q_proj(hidden_states)
        .view(bsz, q_len, self.num_heads, self.head_dim)
        .transpose(1, 2)
    )
    key_states = (
        self.k_proj(hidden_states)
        .view(bsz, q_len, self.num_key_value_heads, self.head_dim)
        .transpose(1, 2)
    )
    value_states = (
        self.v_proj(hidden_states)
        .view(bsz, q_len, self.num_key_value_heads, self.head_dim)
        .transpose(1, 2)
    )  # shape: (b, num_heads, s, head_dim)

    kv_seq_len = key_states.shape[-2]
    if past_key_value is not None:
        kv_seq_len += past_key_value[0].shape[-2]

    cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
    query_states, key_states = apply_rotary_pos_emb(
        query_states, key_states, cos, sin, position_ids
    )

    if past_key_value is not None:
        # reuse k, v
        key_states = torch.cat([past_key_value[0], key_states], dim=2)
        value_states = torch.cat([past_key_value[1], value_states], dim=2)

    past_key_value = (key_states, value_states) if use_cache else None

    # repeat k/v heads if n_kv_heads < n_heads
    key_states = repeat_kv(key_states, self.num_key_value_groups)
    value_states = repeat_kv(value_states, self.num_key_value_groups)

    # Transform the data into the format required by flash attention
    qkv = torch.stack([query_states, key_states, value_states], dim=2)
    qkv = qkv.transpose(1, 3)  # shape: [b, s, 3, num_heads, head_dim]
    key_padding_mask = attention_mask

    if key_padding_mask is None:
        qkv = qkv.reshape(-1, 3, self.num_heads, self.head_dim)
        cu_q_lens = torch.arange(
            0, (bsz + 1) * q_len, step=q_len, dtype=torch.int32, device=qkv.device
        )
        max_s = q_len
        output = flash_attn_unpadded_qkvpacked_func(
            qkv, cu_q_lens, max_s, 0.0, softmax_scale=None, causal=True
        )
        output = output.view(bsz, q_len, -1)
    else:
        qkv = qkv.reshape(bsz, q_len, -1)
        qkv, indices, cu_q_lens, max_s = unpad_input(qkv, key_padding_mask)
        qkv = qkv.view(-1, 3, self.num_heads, self.head_dim)
        output_unpad = flash_attn_unpadded_qkvpacked_func(
            qkv, cu_q_lens, max_s, 0.0, softmax_scale=None, causal=True
        )
        output_unpad = output_unpad.reshape(-1, self.num_heads * self.head_dim)
        output = pad_input(output_unpad, indices, bsz, q_len)

    return self.o_proj(output), None, past_key_value

def apply_rotary_pos_emb_inference(q, k, cos_sin, position_ids):
    gather_indices = position_ids[:, :, None, None]  # [bsz, seq_len, 1, 1]
    gather_indices = gather_indices.repeat(
        1, 1, cos_sin[0].shape[1], cos_sin[0].shape[3]
    )
    bsz = gather_indices.shape[0]
    cos, sin = (
        torch.gather(x.transpose(1, 2).repeat(bsz, 1, 1, 1), 1, gather_indices)
        for x in cos_sin
    )
    q, k = ((x * cos) + (rotate_half(x) * sin) for x in (q, k))
    return q, k


def forward_inference(
    self,
    hidden_states: torch.Tensor,
    attention_mask: Optional[torch.Tensor] = None,
    position_ids: Optional[torch.Tensor] = None,
    past_key_value: Optional[Tuple[torch.Tensor]] = None,
    output_attentions: bool = False,
    use_cache: bool = False,
    padding_mask: Optional[torch.Tensor] = None,
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
    if output_attentions:
        warnings.warn(
            "Output attentions is not supported for patched `LlamaAttention`, returning `None` instead."
        )

    bsz, q_len, _ = hidden_states.size()
    kv_heads = getattr(self, "num_key_value_heads", self.num_heads)

    q, k, v = (
        op(hidden_states).view(bsz, q_len, nh, self.head_dim)
        for op, nh in (
            (self.q_proj, self.num_heads),
            (self.k_proj, kv_heads),
            (self.v_proj, kv_heads),
        )
    )
    # shape: (b, s, num_heads, head_dim)

    kv_seq_len = k.shape[1]
    past_kv_len = 0
    if past_key_value is not None:
        past_kv_len = past_key_value[0].shape[2]
        kv_seq_len += past_kv_len

    cos_sin = self.rotary_emb(v, seq_len=kv_seq_len)
    q, k = apply_rotary_pos_emb_inference(q, k, cos_sin, position_ids)

    if past_key_value is not None:
        assert (
            flash_attn_version >= "2.1.0"
        ), "past_key_value support requires flash-attn >= 2.1.0"
        # reuse k, v
        k = torch.cat([past_key_value[0].transpose(1, 2), k], dim=1)
        v = torch.cat([past_key_value[1].transpose(1, 2), v], dim=1)

    past_key_value = (k.transpose(1, 2), v.transpose(1, 2)) if use_cache else None

    if attention_mask is None:
        output = flash_attn_func(q, k, v, 0.0, softmax_scale=None, causal=True).view(
            bsz, q_len, -1
        )
    else:
        q, indices, cu_q_lens, max_s = unpad_input(q, attention_mask[:, -q_len:])
        # We can skip concat and call unpad twice but seems better to call unpad only once.
        kv, _, cu_k_lens, max_k = unpad_input(
            torch.stack((k, v), dim=2), attention_mask
        )
        output_unpad = flash_attn_varlen_kvpacked_func(
            q,
            kv,
            cu_q_lens,
            cu_k_lens,
            max_s,
            max_k,
            0.0,
            softmax_scale=None,
            causal=True,
        )
        output_unpad = output_unpad.reshape(-1, self.num_heads * self.head_dim)
        output = pad_input(output_unpad, indices, bsz, q_len)

    return self.o_proj(output), None, past_key_value


# Disable the transformation of the attention mask in LlamaModel as the flash attention
# requires the attention mask to be the same as the key_padding_mask
def _prepare_decoder_attention_mask(
    self, attention_mask, input_shape, inputs_embeds, past_key_values_length
):
    # [bsz, seq_len]
    return attention_mask


def _prepare_decoder_attention_mask_inference(
    self, attention_mask, input_shape, inputs_embeds, past_key_values_length
):
    # [bsz, seq_len]
    if past_key_values_length > 0 and attention_mask is not None:
        attention_mask = torch.cat(
            (
                torch.full(
                    (input_shape[0], past_key_values_length),
                    True,
                    dtype=attention_mask.dtype,
                    device=attention_mask.device,
                ),
                attention_mask,
            ),
            dim=-1,
        )

    if attention_mask is not None and torch.all(attention_mask):
        return None  # This uses the faster call when training with full samples


def replace_llama_attn_with_flash_attn(inference=False):
    cuda_major, cuda_minor = torch.cuda.get_device_capability()
    if cuda_major < 8:
        warnings.warn(
            "Flash attention is only supported on A100 or H100 GPU during training due to head dim > 64 backward."
            "ref: https://github.com/HazyResearch/flash-attention/issues/190#issuecomment-1523359593"
        )
    if inference:
        transformers.models.llama.modeling_llama.LlamaModel._prepare_decoder_attention_mask = _prepare_decoder_attention_mask_inference
        transformers.models.llama.modeling_llama.LlamaAttention.forward = forward_inference
    else:
        transformers.models.llama.modeling_llama.LlamaModel._prepare_decoder_attention_mask = (
            _prepare_decoder_attention_mask
        )
        transformers.models.llama.modeling_llama.LlamaAttention.forward = forward