fix check error due to missing cuda device from github test env

jobs-git · jobs-git · commit c0c519b63185 · 2025-06-14T21:10:14.000+08:00
diff --git a/docs/source/tutorials/precompute_sample_usage.ipynb b/docs/source/tutorials/precompute_sample_usage.ipynb
@@ -42,7 +42,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": null,
    "id": "5268f19a",
    "metadata": {},
    "outputs": [],
@@ -70,7 +70,9 @@
     "torch.backends.cuda.matmul.allow_bf16_reduced_precision_reduction = True\n",
     "# torch.backends.cudnn.benchmark = False\n",
     "torch.backends.cuda.matmul.allow_tf32 = True\n",
-    "# torch.backends.cudnn.deterministic = True"
+    "# torch.backends.cudnn.deterministic = True\n",
+    "\n",
+    "device_str = \"cuda\" if torch.cuda.is_available() else \"cpu\""
    ]
   },
   {
@@ -332,7 +334,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": null,
    "id": "3008d01f",
    "metadata": {},
    "outputs": [
@@ -1158,12 +1160,12 @@
     ")\n",
     "\n",
     "baseline_predictions = Baseline().predict(\n",
-    "    val_dataloader, trainer_kwargs=dict(accelerator=\"cuda\"), return_y=True\n",
+    "    val_dataloader, trainer_kwargs=dict(accelerator=device_str), return_y=True\n",
     ")\n",
     "SMAPE()(baseline_predictions.output, baseline_predictions.y)\n",
     "\n",
     "pl.seed_everything(42)\n",
-    "trainer = pl.Trainer(accelerator=\"cuda\", gradient_clip_val=0.1)\n",
+    "trainer = pl.Trainer(accelerator=device_str, gradient_clip_val=0.1)\n",
     "net = NBeats.from_dataset(\n",
     "    training,\n",
     "    learning_rate=3e-2,\n",
@@ -1173,7 +1175,7 @@
     ")\n",
     "\n",
     "pl.seed_everything(42)\n",
-    "trainer = pl.Trainer(accelerator=\"cuda\", gradient_clip_val=0.1)\n",
+    "trainer = pl.Trainer(accelerator=device_str, gradient_clip_val=0.1)\n",
     "net = NBeats.from_dataset(\n",
     "    training,\n",
     "    learning_rate=3e-2,\n",
@@ -1204,7 +1206,7 @@
     "trainer = pl.Trainer(\n",
     "    logger=False,\n",
     "    max_epochs=100,\n",
-    "    accelerator=\"cuda\",\n",
+    "    accelerator=device_str,\n",
     "    enable_model_summary=True,\n",
     "    gradient_clip_val=0.1,\n",
     "    callbacks=[early_stop_callback],\n",
@@ -1319,7 +1321,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 8,
+   "execution_count": null,
    "id": "ec533c49",
    "metadata": {},
    "outputs": [
@@ -2030,12 +2032,12 @@
     ")\n",
     "\n",
     "baseline_predictions = Baseline().predict(\n",
-    "    val_dataloader, trainer_kwargs=dict(accelerator=\"cuda\"), return_y=True\n",
+    "    val_dataloader, trainer_kwargs=dict(accelerator=device_str), return_y=True\n",
     ")\n",
     "SMAPE()(baseline_predictions.output, baseline_predictions.y)\n",
     "\n",
     "pl.seed_everything(42)\n",
-    "trainer = pl.Trainer(accelerator=\"cuda\", gradient_clip_val=0.1)\n",
+    "trainer = pl.Trainer(accelerator=device_str, gradient_clip_val=0.1)\n",
     "net = NBeats.from_dataset(\n",
     "    training,\n",
     "    learning_rate=3e-2,\n",
@@ -2045,7 +2047,7 @@
     ")\n",
     "\n",
     "pl.seed_everything(42)\n",
-    "trainer = pl.Trainer(accelerator=\"cuda\", gradient_clip_val=0.1)\n",
+    "trainer = pl.Trainer(accelerator=device_str, gradient_clip_val=0.1)\n",
     "net = NBeats.from_dataset(\n",
     "    training,\n",
     "    learning_rate=3e-2,\n",
@@ -2076,7 +2078,7 @@
     "trainer = pl.Trainer(\n",
     "    logger=False,\n",
     "    max_epochs=100,\n",
-    "    accelerator=\"cuda\",\n",
+    "    accelerator=device_str,\n",
     "    enable_model_summary=True,\n",
     "    gradient_clip_val=0.1,\n",
     "    callbacks=[early_stop_callback],\n",
diff --git a/examples/precompute_bench_cpu.ipynb b/examples/precompute_bench_cpu.ipynb
@@ -31,7 +31,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": null,
