BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:SmartAttica Τraining modules for SMEs - Module 12 "Inference with
  Transformers & Semantic Search with Sentence Transformers"
DTSTART;VALUE=DATE-TIME:20260622T080000Z
DTEND;VALUE=DATE-TIME:20260622T110000Z
DTSTAMP;VALUE=DATE-TIME:20260612T132429Z
UID:indico-event-209@events.grnet.gr
DESCRIPTION:\n\n\n\nGRNET announces\, in the context of SmartAttica EDIH (
 European Digital Innovation Hub)\, the 12th Module of Τraining modules fo
 r SMEs  with the subject "Inference with Transformers & Semantic Search wi
 th Sentence Transformers". \n\nDate: June 22nd\, 2026\n\nLocation: Onlin
 e via Zoom \n\nPresentation Languages: Greek\, English\n\nInstructors: Ni
 kos Bakas (GRNET)\, Roman Dolgopolyi (GRNET)\n\nDuration: 3 hours\n\nDescr
 iption: This is a hands-on introduction to running open Large Language Mod
 els and turning text into meaning with embeddings. Participants learn how 
 to load a pre-trained model from the HuggingFace Hub\, generate text with 
 streaming\, and control the tokenizer and chat template. The second half c
 overs Sentence Transformers\, showing how text is mapped into vectors and 
 how cosine similarity reveals semantic relationships between words and sen
 tences - the foundation of modern search and retrieval systems.\n\nTarget 
 Audience: This module is designed for SME developers\, technical leads\, a
 nd data scientists who want to incorporate natural language processing (NL
 P) into their projects. It is ideal for those looking to run models in the
 ir own environment and understand the building blocks behind semantic sear
 ch and RAG.\n\nLearning Objectives:\n\n By the end of this module\, parti
 cipants will be able to:\n\n\n	Load and run pre-trained language models fr
 om the HuggingFace Hub using the Transformers library.\n	Generate text wit
 h both streaming and standard inference\, and use the pipeline abstraction
 .\n	Apply chat templates and manage tokenizers\, padding\, and special tok
 ens correctly.\n	Produce embeddings from text using Sentence Transformers.
 \n	Measure semantic similarity with cosine similarity and interpret the re
 sulting vector space. \n\n\nPrerequisites:\n\nParticipants should have:\n
 \n\n	Basic understanding of Python programming.\n	Familiarity with running
  code in Jupyter/Colab notebooks.\n	Interest in NLP applications.\n	Some e
 xperience with machine learning will be helpful.\n\n\nIndicative Content:\
 n\n\n	The Transformers Library. Introduction to HuggingFace and the ecosys
 tem for loading and running open models.\n	Model and Tokenizer Setup. Down
 loading a model from the Hub\, configuring the tokenizer\, and handling pa
 d/EOS tokens.\n	Chat Templates and Messages. Structuring system and user m
 essages and applying the model's chat template.\n	Streaming Inference. Gen
 erating text token-by-token with a streamer for a responsive experience.\n
 	Standard Inference and the Pipeline. Running batch generation and using t
 he high-level pipeline object.\n	Introduction to Embeddings. What embeddin
 gs are and why similar meanings map to nearby vectors.\n	Sentence Transfor
 mers in Practice. Encoding words and sentences into vectors with a compact
 \, fast model.\n	Measuring Similarity. Using cosine similarity to compare 
 texts and visualize a similarity matrix.\n	Summary and Q&A. Key takeaways 
 and open discussion.\n\n\n \n\nThe project is co-funded by the European U
 nion. Views and opinions expressed are however those of the author(s) only
  and do not necessarily reflect those of the European Union or the Europea
 n Commission. Neither the European Union nor the granting authority can be
  held responsible for them. \n\nhttps://events.grnet.gr/event/209/
LOCATION:
URL:https://events.grnet.gr/event/209/
END:VEVENT
END:VCALENDAR