fix for flash attn w mistral w/o sammple packing (#648)

src/axolotl/monkeypatch/mistral_attn_hijack_flash.py

@@ -2,13 +2,17 @@
 # pylint: disable=duplicate-code
 
 import logging
-import math
 from typing import List, Optional, Tuple, Union
 
 import torch
 import transformers
 from einops import rearrange
-from torch import nn
+from flash_attn.bert_padding import pad_input, unpad_input
+from flash_attn.flash_attn_interface import (  # pylint: disable=ungrouped-imports
+    flash_attn_kvpacked_func,
+    flash_attn_varlen_kvpacked_func,
+    flash_attn_varlen_qkvpacked_func,
+)
 from transformers.modeling_outputs import BaseModelOutputWithPast
 from transformers.models.mistral.modeling_mistral import (
     MistralDecoderLayer as OriginalMistralDecoderLayer,
@@ -17,16 +21,6 @@ from transformers.models.mistral.modeling_mistral import apply_rotary_pos_emb, r
 
 from axolotl.monkeypatch.utils import get_cu_seqlens_from_pos_ids
 
-try:
-    from flash_attn.flash_attn_interface import (  # pylint: disable=ungrouped-imports
-        flash_attn_varlen_qkvpacked_func,
-    )
-except ImportError:
-    from flash_attn.flash_attn_interface import (
-        flash_attn_unpadded_qkvpacked_func as flash_attn_varlen_qkvpacked_func,
-    )
-
-
 LOG = logging.getLogger("axolotl.monkeypatch.mistral")
 
 
@@ -108,6 +102,15 @@ def flashattn_forward(
     key_states = repeat_kv(key_states, self.num_key_value_groups)
     value_states = repeat_kv(value_states, self.num_key_value_groups)
 
+    if self.training:
+        # during training q,k,v always have same seqlen
+        assert key_states.shape == query_states.shape
+        is_causal = True
+    else:
+        # turn off FA causal mask after first inference autoregressive iteration
+        # only on first autoregressive step q,k,v have same seqlen
+        is_causal = key_states.shape == query_states.shape
+
     if cu_seqlens is not None and max_seqlen is not None and cu_seqlens.dim() == 1:
         # special handling using sample packing
         qkv = torch.stack(
@@ -120,46 +123,84 @@ def flashattn_forward(
             qkv, cu_seqlens, max_seqlen, 0.0, softmax_scale=None, causal=True
         )
         output = rearrange(output, "(b s) ... -> b s ...", b=bsz)
-        attn_output = output
-        if attn_output.size() != (bsz, q_len, self.num_heads, self.head_dim):
-            raise ValueError(
-                f"`attn_output` should be of size {(bsz, q_len, self.num_heads, self.head_dim)}, but is"
-                f" {attn_output.size()}"
-            )
-        attn_output = rearrange(attn_output, "b s h d -> b s (h d)")
-        attn_weights = None
+    elif query_states.shape == key_states.shape:
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+        qkv_unpad, cu_seqlens_q, max_seqlen_q, _, output_pad_fn = generate_qkv(
+            query_states,
+            key_states,
+            value_states,
+            qkvpacked=True,
+            # We have disabled _prepare_decoder_attention_mask in LlamaModel
+            # the attention_mask should be the same as the key_padding_mask
+            key_padding_mask=attention_mask,
+            query_padding_mask=attention_mask[:, -query_states.size(1) :]
+            if attention_mask is not None
+            else None,
+        )
+        output_unpad = flash_attn_varlen_qkvpacked_func(
+            qkv_unpad,
+            cu_seqlens_q,
+            max_seqlen_q,
+            0.0,
+            softmax_scale=None,
+            causal=is_causal,
+        )
+        output = output_pad_fn(output_unpad)
     else:
-        attn_weights = torch.matmul(
-            query_states, key_states.transpose(2, 3)
-        ) / math.sqrt(self.head_dim)
-        if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
-            raise ValueError(
-                f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is"
-                f" {attn_weights.size()}"
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+        if attention_mask is None or attention_mask.all().item():
+            output = flash_attn_kvpacked_func(
+                query_states,
+                torch.stack([key_states, value_states], 2),
+                causal=is_causal,
             )
-
-        if attention_mask is not None:
-            if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
-                raise ValueError(
-                    f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
-                )
-
-            attn_weights = attn_weights + attention_mask
-
-        # upcast attention to fp32
-        attn_weights = nn.functional.softmax(
-            attn_weights, dim=-1, dtype=torch.float32
-        ).to(query_states.dtype)
-        attn_output = torch.matmul(attn_weights, value_states)
-
-        if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
-            raise ValueError(
-                f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
-                f" {attn_output.size()}"
+        else:
+            (  # pylint: disable=unbalanced-tuple-unpacking
+                q_unpad,
+                kv_unpad,
+                cu_seqlens_q,
+                cu_seqlens_k,
+                max_seqlen_q,
+                max_seqlen_k,
+                _,
+                _,
+                output_pad_fn,
+            ) = generate_qkv(
+                query_states,
+                key_states,
+                value_states,
+                kvpacked=True,
+                key_padding_mask=attention_mask,
+                query_padding_mask=attention_mask[:, -query_states.size(1) :]
+                if attention_mask is not None
+                else None,
             )
+            if q_unpad.dtype != kv_unpad.dtype:
+                kv_unpad = kv_unpad.to(q_unpad.dtype)
+            output_unpad = flash_attn_varlen_kvpacked_func(
+                q_unpad,
+                kv_unpad,
+                cu_seqlens_q,
+                cu_seqlens_k,
+                max_seqlen_q,
+                max_seqlen_k,
+                0.0,
+                softmax_scale=None,
+                causal=is_causal,
+            )
+            output = output_pad_fn(output_unpad)
 
