RLHF a LLM in <50 lines of Python
223 points by patelajay285 1 year ago | 66 comments- mk_stjames 1 year agoI feel the preparation and loading of the dataset has been abstracted too far away. I have no idea what type of data format I need or how it is loaded for this (it is using a pre-prepared huggingface dataset?). If I have local data how should it be loaded? What does that even look like? Is it expecting some sort of JSON?
When you get so far as to abstracting every step to loading a one-liner from huggingface, including the downloading of a prepared dataset with no example of doing the same on custom local dataset, you've abstracted too far to be useful for anyone other than the first user.
- patelajay285 1 year agoThanks for the question. This is built for ML researchers, so in examples we use the defacto source for datasets researchers often use, HF Hub.
However, there is a lot of documentation on the site to help guide users. This documentation page shows you can load in data via local datasets as well. For example, JSON, CSV, text files, a local HF Dataset folder, or even from a Python `dict` or `list`:
https://datadreamer.dev/docs/latest/datadreamer.steps.html#t...
We'll definitely keep improving documentation, guides, and examples. We have a lot of it already, and more to come! This has only recently become a public project :)
If anyone has any questions on using it, feel free to email me directly (email on the site and HN bio) for help in the meantime.
- mk_stjames 1 year agoI did glance at the docs first before commenting but I was looking in 'datasets' to try and understand importing a potential CSV/JOSN etc and all I saw was verbage on accessing the output.
I would not have guessed that the base input data processing would have been filed under 'steps'. But now I kinda see how you are working, but I admit I'm not the target audience.
If you want this to really take off for people outside of a very, very specific class of researchers... setup an example on your landing page that calls to a local JSON of user prompts/answers/rejects finetuning a llama model with your datadreamer.steps.JSONDataSource into the loader. Or, a txt file with the system/user/assistant prompts tagged and examples given. Yes, your 'lines of code' for your frontpage example may grow a bit!
Maybe there are a lot of 'ML researchers' that are used to the type of super-abstract OOP API, load-it-from-huggingface-scheme-people you are targeting but also know that there are a ton that aren't.
- patelajay285 1 year agoThat's totally fair and good feedback, it's hard to support everyone's use cases simultaneously, but from my own research and other researchers we collaborate with, this solves and streamlines the right set of problems, but we want to make this as broadly useful as possible. Always happy to chat more / provide support if you would like, feel free to reach out if you want to try it and run into any sharp edges I could help make easier.
- patelajay285 1 year ago
- mk_stjames 1 year ago
- 1 year ago
- ganeshkrishnan 1 year agoThe dataset is here I presume: https://huggingface.co/datasets/Intel/orca_dpo_pairs
you can look at the samples. Mostly its questions and accepted/rejected answers.
- patelajay285 1 year ago
- jerpint 1 year agoI don’t understand the obsession of LOC for wrappers - it’s the whole point of a wrapper. It makes it much easier for the user at the expense of making it less hackable
Title should be instead “Library for low-code RLHF in python”
- vvrm 1 year agoAnother problem with the title: the article is about DPO, which doesn’t do reinforcement learning. So not RLHF. I guess RLHF has more of a name recognition than DPO.
- janalsncm 1 year agoHonestly a much bigger problem than LOC. It’s a completely different algorithm.
- patelajay285 1 year agoThis was discussed in another comment, DPO is pretty much strictly better than RLHF + PPO, and far more stable when training. Yes, DPO is not technically "RL", but it's semantics for the most part. DataDreamer does support PPO training if you want, but it's so unstable, it's a less popular choice now.
- antonvs 1 year agoIn the DPO paper linked from the OP page, DPO is described as "a simple RL-free algorithm for training language models from preferences." So as you say, "not technically RL."
Given that, shouldn't the first sentence on the linked page end with "...in a process known as DPO (...)" ? Ditto for the title.
It sounds like you're saying that the terms RL and RLHF should subsume DPO because they both solve the same problem, with similar results. But they're different techniques, and there are established terms for both of them.
- vvrm 1 year ago> DPO is pretty much strictly better than RLHF + PPO
Out of genuine curiosity, do you have any pointers/evidence to support this. I know that some of the industry leading research labs haven't switched over to DPO yet, in spite of the fact that DPO is significantly faster than RLHF. It might just be organizational inertia, but I do not know. I would be very happy if simpler alternatives like DPO were as good as RLHF or better, but I haven't seen that proof yet.
