Merge pull request #299 from OpenAccess-AI-Collective/flash-attention-2

Flash attention 2

docker/Dockerfile-base

@@ -40,7 +40,7 @@ ARG TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6+PTX"
 
 RUN git clone https://github.com/Dao-AILab/flash-attention.git && \
     cd flash-attention && \
-    git checkout v1.0.9 && \
+    git checkout 9ee0ff1  && \
     python3 setup.py bdist_wheel && \
     cd csrc/fused_dense_lib && \
     python3 setup.py bdist_wheel && \

src/axolotl/flash_attn.py

@@ -8,7 +8,7 @@ import torch
 import transformers
 from einops import rearrange
 from flash_attn.bert_padding import pad_input, unpad_input
-from flash_attn.flash_attn_interface import flash_attn_unpadded_qkvpacked_func
+from flash_attn.flash_attn_interface import flash_attn_varlen_qkvpacked_func
 from transformers.models.llama.modeling_llama import apply_rotary_pos_emb
 
 
@@ -79,7 +79,7 @@ def forward(
             dtype=torch.int32,
             device=qkv.device,
         )
-        output = flash_attn_unpadded_qkvpacked_func(
+        output = flash_attn_varlen_qkvpacked_func(
             qkv, cu_q_lens, max_s, 0.0, softmax_scale=None, causal=True
         )
         output = rearrange(output, "(b s) ... -> b s ...", b=bsz)
@@ -95,7 +95,7 @@ def forward(
             three=3,
             h=nheads,
         )
-        output_unpad = flash_attn_unpadded_qkvpacked_func(
+        output_unpad = flash_attn_varlen_qkvpacked_func(
             x_unpad,
             cu_q_lens,
             max_s,