Naveen Arulselvan

❝ How do you taste your Wine? With the right Glass, served from the right Bottle. How do you utilize your LLM? With the right Task, served from the right Compute environment. ❞ If the quote above caught your interest, check out my blog post that was featured on the Toyota Connected India Medium channel last week: Leveraging LLMs in Cloud - 🔗 https://lnkd.in/gfHsdevH --- I hope this blog adds a bit of amusement for the seasoned AI/ML engineers 🧠👨‍💻 with the carefully crafted images and perhaps even the silly wine analogies 😉. Whether you're a cloud engineer ☁️ 👨‍💻, a product owner of cloud-based apps, or someone new to Generative AI in the cloud, this blog offers a detailed, beginner-friendly look into the selection and deployment of LLMs in cloud environments. --- A heartfelt thanks 🙏 to my colleagues at Toyota Connected India for taking the time to review, shape this piece and even share a creative writeup like below in the original post 🫡 👇 🤔 Curious to dive in? Click the link below to explore the blog on Medium! 🌐 🔗 https://lnkd.in/gfHsdevH

18

This how you approximately calculate GPU memory required for any LLM. Read “How Much GPU Memory is Needed to Serve a Large Language Model (LLM)?“ by Mastering LLM (Large Language Model) on Medium: https://lnkd.in/dh9V5-8s

6

AEs (Variational Autoencoders) and AEs (Autoencoders) are neural network architectures used for dimensionality reduction and data compression. They share a similar structure with an encoder and a decoder, but they differ in how they handle the latent space, which is a compressed representation of the input data. Here is a good article explaining the difference between AEs and VAEs. #AE #VAE #neuralnetworks #deeplearning #computervision I like to share my Data Science journey with my LinkedIn friends… I believe learning everyday is part of life. Follow me if you like my posts. keep reading. keep learning.

9

Last 2.5 weeks have been a grind. Watched 9 lectures of Prof. Mitesh Khapra deep learning course, plus assignments from iitm bs and nptel. Ive also been referring to: https://lnkd.in/dfpNYN7V, https://d2l.ai/ and https://lnkd.in/dan7PvAG. Excellent books. Got through 4 key papers—𝐃𝐫𝐨𝐩𝐨𝐮𝐭, 𝐁𝐚𝐭𝐜𝐡 𝐍𝐨𝐫𝐦, 𝐋𝐞𝐍𝐞𝐭, 𝐀𝐥𝐞𝐱𝐍𝐞𝐭. Spent an entire day on batch norm backprop calculus (spoiler: didn’t get it), but managed to make progress the next day. Also implemented 𝐋𝐞𝐍𝐞𝐭 and 𝐀𝐥𝐞𝐱𝐍𝐞𝐭 in PyTorch. Links below to the code and my notes: 𝐀𝐥𝐞𝐱𝐍𝐞𝐭: https://lnkd.in/dcFB7pX3 𝙸𝚖𝚙𝚕𝚎𝚖𝚎𝚗𝚝𝚊𝚝𝚒𝚘𝚗: https://lnkd.in/dDHNqjYD 𝐋𝐞𝐍𝐞𝐭: https://lnkd.in/dHQ8n_Hn 𝙸𝚖𝚙𝚕𝚎𝚖𝚎𝚗𝚝𝚊𝚝𝚒𝚘𝚗: https://lnkd.in/dVZmFZaW Deep learning’s a crazy mix of fun, a little bit of frustration, and a ton of focus :) Planning to dive into more vision papers this week before I take on Lecture 10....

14

Transformers Transformers has been a major milestone in the field of NLP and heavily used in Generative AI. There are 3 variants of transformer based models and they are 1. Encoder-only 2. Decoder-only 3. Encoder-Decoder Encoder only models: These are also called as autoencoders and these are pretrained using a technique called masked language modeling. A text with a random masked token is sent to the model to predict the masked token. For example, consider the text "If you don't stop at the sign, you will get a ticket" The training input that will be passed on the encoder model is "If you don't _____ at the sign, you will get a ticket. The model is expected to predict the token (word) "stop" These models use bi-directional representations of the input to better understand the full context of a token. Examples of encoder models: BERT Family models Decoder only models: These models are called autoregressive models and are pretrained using a technique called causal language modeling. These models predict the next token using the previous tokens. For example, consider the text "If you don't stop at the sign, you will get a ticket" The training input that will be passed on the encoder model is "If you don't ______" The model will still try to predict the word "stop", but only based on previous tokens. These models are used for generative tasks, including question answering. Examples of decoder only models: GPT Family, Falcon, LLama models Encoder decoder models: These models are called sequence to sequence models and the pretraining objective varies from model to model. For example the popular FLAN - T5 uses a consecutive multitoken masking called span corruption. For example, consider the text "If you don't stop at the sign, you will get a ticket" The training input that will be passed on the encoder model is "If you don't _____ _______ the sign, you will get a ticket. The model will try to predict the tokens "stop at" These models are good at translation tasks. Examples: T5 family In all the above explanations we just took one sentence of text. The LLMs you see in the market are trained on huge volumes of text available over the internet. #LLM #encoder #decoder #transformer #genai

3

This guy needs to learn to stop judging people for doing their experiments. I have never met an MLE who doesn't understand learning curves. maybe he doesn't understand double descent phenomenon. And yes, it also happens for epochs, see the link to my post below for epoch-wise double descent. Quoting from openAI blog on deep double descent https://lnkd.in/g-BNqcfD : "performance first improves, then gets worse, and then improves again with increasing model size, data size, or training time." I wrote a post on epoch-wise double descent phenomenon here https://lnkd.in/ga9c_dGM Deep Double Descent: https://lnkd.in/g-BNqcfD More of it: https://lnkd.in/grAb8_Rw Visit our website @ https://lnkd.in/gm_rfz5R

9

6 Comments