# int8
--reset
--dir=FWD_B,FWD_D
--dt=s8:s8:s32,s8:s8:s8,s8:s8:u8,u8:s8:s32,u8:s8:s8,u8:s8:u8
--attr-scales=src:common:0.25+wei:per_oc+dst:common:2.25
--attr-post-ops=sum:0.5,add:s8:per_tensor,linear:0.5:1.5:2, \
                add:f32:per_oc+sum:0.25+relu:0.5+prelu
--mb=2 --batch=set_gpu
--mb=0 --batch=shapes_0d_gpu

--attr-scales=src:common:0.25+wei:common:0.5+dst:common:2.25
--attr-post-ops=sum:0.5,add:u8:per_tensor,linear:-0.5:0.5:0.5, \
                add:f32:per_oc+abs+sum:0.5+prelu
--batch=shapes_1d

# f32
--reset
--dt=f32
--mb=2
--dir=FWD_B,BWD_D,BWD_WB
--batch=set_gpu

# f32 + post_ops
--reset
--dt=f32
--dir=FWD_B,FWD_D
--attr-post-ops=sum:2+relu:3
--mb=2 --batch=set_gpu
--mb=0 --batch=shapes_0d_gpu

--attr-post-ops=sum:2,linear:3:2,add:f32,prelu:per_oc, \
                sum+add:s8:per_tensor+linear:0.5:1.5:2
--batch=shapes_1d

# bf16
--reset
--dt=bf16
--dir=FWD_B,BWD_D,BWD_WB
--batch=shapes_1d

# f16
--reset
--dt=f16
--dir=FWD_B,BWD_D,BWD_WB
--batch=shapes_1d


# f16
--reset
--mb=0,2
--dt=f16,f16:f16:s8,f16:f16:u8
--dir=FWD_B,FWD_D
--attr-post-ops=linear:1:2,add:f16,add:s8:per_tensor+linear:2:3:0.5,prelu:per_oc
--batch=shapes_1d

# bf16 + f32
--attr-post-ops=

--dir=FWD_B
--dt=bf16:bf16:f32
--batch=shapes_1d

--dir=BWD_D
--dt=f32:bf16:bf16
--batch=shapes_1d

--dir=BWD_WB
--dt=bf16:f32:bf16
--batch=shapes_1d

# f16 + f32

--dir=FWD_B
--dt=f16:f16:f32
--batch=shapes_1d

--dir=BWD_D
--dt=f32:f16:f16
--batch=shapes_1d

--dir=BWD_WB
--dt=f16:f32:f16
--batch=shapes_1d


# bf16 + post_ops
--mb=0
--dir=FWD_B
--dt=bf16,bf16:bf16:f32
--attr-post-ops=sum:0.5,tanh,add:bf16,add:f32:per_tensor+relu:0.5+prelu
--batch=shapes_1d

# diff mem tags
--batch=harness_ip_tag_gpu

# regression tests
--batch=harness_ip_regression

# Test CI in Nightly
--reset
--batch=test_ip_ci

# Test layers of some key and ext GPU DL Frameworks
--reset
--batch=option_set_fwks_key_gpu
--reset
--batch=option_set_fwks_ext_gpu
