Abby Soni

Understand Code and know human behaviour

[Abby Soni @AbbyReact]

https://t.co/22ummcK8Z0

[Abby Soni @AbbyReact]

Enginnering team with claude code

https://t.co/y5xYzgNflp

[Alok Kumar @Alokkumarzz]

MY FRIEND IS STILL PAYING FOR GMAIL STORAGE.

I told him to do this before.
He went from 14.9/15 GB to 6 GB in a single afternoon.

Hope this helps you ↓

[Abby Soni @AbbyReact]

Windows rescue pendrive

https://t.co/aumWPsEkFh

[Abby Soni @AbbyReact]

Kal hara kasht hara

https://t.co/hux3yf7fPh

[Abby Soni @AbbyReact]

Check if second hand phone is stolen

https://t.co/XRqc2qcp8t

[Abby Soni @AbbyReact]

Must have things with claude to enhance the results

https://t.co/gIIKzyiGMP

[Abby Soni @AbbyReact]

https://t.co/GtNXFZyaZj

[Abby Soni @AbbyReact]

Professional cheque printing

https://t.co/3eIbRZB1BJ

[Abby Soni @AbbyReact]

Professional cheque printing

https://t.co/3eIbRZB1BJ

[Abby Soni @AbbyReact]

https://t.co/Uni4dylCVF

[Abby Soni @AbbyReact]

https://t.co/GtNXFZyaZj

[Abby Soni @AbbyReact]

Ai agent main cmponents

https://t.co/uAl4X7u0hu

[Abby Soni @AbbyReact]

Do this every morning
https://t.co/MGjtoJYI7Q

[Abby Soni @AbbyReact]

Proxy server n exam centre

https://t.co/oLfKnU8S4E

[Abby Soni @AbbyReact]

Create video replicas of amit kilhor

https://t.co/jCUXFZMk7e

[Abby Soni @AbbyReact]

Apply for trademark

https://t.co/Vnoi88MnZy

[Abby Soni @AbbyReact]

https://t.co/x4beNVizOU

Block website on laptop using notepad

[Abby Soni @AbbyReact]

Agent testing
https://t.co/90owmuuchI

[Abby Soni @AbbyReact]

Train llm model easily -1

[Tech with Mak @techNmak]

Fine-tuning an LLM used to require a PhD. Now it takes 3 clicks.

We have been gatekept by complex Python scripts and "CUDA OOM" errors for too long.

I found the tool that democratizes custom models. It’s called LlamaFactory.

It is the "Automatic1111" for Large Language Models.

The "Cheat Code" => It provides a WebUI (LlamaBoard) that lets you visually configure, train, and evaluate models without writing a single line of code.

This unlocks "Zero-Code" Fine-Tuning for everyone:

1./ Day-0 Model Support
It supports 100+ Models out of the box. Llama 3 (and 4!), Mistral, Qwen, Gemma, DeepSeek, Phi-3... if it's on Hugging Face, it's likely here.

2./ Advanced Techniques Made Simple
Want to run QLoRA (4-bit tuning) on a consumer GPU? Want to align your model with DPO or ORPO? It’s just a dropdown menu. No complex math required. It even integrates Unsloth for 2x faster training.

3./ Visual Monitoring
Stop staring at terminal logs. Watch your loss curves and evaluate model outputs in real-time directly inside the browser (LlamaBoard).

It runs locally, on Colab, or on your server.

Stop writing training loops. Start building models.