    "id": "5268f19a",
    "metadata": {},
    "outputs": [],
@@ -52,14 +52,7 @@
     "from pytorch_forecasting.data.examples import generate_ar_data\n",
     "from pytorch_forecasting.metrics import MAE, SMAPE, MQF2DistributionLoss, QuantileLoss\n",
     "\n",
-    "torch.set_float32_matmul_precision(\"medium\")\n",
-    "torch.backends.cuda.matmul.allow_tf32 = True\n",
-    "torch.backends.cudnn.enabled = True\n",
-    "torch.backends.cudnn.allow_tf32 = True\n",
-    "torch.backends.cuda.matmul.allow_bf16_reduced_precision_reduction = True\n",
-    "# torch.backends.cudnn.benchmark = False\n",
-    "torch.backends.cuda.matmul.allow_tf32 = True\n",
-    "# torch.backends.cudnn.deterministic = True"
+    "torch.set_float32_matmul_precision(\"medium\")"
    ]
   },
   {
@@ -275,14 +268,12 @@
     "\n",
     "completion_time = end_time - start_time\n",
     "\n",
-    "print(f\"Time: {completion_time:.6f} seconds\")\n",
-    "\n",
-    "torch.cuda.empty_cache()"
+    "print(f\"Time: {completion_time:.6f} seconds\")"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": null,
    "id": "b1a183e2",
    "metadata": {},
    "outputs": [
@@ -302,13 +293,12 @@
     "baseline_predictions = Baseline().predict(\n",
     "    val_dataloader, trainer_kwargs=dict(accelerator=\"cpu\"), return_y=True\n",
     ")\n",
-    "SMAPE()(baseline_predictions.output, baseline_predictions.y)\n",
-    "torch.cuda.empty_cache()"
+    "SMAPE()(baseline_predictions.output, baseline_predictions.y)"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": null,
    "id": "2e696d76",
    "metadata": {},
    "outputs": [
@@ -336,9 +326,7 @@
     "    backcast_loss_ratio=0.0,\n",
     "    hidden_size=64,\n",
     "    optimizer=\"AdamW\",\n",
-    ")\n",
-    "\n",
-    "torch.cuda.empty_cache()"
+    ")"
    ]
   },
   {
@@ -351,7 +339,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 7,
+   "execution_count": null,
    "id": "e1c56874",
    "metadata": {},
    "outputs": [
@@ -424,16 +412,13 @@
     "fig.show()\n",
     "net.hparams.learning_rate = res.suggestion()\n",
     "\n",
-    "torch.cuda.synchronize()\n",
     "end_time = time.time()\n",
     "\n",
     "completion_time = end_time - start_time\n",
     "\n",
     "time_hyperopt.append(completion_time)\n",
     "\n",
-    "print(f\"Time: {completion_time:.6f} seconds\")\n",
-    "\n",
-    "torch.cuda.empty_cache()"
+    "print(f\"Time: {completion_time:.6f} seconds\")"
    ]
   },
   {
@@ -446,7 +431,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 8,
+   "execution_count": null,
    "id": "b4bd82cf",
    "metadata": {},
    "outputs": [
@@ -629,16 +614,14 @@
     "    train_dataloaders=train_dataloader,\n",
     "    val_dataloaders=val_dataloader,\n",
     ")\n",
-    "torch.cuda.synchronize()\n",
+    "\n",
     "end_time = time.time()\n",
     "\n",
     "completion_time = end_time - start_time\n",
     "\n",
     "time_train.append(completion_time)\n",
     "\n",
-    "print(f\"Time: {completion_time:.6f} seconds\")\n",
-    "\n",
-    "torch.cuda.empty_cache()"
+    "print(f\"Time: {completion_time:.6f} seconds\")"
    ]
   },
   {
@@ -724,14 +707,12 @@
     "\n",
     "completion_time = end_time - start_time\n",
     "\n",
-    "print(f\"Time: {completion_time:.6f} seconds\")\n",
-    "\n",
-    "torch.cuda.empty_cache()"
+    "print(f\"Time: {completion_time:.6f} seconds\")"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
+   "execution_count": null,
    "id": "294470a5",
    "metadata": {},
    "outputs": [
@@ -751,13 +732,12 @@
     "baseline_predictions = Baseline().predict(\n",
     "    val_dataloader, trainer_kwargs=dict(accelerator=\"cpu\"), return_y=True\n",
     ")\n",
-    "SMAPE()(baseline_predictions.output, baseline_predictions.y)\n",
-    "torch.cuda.empty_cache()"
