Merge pull request graviraja#18 from graviraja/metric_fix

updated metric calls

Author: graviraja

week_1_wandb_logging/model.py

@@ -19,7 +19,8 @@ def __init__(self, model_name="google/bert_uncased_L-2_H-128_A-2", lr=3e-5):
  model_name, num_labels=2
  )
  self.num_classes = 2
- self.accuracy_metric = torchmetrics.Accuracy()
+ self.train_accuracy_metric = torchmetrics.Accuracy()
+ self.val_accuracy_metric = torchmetrics.Accuracy()
  self.f1_metric = torchmetrics.F1(num_classes=self.num_classes)
  self.precision_macro_metric = torchmetrics.Precision(
  average="macro", num_classes=self.num_classes
@@ -42,7 +43,7 @@ def training_step(self, batch, batch_idx):
  )
  # loss = F.cross_entropy(logits, batch["label"])
  preds = torch.argmax(outputs.logits, 1)
- train_acc = self.accuracy_metric(preds, batch["label"])
+ train_acc = self.train_accuracy_metric(preds, batch["label"])
  self.log("train/loss", outputs.loss, prog_bar=True, on_epoch=True)
  self.log("train/acc", train_acc, prog_bar=True, on_epoch=True)
  return outputs.loss
@@ -55,7 +56,7 @@ def validation_step(self, batch, batch_idx):
  preds = torch.argmax(outputs.logits, 1)
 
  # Metrics
- valid_acc = self.accuracy_metric(preds, labels)
+ valid_acc = self.val_accuracy_metric(preds, labels)
  precision_macro = self.precision_macro_metric(preds, labels)
  recall_macro = self.recall_macro_metric(preds, labels)
  precision_micro = self.precision_micro_metric(preds, labels)

week_2_hydra_config/model.py

@@ -21,7 +21,8 @@ def __init__(self, model_name="google/bert_uncased_L-2_H-128_A-2", lr=3e-5):
  model_name, num_labels=2
  )
  self.num_classes = 2
- self.accuracy_metric = torchmetrics.Accuracy()
+ self.train_accuracy_metric = torchmetrics.Accuracy()
+ self.val_accuracy_metric = torchmetrics.Accuracy()
  self.f1_metric = torchmetrics.F1(num_classes=self.num_classes)
  self.precision_macro_metric = torchmetrics.Precision(
  average="macro", num_classes=self.num_classes
@@ -44,7 +45,7 @@ def training_step(self, batch, batch_idx):
  )
  # loss = F.cross_entropy(logits, batch["label"])
  preds = torch.argmax(outputs.logits, 1)
- train_acc = self.accuracy_metric(preds, batch["label"])
+ train_acc = self.train_accuracy_metric(preds, batch["label"])
  self.log("train/loss", outputs.loss, prog_bar=True, on_epoch=True)
  self.log("train/acc", train_acc, prog_bar=True, on_epoch=True)
  return outputs.loss
@@ -57,7 +58,7 @@ def validation_step(self, batch, batch_idx):
  preds = torch.argmax(outputs.logits, 1)
 
  # Metrics
- valid_acc = self.accuracy_metric(preds, labels)
+ valid_acc = self.val_accuracy_metric(preds, labels)
  precision_macro = self.precision_macro_metric(preds, labels)
  recall_macro = self.recall_macro_metric(preds, labels)
  precision_micro = self.precision_micro_metric(preds, labels)

week_3_dvc/model.py

@@ -21,7 +21,8 @@ def __init__(self, model_name="google/bert_uncased_L-2_H-128_A-2", lr=3e-5):
  model_name, num_labels=2
  )
  self.num_classes = 2
- self.accuracy_metric = torchmetrics.Accuracy()
+ self.train_accuracy_metric = torchmetrics.Accuracy()
+ self.val_accuracy_metric = torchmetrics.Accuracy()
  self.f1_metric = torchmetrics.F1(num_classes=self.num_classes)
  self.precision_macro_metric = torchmetrics.Precision(
  average="macro", num_classes=self.num_classes
@@ -44,7 +45,7 @@ def training_step(self, batch, batch_idx):
  )
  # loss = F.cross_entropy(logits, batch["label"])
  preds = torch.argmax(outputs.logits, 1)
- train_acc = self.accuracy_metric(preds, batch["label"])
+ train_acc = self.train_accuracy_metric(preds, batch["label"])
  self.log("train/loss", outputs.loss, prog_bar=True, on_epoch=True)
  self.log("train/acc", train_acc, prog_bar=True, on_epoch=True)
  return outputs.loss
@@ -57,7 +58,7 @@ def validation_step(self, batch, batch_idx):
  preds = torch.argmax(outputs.logits, 1)
 
