Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete

huggingface.co

Post reply on HN
Print view View original

141–150 of 160 posts

Re: Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete

#141

Post by cmrdporcupine »

kevinlu1248 wrote: post #135
cmrdporcupine wrote: post #92

I've been trying my hands at implementing an emacs package for inline completions with this. I have it mostly working and performance is good enough but I haven't been blown away by the quality of its suggestions unfortunately. Which I guess is expected from a 1.5B model. I'd love to see them making a larger model in the 10-20b range maybe? I know most people wouldn't be able to run that on their machines, but some c…

Are you using the right format? https://huggingface.co/sweepai/sweep-next-edit-1.5B/blob/mai...

Yea, I tweaked it a bunch to try to follow what was described there

Re: Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete

#142

Post by kevinlu1248 »

zarzavat wrote: post #98
genghisjahn wrote: post #91

Earlier quoted context omitted.

I’m in the “write new stuff with cc and get great code.” Of course I’ll be told I don’t really know what I’m doing. That I just don’t know the difference between good and bad quality code. Sigh.

The best code is no code.[0] The main "issue" I have with Claude is that it is not good at noticing when code can be simplified with an abstraction. It will keep piling on lines until the file is 3000 lines long. You have to intervene and suggest abstractions and refactorings. I'm not saying that this is a bad thing . I don't want Claude refactoring my code (GPT-5 does this and it's very annoying). Claude is a junior…

Definitely agree here, have had so many cases where I would like ask Claude for XYZ, then ask for XYZ again but with a small change. Instead of abstracting out the common code it would just duplicate the code with the small change.

Re: Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete

#143

Post by kevinlu1248 »

sitkack wrote: post #76
qorrect wrote: post #73

Earlier quoted context omitted.

Glad to hear I'm not alone, the latest releases of JetBrains have been so bad I finally cancelled my subscription. VSCode has been a nice surprise, "its giving old emacs" as the kids would say.

I am curious about how both of you think Jetbrains is dropping the ball so much that you are no longer buying the tool. You are still using it but no longer getting updates?

I've done some testing before and many of the new Jetbrains internal plugins cause memory leaks which really lags down my IDE...

Re: Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete

#144

Post by kevinlu1248 »

KronisLV wrote: post #117
cmrdporcupine wrote: post #89

Earlier quoted context omitted.

I use free RustRover for my open source work. I have not purchased a (new) license for my commercial work, because I haven't been getting as much value as it since my flow has switched to primarily agentic. Mainly, they're pushing Junie and it just isn't that good or compelling, when faced off against the competition. The key thing for me is that I think they had an opportunity here to really rethink how LLMs could i…

> I have not purchased a (new) license for my commercial work, because I haven't been getting as much value as it since my flow has switched to primarily agentic. I still buy a personal Ultimate license because I want to see them succeed even if like 80% of my time is spent either in a CLI or Visual Studio Code (for quicker startup and edits), a bit unfortunate that Fleet never got to be really good but oh well.

Also wish Fleet took off, not a fan of installing a new IDE for every separate repo that's in a different language

Re: Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete

#145

Post by keepamovin »

kevinlu1248 wrote: post #127
keepamovin wrote: post #48

This is so cool. What is the second order effect of model training becoming democratized? And local models becoming the norm? Tasks like agentic work are well handled by current AI as long as you know what you're doing and can stress the agent against tests/spec, etc. I am thinking that one effect is: - it will become normal for meta-models to train a model specific to a particular task/product. Also, differently, I'…

Personally, I think usable AI is more valuable than simply more intelligence. Many of the labs are pushing towards models that are 1% better on CodeForces and AIME if you just let it think and use tools for hours, instead of more user-friendly models with better coding habits, like writing shorter and more modular code.

Totally this. But the corp labs have incentives to keep researching per investors and staffing load, so they have to show work.

I guess a nice advantage of backwardness here is that economic opportunities exist for those who can solve pain points in the use of existing intel. Older models often do almost as well at agentic tasks in reality, can probably go further.

Still, AGI should remove a lot of this making it redundant, and it will then be more about the intel than the tooling. But an opportunity exists now. We may not have widespread AGI until 8 - 10 years later, so plenty of money to be made in the meantime.

Re: Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete

#146

Post by ttoinou »

kevinlu1248 wrote: post #132
ttoinou wrote: post #50

Wow, I can even chat about C code with that model with LM Studio on my Macbook at 200 tokens per seconds

Haha, we never trained it for chat but I would bet it works regardless. Also that's crazy, M4 Mac?