+    "SMAPE()(baseline_predictions.output, baseline_predictions.y)"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 11,
+   "execution_count": null,
    "id": "50d6725e",
    "metadata": {},
    "outputs": [
@@ -785,9 +765,7 @@
     "    backcast_loss_ratio=0.0,\n",
     "    hidden_size=64,\n",
     "    optimizer=\"AdamW\",\n",
-    ")\n",
-    "\n",
-    "torch.cuda.empty_cache()"
+    ")"
    ]
   },
   {
@@ -800,7 +778,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 12,
+   "execution_count": null,
    "id": "04c933d8",
    "metadata": {},
    "outputs": [
@@ -873,16 +851,13 @@
     "fig.show()\n",
     "net.hparams.learning_rate = res.suggestion()\n",
     "\n",
-    "torch.cuda.synchronize()\n",
     "end_time = time.time()\n",
     "\n",
     "completion_time = end_time - start_time\n",
     "\n",
     "time_hyperopt.append(completion_time)\n",
     "\n",
-    "print(f\"Time: {completion_time:.6f} seconds\")\n",
-    "\n",
-    "torch.cuda.empty_cache()"
+    "print(f\"Time: {completion_time:.6f} seconds\")"
    ]
   },
   {
@@ -895,7 +870,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 13,
+   "execution_count": null,
    "id": "31d3cc8f",
    "metadata": {},
    "outputs": [
@@ -1079,16 +1054,14 @@
     "    train_dataloaders=train_dataloader,\n",
     "    val_dataloaders=val_dataloader,\n",
     ")\n",
-    "torch.cuda.synchronize()\n",
+    "\n",
     "end_time = time.time()\n",
     "\n",
     "completion_time = end_time - start_time\n",
     "\n",
     "time_train.append(completion_time)\n",
     "\n",
-    "print(f\"Time: {completion_time:.6f} seconds\")\n",
-    "\n",
-    "torch.cuda.empty_cache()"
+    "print(f\"Time: {completion_time:.6f} seconds\")"
    ]
   },
   {
diff --git a/examples/precompute_bench_gpu.ipynb b/examples/precompute_bench_gpu.ipynb
@@ -35,11 +35,25 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": null,
    "id": "5268f19a",
    "metadata": {},
    "outputs": [],
    "source": [
+    "import torch\n",
+    "\n",
+    "device_str = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "\n",
+    "class StopExecution(Exception):\n",
+    "    def _render_traceback_(self):\n",
+    "        return []\n",
+    "\n",
+    "\n",
+    "if device_str != \"cuda\":\n",
+    "    print(\"CUDA not available. Aborting notebook execution.\")\n",
+    "    raise StopExecution\n",
+    "\n",
     "import warnings\n",
     "\n",
     "warnings.filterwarnings(\"ignore\")\n",
@@ -49,7 +63,6 @@
     "import matplotlib.pyplot as plt\n",
     "import numpy as np\n",
     "import pandas as pd\n",
-    "import torch\n",
     "\n",
     "from pytorch_forecasting import Baseline, NHiTS, TimeSeriesDataSet\n",
     "from pytorch_forecasting.data import NaNLabelEncoder\n",
diff --git a/examples/precompute_bench_gpu_small_data.ipynb b/examples/precompute_bench_gpu_small_data.ipynb
@@ -35,11 +35,25 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": null,
    "id": "5268f19a",
    "metadata": {},
    "outputs": [],
    "source": [
+    "import torch\n",
+    "\n",
+    "device_str = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "\n",
+    "class StopExecution(Exception):\n",
+    "    def _render_traceback_(self):\n",
+    "        return []\n",
+    "\n",
+    "\n",
+    "if device_str != \"cuda\":\n",
+    "    print(\"CUDA not available. Aborting notebook execution.\")\n",
+    "    raise StopExecution\n",
+    "\n",
     "import warnings\n",
     "\n",
     "warnings.filterwarnings(\"ignore\")\n",
@@ -49,7 +63,6 @@
     "import matplotlib.pyplot as plt\n",
     "import numpy as np\n",
     "import pandas as pd\n",
-    "import torch\n",
     "\n",
     "from pytorch_forecasting import Baseline, NHiTS, TimeSeriesDataSet\n",
     "from pytorch_forecasting.data import NaNLabelEncoder\n",
