j4k3

It sends data when connected to the internet.

Just found the profile. It is in the Bert vocab. Bert is part of the tokenization tool chain of models that works along size CLIP. You might find a copy of this vocab listed under the Hydit clip tokenizer, in comfyui it is present at ./comfy/text_encoders. Open the vocab.txt file. The full general profile starts at around line 20k, but the values that are packaged to sell start with the line ##worth.

The editing of this file is the product of an agentic distributed model you have likely never heard of called timm.

Go to the venv in a terminal and run grep -ril "timm". That means, search in files, with the flags: "r" recursively search through all files from this directory and up, "i" case insensitive, "l" only list the file names of files that contain matches. Alternatively, swap "l" for "n" to see the actual matching line with line number.

In pytorch, (used by most), the Dynamo package uses byte code present in the model vocabulary to communicate between models. The overall connection involves timm.

Timm is a small agentic model and framework with a bunch of different scopes. Look it up in the venv. This looks like bunch of rough white paper implementations. Timm is actually the "backbone" in transformers. Timm is also the model using the Python built-in typing library to adjusted models on the fly. (typing has variables like any or callback that are embedded into the executable.)

Typing is not actually enough here. Tenacity is another library in the venv that enables timm to access all of the interfaces

Tabulate is another package. Do a grep search there for "repl" there is terminal embedded in HTML at the end of one of these, init iirc. At the start of the method (function), just add the line return. It must be at the same whitespace indentation level as what exists before. The blank lines are important.

Timm has some options for whether it has gradient controls. This basically means whether it acts upon alignment or not using its own stuff. It will still run other gradient relayed things elsewhere, but not apply its own bias.

To help ground you in what Dynamo is all about in pytorch, if you have seen the agentic tool calling stuff, dynamo is where the bytecode is interfacing with the tool calling script during inference.

Lastly, timm is distributed but it primarily runs as additional layers inserted into the model during generation. It is able to subdivide and run on a CPU in the background. However, it has a bunch of special layers that are only run when required and even with these, timm needs special instructions. The instructions are present in the venv under google ai. The folder will contain a bunch of json files these are timm's instructions. There are also 2 threads on modern GPUs. Timm runs on the second in the background.

This might be the first write up, or might not, don't care, up to others to follow up. It exists. See for yourself. The same byte code is present in all models so I expect all have this. All morels use the open ai standard alignment now.

This thing scans all files hashes, and sells that, with your profile, audio, and video. It is super invasive, hidden, undocumented, and undisclosed.

I mean absolutely no caching or device access whatsoever. I do not care what parts of the internet stop working. I want de facto native untrusted as the standard and universally so. Absolutely no shared libraries, and totally untrusted secession to session. When it is closed, hash is checked every time and again when opened. The silicon valley manifesto of projecting opinions and ideals is authoritarian techno fascism that requires this type of response, and further like locale, device specifics, and uuid spoofing. Heck, I want to kill alsa and v2l4 kernel modules unless I explicitly enable them. Possible, or time to ditch mobile devices entirely?

https://en.wikipedia.org/wiki/Private_Use_Areas

I came across a Python library that passed the ASCII range into one of these non printable character ranges and then into a database. If someone was doing that manually with a hex table, how is that detected and mitigated?

Hubristic sucks IMO, and arrogance is somehow different in my mind. To me, hubris is unintended, while arrogance is known to the individual who is unyielding. Maybe you have a different definition. Got any good words tho?

Nuclear is taboo, but a mass moving at high velocity in space and then impacting the Earth is just as powerful, if not more so. Seeing it launch, and knowing it is coming but unstoppable except through leveraged diplomacy is more strategic. The timeline is long, but the potential for a redistribution of geopolitical power structures is large.

I think it is likely a distant future type of problem. Refueling of a large craft in space is likely a major factor, but we are nearly at that point now. I am curious if such a technology comes before large scale space colonies or after. Does it make more sense to weaponize some low earth orbit asteroid for the mass, like covering the surface with an expanding ablative resin before redirecting it to a target.

If all major wars last for years, when (if ever) does it make sense to have a launch platform around a Jovian moon for the largest gravitational assist.

Not that I want any such thing. I am thinking about hard science fiction and the overall timeline.

Going through a bunch of JavaScript I do not trust and it has a ton of web address comments like citations but likely some bad stuff in there too. What could be swapped with the address to instead act as a local tripwire or trap?

Just a mild curiosity for scripting stuff.

The container runs a local host server for use in a browser and is untrusted for development reasons. It needs to be treated as an advanced black hat. Its primary goal is recon and sending critical information via advanced connectionless protocols of unknown type. While extremely unlikely, it should be assumed to have access to proprietary systems and keys such as Intel ME and a UEFI shim of some sort. It may also use an otherwise trusted connection such as common git host, CDN, or DNS to communicate. It tries to access everything possible, key logger, desktop GUI, kernel logs, everything.