M4 Max 128GB yeah

Re: Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete

#147

Post by Semaphor »

At least for C#, the quality of the cloud offering is rather mediocre, so I don’t expect this model to be that useful there. It’s very overeager, suggesting tons of stuff that I never accepted because it made no sense. It’s also producing bad code, wanting me to use `.Result` for async calls instead of simply await-ing.

Re: Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete

#148

Post by dainiusse »

notsylver wrote: post #61

I've been waiting for something like this for ages. Cursor making me pay $20/month when all I use from it is autocomplete was always a little annoying, especially as they changed the UI to push agents more and it got in the way. I was even considering doing it myself but wasn't sure about gambling on models small enough to run locally being smart enough to do anything useful. I threw together a vscode extension to ru…

Do you have anything to share? Would be curious trying it out

Re: Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete

#149

Post by Imustaskforhelp »

mika6996 wrote: post #99
Imustaskforhelp wrote: post #97

Earlier quoted context omitted.

Edit: I asked chatgpt and just tinkered around till I found a setting which could work { "agent": { "default_model": { "model": "hf.co/sweepai/sweep-next-edit-1.5B:latest" } }, "inline_completion": { "default_provider": { "model": "hf.co/sweepai/sweep-next-edit-1.5B" } }, "chat_panel": { "default_provider": { "model": "hf.co/sweepai/sweep-next-edit-1.5B" } } } Then go on the down bottom AI button or that gemini like…

You sure this works? inline_completion and chat_panel give me "Property inline_completion is not allowed." - not sure if this works regardless?

I really don't know, I had asked chatgpt to create it and earlier it did give me a wrong one & I had to try out a lot of things and how it worked on my mac

I then pasted that whole convo into aistudio gemini flash to then summarize & give you the correct settings as my settings included some servers and their ip's by the zed remote feature too

Sorry that it didn't work. I um again asked from my working configuration to chatgpt and here's what I get (this may also not work or something so ymmv)

{ "agent": { "default_model": { "provider": "ollama", "model": "hf.co/sweepai/sweep-next-edit-1.5B:latest" }, "model_parameters": [] },

  "ui_font_size": 16,
  "buffer_font_size": 15,

  "theme": {
    "mode": "system",
    "light": "One Light",
    "dark": "One Dark"
  },

  // --- OLLAMA / SWEEP CONFIG ---
  "openai": {
    "api_url": "http://localhost:11434/v1",
    "low_latency_mode": true
  },

  //  TAB AUTOCOMPLETE (THIS IS THE IMPORTANT PART)
  "inline_completion": {
    "default_provider": {
      "name": "openai",
      "model": "hf.co/sweepai/sweep-next-edit-1.5B"
    }
  },

  //  CHAT SIDEBAR
  "chat_panel": {
    "default_provider": {
      "name": "openai",
      "model": "hf.co/sweepai/sweep-next-edit-1.5B"
    }
  }
}

Re: Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete

#150

Post by ihales »

theophaniel wrote: post #80
woile wrote: post #79

Earlier quoted context omitted.

I've been using it with the Zed editor and it works quite well! Congrats. This kind of AI are the ones I like and I'm looking to run in my workstation.

Could you give the gist / config on how you made it work with Zed ?

I had Claude add it as an edit-prediction provider (running locally on llama.cpp on my Macbook Pro). It's been working well so far (including next-edit prediction!), though it could use more testing and tuning. If you want to try it out you can build my branch: https://github.com/ihales/zed/tree/sweep-local-edit-predicti...

If you have llama.cpp installed, you can start the model with `llama-server -hf sweepai/sweep-next-edit-1.5B --port 11434`

Add the following to your settings.json:

```

  "features": {
    "edit_prediction_provider": { "experimental": "sweep-local" },
  },
  "edit_predictions": {
    "sweep_local": {
      "api_url": "http://localhost:11434/v1/completions",
    },
  }
```

Other settings you can add in `edit_predictions.sweep_local` include:

- `model` - defaults to "sweepai/sweep-next-edit-1.5B"

- `max_tokens` - defaults to 2048

- `max_editable_tokens` - defaults to 600

- `max_context_tokens` - defaults to 1200

I haven't had time to dive into Zed edit predictions and do a thorough review of Claude's code (it's not much, but my rust is... rusty, and I'm short on free time right now), and there hasn't been much discussion of the feature, so I don't feel comfortable submitting a PR yet, but if someone else wants to take it from here, feel free!

Post reply on HN