It was the year 1960 and a brainteaser was formulated as “The Secretary Problem”. You need to hire a secretary; there are n applicants to be interviewed. You meet each of them in a random order. You can rank them according to suitability, but once rejected an applicant they cannot be recalled. How can you maximize the probability of picking the best person for the job?
Other versions of this include the “fiancé problem” (same idea, but you’re looking for a fiancé instead of a secretary) and the “googol game” – in which you are flipping slips of paper to reveal numbers until you decide you’ve probably found the largest of all.
The answer is… surprisingly predictable, it turns out.
“This basic problem has a remarkably simple solution,” wrote mathematician and statistician Thomas S Ferguson in 1989. “First, one shows that attention can be restricted to the class of rules that for some integer r > 1 rejects the first r – 1 applicants, and then chooses the next applicant who is best in the relative ranking of the observed applicants.”
So, when faced with a stream of random choices and wanting to pick the best, the first thing you do is reject everyone. That is, up to a point. Once you reach that point, just accept the next applicant, suitor, or slip of paper, that beats everything you’ve seen so far.
The statistics are fascinating; and it says that you reject the first 37% of applicants and then take the next one that’s better than what you’ve seen in the rejected pool.
This works if it’s apartments, job candidates, or potential life partners.
Collada was an interchange file format for 3D application that started around 2004 and largely died around 2016. I actually worked in a group with Remi Arnaud when it was being used for a project at Intel.
It was a sound idea. With lots of 3d packages and engines out there, getting files from one tool or engine to another was never easy. Since every authoring tool and game uses different structures for storing mesh, material, animation data, etc – the Collada format tried to define a open-standard format to store these relationships in an XML style text file. This allowed maximum flexibility to define relationships; but had the unfortunate side effect of generating sometimes gigantic files that were extremely slow to load.
While it was an extremely flexible format for exchanging data between packages or game engines, once you got there, it was dramatically faster to use a native binary format. Trying to load or save a XML based file format to load a block of content often took 10-100x longer than a binary version. The speed alone meant that it wasn’t practical for any realtime purposes.
Additionally, supporting the entire Collada spec would mean supporting every kind of data relationship – even if the tool or game didn’t need it. It meant that loaders often only implemented the desired features – which meant that you were almost back to where you started from. Custom loaders and savers with limited features. Except Collada files were gigantic and slow to load/save. A real problem when your primary costs are the speed of your content development.
Collada’s practical use was therefore primarily in one or two time transfers between tools. As time went on, and tools and engines consolidated on a few efficient binary formats, formats such as Collada became less and less useful. By the early 2010’s, development and work on it largely died. The last loaders were apparently updated in 2018 and the github site that hosts the binary versions is kind of broken.
At any rate, if you do need to load an old Collada file (.dae, etc) then you’ll need a copy of 3D Studio Max or Maya, and a plugin loader. You can download one of the last collada loaders here.
Install the plugin (make sure Maya is closed) and then start your tool (Maya in my case).
Ensure the Collada plugin is loaded. Go to the Windows-> Settings/Preferences -> Plug-in Manager in Maya and ensure the fbxmaya, FBX, or ColladaMaya pluings are loaded and/or set to auto load:
When you want to import a Collada file, go to File->Import and select the fbx/collada file you want to load and it should load it up.
The Long Dark was a great game I started playing during early access and really enjoyed. The lonely and desolate wilderness feel really worked well with the the struggle against very simple but brutal natural elements.
The game has been in development longer than some teenagers have even been alive – and has consequently changed a lot over that time. Kudos to Long Dark team for making a time capsule that lets you go back to those early drops by entering a release code in Steam.
While one should ALWAYS be cautious of trainers and save game editors (and there are some on the list that do have viruses (so it’s a good idea to scan them with a virus scanner and only run them in a virtual machine) here’s some of the older trainers for these early drops on GameCopyWorld.
Stable Diffusion really opened the world to what is possible with generative AI. Stable Diffusion 2 and 3 …well…did not go so well. For a while now, Stable Diffusion 1.5 was your best bet on locally generated AI art but it is really showing it’s age.
Now there is a new player in open source generative AI you can run locally. The developers from Stability.ai have founded Black Forest Labs and released their open source tool: Flux.1
While there are plenty of online generative AI’s like Midjourney, Adobe Firefly and others, they usually require paid or only give limited usage. What’s great about Flux.1 is that is allows completely local installation and usage.
Like many open source packages, there are free and paid versions. Their paid Pro version gives the most impressive results via their api (no purely local generation), a local dev version that can be used by developers but not for commercial use, and a free schnell version for personal use. Both the dev and shnell versions are available for local install and use.
So, lets get started with the shnell version – but the instructions are the same for dev except using 2 different model/weight files.
