def run_b(self, mod, inputs):
mod.eval()
try:
mod = self.lower_fn(mod, inputs, self.max_batch_size)
output = mod(*inputs)
except RuntimeError as e:
raise net_min_base.FxNetMinimizerRunFuncError(
f"Encounter an error when processing \n{mod.graph}\n {e}")
else:
return output
def run_b(self, mod, inputs):
"""
Here we actually run mod on TensorRT return TensorRT result.
"""
mod.eval()
try:
mod = lower_mod_to_trt(mod, inputs)
output = mod(*inputs)
except RuntimeError as e:
raise net_min_base.FxNetMinimizerRunFuncError(
f"Encounter an error when processing \n{mod.graph}\n {e}")
else:
return output