【PPSCI Export&Infer No.23】viv

[smallpoxscattered]

PR types

Others

PR changes

Others

Describe

• 【快乐开源】为 PaddleScience 案例添加 export 和 inference 功能 #788
  因为需要将equation导出 就在 ComposedNode 加入了paddle.nn.LayerList，不然导出找不到里面的参数
  equation导出时，模型裁剪也会出问题，我就先跳过了

[paddle-bot]

Thanks for your contribution!

[HydrogenSulfate]

感谢提交PR，辛苦修改一下

[HydrogenSulfate]

input_keys可以使用omegaconf的引用语法：input_keys: ${MODEL.input_keys}，保持跟MODEL.inputs_keys字段一致

[HydrogenSulfate]

我尝试优化了下你的代码，可以按照如下修改，配合这个PR就能跑通了： #835

Suggested change

	def export(cfg: DictConfig):
	# set model
	model = ppsci.arch.MLP(**cfg.MODEL)
	# initialize equation
	equation = {"VIV": ppsci.equation.Vibration(2, -4, 0)}
	# initialize solver
	solver = ppsci.solver.Solver(
	model,
	equation=equation,
	pretrained_model_path=cfg.INFER.pretrained_model_path,
	)
	# Convert equation to func
	funcs = ppsci.lambdify(
	solver.equation["VIV"].equations["f"],
	solver.model,
	list(solver.equation["VIV"].learnable_parameters),
	)
	def wrap_prediction_to_dict(instance, func):
	def wrapper(instance, *args, **kwargs):
	result = func(*args, **kwargs)
	return {"f": result}
	if hasattr(func, "__func__"):
	wrapper.__func__ = func.__func__
	return wrapper
	def wrap_forward_methods(instance):
	instance.input_keys = cfg.MODEL.input_keys
	instance.output_keys = ["f"]
	for attr_name in dir(instance):
	if attr_name == "forward":
	attr = getattr(instance, attr_name)
	setattr(instance, attr_name, wrap_prediction_to_dict(instance, attr))
	return instance
	eqn = wrap_forward_methods(funcs)
	# Combine the two instances
	models = ppsci.arch.ModelList((solver.model, eqn))
	# export models
	from paddle.static import InputSpec
	input_spec = [
	{key: InputSpec([None, 1], "float32", name=key) for key in model.input_keys},
	]
	from paddle import jit
	jit.enable_to_static(True)
	static_model = jit.to_static(
	models,
	input_spec=input_spec,
	full_graph=True,
	)
	jit.save(static_model, cfg.INFER.export_path, skip_prune_program=True)
	jit.enable_to_static(False)
	def export(cfg: DictConfig):
	from paddle import nn
	from paddle.static import InputSpec
	# set model
	model = ppsci.arch.MLP(**cfg.MODEL)
	# initialize equation
	equation = {"VIV": ppsci.equation.Vibration(2, -4, 0)}
	# initialize solver
	solver = ppsci.solver.Solver(
	model,
	equation=equation,
	pretrained_model_path=cfg.INFER.pretrained_model_path,
	)
	# Convert equation to func
	f_func = ppsci.lambdify(
	solver.equation["VIV"].equations["f"],
	solver.model,
	list(solver.equation["VIV"].learnable_parameters),
	)
	class Wrapped_Model(nn.Layer):
	def __init__(self, model, func):
	super().__init__()
	self.model = model
	self.func = func
	def forward(self, x):
	model_out = self.model(x)
	func_out = self.func(x)
	return {**model_out, "f": func_out}
	solver.model = Wrapped_Model(model, f_func)
	# export models
	input_spec = [
	{key: InputSpec([None, 1], "float32", name=key) for key in model.input_keys},
	]
	solver.export(input_spec, cfg.INFER.export_path, skip_prune_program=True)

[HydrogenSulfate]

这里可以改为：https://paddle-org.bj.bcebos.com/paddlescience/models/viv/viv_pretrained.pdparams，我在 https://github.com/PaddlePaddle/PaddleScience/pull/834/files#diff-68effb6a6b046bfbf5b6cb8b843f284e8779eadfa5977b66870a783d6d61ebe7R105-R118 支持了自动下载方程参数文件的功能

[HydrogenSulfate]

"f"双引号

[HydrogenSulfate]

这两行在pretrained_model_path改完之后就可以删除了

[smallpoxscattered]

好的好的 改好了 辛苦老师了

[HydrogenSulfate]

LGTM