未验证 提交 07b75a88 编写于 作者: K Kaipeng Deng 提交者: GitHub

fix switch_ir_optim in deploy/infer (#2230)

上级 79e2436a
...@@ -316,7 +316,11 @@ def load_predictor(model_dir, ...@@ -316,7 +316,11 @@ def load_predictor(model_dir,
# initial GPU memory(M), device ID # initial GPU memory(M), device ID
config.enable_use_gpu(200, 0) config.enable_use_gpu(200, 0)
# optimize graph and fuse op # optimize graph and fuse op
config.switch_ir_optim(True) # FIXME(dkp): ir optimize may prune variable inside graph
# and incur error in Paddle 2.0, e.g. in SSDLite
# FCOS model, set as False currently and should
# be set as True after switch_ir_optim fixed
config.switch_ir_optim(False)
else: else:
config.disable_gpu() config.disable_gpu()
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册