Kalyan KS

Llama-OCR : Llama-based Open-Source OCR Tool

Llama-OCR is an open-source OCR tool based on Llama, the most popular open-source LLM.

Llama-OCR is built on the top of Llama 3.2 vision LLM powered by TogetherAI API.

Specifically, Llama-OCR is a wrapper on top of the Llama 3.2 vision model.

*Llama-OCR details in the comments

#ocr #llama #llms #nlproc #deeplearning #ocrtool #opticalcharacterrecognition

AI Agents Stack

AI Agents are advanced computer programs that use LLMs to automate complex tasks.

AI Agents Stack includes
- Vertical Agents
- Hosting & Serving
- Observability
- Agent Frameworks
- Memory
- Tool libraries
- Storage
- Model Serving

Pic credits: Adam Silverman

#aiagents #agents #llms #nlproc

Introduction to Large Language Models

This course covers

[1] Introduction to Transformer Architecture
[2] Attention is all you need
[3] Self-Attention
[4] Multi-headed attention
[5] Teacher Forcing and Masked attention
[6] Positional Encoding
[7] Zooming into Decoder layer
[8] Sinusoidal encoding
[9] Batch normalization
[10] Layer normalization
[11] Language Modelling
[12] Transformers for language modelling
[13] Causal Language Modelling
[14] Generative Pre-trained Transformer
[15] Pre-training and Fine Tuning
[16] Decoding Strategies (Greedy and Beam Search)
[17] Sampling Strategies (Top-k and Top-p)
[18] Sub-Word Embeddings (BPE, WordPiece etc.)
[19] BERT, GPT, BART models
[20] Fine-tuning Strategies
[21] Fast Attention, Sparse Attention and Local Attention
.................

This course is offered by Prof Mitesh Khapra from IIT Madras.

#llms #course #onlinelearning #nlproc #deeplearning