Posts

AutoGPT stuck repeating the same task loop Last night I let AutoGPT try a simple research job. It found the first website, wrote a plan, then looped on the same “visit site” action forever. INFO Running with goal: Research Rust async examples THOUGHT Visit https://tokio.rs/ ACTION browser.open_url RESULT Page loaded THOUGHT Visit https://tokio.rs/ ... Docker image significantgravitas/autogpt:0.4.2, Python 3.10.11. Running on WSL2 Ubuntu 22.04. OPENAI_API_KEY set, no proxy.

Hugging Face Transformers crawling on WSL2 with CUDA Ran a simple AutoModelForCausalLM generate on a T4 but got one token every five seconds. from transformers import AutoTokenizer, AutoModelForCausalLM import torch, time model = AutoModelForCausalLM.from_pretrained("gpt2") tokenizer = AutoTokenizer.from_pretrained("gpt2") t0 = time.time() model.generate(**tokenizer("hello", return_tensors="pt").to("cuda")) print(time.time() - t0) # ~5 seconds for a single token GPU showed zero utilisation.

langchain load_tools() throws KeyError for ‘requests’ Yesterday I was wiring up a quick agent and called: from langchain.agents import load_tools tools = load_tools(["requests"]) The import blew up with: File ".../langchain/agents/load_tools.py", line 78, in load_tools tool = TOOL_NAME_TO_CLASS[name] KeyError: 'requests' Python 3.11, langchain 0.0.259, running on Windows 11 PowerShell.

OpenAI API returns 429 despite backoff Yesterday I was trying to run a batch job through the completions endpoint using text-davinci-003. Got this back immediately: { "error": { "message": "You exceeded your current quota, please check your plan and billing details.", "type": "insufficient_quota", "param": null, "code": "insufficient_quota" } } No Retry-After header. First request failed. Slowing down didn’t help.

I use AWS CodeBuild to build and then deploy this Hugo static site to s3. The site is also sat behind Amazon’s CloudFront CDN and after a sucessful deployment I wanted to delete the cache automatically so I could see the changes from the build immedietly.

I was recently doing some load testing for a client and when uploading our Taurus YAML tests into the Blazemeter console we recieved the following error: BlazeMeter does not support the inline scenario configuration. Please make sure you configure the scenarios outside the execution section.

I have been playing around with .NET Core and Blazor recently to create a client side application. When it came to deploying the site I thought I would try using something different, in this case it was [Amazon’s Amplify] (https://aws.amazon.com/amplify/) which appears to be Amazon’s answer to Netlify.

This is a quick cheat sheet for the steps involved to create a manual pipeline to deploy a static site with CDN and HTTPS to AWS automatically on commit, the simple solution nowadays is to use [Amazon’s Amplify] (https://aws.amazon.com/amplify/) but that wasn’t around back when I had to do this, but you might also need it if you have some non standard build requirements.

I needed to use a custom docker image for some Dotnet Core 3 work with Amazon CodeBuild and hit into a few issues that took a bit of digging around to find, so thought I’d leave the here should anyone else need the same info.

This is a dumping ground for software development quotes I have collected over the years that resonated with me (work in progress): Clear is better than clever Software engineering is what happens to programming when you add time and other programers.