@Alen_MN_357 Ne možeš ništa sa 10-15 sati. Skinuo sam gomile podkasta, fokusirao se na voditelja, očistio praznine da ostane samo govor, trenirao danima i opet to nije to. Mislim da ne vrijedi truda.
To be honest, I used to refresh a YouTube tab with a “Blazor” search open every single day. Not so much anymore, because I spend all my time coding.
Coding now takes up all the free time I used to spend on other things.
Instead of making our lives easier, I feel like AI coding has completely taken over our lives 🙂
@NullBurn Just spent a couple of hours testing this, and all I could get was 33-39 tok/s. Much slower then TensorFold. Really disappointed by this misinformation.
@theo I tried T3 before, but I didn't like it at the time, so I started using Herdr instead. I decided to give T3 another try two days ago, and now I'm hooked. It's so good!
No, of course not. I'm talking about the Web functionality, so that I can add "?w=100" to an image URL and get a 100-pixel-wide image that was resized by some other free library and saved to the cache. Something similar is now offered by Cloudflare and some other CDN providers.
I think you've done a really great job and I highly appreciate your work, but I simply can't justify paying $800 a year for what I described above, because in my case that's literally all I use - "?preset=thumb". And I'd love to be able to pay something in line with my needs and keep using the existing library.
I didn't know about that. If I remember correctly, it was free when I started using it. I read a post on some site (back when people still read blog posts) and installed the NuGet package. I didn't pay attention to any of that until I read the post here on X about version 4.0. For years, my whole process was right-click and update all NuGet packages, and then I see a post saying there's a commercial license in v4. The level of functionality I need I could build in a few hours, but the mere thought of touching any part of a site that has been running without issues for years leads me to the conclusion that I'd rather pay for some kind of upgrade and simply not change anything. Mostly out of gratitude.
I believe that most of the 95% of people still downloading version 3 also don't know that a commercial license exists there too, or they simply don't want to deal with free license applications.
The company has revenue of over $1 million, but not from that site, and I can't justify a monthly license cost just for resizing photos. I'd still like to pay something, given that I've been using ImageSharp.Web for free for years. Right now I pay nothing and I'm not asking for anything new for free.
I'm just mentioning this as an example, in case there are other developers in a similar situation who would happily move to the new version (which, in my case, realistically doesn't bring any new value since resizing will work the same) with a one-time upgrade license, instead of the $80/month, i.e. $800/year. But I understand it might be too small an amount for anyone to bother with.
@ChrizBogota@TechMDAI That’s exactly how I set them up as well: two pairs of two Sparks, with a LiteLLM proxy in front of them doing session-level load balancing so I always get maximum performance.
@Blackwellboy@ViC305@Blackfrost_AI So what exactly is the article saying? That Mia is an avatar? And that's a bad thing?
Honestly, if it weren't for her recipes, I would never have gotten into the world of LocalAI or bought 4xSpark while they were still affordable. Thanks @MiaAI_lab
@0xSero Tried your Inkling-Small 2x Spark recipe - it works for me, but only with --enforce-eager. Can't build the pinned image because image-root/ isn't in the repo, and my own SM121 paged-KV fix crashes during CUDA graph capture. Any chance you could publish link for it ?