tanner/min-dalle-test
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

dtype dropdown in colab

Browse Source
This commit is contained in:
Brett Kuprel 2022-07-10 13:23:42 -04:00
parent 1ffdef9a56
commit cfb9f60b6e
2 changed files with 4 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
README.md vendored
View File

@ -6,7 +6,7 @@
   
[![Discord](https://img.shields.io/discord/823813159592001537?color=5865F2&logo=discord&logoColor=white)](https://discord.com/channels/823813159592001537/912729332311556136) [![Discord](https://img.shields.io/discord/823813159592001537?color=5865F2&logo=discord&logoColor=white)](https://discord.com/channels/823813159592001537/912729332311556136)
This is a fast, minimal port of Boris Dayma's [DALL·E Mega](https://github.com/borisdayma/dalle-mini). It has been stripped down for inference and converted to PyTorch. The only third party dependencies are numpy, requests, pillow and torch. This is a fast, minimal port of [DALL·E Mega](https://github.com/borisdayma/dalle-mini). It has been stripped down for inference and converted to PyTorch. The only third party dependencies are numpy, requests, pillow and torch.
To generate a 4x4 grid of DALL·E Mega images it takes: To generate a 4x4 grid of DALL·E Mega images it takes:
- 89 sec with a T4 in Colab - 89 sec with a T4 in Colab

5
min_dalle.ipynb vendored
View File

@ -102,7 +102,7 @@
}, },
"source": [ "source": [
"### Load Model\n", "### Load Model\n",
"Float32 is faster but uses more GPU memory. Change the `grid_size` to 3 or less if using float32." "`float32` is faster than `float16` but uses more GPU memory. Change the `grid_size` to 3 or less if using `float32`."
] ]
}, },
{ {
@ -128,13 +128,14 @@
} }
], ],
"source": [ "source": [
"dtype = \"float32\" #@param [\"float32\", \"float16\", \"bfloat16\"]\n",
"from IPython.display import display, update_display\n", "from IPython.display import display, update_display\n",
"from math import log2\n", "from math import log2\n",
"import torch\n", "import torch\n",
"from min_dalle import MinDalle\n", "from min_dalle import MinDalle\n",
"\n", "\n",
"model = MinDalle(\n", "model = MinDalle(\n",
" dtype=torch.float16,\n", " dtype=getattr(torch, dtype),\n",
" is_mega=True, \n", " is_mega=True, \n",
" is_reusable=True\n", " is_reusable=True\n",
")" ")"