  # Metrics
- valid_acc = self.accuracy_metric(preds, labels)
+ valid_acc = self.val_accuracy_metric(preds, labels)
  precision_macro = self.precision_macro_metric(preds, labels)
  recall_macro = self.recall_macro_metric(preds, labels)
  precision_micro = self.precision_micro_metric(preds, labels)

week_4_onnx/model.py

@@ -21,7 +21,8 @@ def __init__(self, model_name="google/bert_uncased_L-2_H-128_A-2", lr=3e-5):
  model_name, num_labels=2
  )
  self.num_classes = 2
- self.accuracy_metric = torchmetrics.Accuracy()
+ self.train_accuracy_metric = torchmetrics.Accuracy()
+ self.val_accuracy_metric = torchmetrics.Accuracy()
  self.f1_metric = torchmetrics.F1(num_classes=self.num_classes)
  self.precision_macro_metric = torchmetrics.Precision(
  average="macro", num_classes=self.num_classes
@@ -44,7 +45,7 @@ def training_step(self, batch, batch_idx):
  )
  # loss = F.cross_entropy(logits, batch["label"])
  preds = torch.argmax(outputs.logits, 1)
- train_acc = self.accuracy_metric(preds, batch["label"])
+ train_acc = self.train_accuracy_metric(preds, batch["label"])
  self.log("train/loss", outputs.loss, prog_bar=True, on_epoch=True)
  self.log("train/acc", train_acc, prog_bar=True, on_epoch=True)
  return outputs.loss
@@ -57,7 +58,7 @@ def validation_step(self, batch, batch_idx):
  preds = torch.argmax(outputs.logits, 1)
 
  # Metrics
- valid_acc = self.accuracy_metric(preds, labels)
+ valid_acc = self.val_accuracy_metric(preds, labels)
  precision_macro = self.precision_macro_metric(preds, labels)
  recall_macro = self.recall_macro_metric(preds, labels)
  precision_micro = self.precision_micro_metric(preds, labels)

week_5_docker/model.py

@@ -21,7 +21,8 @@ def __init__(self, model_name="google/bert_uncased_L-2_H-128_A-2", lr=3e-5):
  model_name, num_labels=2
  )
  self.num_classes = 2
- self.accuracy_metric = torchmetrics.Accuracy()
+ self.train_accuracy_metric = torchmetrics.Accuracy()
+ self.val_accuracy_metric = torchmetrics.Accuracy()
  self.f1_metric = torchmetrics.F1(num_classes=self.num_classes)
  self.precision_macro_metric = torchmetrics.Precision(
  average="macro", num_classes=self.num_classes
@@ -44,7 +45,7 @@ def training_step(self, batch, batch_idx):
  )
  # loss = F.cross_entropy(logits, batch["label"])
  preds = torch.argmax(outputs.logits, 1)
- train_acc = self.accuracy_metric(preds, batch["label"])
+ train_acc = self.train_accuracy_metric(preds, batch["label"])
  self.log("train/loss", outputs.loss, prog_bar=True, on_epoch=True)
  self.log("train/acc", train_acc, prog_bar=True, on_epoch=True)
  return outputs.loss
@@ -57,7 +58,7 @@ def validation_step(self, batch, batch_idx):
  preds = torch.argmax(outputs.logits, 1)
 
  # Metrics
- valid_acc = self.accuracy_metric(preds, labels)
+ valid_acc = self.val_accuracy_metric(preds, labels)
  precision_macro = self.precision_macro_metric(preds, labels)
  recall_macro = self.recall_macro_metric(preds, labels)
  precision_micro = self.precision_micro_metric(preds, labels)

week_6_github_actions/model.py

@@ -21,7 +21,8 @@ def __init__(self, model_name="google/bert_uncased_L-2_H-128_A-2", lr=3e-5):
  model_name, num_labels=2
  )
  self.num_classes = 2
- self.accuracy_metric = torchmetrics.Accuracy()
+ self.train_accuracy_metric = torchmetrics.Accuracy()
+ self.val_accuracy_metric = torchmetrics.Accuracy()
  self.f1_metric = torchmetrics.F1(num_classes=self.num_classes)
  self.precision_macro_metric = torchmetrics.Precision(
  average="macro", num_classes=self.num_classes
@@ -44,7 +45,7 @@ def training_step(self, batch, batch_idx):
  )
  # loss = F.cross_entropy(logits, batch["label"])
  preds = torch.argmax(outputs.logits, 1)
- train_acc = self.accuracy_metric(preds, batch["label"])
+ train_acc = self.train_accuracy_metric(preds, batch["label"])
  self.log("train/loss", outputs.loss, prog_bar=True, on_epoch=True)
  self.log("train/acc", train_acc, prog_bar=True, on_epoch=True)
  return outputs.loss
@@ -57,7 +58,7 @@ def validation_step(self, batch, batch_idx):
  preds = torch.argmax(outputs.logits, 1)
 
