Colab shell Terminal, ffmpeg, DeepSeek

For a disposable Linux command-line shell VM:

1. Load/create a Colab notebook,
2. Change the runtime type to 'T4' if you want to use an LLM (AI), 
3. Paste the following into a cell and run it:

XTerm

!pip install colab-xterm

%load_ext colabxterm

%xterm

An interactive terminal will start.

Notice that you start in the /content folder, to which you can upload and download files through the left-menu. 

Cool stuff you can do with this free compute from the comfort of your potato PC include:

ffmpeg

Colab notebooks come with many pre-installed applications that you can just run. For instance, ffmpeg -i input.mp4 -r 30 -c:v libx264 -c:a aac output.mp4 will transcode a video file that you have presumably uploaded.

Running ffmpeg this way is superior to running it through a code shell because you get progress indicators. It is also better than running it on your local PC because even through your computer can handle it, you can do other tasks without worrying about the jobs interfering with each other.

(Remember: option/flag position is important in ffmpeg)

DeepSeek Uncensored/Abliterated 

I want to try the DeepSeek LLM without creating an account. I also want to have controversial conversations without my edge-lord prompts getting rebuffed.

Soon after DeepSeek was open sourced some clever folks taught it to uncensor itself (abliteration).

Ollama provides a standard interface to LLMs.

You can chat with abliterated DeepSeek in the terminal by running:

curl -fsSL https://ollama.com/install.sh | sh

ollama serve & ollama run huihui_ai/deepseek-r1-abliterated:7b

References

• DeepSeek R1 on Google Colab
• DeepSeek Abliterated