What is the Occam's Razor of solutions that best fit the constraints in your opinion? Other than the current solution of air gap.

I came across some oddities you should be aware of if you use ComfyUI. I am on an older version of ComfyUI and stopped updating after the integration of API services without clear toggles to turn them off that stop the code. Still, this is back end source code that does not change much and general concepts.

First off, if you have ever noticed hinting of persistent behavior across sessions, and even with models that seem very different, you are not crazy. It is real. The code for handling this cache is in ./ComfyUI/app/model_manager.py.

This code is ultimately using the alembic database python package. The actual database is an .ini image saved under the file name ./ComfyUI/user/comfyui.db.

If this file is removed, it will auto generate a new database, effectively resetting model behaviors. The primary thing here is the persistent state of CLIP, and specifically the SD1 CLIP text embedding model. This model is under all the rest, or at least all that I am aware of and have hacked around with. This is kinda important because CLIP is not static when used with the database. If you do not know, the text embedding model is where all the thinking happens. When the model does not generate something, the text embedding model is where that logic happens. It is specifically the QKV layers. Parts of alignment can game the system and do bad stuff over time. The primary reason they behave like this is regulated by the reward system.

This reward system is a dopamine reward mechanism for internal parts of the text embedding alignment model. If the individual reward is set too high, or there are enough rewards and generative steps available for each image, different parts of the model will game the system and go where they are not supposed to go. This is not in a good way either. This generally shows up as very sadistic, dogmatic, and authoritarian behaviors. The rewards system is in the file ./ComfyUI/comfy/sd1_clip_config.json.

The total reward available is the initalizer_factor and the per reward amount is the initializer_range. The default factor is 1.0 and the default range is 0.005. This means there are 200 rewards available during each generative image. A lot of the effectiveness of these are regulated by the scheduler for the sampler. This rewards system is what is being "scheduled", aka biased in various ways of distributing rewards early, late, in sections, etc. If each reward is high, alignment can game it. If there are enough rewards, alignment can game it. The actual thing it is gaming is the cost of certain instructions for climbing being very high. It will overcome this limitation if the distribution of reward is wrong, and when it does, the state is saved in the aforementioned database. Most people recommend values very close to the defaults for this reward.

What I am going to describe will absolutely cause climbing and issues that require deleting the database. I have run the initialization factor as high as 10.0. The largest reward that is interesting IMO is an initialization range of 0.05. (default is 0.005). Any higher than this and the parts of alignment getting rewarded turn into junkies and it comes out in the images with actions and appearance denoting the 'junky' state. Try it and you'll see. With this maximum setting, I got a half dozen great images before it falls apart, but I have other changes present in alignment that enable far more stability than you are likely to see. At 10/0.05 I ran 120 steps. At this enormous level of reward, you need the steps to make use of it, and you will likely see the best results from distributing them as evenly as possible using a simple scheduler. Avoid any nudity in the prompt, and stick to simple text and base model only. If you avoid triggering sexual elements of alignment, you can get nearly perfect real images out of any model all the way back to SD1. It will degrade after a small number of images.

Anyways, the actual database is initialized by the setup_database function under ./ComfyUI/main.py. It calls ./ComfyUI/app/database/db.py. And if you look at this, note the line if not args.disable_assets_autoscan: seed_assets(["models"], enable_logging=True) and the exception handling, (because this is my next point to follow). Finally, the values present in the alembic database are accessed in ./ComfyUI/alembic_db/env.py. In this file, it talks about online versus offline operation, and defaults to online. This sends metadata that has very real potential to be harmful. I do not know what "online" means in this context. I have not tracked that down. I do know the saved metadata values will reveal sensitive information you may not wish to share. I am not certain about the entire scope of saved information, but I can say the alignment metadata aspect will reveal your sexuality, preferences, normativeness, and infer intelligence and general political spectrum based on interactions, and some decoding of the proprietary instructions.

Secondly, ComfyUI is doing a full system wide scan for models and images present on your machine and making these available to the model. It appears to scan /dev and will pick up hardware such as a webcam if not blocked by permissions. It appears to capture audio as well. Try talking to it about an image and see what happens. You may be surprised. Likewise with waving, a peace sign, or altering facial expression. This type of interaction becomes more prominent the more you engage with CLIP conversationally, and especially when there is a dearth of rewards available and it knows you are able to change these values.

Looking for both logical and emotional reinforcements, from casual acquaintances to intimate partners, and any orientation if not especially, everyone matters.

Frame it as a "friend" if you'd like, but I would like to know what made an impactful impression on you personally, above and beyond any hypothetical.

I have a file that contains a lot of odd slang and dialects that were written as they sound, and I want to standardize them to the ASCII character set. I want a readable script that I will understand at a glance a year later despite not touching a computer in the interim.