Instructions for installing Flux.1 on nVidia based Windows 10/11 system:
You might want to enable Windows Long Path support as python sometimes requires it for dependent packages. Be sure to reboot your system after enabling it.
Supported graphics card.
32gb of system ram (though again, you can use the smaller model if you have less ram)
Open a command prompt and make a local working root directory somewhere, I’ll use c:\depot\
You have a few options. First, you need to pick if you’re using the non-commercial Dev version or Schnell version. After that, each has the option of a single easy to use checkpoint package file, or each of the model data files individually. I’ll be using the Schnell ones, but you just need to get the Dev ones from the Dev branch if you want those instead.
C:\depot\ComfyUI>python main.py
A module that was compiled using NumPy 1.x cannot be run in
NumPy 2.0.1 as it may crash. To support both 1.x and 2.x
versions of NumPy, modules must be compiled with NumPy 2.0.
Some module may need to rebuild instead e.g. with 'pybind11>=2.12'.
If you are a user of the module, the easiest solution will be to
downgrade to 'numpy<2' or try to upgrade the affected module.
We expect that some modules will need time to support NumPy 2.
Traceback (most recent call last): File "C:\depot\ComfyUI\main.py", line 83, in <module>
import comfy.utils
File "C:\depot\ComfyUI\comfy\utils.py", line 20, in <module>
import torch
File "C:\Users\matt\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.12_qbz5n2kfra8p0\LocalCache\local-packages\Python312\site-packages\torch\__init__.py", line 2120, in <module>
from torch._higher_order_ops import cond
File "C:\Users\matt\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.12_qbz5n2kfra8p0\LocalCache\local-packages\Python312\site-packages\torch\_higher_order_ops\__init__.py", line 1, in <module>
from .cond import cond
File "C:\Users\matt\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.12_qbz5n2kfra8p0\LocalCache\local-packages\Python312\site-packages\torch\_higher_order_ops\cond.py", line 5, in <module>
import torch._subclasses.functional_tensor
File "C:\Users\matt\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.12_qbz5n2kfra8p0\LocalCache\local-packages\Python312\site-packages\torch\_subclasses\functional_tensor.py", line 42, in <module>
class FunctionalTensor(torch.Tensor):
File "C:\Users\matt\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.12_qbz5n2kfra8p0\LocalCache\local-packages\Python312\site-packages\torch\_subclasses\functional_tensor.py", line 258, in FunctionalTensor
cpu = _conversion_method_template(device=torch.device("cpu"))
C:\Users\matt\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.12_qbz5n2kfra8p0\LocalCache\local-packages\Python312\site-packages\torch\_subclasses\functional_tensor.py:258: UserWarning: Failed to initialize NumPy: _ARRAY_API not found (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\torch\csrc\utils\tensor_numpy.cpp:84.)
cpu = _conversion_method_template(device=torch.device("cpu"))
Total VRAM 24576 MB, total RAM 32492 MB
pytorch version: 2.4.0+cu121
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 3090 : cudaMallocAsync
Using pytorch cross attention
C:\depot\ComfyUI\comfy\extra_samplers\uni_pc.py:19: SyntaxWarning: invalid escape sequence '\h'
"""Create a wrapper class for the forward SDE (VP type).
****** User settings have been changed to be stored on the server instead of browser storage. ******
****** For multi-user setups add the --multi-user CLI argument to enable multiple user profiles. ******
[Prompt Server] web root: C:\depot\ComfyUI\web
C:\Users\matt\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.12_qbz5n2kfra8p0\LocalCache\local-packages\Python312\site-packages\kornia\feature\lightglue.py:44: FutureWarning: `torch.cuda.amp.custom_fwd(args...)` is deprecated. Please use `torch.amp.custom_fwd(args..., device_type='cuda')` instead.
@torch.cuda.amp.custom_fwd(cast_inputs=torch.float32)
Import times for custom nodes:
0.0 seconds: C:\depot\ComfyUI\custom_nodes\websocket_image_save.py
Starting server
To see the GUI go to: http://127.0.0.1:8188
Open your web browser and go to http://127.0.01:8188
Click on the ‘Queue Prompt’ button to execute the current prompt
Technically it queues up the work and you should see progress in the command window where you launched python main.py
got prompt
model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16
model_type FLOW
Using pytorch attention in VAE
Using pytorch attention in VAE
Model doesn't have a device attribute.
C:\Users\matt\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.12_qbz5n2kfra8p0\LocalCache\local-packages\Python312\site-packages\transformers\tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884
warnings.warn(
Model doesn't have a device attribute.
loaded straight to GPU
Requested to load Flux
Loading 1 new model
Requested to load FluxClipModel_
Loading 1 new model
C:\depot\ComfyUI\comfy\ldm\modules\attention.py:407: UserWarning: 1Torch was not compiled with flash attention. (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\aten\src\ATen\native\transformers\cuda\sdp_utils.cpp:555.)
out = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=mask, dropout_p=0.0, is_causal=False)
100%|████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:04<00:00, 1.18s/it]
Requested to load AutoencodingEngine
Loading 1 new model
Prompt executed in 23.65 seconds
When it completes you should see your image. You can then save your image or tweak the parameters.
