Yogesh Jadhav’s Post

With Multilingual Image Generation & Text Prompt, I've been exploring the fascinating intersection of language and art. Using stable diffusion models and Google Translate, I can generate stunning images from text prompts in various languages! 🎨🌍 https://lnkd.in/gHeXJtgm #AI #ImageGeneration #LanguageArt

🤖Agent Types 🔗 As incredible as the agents are, it is equally important to choose the appropriate agent type for our use case. 1️⃣ Zero-shot ReAct Without being trained on sepecific data Zero-shot ReAct Agent has the ability to create realistic contexts. It can be used for tasks involving generation of creative text formats, language translation, and generating different types of creative content. agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION 2️⃣ Conversational ReAct Designed for conversational use, it incorporates React framework to determine which tool to use couple with memory to remember the context of the previous interactions (conversation buffer memory). agent="conversational-react-description" 3️⃣ ReAct Docstore These Agents utilizes React framework to communicate with a docstore (cloud-based digital document archive and retrieval system). Uses a combination of lookup tool and search tool agent="react-docstore" 4️⃣ Self-ask with Search This Agent utilizes the Intermediate Answer tool (which can do wikipedia searches) for self asking questions. agent="self-ask-with-search" 🍀Pro Tip: Make use of - Verbose= True to see the "Chain of thought" series the agent goes through.  #Agents #autonomousagents #AgentTypes #AI #artificialintelligence #Langchain #LangChainAgents

🚀 Exciting news in the world of AI and ML! A new study introduces a Partial Relevance Enhanced Model (PREM) to improve Video Corpus Moment Retrieval (VCMR). The model addresses the partial relevance between query and video, enhancing both video retrieval and moment localization. Experimental results demonstrate that PREM outperforms existing methods, achieving a new state-of-the-art in VCMR. Stay ahead of the game with this cutting-edge development in AI and ML! #AI #ML #VCMR #DataScience #TechInnovation

This Bengali Sentence Error Correction Model is trained on 1.3 M Bengali incorrect-correct sentence pairs and archives a score of BLEU: 0.443, CER: 0.159, WER: 0.406, and Meteor: 0.655 when tested on unseen data. For a closer look at the model in action, check out the live demo here: https://lnkd.in/gSsh7u6x The full training process and methodology can be reviewed on GitHub: https://lnkd.in/gzeDC_TK The model was trained on Google Colab for a single epoch, spanning 6500 steps with a batch size of 128. I believe that expanding the dataset and extending the training duration could significantly enhance the model’s performance and utility. That would be the next step for this model. Try this model and let me know your experience. https://lnkd.in/gb6iPqwQ #BengaliNLP #MachineLearning #NaturalLanguageProcessing #Finetune #LLM #DataScience #AI #DeepLearning #HuggingFace #ErrorCorrectionModel

This is a startling example of why automated quality assurance tools may not work in certain languages or use cases. Many tools are fine tuned for English language and American spelling. They may fail in British spelling, sometimes. Their performance in other languages is questionable. Translation agencies and project managers need better training on AI enabled tools. Running a pilot project is always a good idea before obliging professional translators to use a new tool or function. In this case, I am sorry for the translator colleague. If I were the translation agency, I would pay her extra or send a gift to help compensating for the time lost and more importantly for the emotional damage.

With the holidays upon many of us, I wanted to add some holiday reading materials for those using LLMs. The article below (by Natalia Kuzminykh here at Winder.AI) is about Calculating Large Language Model Token Counts. This article is a practical guide that aims to demystify the concept of token counts in LLMs, detailing how they influence model performance and interaction. Starting with the basics, it defines a token and how it’s used for text processing. The paper then moves to the tokenization process, examining different models and the tools used for token counting. Finally, it looks at both manual and automated methods for calculating tokens, highlighting their roles in larger datasets. https://lnkd.in/e8Us_p-y Happy Reading! #LLM, #ContextWindows, #TokenCounts, #LanguageModels, #AI