-        attn_output = attn_output.transpose(1, 2).contiguous()
-        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+    attn_output = output
+    if attn_output.size() != (bsz, q_len, self.num_heads, self.head_dim):
+        raise ValueError(
+            f"`attn_output` should be of size {(bsz, q_len, self.num_heads, self.head_dim)}, but is"
+            f" {attn_output.size()}"
+        )
+    attn_output = rearrange(attn_output, "b s h d -> b s (h d)")
 
     attn_output = self.o_proj(attn_output)
 
@@ -169,6 +210,105 @@ def flashattn_forward(
     return attn_output, attn_weights, past_key_value
 
 
+# based on https://github.com/Dao-AILab/flash-attention/blob/364a5b/tests/test_flash_attn.py#L38
+def generate_qkv(
+    q,
+    k,
+    v,
+    query_padding_mask=None,
+    key_padding_mask=None,
+    kvpacked=False,
+    qkvpacked=False,
+):  # pylint: disable=invalid-name,unnecessary-lambda-assignment
+    """
+    Arguments:
+        q: (batch_size, seqlen_q, nheads, d)
+        k: (batch_size, seqlen_k, nheads_k, d)
+        v: (batch_size, seqlen_k, nheads_k, d)
+        query_padding_mask: (batch_size, seqlen), bool
+        key_padding_mask: (batch_size, seqlen), bool
+    """
+    assert not (kvpacked and qkvpacked)
+    batch_size, seqlen_q, nheads, d = q.shape
+    _, seqlen_k, nheads_k, _ = k.shape
+    assert k.shape == (batch_size, seqlen_k, nheads_k, d)
+    assert v.shape == (batch_size, seqlen_k, nheads_k, d)
+
+    if query_padding_mask is not None:
+        q_unpad, indices_q, cu_seqlens_q, max_seqlen_q = unpad_input(
+            q, query_padding_mask
+        )
+
+        output_pad_fn = lambda output_unpad: pad_input(  # noqa: E731
+            output_unpad, indices_q, batch_size, seqlen_q
+        )
+
+    else:
+        q_unpad = rearrange(q, "b s h d -> (b s) h d")
+        cu_seqlens_q = torch.arange(
+            0,
+            (batch_size + 1) * seqlen_q,
+            step=seqlen_q,
+            dtype=torch.int32,
+            device=q_unpad.device,
+        )
+        max_seqlen_q = seqlen_q
+
+        output_pad_fn = lambda output_unpad: rearrange(  # noqa: E731
+            output_unpad, "(b s) h d -> b s h d", b=batch_size
+        )
+
+    if key_padding_mask is not None:
+        k_unpad, _, cu_seqlens_k, max_seqlen_k = unpad_input(k, key_padding_mask)
+        v_unpad, _, _, _ = unpad_input(v, key_padding_mask)
+    else:
+        k_unpad = rearrange(k, "b s h d -> (b s) h d")
+        v_unpad = rearrange(v, "b s h d -> (b s) h d")
+        cu_seqlens_k = torch.arange(
+            0,
+            (batch_size + 1) * seqlen_k,
+            step=seqlen_k,
+            dtype=torch.int32,
+            device=k_unpad.device,
+        )
+        max_seqlen_k = seqlen_k
+
+    if qkvpacked:
+        assert nheads == nheads_k
+        qkv_unpad = torch.stack([q_unpad, k_unpad, v_unpad], dim=1)
+        qkv = torch.stack([q, k, v], dim=2)
+        return (qkv_unpad, cu_seqlens_q, max_seqlen_q, qkv, output_pad_fn)
+
+    if kvpacked:
+        kv_unpad = torch.stack([k_unpad, v_unpad], dim=1)
+        kv = torch.stack([k, v], dim=2)
+        return (
+            q_unpad,
+            kv_unpad,
+            cu_seqlens_q,
+            cu_seqlens_k,
+            max_seqlen_q,
+            max_seqlen_k,
+            q,
+            kv,
+            output_pad_fn,
+        )
+
+    return (
+        q_unpad,
+        k_unpad,
+        v_unpad,
+        cu_seqlens_q,
+        cu_seqlens_k,
+        max_seqlen_q,
+        max_seqlen_k,
+        q,
+        k,
+        v,
+        output_pad_fn,
+    )
+
+
 def mistral_model_forward(
     self,
     input_ids: torch.LongTensor = None,