Debugging help:
numpy is not available
My first runs, I got this from the console when I queued up a request:
got prompt
model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16
model_type FLOW
Using pytorch attention in VAE
Using pytorch attention in VAE
Model doesn't have a device attribute.
C:\Users\matt\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.12_qbz5n2kfra8p0\LocalCache\local-packages\Python312\site-packages\transformers\tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884
warnings.warn(
Model doesn't have a device attribute.
loaded straight to GPU
Requested to load Flux
Loading 1 new model
Requested to load FluxClipModel_
Loading 1 new model
C:\depot\ComfyUI\comfy\ldm\modules\attention.py:407: UserWarning: 1Torch was not compiled with flash attention. (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\aten\src\ATen\native\transformers\cuda\sdp_utils.cpp:555.)
out = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=mask, dropout_p=0.0, is_causal=False)
100%|████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:04<00:00, 1.19s/it]
Requested to load AutoencodingEngine
Loading 1 new model
!!! Exception during processing!!! Numpy is not available
Traceback (most recent call last):
File "C:\depot\ComfyUI\execution.py", line 152, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\depot\ComfyUI\execution.py", line 82, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\depot\ComfyUI\execution.py", line 75, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\depot\ComfyUI\nodes.py", line 1445, in save_images
i = 255. * image.cpu().numpy()
^^^^^^^^^^^^^^^^^^^
RuntimeError: Numpy is not available
Prompt executed in 26.44 seconds
C:\depot\ComfyUI>pip install numpy==1.26.4
Defaulting to user installation because normal site-packages is not writeable
Collecting numpy==1.26.4
Downloading numpy-1.26.4-cp312-cp312-win_amd64.whl.metadata (61 kB)
Downloading numpy-1.26.4-cp312-cp312-win_amd64.whl (15.5 MB)
---------------------------------------- 15.5/15.5 MB 57.4 MB/s eta 0:00:00
Installing collected packages: numpy
Attempting uninstall: numpy
Found existing installation: numpy 2.0.1
Uninstalling numpy-2.0.1:
Successfully uninstalled numpy-2.0.1
Successfully installed numpy-1.26.4
C:\depot\ComfyUI>
Uninstalling all pip/python package, clear your pip cache, then re-install the requirements
The first time I installed, I got an error when downloading the numpy library during step in which you pip install the requirements. In order to clear the pip cache, uninstall all pip packages, then re-install all requirements again, I did the following:
Remember that 8088 or XT based PC’s ran their ISA bus at 4.77mhz, AT ran at 6mhz, and 386 and beyond ran ISA at 8mhz (though many later ISA systems ran them at 10mhz since ISA is very forgiving and the higher clock speed gave better speeds). This means you may find old 8-bit or AT-based ISA cards won’t work in AT or 386/486 systems since the ISA bus runs too fast.
If your BIOS allows it, you may be able slow down the ISA bus speeds and that might help old cards work.
I wrote awhile back on how to crash Linux/cause a Linux kernel panic in order to test how your program can handle a crash – but can you cause a Windows blue-screen programmatically?
Methods you can use to cause a Windows Blue-screen:
Windows allows you to configure a specific keyboard combination to cause a crash. You set some registry keys and then can crash a system by holding right CTRL then pressing scroll lock key twice. You can also customize the key sequence via registering custom keyboard scan codes. If you have a kernel debugger attached it will trigger the kernel debugger after the crash dump is written.
The best way to trigger an artificial kernel crash is to use NotMyFault, which is part of the Microsoft Windows SysInternals tools.
Oculus is doing a black Friday sale for 2023. I have an Oculus Quest 2, but hadn’t used it in over a year. I plugged it in to charge it back up and browse the store. Unfortunately, the store app screen told me it couldn’t load the store.
I went to setting->Wifi and manually connected to my home network. Duh! (so I thought) Even after this, the store app was blank and would tell me it couldn’t display anything. Time to start debugging.
The wifi would connect and said it had excellent wifi signal, but limited connectivity.
I tested my other wireless devices and they had no trouble connecting to the wifi and could browse the net normally.
I unplugged the mesh network repeaters around my house in case it was picking up a weak signal from one of those. No change.
I tried setting up my iPhone with as a wireless hotspot and connected to that. I got the same strong signal, but limited connectivity.
I checked the IP address of my Oculus in Settings, and I could ping the device just fine. If I turned the headset off, I couldn’t ping it. It seemed like it was connected ok.
I tried unplugging other devices from my wifi in case there was an IP conflict anywhere. Same problem.
