Anditty

[Question]: 如何把预训练模型转换成inference_model保存下以供推理使用

1

### 请提出你的问题 我正在尝试PaddleNLP/examples/code_generation/codegen/中的例子，但是我发现这个模型是通过paddle.save保存的训练模型，但是我想把该codegen模型转换成一个inference mode部署我的推理服务，我不知道如何将该模型转换成inference mode。

[BUG] CGNN run() Wrong way to calculate the score

1

Code: ```python dataloader = DataLoader(dataset, batch_size=self.batch_size, shuffle=True, drop_last=True, num_workers=dataloader_workers) with trange(train_epochs + test_epochs, disable=not verbose) as t: for epoch in t: for i, data in enumerate(dataloader): optim.zero_grad() generated_data = self.forward()...

bug: estimate_ate() will return nan if backdoor is null.

2

For example, there is a dag looks like: [('T','B'),('B','Y'),('Z','Y'),('Z','T')] Then if we want to use estimate_ate() to get the ate from T to Y. Then estimate_ate() will return nan since...

How should I calculate P(Y|do(X),Z∈{1,2,3}) ？

2

## Subject of the issue In class `pgmpy.inference.CausalInference.CausalInference` , I can use function `query(variables, do=None, evidence=None, adjustment_set=None, inference_algo='ve', show_progress=True, **kwargs)` to calculate P(Y|do(X),Z=1), but I can't calculate P(Y|do(X),Z∈{1,2,3}) bucause the...

bug: get empty state dict

1

I follow the step in README, but I get the empty state dict. Here is the code and the output: code: ```python trainer = Trainer(model=model, tokenizer=tokenizer, args=training_args, **data_module) print(trainer.model.state_dict()['model.layers.30.mlp.gate_proj.weight']) print('training')...

请问在推理阶段，为什么需要对input部份计算注意力呢？

1

https://github.com/bojone/rerope/blob/3710a359ae0a80aad98312c9194898fb7312dcf3/rerope_patch.py#L75C5-L75C10 请问在这段代码中，为什么在推理阶段需要对input部份计算注意力，为什么不能直接使用 if q_len == 1 中的逻辑呢？