use resnet50 to train a cifar10 classifier

dvclive/train/metrics.json

@@ -0,0 +1,7 @@
+{
+    "train_loss": 2.2422571182250977,
+    "train_acc": 0.7347080707550049,
+    "valid_loss": 2.3184266090393066,
+    "valid_acc": 0.7381500005722046,
+    "step": 4
+}

dvclive/train/params.yaml

@@ -0,0 +1,5 @@
+data_dir: data/processed
+epochs: 5
+batch_size: 256
+learning_rate: 5e-05
+epoch: 4

dvclive/train/plots/metrics/train_acc.tsv

@@ -0,0 +1,6 @@
+step	train_acc
+0	0.6712241768836975
+1	0.6976224184036255
+2	0.7157850861549377
+3	0.7277812957763672
+4	0.7347080707550049

dvclive/train/plots/metrics/train_loss.tsv

@@ -0,0 +1,6 @@
+step	train_loss
+0	3.0726168155670166
+1	2.7409346103668213
+2	2.5224294662475586
+3	2.364570140838623
+4	2.2422571182250977

dvclive/train/plots/metrics/valid_acc.tsv

@@ -0,0 +1,6 @@
+step	valid_acc
+0	0.6918894052505493
+1	0.7131190896034241
+2	0.7261338233947754
+3	0.7339118123054504
+4	0.7381500005722046

dvclive/train/plots/metrics/valid_loss.tsv

@@ -0,0 +1,6 @@
+step	valid_loss
+0	2.890321969985962
+1	2.669679880142212
+2	2.5183584690093994
+3	2.4061686992645264
+4	2.3184266090393066

dvclive/train/report.md

@@ -0,0 +1,21 @@
+# DVC Report
+
+params.yaml
+
+| data_dir       |   epochs |   batch_size |   learning_rate |   epoch |
+|----------------|----------|--------------|-----------------|---------|
+| data/processed |        5 |          256 |           5e-05 |       4 |
+
+metrics.json
+
+|   train_loss |   train_acc |   valid_loss |   valid_acc |   step |
+|--------------|-------------|--------------|-------------|--------|
+|      2.24226 |    0.734708 |      2.31843 |     0.73815 |      4 |
+
+![static/valid_loss](static/valid_loss.png)
+
+![static/train_acc](static/train_acc.png)
+
+![static/valid_acc](static/valid_acc.png)
+
+![static/train_loss](static/train_loss.png)