  # Metrics
- valid_acc = self.accuracy_metric(preds, labels)
+ valid_acc = self.val_accuracy_metric(preds, labels)
  precision_macro = self.precision_macro_metric(preds, labels)
  recall_macro = self.recall_macro_metric(preds, labels)
  precision_micro = self.precision_micro_metric(preds, labels)

week_7_ecr/model.py

@@ -21,7 +21,8 @@ def __init__(self, model_name="google/bert_uncased_L-2_H-128_A-2", lr=3e-5):
  model_name, num_labels=2
  )
  self.num_classes = 2
- self.accuracy_metric = torchmetrics.Accuracy()
+ self.train_accuracy_metric = torchmetrics.Accuracy()
+ self.val_accuracy_metric = torchmetrics.Accuracy()
  self.f1_metric = torchmetrics.F1(num_classes=self.num_classes)
  self.precision_macro_metric = torchmetrics.Precision(
  average="macro", num_classes=self.num_classes
@@ -44,7 +45,7 @@ def training_step(self, batch, batch_idx):
  )
  # loss = F.cross_entropy(logits, batch["label"])
  preds = torch.argmax(outputs.logits, 1)
- train_acc = self.accuracy_metric(preds, batch["label"])
+ train_acc = self.train_accuracy_metric(preds, batch["label"])
  self.log("train/loss", outputs.loss, prog_bar=True, on_epoch=True)
  self.log("train/acc", train_acc, prog_bar=True, on_epoch=True)
  return outputs.loss
@@ -57,7 +58,7 @@ def validation_step(self, batch, batch_idx):
  preds = torch.argmax(outputs.logits, 1)
 
  # Metrics
- valid_acc = self.accuracy_metric(preds, labels)
+ valid_acc = self.val_accuracy_metric(preds, labels)
  precision_macro = self.precision_macro_metric(preds, labels)
  recall_macro = self.recall_macro_metric(preds, labels)
  precision_micro = self.precision_micro_metric(preds, labels)

week_8_serverless/model.py

@@ -21,7 +21,8 @@ def __init__(self, model_name="google/bert_uncased_L-2_H-128_A-2", lr=3e-5):
  model_name, num_labels=2
  )
  self.num_classes = 2
- self.accuracy_metric = torchmetrics.Accuracy()
+ self.train_accuracy_metric = torchmetrics.Accuracy()
+ self.val_accuracy_metric = torchmetrics.Accuracy()
  self.f1_metric = torchmetrics.F1(num_classes=self.num_classes)
  self.precision_macro_metric = torchmetrics.Precision(
  average="macro", num_classes=self.num_classes
@@ -44,7 +45,7 @@ def training_step(self, batch, batch_idx):
  )
  # loss = F.cross_entropy(logits, batch["label"])
  preds = torch.argmax(outputs.logits, 1)
- train_acc = self.accuracy_metric(preds, batch["label"])
+ train_acc = self.train_accuracy_metric(preds, batch["label"])
  self.log("train/loss", outputs.loss, prog_bar=True, on_epoch=True)
  self.log("train/acc", train_acc, prog_bar=True, on_epoch=True)
  return outputs.loss
@@ -57,7 +58,7 @@ def validation_step(self, batch, batch_idx):
  preds = torch.argmax(outputs.logits, 1)
 
  # Metrics
- valid_acc = self.accuracy_metric(preds, labels)
+ valid_acc = self.val_accuracy_metric(preds, labels)
  precision_macro = self.precision_macro_metric(preds, labels)
  recall_macro = self.recall_macro_metric(preds, labels)
  precision_micro = self.precision_micro_metric(preds, labels)

week_9_monitoring/model.py

@@ -21,7 +21,8 @@ def __init__(self, model_name="google/bert_uncased_L-2_H-128_A-2", lr=3e-5):
  model_name, num_labels=2
  )
  self.num_classes = 2
- self.accuracy_metric = torchmetrics.Accuracy()
+ self.train_accuracy_metric = torchmetrics.Accuracy()
+ self.val_accuracy_metric = torchmetrics.Accuracy()
  self.f1_metric = torchmetrics.F1(num_classes=self.num_classes)
  self.precision_macro_metric = torchmetrics.Precision(
  average="macro", num_classes=self.num_classes
@@ -44,7 +45,7 @@ def training_step(self, batch, batch_idx):
  )
  # loss = F.cross_entropy(logits, batch["label"])
  preds = torch.argmax(outputs.logits, 1)
- train_acc = self.accuracy_metric(preds, batch["label"])
+ train_acc = self.train_accuracy_metric(preds, batch["label"])
  self.log("train/loss", outputs.loss, prog_bar=True, on_epoch=True)
  self.log("train/acc", train_acc, prog_bar=True, on_epoch=True)
  return outputs.loss
@@ -57,7 +58,7 @@ def validation_step(self, batch, batch_idx):
  preds = torch.argmax(outputs.logits, 1)
 
  # Metrics
- valid_acc = self.accuracy_metric(preds, labels)
+ valid_acc = self.val_accuracy_metric(preds, labels)
  precision_macro = self.precision_macro_metric(preds, labels)
  recall_macro = self.recall_macro_metric(preds, labels)
  precision_micro = self.precision_micro_metric(preds, labels)