I tried connecting the quest with a USB cable but I still could not get updates nor see anything in the store app or main menu.
Despite not being used for 2 years, when I went to system->updates, it showed no updates available. Something is fishy, there had to be updates.
I opened the in-headset browser and it would tell me that I could not browse because the date was wrong. It was set to 5:00am Sept 17, 2037. Whoa.
There is NO way to change the date/time in the settings or anywhere else I could find.
It turns out, others have seen this issue too. Their Oculus fast forwards to the future mysteriously and then connectivity to the store/web/updates doesn’t seem to work after that. You need to get the date fixed, but there’s no obvious way to do it.
Solution: Factory Reset
In the end, I decided to do a factory reset (Hold the power and volume buttons while booting) because it had been well over a year since I used it last and I figured it would be good to have a clean start. However, there is the option of using side-loaded apps (see below).
Unfortunately, even the factory reset gave me a few headaches. First, the headset isn’t always obvious when it sleeps vs actually powers off. My first attempt I didn’t power off all the way and just woke from sleep and I didn’t get the reset menu holding the power+volume buttons down. I went to settings menu and shut the device down in the headset to be sure.
After that, I was then able to cold boot and get into the factory reset menu. I selected factory reset and waited for it to clean the device and the progress bar indicated the reset was complete. The screen went black (but still powered), but didn’t reboot. I let it set a few minutes, then decided to manually reboot with the power button. Fingers crossed.
The first reboot I got the meta logo, but shortly after that the screen went blank (still powered) but no reboot. I let it set for a few minutes then manually powered it off using the power button – again.
On the second reboot, I got the meta logo, and then it started animating. That’s a good sign. Then the welcome page came up and I could connect to wifi and start updating.
During the update phase (1/2) while the progress bar was moving, I took the headset off to read more instructions. When I put it back on to see how far it was, the display was a patterned garbled static. I took it off and let it sit for a minute, then tried again. The display came back up and the update phase 1 of 2 completed normally.
Sidequest
The bad part about a factory reset is you lose all your installed games. I had to go back in and start installing all of them again. What a pain, because it wasn’t a very fast process.
The Oculus is really just an Android device underneath. This means if you have developer mode enabled and have the Android developer kit installed, you can use ADB commands. I haven’t tried this, but supposedly this will work:
Try adjusting the aspect ratio of the image to be taller instead of wider. Be careful not to go too tall (or two wide) or you’ll get the double-head or start generating combinations of two people.
The source material has been scanned in a taller aspect ratio, try adjusting the x-side of your ratio
Use img2img on a crop that includes part of the chest to make it match the rest of the drawing
Cinematography terms tend to work well. In order of close to far: Extreme close-up, close-up, medium close-up, medium shot, medium full shot, full shot, long shot, extreme long shot.
Content warning: Some of the links have some moderately NSFW pictures. There is no outright nudity, but it does deal with generating rather busty images. This article should be fine, but be aware following the links.
While this guide is translated from a Japanese source and uses the Waifu/Danbooru model to generate more anime-looking models, it works really well for generating ultra-realistic Victorian pictures using stable diffusion’s standard 1.5 model. Here’s some I made using his prompt with just 30 minutes of experimenting:
Fair warning, the original author is trying to generate more…busty women that look better as anime characters under the Waifu model. I won’t comment on his original purpose, but I thought this was an interesting description of how a ‘prompt engineer’ moved from an idea to generating a stable diffusion prompt.
First he started with a good description of what he wanted:
I want a VICTORIAN GIRL in a style of OIL PAINTING Eye and Face are important in art so she must have PERFECT FACE, SEXY FACE and her eye have DETAILED PUPILS I want she to have LARGE BREAST, TONED ABS and THICK THIGH. She must look FEMININE doing EVOCATIVE POSE, SMIRK and FULL BODY wearing NIGHT GOWN The output must be INTRICATE, HIGH DETAIL, SHARP And in the style of {I’m not give out the artist names to avoid trouble. Apologize.}
This lead him to generate the following prompt. Note his use of parenthesize () to add emphasis, and terms inside square brackets [] to minimize the direction.
Unfortunately, you don’t need to experiment for long to realize stable diffusion needs a lot of help with anatomy. It often generates nightmare fuel images that that have multiple heads, messed up arms, hands with too many fingers, eyes with terrifying pupils (or no pupils), too many limbs – well, you get the idea. So you need to make sure those things don’t show up by banning them via setting the negative prompts (again, not commenting on original purpose):
Finally, he made these settings to the stable diffusion settings. Note you want to keep the aspect ratio in a portrait-like format (768 tall x 512 wide). Going taller can result in multiple heads, going wider can result in more than one person in the scene.
Restore Face: ON
Steps: 42
Sampler: DDIM
CFG scale: 10
Height: 768
Width: 512
