Instructions to use msherry/pakflood-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use msherry/pakflood-lora with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
notebook steps
Browse files
PakFlood_LoRA_Colab_demo.ipynb
CHANGED
|
@@ -5,9 +5,11 @@
|
|
| 5 |
"metadata": {},
|
| 6 |
"source": [
|
| 7 |
"# PakFlood-LoRA live demo (Google Colab)\n",
|
| 8 |
-
"Runs the Gradio app of the PakFlood-LoRA model (TerraMind v1 base + rank-32 LoRA + UNet, trained on GEOID-Flood) and
|
| 9 |
"\n",
|
| 10 |
-
"
|
|
|
|
|
|
|
| 11 |
]
|
| 12 |
},
|
| 13 |
{
|
|
@@ -45,9 +47,15 @@
|
|
| 45 |
}
|
| 46 |
],
|
| 47 |
"metadata": {
|
| 48 |
-
"kernelspec": {
|
| 49 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 50 |
},
|
| 51 |
"nbformat": 4,
|
| 52 |
"nbformat_minor": 5
|
| 53 |
-
}
|
|
|
|
| 5 |
"metadata": {},
|
| 6 |
"source": [
|
| 7 |
"# PakFlood-LoRA live demo (Google Colab)\n",
|
| 8 |
+
"Runs the Gradio app of the PakFlood-LoRA model (TerraMind v1 base + rank-32 LoRA adapter + UNet decoder, trained on GEOID-Flood) and prints a **public link** that works while this notebook keeps running (Colab allows up to 72 h; keep this tab open).\n",
|
| 9 |
"\n",
|
| 10 |
+
"**Steps:** 1) optional but faster: Runtime \u2192 Change runtime type \u2192 T4 GPU. 2) Runtime \u2192 Run all. 3) Wait for cell 1 (about 3 to 5 min) and cell 2 (about 1 min). 4) Cell 3 prints `Running on public URL: https://....gradio.live`: open that link, pick a Pakistani test tile and press Run.\n",
|
| 11 |
+
"\n",
|
| 12 |
+
"Model and samples: https://huggingface.co/msherry/pakflood-lora"
|
| 13 |
]
|
| 14 |
},
|
| 15 |
{
|
|
|
|
| 47 |
}
|
| 48 |
],
|
| 49 |
"metadata": {
|
| 50 |
+
"kernelspec": {
|
| 51 |
+
"display_name": "Python 3",
|
| 52 |
+
"language": "python",
|
| 53 |
+
"name": "python3"
|
| 54 |
+
},
|
| 55 |
+
"language_info": {
|
| 56 |
+
"name": "python"
|
| 57 |
+
}
|
| 58 |
},
|
| 59 |
"nbformat": 4,
|
| 60 |
"nbformat_minor": 5
|
| 61 |
+
}
|