refactor apis (#184)

gradsflow · May 18, 2022 · 74eced1 · 74eced1
1 parent 546bad8
commit 74eced1
Show file tree

Hide file tree

Showing 3 changed files with 66 additions and 14 deletions.
diff --git a/examples/nbs/02-TextClassification.ipynb b/examples/nbs/02-TextClassification.ipynb
@@ -42,7 +42,11 @@
    "source": [
     "from flash.core.data.utils import download_data\n",
     "from flash.text import TextClassificationData\n",
-    "from gradsflow import AutoTextClassifier"
+    "from gradsflow import AutoTextClassifier\n",
+    "\n",
+    "import ray\n",
+    "\n",
+    "ray.init(address=\"auto\")"
    ]
   },
   {
@@ -63,7 +67,7 @@
     "    \"sentiment\",\n",
     "    train_file=\"data/imdb/train.csv\",\n",
     "    val_file=\"data/imdb/valid.csv\",\n",
-    "    batch_size=4,\n",
+    "    batch_size=16,\n",
     ")"
    ]
   },
@@ -74,31 +78,77 @@
    "metadata": {
     "pycharm": {
      "name": "#%%\n"
-    }
+    },
+    "tags": []
    },
    "outputs": [],
    "source": [
     "suggested_conf = dict(\n",
-    "    optimizers=[\"adam\"],\n",
-    "    lr=(5e-4, 1e-3),\n",
+    "    optimizer=[\"adam\", \"adamw\"],\n",
+    "    lr=(4e-3, 1e-2),\n",
     ")\n",
     "\n",
     "model = AutoTextClassifier(\n",
     "    datamodule,\n",
-    "    suggested_backbones=[\"sgugger/tiny-distilbert-classification\"],\n",
+    "    suggested_backbones=[\"prajjwal1/bert-medium\"],\n",
     "    suggested_conf=suggested_conf,\n",
-    "    max_epochs=2,\n",
+    "    max_epochs=1,\n",
     "    optimization_metric=\"val_accuracy\",\n",
+    "    n_trials=1,\n",
     ")\n",
     "\n",
     "print(\"AutoTextClassifier initialised!\")\n",
     "model.hp_tune()"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "4a9c19a3",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model.analysis.dataframe()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "d0b96bd0-3949-4dd5-b6d9-4fa6ac213e30",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from flash import Trainer\n",
+    "trainer = Trainer()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "6a95bc01-4291-49a8-bea7-c76ba64bf42a",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "trainer.validate(model.model, datamodule=datamodule)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "b5b20a6f-9efd-493e-94ea-ce00e8879451",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "ray.shutdown()"
+   ]
   }
  ],
  "metadata": {
+  "interpreter": {
+   "hash": "e6697cd4c0f4f58297a92a2dfda85db933b7e27cf6bc19e3dafb7e93fff75254"
+  },
   "kernelspec": {
-   "display_name": "Python 3",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
    "name": "python3"
   },
@@ -112,7 +162,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.7.10"
+   "version": "3.9.12"
   }
  },
  "nbformat": 4,

diff --git a/examples/src/tasks/text_classification.py b/examples/src/tasks/text_classification.py
@@ -14,16 +14,17 @@
 )
 
 suggested_conf = dict(
-    optimizers=["adam"],
+    optimizers=["adam", "adamw"],
     lr=(5e-4, 1e-3),
 )
 
 model = AutoTextClassifier(
     datamodule,
-    suggested_backbones=["sgugger/tiny-distilbert-classification"],
+    suggested_backbones=["prajjwal1/bert-medium"],
     suggested_conf=suggested_conf,
-    max_epochs=2,
+    max_epochs=1,
     optimization_metric="val_accuracy",
+    n_trials=4,
 )
 
 print("AutoTextClassifier initialised!")

diff --git a/gradsflow/autotasks/autoclassification/text/text.py b/gradsflow/autotasks/autoclassification/text/text.py
@@ -37,6 +37,7 @@ class AutoTextClassifier(AutoClassifier):
                 "sentiment",
                 train_file="data/imdb/train.csv",
                 val_file="data/imdb/valid.csv",
+                batch_size=4,
             )
 
             model = AutoTextClassifier(datamodule,
@@ -66,8 +67,8 @@ class AutoTextClassifier(AutoClassifier):
         "sgugger/tiny-distilbert-classification",
     ]
 
-    def __init__(self, *args, **kwargs):
-        super().__init__(*args, **kwargs)
+    def __init__(self, *args, max_steps=-1, **kwargs):
+        super().__init__(*args, max_steps=max_steps, **kwargs)
         meta = self.auto_dataset.meta
         self.num_classes = meta.get("num_labels") or meta.get("num_classes")
         logger.debug(f"num_classes = {self.num_classes}")