Guide us how to fine-tune LLM : PEFT, LoRA (2 days, no dev work req)
Budget: $30 – $250 USD
We seek an excellent developer with significant experience training LLMs using different methods.
You must have done this work before. Please ensure that your knowledge, skills, and experience are up-to-date to minimise the risk of this project going wrong or taking more time than it should.
* Scope *
1. Guide the CEO how to fine-tune LLMs
Work with us step-by-step on live video calls with screen sharing throughout the process. Educate us on what you've learned so far and guide on us what is good/not good based on our needs
We will try three different ways:
PEFT + LoRA
QUORA
RLHF
2. Help us understand how to prep our own datasets to enable fine-tuning
3. Provide training and video walkthroughs of any areas that we cannot cover on the live calls together (where requested)
* Out of Scope
1. Any development work - we have the solutions already
2. Documentations - unless you are being proactive in providing exceptional value
* Expected timeline: 1-2 days *
* Our Tech *
Very flat on purpose
Windows, Google Cloud, Google Suite, Python, Langchain, MongoDB
* Standard Critical Expectations For All Dev Roles*
1. Fully understand the requirements
2. Provide the solutions / work issues out as a team
3. Any code must meet our standards (DRY, adheres to best practices, well commented, modular and dynamic)
4. Create short walkthroughs of solutions and the code (if applicable)
5. Available for video calls and live working sessions if needed via shared screen
6. Futureproof as possible (i.e. you've thought about how the solution could break and either come up with creative solutions and/or clearly identified and prepared us for areas at risk with suggestions to fix).
* Out of scope *
1. Politics
2. Pointless admin work
3. Long project documentation
* The environment *
1. Value-based: No pointless work that will be thrown away
2. Fast Paced: We create value and deliver in hours, days and sometimes 1-2 weeks - Never months!
3. Enjoyable: No politics, no BS, no blame culture, no stress
4. Rewarding : Trust + support + responsibilities + mentorship + bonuses (if you exceed expectations).
What we can offer you
--------------------------
1. Opportunities to work with famous external clients (e.g. Disney, Ernst & Young, Credit Suisse)
You can pick contracts that work for you without politics, stress or drama. We manage all of that for you.
2. Multiple internal project opportunities
Become our long-term preferred supplier. We offer you projects that match your skill set and interest.
3. Great working environment
Be welcomed into our happy, supportive and helpful family environment (read our Upwork reviews).
4. Bonuses!
We value great people.
How we reward
--------------------
We do everything to help our people become the best and reward the best people to keep them happy.
We reward people based on 4 levels of performance:
STAR SUPPLIER
IF YOU: Deliver work on time, meet acceptance criteria & meet basic company expectations [e.g. attend a few calls, help others, be dependable & proactive]
WE REWARD: 5* Rating + Agreed full contract rate paid + Guaranteed bonus + Long-term partnership + recommendation to global clients + invitations to work with us on big Fortune 500 clients (you will be protected from politics by us)
GOOD SUPPLIER
IF YOU: Deliver your work as promised on time, meet acceptance criteria
WE REWARD: 5* Rating + Agreed full contract rate paid
ONE-OFF SUPPLIER
IF YOU: Deliver your work as promised, meet acceptance criteria (but work is late)
WE REWARD: Agreed full contract rate paid only
POOR SUPPLIER
IF YOU: Don't meet acceptance criteria or become unresponsive, unreliable or difficult.
WE REWARD: Lower rating and, or payment to be negotiated.
Explain the kind of work you: 1) enjoy doing, 2) best at (you might be good at something but not enjoy it), 3) want to learn. Also, explain what type of work environment do you enjoy the most and why? (We 'try' to give people what they enjoy when we can)
Explain the kind of work you: 1) do not enjoy, 2) weakest in, 3) want to avoid if possible. Also - explain the kind of environment do you dislike and why. (We will try to ensure that we suit your needs. The happier you are, the better you'll perform)
What's your availability between 7am - 5 pm UK time, Mon-Sat? This is a flexible deadline-driven job. What are your restrictions (e.g. due to other clients, studies, responsibilities)? Pls do NOT apply if you have other jobs.
Score your capability and *interest 1-5 (1= not interested or not confident) for. A)Python, B) Langchain, C) Fine-tuning LLMs, D)Preparing training dataset, Score format: capability/interest. e.g A = 2/5, B=3/5
What is your experience with finetuning LLMs using: 1) PEFT, 2) LoRA, 3) QORA, 4) RHFL?