I went through the Gemini’s technical report and here’s 4 of the most interesting things 1) Gemini can understand audio and the pro (GPT3.5 equivalent) model outperforms Whisper. It also seems to work better on proper nouns and rarer words 2) Reduce hallucinations by fine tuning on 3 aspects - Attribution: If it is given context such as text from a book, it stays truthful to the context - Closed book response generation: If there is no context and asked a question about factuality, it is also trained to not hallucinate information - Hedging: If it is given a question with no possible answer, it should admit that it cannot provide that response 3. Given the news we received earlier that Gemini would be delayed to Jan 2024 because they wanted to improve it’s performance on non-english languages, I was expecting its performance in language translation to be much better than GPT4 But Gemini doesn’t seem that much better at language translation compared to GPT4. It’s Ultra model is about .2-1.2% accuracy better than GPT-4 on translation and .2% worse than GPT4 when translating to English 4. Interleaving text and image generation. You can create charts with text on one side and images on the other side #ai #artficialintelligence #ml #gemini

Ever wondered how machines understand human language? Spoiler alert: It's not just about the words! Here I am going to try my best to discuss an important concept of the "Transformer Model" in simple terms. Recently I have been exploring the realm of AI and ML and I found an important foundation-level concept of the "Transformer Model" which was introduced in the article "Attention Is All You Need." Published in 2017 by researchers Ashish Vaswani and others. Unlike us, machines don't get context by default. They need a little help, and that's where the Transformer shines. It ditches the old sequential playbook for something smarter: attention mechanisms. Here's the deal: The Transformer model doesn't just read words; it understands how each word connects to another. It's like having a superpower that highlights the important stuff while reading. And it's not a lone ranger. It has two sidekicks, the Encoder and Decoder, each packed with layers that act like mini-brains. They work together to make sense of language, with a special trick up their sleeve called multi-head attention. This lets the model multitask, capturing different nuances of word relationships simultaneously. But wait, there's more! Feed-forward networks step in after the attention magic, piecing together complex patterns like a friend who helps you solve a puzzle. And when it comes to learning languages, the Transformer is a quick study. It's efficient in training and inference, making it a whiz at tasks like translating languages on cutting-edge hardware. Imagine having a buddy who's not just bilingual but multilingual, grasping the full context and vibe of any conversation. That's the Transformer model for you – a language genius in the digital realm. Drop a like, share your thoughts in the comments, or hit that repost button to keep the conversation going! This was also discussed in Nikhil Kamath's Podcast where Varun Mayya mentioned this. #MachineLearning #ArtificialIntelligence #productmanagement

Evaluate anything you want | Creating advanced evaluators with LLMs 1. The Importance of Evaluating Language Models 1.1. Understanding the capabilities and limitations of language models is crucial for aligning with business objectives. 1.2. Standard metrics such as perplexity, BLEU scores, and sentence distance often fail to capture the subtle nuances in real-world applications. 1.3. LLMs use 'black box' metrics to assess the quality of generated text using large language models themselves. 2. Building and Running Evaluation Examples 2.1. Build chatbots and RAG using the LangChain framework, as it incorporates simple evaluation functionalities. 2.2. Demonstrate how to create custom evaluators through detailed code examples. 2.3. Discuss the implementation of translation quality assessment and context relevance evaluation. 3. Strategies for Creating Assessment Prompts 3.1. Establishing assessment criteria and defining a numerical scoring scale are key to success. 3.2. Require reasoning behind scores to gain deeper insight into the assessment logic. 3.3. Provide queries and context for reference, and demand a strict response format for ease of parsing. 4. Implementation and Optimization of Evaluation Chains 4.1. Implement a basic evaluation chain class for parsing output scores and reasons. 4.2. Consider the randomness of evaluations by running asynchronously and averaging scores. 4.3. Integrate frameworks to leverage their maximum benefits (optional). 5. Case Studies in Practical Assessment 5.1. Assessment of translation chains from English to French, identifying and reasonably explaining issues. 5.2. Contextual relevance assessment effectively identifies information unrelated to the query. 5.3. Results visualization and experimental tracking through the Langsmith platform. 6. Conclusions and Practical Applications 6.1. Customizing control model performance allows companies to build AI systems that align with their unique business objectives. 6.2. Encourage experimentation and the creation of custom evaluators for specific use cases. 6.3. All code is available on GitHub, facilitating practical application and further development. #LLMPerformance #CustomEvaluators #LangChainTech #TranslationQuality #RealTimeFeedback