Merge branch 'master' into wengshiy/accelerator_align

microsoft · May 15, 2024 · 3562b68 · 3562b68
2 parents 93b8b00 + ebf82e8
commit 3562b68
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 2 deletions.
diff --git a/MANIFEST.in b/MANIFEST.in
@@ -2,7 +2,7 @@ include *.txt README.md
 include deepspeed/inference/v2/kernels/ragged_ops/libs/*.so
 include deepspeed/inference/v2/kernels/cutlass_ops/libs/*.so
 recursive-include requirements *.txt
-recursive-include deepspeed *.cpp *.h *.cu *.hip *.tr *.cuh *.cc *.json
+recursive-include deepspeed *.cpp *.h *.hpp *.cu *.hip *.tr *.cuh *.cc *.json
 recursive-include csrc *.cpp *.h *.hpp *.cu *.tr *.cuh *.cc
 recursive-include op_builder *.py
 recursive-include benchmarks *.py

diff --git a/deepspeed/module_inject/auto_tp.py b/deepspeed/module_inject/auto_tp.py
@@ -307,7 +307,7 @@ def tp_parser(model):
                 # Mixtral-7x8b used w2*act(w1*w3) linear. need to replace w2 to linearallreduce.
                 elif 'w2' in layer and 'Mixtral' in str(type(module)):
                     gem_list = gem_list + [layer]
-                elif "self_attn.dense" in layer and "Phi" in str(type(module)):
+                elif 'self_attn.dense' in layer and 'Phi' in str(type(module)):
                     gem_list = gem_list + [layer]
 
             layer_list = []