Budget: $200 for the project
You must have done this work before. Please ensure that your knowledge, skills, and experience are up-to-date to minimise the risk of this project going wrong or taking more time than it should.
* Scope *
1. Guide the CEO how to fine-tune LLMs
Work with us step-by-step on live video calls with screen sharing throughout the process. Educate us on what you've learned so far and guide on us what is good/not good based on our needs
We will try three different ways:
PEFT + LoRA
QUORA
RLHF
2. Help us understand how to prep our own datasets to enable fine-tuning
3. Provide training and video walkthroughs of any areas that we cannot cover on the live calls together (where requested)
* Out of Scope
1. Any development work - we have the solutions already
2. Documentations - unless you are being proactive in providing exceptional value
* Expected timeline: 1-2 days *
* Our Tech *
Very flat on purpose
Windows, Google Cloud, Google Suite, Python, Langchain, MongoDB
* Standard Critical Expectations For All Dev Roles*
1. Fully understand the requirements
2. Provide the solutions / work issues out as a team
3. Any code must meet our standards (DRY, adheres to best practices, well commented, modular and dynamic)
4. Create short walkthroughs of solutions and the code (if applicable)
5. Available for video calls and live working sessions if needed via shared screen
6. Futureproof as possible (i.e. you've thought about how the solution could break and either come up with creative solutions and/or clearly identified and prepared us for areas at risk with suggestions to fix).
* Out of scope *
1. Politics
2. Pointless admin work
3. Long project documentation
* The environment *
1. Value-based: No pointless work that will be thrown away
2. Fast Paced: We create value and deliver in hours, days and sometimes 1-2 weeks - Never months!
3. Enjoyable: No politics, no BS, no blame culture, no stress
4. Rewarding : Trust + support + responsibilities + mentorship + bonuses (if you exceed expectations).
What we can offer you
--------------------------
1. Opportunities to work with famous external clients (e.g. Disney, Ernst & Young, Credit Suisse)
You can pick contracts that work for you without politics, stress or drama. We manage all of that for you.
2. Multiple internal project opportunities
Become our long-term preferred supplier. We offer you projects that match your skill set and interest.
3. Great working environment
Be welcomed into our happy, supportive and helpful family environment (read our Upwork reviews).
4. Bonuses!
We value great people.
How we reward
--------------------
We do everything to help our people become the best and reward the best people to keep them happy.
We reward people based on 4 levels of performance:
STAR SUPPLIER
IF YOU: Deliver work on time, meet acceptance criteria & meet basic company expectations [e.g. attend a few calls, help others, be dependable & proactive]
WE REWARD: 5* Rating + Agreed full contract rate paid + Guaranteed bonus + Long-term partnership + recommendation to global clients + invitations to work with us on big Fortune 500 clients (you will be protected from politics by us)
GOOD SUPPLIER
IF YOU: Deliver your work as promised on time, meet acceptance criteria
WE REWARD: 5* Rating + Agreed full contract rate paid
ONE-OFF SUPPLIER
IF YOU: Deliver your work as promised, meet acceptance criteria (but work is late)
WE REWARD: Agreed full contract rate paid only
POOR SUPPLIER
IF YOU: Don't meet acceptance criteria or become unresponsive, unreliable or difficult.
WE REWARD: Lower rating and, or payment to be negotiated.
Explain the kind of work you: 1) enjoy doing, 2) best at (you might be good at something but not enjoy it), 3) want to learn. Also, explain what type of work environment do you enjoy the most and why? (We 'try' to give people what they enjoy when we can)
Explain the kind of work you: 1) do not enjoy, 2) weakest in, 3) want to avoid if possible. Also - explain the kind of environment do you dislike and why. (We will try to ensure that we suit your needs. The happier you are, the better you'll perform)
What's your availability between 7am - 5 pm UK time, Mon-Sat? This is a flexible deadline-driven job. What are your restrictions (e.g. due to other clients, studies, responsibilities)? Pls do NOT apply if you have other jobs.
Score your capability and *interest 1-5 (1= not interested or not confident) for. A)Python, B) Langchain, C) Fine-tuning LLMs, D)Preparing training dataset, Score format: capability/interest. e.g A = 2/5, B=3/5
What is your experience with finetuning LLMs using: 1) PEFT, 2) LoRA, 3) QORA, 4) RHFL?
Budget: $200 for the project