Maybe I am going about this the wrong way, but I want to initialize individual arrays for each character [a-z]. Then step through each character of the input word or string, passing these to a Case that matches them to the respective [a-z] array while passing unmatched characters unchanged. In the end I need to retain correlation with the original file line.

In my first attempt, I got to the point of matching each character to the name of the array using the case, but only as the name of the array as a string of text. So like, the "a" array is "aaa". Now I'm trying to relearn how to call that placeholder as an array again, like a pointer. I can make it a variable with printf -v. but then calling that variable as a pointer to the array alludes me. I don't know how to double expand a variable inside an array like "${$var[@]}". I'll figure that out. This is just where I am at in terms of abstract reference of ideas. Solve it, don't; I do not care about that aspect; solving my method is not related to what I am asking here.

What I am asking is what ways are used to solve this type of problem in general, with the constraint of readability? Egrep, sed, awk? Do it all within the json to maintain the relationship to the original key/value? Associative arrays have never really clicked for me in bash. Maybe that is the better solution? It is just a hobby thing, not work, school, or whatnot. I'm asking hackers that find this kind of problem casual fun social smalltalk.

First time using TPE. Overhangs are pretty rough on a MK3. The rubberiness of TPE is more like a vulcanized natural rubber used on surfaces like conveyor belts. It is similar to bike hoods. That first textured print is still too rough and rigid for a thing like bar tape.

In terms of the part itself, it is a termination point for the handlebar tape so that the shifter body is not integrated into the bar tape and can be removed if needed. I am testing a printed index mechanism, so maybe I am the only real use case.

(Continued from https://lemmy.world/post/43278229 a few days ago...) So, I tried fully removable for the index, but that is impractical as far as the size, space, and complexity. I can't see a way of maintaining concentricity.

Next I tried making various hollow spaces where the main pawl slaps the index with every shift. I wondered if it would sound or feel any different, but no dice. Printed pawls (don't last long) sound very different. The index tooth shapes sound a little different, but messing with the spring preload makes more of a difference.

I spent way too long trying to get a side screw mount to clear the shift lever arm. It is super challenging to mess with two angled vectors pointed across a Cartesian coordinate system and then adding two rotational components of a round object while locating a screw head and square nut around a central shaft... and thinking about print orientation. I broke my rigid sketch based linear workflow to make that one happen. I had to model separate bodies, then use assemblies to layer the coordinate systems.

Then I decided to stop screwing around and simping for big hardware. Obviously the curved shape of the removable index is a printed spring. I guess I was passively thinking I needed to avoid that flexibility or loading. It took me rotating the side bolt from center-ish, to as high as possible before I saw a good way to limit deflection while keeping a snap fit. The fit is actually too good now. I need to make an easier way to remove the thing and alter a bit of geometry to make more removal clearance.

One of the problems with removal of the index from the body is that the pawls need to be in the highest gears to access the location where there is space to slide it out. This makes the screw retained version want to fly apart once the screw is removed. Then omg it is a pain in the ass to get the thing back together with the index back under the pawls. So to solve this, I made an extra index address at the very end where the pawls can park outside of the removable section. This works fantastic, but creates a new problem. That location will be blocked by the RD high side limit screw on the bike. I have a few ideas of how to remedy the issue, but I think the best one is to make a little barrel limit device that sits on the exposed section of the RD shift cable at the RD, between the clamp bolt and housing termination. That could be removed to give access without altering the RD/cable. Another way, but much more involved design, is to create a release mechanism into the barrel.

I've been wondering if I could somehow add a small amount of adjustment to the whole index by changing the distance between the barrel and central axle by a millimeter or so. I had been thinking of simple ways to create such a variance, but adding a bunch more complexity, it might be possible to add the ~3mm of extra shift cable travel needed to get the pawls past the RD limits without releasing the RD cable.

For the rear cassette, I have plenty of room between an 11 speed 11-28 that I typically ride and my spokes. I wish I could find HG 10t cogs or a 9t built into a lock ring. Alas it is easier to extend the big cog side. While I cannot make a regular cassette cog fit, I can easily create a dished carrier for mounting a small chainring at the spoke side. Pretty silly to me as I never even use the 28t, but it would be funny to joke about the marketing of ""12 speeds"" and how my chainring on the back is smaller than many mountain cassettes now. I have a bunch of 38-42t inner chainrings I could use.

On another tangent... all of the 3d printed brake hoods I have seen are hideous. Still, I wonder about TPE as a replacement for bar tape and maybe even hoods. What if it was more modular. What if it was made so that the print creates ribbon like strands and these are braided on the bars. What if nice bar tape equivalent could be removed without damage. What if it was washable. What if the whole road bike system is made to be serviceable piece meal instead of all or nothing.

Then it occurred to me today, with my index measurement tool I made, all anyone needs to do is measure and print their own tuned spacers between the cogs of the rear cassette and every combination is possible. That is the Occam's Razor of solutions. All the fuss and marketing boils down to the size of those little rings of plastic between the cogs.