{"slug": "llm-day-1-intro", "title": "LLM - Day 1 - Intro", "summary": "A developer has published a beginner-level explainer on large language models, covering core concepts such as weights, parameters, transformer architecture, tokenization, context windows, and sampling controls like temperature, top-k, and top-p. The writeup also distinguishes model files from the tools used to run them locally, including llama.cpp, Ollama, and LM Studio, and notes that fine-tuning wraps a model rather than updating its weights.", "body_md": "## \n  \n  \n  Large Language Model\n\n- Its a model\n- what is a model ? its a equation.\n- Then what is equation ? y = mx + c or y = m1x2 + m2x3 + m3x2\n- Models are nothing but WEIGHTS.\n- Take an example Eg: gpt-oss-120b , for this model they would have used data with TRANSFORMER ARCHITECTURE and we are getting some output that is called MODEL.\nm --> weights ( we can adjust the m and get many desierd outputs )\n\n- Model is floating point storage.\n- சில பேரு சொல்லுவாங்க இது வந்து 120B Paramter model , what it means ? These Parameters are nothing but how many weights I am having inside the model.\nநான் எந்த அளவுக்கு dataவை கொடுத்து train பண்ணி இருக்கேன்னு அர்த்தம். In other words , CONTEXT அதிகம் .\n- Can we run this 120B Paramter in machine ? \n--> Where the OS sits ? C drive or /boot , initially it will be in hard disc , once the machine starts , it will go to RAM Running memory போய் உட்கார்ந்து கொல்லும். So to run the model we need RAM and also the CPU , so these two decides which model to run. Eg., in laptop we can run 7B parameter.\n- In local like laptop we can run eg., llama cpp, ollama , LM Studio . இது எல்லாம் modelஆ ? No , these are softwares/tools to run the model.\n- Open model are called like \" Open weights model \". Eg., chatgpt model , gemini model. ( Referring to Open source )\n\n**Why I Need a model ?**\n\n1. LLM is a next word predictor.\nEg., Hi How ......... இதுக்கப்புறம் என்ன வரலாம் like are you , is your life , is the food . இது எல்லாம் ஒரு possibilities ( It can have n no of possibilites ).\nMORE THE PRESENCE , MORE POSSIBILITY OF OCCURING.\nநீங்க ஒரு வார்த்தையை குடுக்குறீங்க , அதுக்கு அடுத்த வார்த்தை இது கண்டுபிடித்து தர போது . Then this will go as input , like wise it will generate.\n2. All each possibilities will have SCORE , entire score will like between [ 0 - 100 ] or [ 0 - 1 ] . \n3. If you are not giving any controlling parameter , then it will chose the highest SCORE.\n4. If you want to control the LLM , at that place . We have controlling parameters. Namely :\n\n*Temparature*\n\n1. 0-1\n2. controls how random or safe the model choices.\n3. If the TEMP is low like ( 0-0.3 ) then it will ACT BASED ON THE FACTS. Likely having max score / most occuring one.\n4. If the TEMP is high like ( 0.7 - 1+ ) , model is willing to choice less likely words. Eg., in storing telling or in creating writing. It will choose less likely words from dataset which is trained on.\n\n*Top-k*\n\n1. Along with temperature its used.\n2. Model only considers , Top k most items. \n3. K --> Elements\n\n*Top-P*\n\n1. Also its called Sampling methods.\n2. It will add the score and then it will work.\n3. Along with temperature its used.\n\nஇப்ப Top - pகு 90 % value  சொல்லிட்டாங்க .  இப்ப அது எப்படி work ஆகும் நா. Each prediction will have a score eg., first prediction  0.70 ,  அடுத்த prediction 0.15 , அடுத்த prediction 0.30 . It will add the prediction score which is equal to 90 % ( more or less ) இத வச்சு தான் அந்த output கொடுக்கும்.\n\n**Tokens**\n\n1. words ?\n2. which are \" Not complete \".\n3. Entire text into pieces.\n4. Tokenization ? Process of chopping your text into tokens.\n5. Each token will be assigned numbers.\n6. Take it like a translator. \n\n**Context Window**\n\n1. Model ஓட short memory.\n2. Short term memory.\n3. Max of tokens a model can see in a given time.\n4. ஒரு timeல  ஒரு model லால , எவ்வளவு விஷயத்தை பார்க்க முடியும்.\n\n**TBD  :**\n\n1. What is dataset ?\n2. Chunkings ?\n\n**Notes :**\n\n1. If I am going to use a model , then I need to unzip it , for this i need a running memory so we need RAM. Now , i need process the data , then CPU & GPU ( training a& usage ) are required.\n2. Weigts will be like FOLATING POINTS or EMBEDDINGS.\n3. \n[https://groq.com/](https://groq.com/) --> Its a tool --> They give free models for usage.\n\n1. Fine Tuning is not updating the model or touching the weights. We use like rapper around the model thats called Fine Tunning.\n2. Markov Chain --> this is used in next word prediction in gmail , word etc .\n3. Harry Porter and context window .", "url": "https://wpnews.pro/news/llm-day-1-intro", "canonical_source": "https://dev.to/technonotes/llm-day-1-lets-learn-ai-4p9l", "published_at": "2026-09-17 19:19:41+00:00", "updated_at": "2026-09-17 19:52:59.887692+00:00", "lang": "en", "topics": ["large-language-models", "natural-language-processing", "ai-tools", "generative-ai"], "entities": ["GPT-OSS-120B", "llama.cpp", "Ollama", "LM Studio", "Groq"], "alternates": {"html": "https://wpnews.pro/news/llm-day-1-intro", "markdown": "https://wpnews.pro/news/llm-day-1-intro.md", "text": "https://wpnews.pro/news/llm-day-1-intro.txt", "jsonld": "https://wpnews.pro/news/llm-day-1-intro.jsonld"}}