Home
Fine Tuning Open Source LLM

Sessions 1 day
Duration 7.5 hrs
Level Beginner
Assessment NA

Venue

12 Woodlands Square #07-85/86/87 Woods Square Tower 1, Singapore 737715. 5 mins walk from Woodlands (NS9) MRT station.

The venue is disabled-friendly.

Download Course Brochure

Certificate of Achievement from Tertiary Infotech Academy Pte Ltd - Upon meeting at least 75% attendance and passing the assessment(s), participants will receive a Certificate of Achievement from Tertiary Infotech Academy Pte Ltd.

Fine Tuning Open Source LLM

Course Code: C922

4 Review(s) | Add Your Review

Every prompt you send to a commercial AI model carries three costs: the per-token bill, the latency, and the fact that your data leaves your building. Fine tuning an open source LLM removes all three. Models like Llama, Mistral and Qwen are now strong enough that a well-tuned 7B or 8B model, trained on your own examples, will beat a general-purpose frontier model at your specific task — your tone, your terminology, your formats — while running on hardware you control.

In this hands-on 1-day course, you will take an open source model end to end: choosing the right base model for the job, building and cleaning a training dataset from your own documents and transcripts, and running efficient fine tunes with LoRA and QLoRA that train on a single GPU instead of a cluster. You will then evaluate the result honestly — measuring against the base model, watching for overfitting and catastrophic forgetting — and quantise and deploy the finished model behind an API you own.

You will leave with a fine-tuned model, a repeatable training pipeline and a clear-eyed view of when fine tuning is the right answer and when prompting or RAG would serve you better — the judgement that separates a working private model from an expensive experiment. Ideal for developers, data scientists, ML engineers and technical leads evaluating open source AI for production use.

No funding is available for this course

For WSQ funding, please checkout the details at WSQ - Generative AI Model Development and Fine Tuning

Course Fee

$350.00 (GST-exclusive)

$381.50 (GST-inclusive)

Mode of Training

Physical Classroom Training
Synchronised Teaching using ZOOM
Corporate Training at Company Premise

Course Date * Course Date

Course Time * Course Time: No. of Participants

Sponsorship

Self-Sponsored
Employer-Sponsored

Additional Message

* Required Fields

Please bring your own laptop for hands-on training. If you don't have laptop, we can provide spare laptop for training use.

We provide free consultation related to the subject matter after the course.
Please email your queries to enquiry@tertiaryinfotech.com and we will forward your queries to the subject matter experts.

You can register your interest without upfront payment. There is no penalty for withdrawal of the course before the class commences.
We reserve the right to cancel or re-schedule the course due to unforeseen circumstances. If the course is cancelled, we will refund 100% for any paid amount.
Note the venue of the training is subject to changes due to availability of the classroom.

Course Details

Topic 1 Open Source LLM Landscape

Llama, Mistral, Qwen and the open model ecosystem
Model sizes, licences and hardware requirements
Fine tuning vs prompting vs RAG: choosing the right approach
Setting up your training environment

Topic 2 Preparing Your Training Data

Building a dataset from your own documents and transcripts
Instruction, chat and completion formats
Cleaning, deduplication and train/test splits
How much data you actually need

Topic 3 Fine Tuning with LoRA and QLoRA

Full fine tuning vs parameter-efficient methods
LoRA and QLoRA explained
Running a training job and reading the loss curves
Hyperparameters, overfitting and catastrophic forgetting

Topic 4 Evaluation and Deployment

Evaluating against the base model
Quantisation for smaller, faster inference
Serving your model behind an API
Cost, privacy and maintenance in production

Course Info

Promotion Code

Your will get 10% discount voucher for 2nd course onwards if you write us a Google review.

Minimum Entry Requirement

Knowledge and Skills

Able to operate using computer functions
Minimum 3 GCE ‘O’ Levels Passes including English or WPL Level 5 (Average of Reading, Listening, Speaking & Writing Scores)

Attitude

Positive Learning Attitude
Enthusiastic Learner

Experience

Minimum of 1 year of working experience.

Target Age Group: 18-65 years old

Minimum Software/Hardware Requirement

Software:

Hardware: Window or Mac Laptops

Job Roles

DevOps Engineer
Cloud Solutions Architect
Software Developer
Infrastructure Engineer
Site Reliability Engineer (SRE)
Systems Administrator
IT Operations Specialist
Platform Engineer
Microservices Developer
Release Manager
Network Administrator (focusing on cloud-native applications)
Continuous Deployment Specialist
Backend Developer
Cloud Security Specialist
Container Orchestration Manager

Trainers

Truman Ng is a ACTA certified trainer that graduated with Bachelor Degree in Electrical Engineering from NUS in year 2002. He designed Artificial Intelligence (AI) controller for DC-DC Power Convertor by using Fuzzy Logic and Neural Network (NN) as his university Final Year Project.
Truman has over 15 years project experiences across Database & Web Design, PLC machinery, Data Center Design , Structure Cabling System(SCS) and Enterprise Network Design and Implementation. He used to be a network architect for Hewlett Packard, working with a group of virtual team from the US in handling network design and projects in the States.
Truman is the founder of Nexplore (S) Pte Ltd. He provides solutions of Cloud SaaS, IaaS & PaaS and Software Defined Network (SDN), VoIP and Internet Security. He was engaged by Huawei Global Training Center to provide 60+ consultations and trainings for Internet Service Providers(ISP) from Malaysia, Singapore, Brunei, Philippines, Australia, Poland, Iran, South Africa, Swaziland, Cote Dlvoire, Syria, Uzbekistan, New Zealand and countries over the world.As achievement, Truman has successfully completed 100+ IT network projects for Bank, Hotel and Factory within 5 years.Truman is certified in PMP, Cisco CCNP, CCIP, CCDP, HP Ase and Huawei HCNP, HCIE R&S, HCNA Cloud, HCNA Security, etc.

Review

Customer Reviews (4)

will recommend Review by Course Participant/Trainee

1. Do you find the course meet your expectation?
2. Do you find the trainer knowledgeable in this subject?
3. How do you find the training environment

. (Posted on 12/30/2021)

might consider Review by Course Participant/Trainee

1. Do you find the course meet your expectation?
2. Do you find the trainer knowledgeable in this subject?
3. How do you find the training environment

. (Posted on 12/30/2021)

Clear explanation Review by Course Participant/Trainee

1. Do you find the course meet your expectation?
2. Do you find the trainer knowledgeable in this subject?
3. How do you find the training environment

Clear explanation (Posted on 12/30/2021)

Might Consider Review by Course Participant/Trainee

1. Do you find the course meet your expectation?
2. Do you find the trainer knowledgeable in this subject?
3. How do you find the training environment

Would be good to signup for Github, docker hub and aws account before we come for the course to avoid obstacle on signup. Forward the required notes for our reading before the course started.

Will be good to go through some Ubuntu basic command needed for the class. Simply typing some of the commands without knowing what is the command is doing. Also can include comment beside each command for reference. Still very blurred after the course. Course schedule too tight from 10 to 5pm. WIll be good to start early maybe before 9am. (Posted on 11/13/2019)