回答此问题可获得 20 贡献值,回答如果被采纳可获得 50 分。
<p><a href="https://pytorch-lightning.readthedocs.io/en/0.8.3/metrics.html#tensormetric" rel="nofollow noreferrer">The official doc</a>仅限国家</p>
<pre><code>>>> from pytorch_lightning.metrics import ConfusionMatrix
>>> target = torch.tensor([1, 1, 0, 0])
>>> preds = torch.tensor([0, 1, 0, 0])
>>> confmat = ConfusionMatrix(num_classes=2)
>>> confmat(preds, target)
</code></pre>
<p>这并没有说明如何在框架中使用度量</p>
<p>我的尝试(方法不完整,仅显示相关部分):</p>
<pre><code>def __init__(...):
self.val_confusion = pl.metrics.classification.ConfusionMatrix(num_classes=self._config.n_clusters)
def validation_step(self, batch, batch_index):
...
log_probs = self.forward(orig_batch)
loss = self._criterion(log_probs, label_batch)
self.val_confusion.update(log_probs, label_batch)
self.log('validation_confusion_step', self.val_confusion, on_step=True, on_epoch=False)
def validation_step_end(self, outputs):
return outputs
def validation_epoch_end(self, outs):
self.log('validation_confusion_epoch', self.val_confusion.compute())
</code></pre>
<p>在第0个纪元之后,这将给出</p>
<pre><code> Traceback (most recent call last):
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 521, in train
self.train_loop.run_training_epoch()
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\training_loop.py", line 588, in run_training_epoch
self.trainer.run_evaluation(test_mode=False)
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\trainer.py", line 613, in run_evaluation
self.evaluation_loop.log_evaluation_step_metrics(output, batch_idx)
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\evaluation_loop.py", line 346, in log_evaluation_step_metrics
self.__log_result_step_metrics(step_log_metrics, step_pbar_metrics, batch_idx)
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\evaluation_loop.py", line 350, in __log_result_step_metrics
cached_batch_pbar_metrics, cached_batch_log_metrics = cached_results.update_logger_connector()
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\connectors\logger_connector\epoch_result_store.py", line 378, in update_logger_connector
batch_log_metrics = self.get_latest_batch_log_metrics()
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\connectors\logger_connector\epoch_result_store.py", line 418, in get_latest_batch_log_metrics
batch_log_metrics = self.run_batch_from_func_name("get_batch_log_metrics")
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\connectors\logger_connector\epoch_result_store.py", line 414, in run_batch_from_func_name
results = [func(include_forked_originals=False) for func in results]
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\connectors\logger_connector\epoch_result_store.py", line 414, in <listcomp>
results = [func(include_forked_originals=False) for func in results]
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\connectors\logger_connector\epoch_result_store.py", line 122, in get_batch_log_metrics
return self.run_latest_batch_metrics_with_func_name("get_batch_log_metrics",
*args, **kwargs)
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\connectors\logger_connector\epoch_result_store.py", line 115, in run_latest_batch_metrics_with_func_name
for dl_idx in range(self.num_dataloaders)
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\connectors\logger_connector\epoch_result_store.py", line 115, in <listcomp>
for dl_idx in range(self.num_dataloaders)
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\trainer\connectors\logger_connector\epoch_result_store.py", line 100, in get_latest_from_func_name
results.update(func(*args, add_dataloader_idx=add_dataloader_idx, **kwargs))
File "C:\code\EPMD\Kodex\Templates\Testing\venv\lib\site-packages\pytorch_lightning\core\step_result.py", line 298, in get_batch_log_metrics
result[dl_key] = self[k]._forward_cache.detach()
AttributeError: 'NoneType' object has no attribute 'detach'
</code></pre>
<p>培训前,它确实通过了健全性验证检查</p>
<p>失败发生在<code>validation_step_end</code>中的返回上。对我来说没什么意义</p>
<p>使用mertics的完全相同的方法可以精确地工作</p>
<p><strong>如何获得正确的混淆矩阵?</strong></p>