- antonvs 1 year ago
- janalsncm 1 year ago
- behnamoh 1 year agoThis. If I'm the type of person who wants to do RLHF, then I'm the type of person who wants control and doesn't like delegating it to imported libraries.
- patelajay285 1 year agoThis is built for ML researchers out of an academic lab. There's a ton of functionality in the library (beyond RLHF and alignment) that ML researchers do every day to write papers and run experiments that the library helps abstract and make repeatable and usable.
Unless your research hypothesis is specifically around improving or changing RLHF, it's unlikely you should be implementing it from scratch. Abstractions are useful for a reason. The library is quite configurable to let you tune any knobs you would want.
- patelajay285 1 year ago
- patelajay285 1 year agoThis is developed for researchers, so I assure it’s very hackable and configurable. ;-) but appreciate the feedback on the title!
- tgsovlerkhgsel 1 year agoHonestly the amount of complicated boilerplate that you're supposed to write from scratch every time you do something with ML (in some of the major frameworks) deterred me from touching anything ML-related for a long time.
As far as I understand, what the training loop is supposed to be doing is pretty static and you don't need to understand most of it in order to "do ML", but at the same time it's full of complicated things to get right (which would be much easier to understand when controlled through well defined parameters instead of mixing boilerplate and config).
- brigadier132 1 year agoI always appreciate these projects because I just dive into the code itself and copy out what I need once the wrapper becomes too much of a burden.
- patelajay285 1 year agoThat’s totally valid and something we would even encourage! This project is for researchers so if there is a point where the abstraction is no longer useful, by all means configure, or subclass, or copy code.
- rovr138 1 year agoOf course. And they're not saying they don't have a place.
They're saying why does it matter if it's 50 vs 60 or even 100. It's a wrapper, which should be less lines. That's the whole point. Abstracting things even further and making assumptions.
Of course you can use them. Of course you can remove them after and use the underlying code. But the LOC shouldn't be the important part of it
- patelajay285 1 year ago
- verticalscaler 1 year agoYes you do. Most casuals are downright afraid of code. This messaging is meant to make the project more approachable.
Kind of like everybody knows the pop-science around e = mc^2 but most are completely oblivious that it takes a bunch of whiteboards to derive it and what all that actually means.
No pithy formula no way for the actual ideas to spread to the mainstream for you to somehow hear about it.
- antonvs 1 year agoThis reminds me of the advice Stephen Hawking's publisher gave him, which was that every equation he included in his book, A Brief History of Time, would cut the sales of the book in half. As a result the only equation that ended up in the book was E=mc^2.
- antonvs 1 year ago
- vvrm 1 year ago
- lopkeny12ko 1 year agoIt's not 50 lines of code if all the real work is done by importing a library...
That's like saying, I can solve any problem in 2 lines of code. I'll publish a library for it first, then:
import foo; foo.do_the_thing()
Magic!
- antonvs 1 year agoSoftware developers hate this one simple trick!
- peab 1 year agodid people say the same thing when assembly code got abstracted away?
- orblivion 1 year agoThere's levels of abstraction. "lines of Python" to me roughly means the standard library.
- sciolist 1 year agoAgreed. If it was "RLHF a LLM in <50 lines" I feel it would be clear, but "lines of Python" implies something different.
- sciolist 1 year ago
- skelpmargyar 1 year agoImporting a library is not abstraction any more than closing your eyes is abstracting the world to black.
- __loam 1 year agoThe point of libraries is abstraction. You get to use the library without writing the library or knowing the underlying implementation.
- __loam 1 year ago
- orblivion 1 year ago
- therealdrag0 1 year ago[relevant xkcd]
- antonvs 1 year ago
- patelajay285 1 year agoHi everyone, there are no easy tools for synthetic data generation or training and aligning LLMs simply in Python. Most of the stuff out there are messy adhoc scripts.
DataDreamer is an open source Python package with a nice API from the University of Pennsylvania that does all this that we’re actively developing. Will be here to answer questions.
- baggiponte 1 year agoThe API looks nice, congratulations. Will experiment with it. One small silly question: why did you choose to specify the dependencies inside the src dir with the requirements format - rather than inside the pyproject?
- patelajay285 1 year agoThanks! It makes it easier to run with the existing run scripts I have on our large university GPU cluster. :) no other reason
- patelajay285 1 year ago
- baggiponte 1 year ago
- g4zj 1 year agoVery cool, but I can't help but feel like titles that reference low-LOC are a bit clickbait-y when nearly all the heavy lifting is done by imported libraries.
- patelajay285 1 year agoAppreciate the feedback on the title, this is developed for ML researchers, so I assure there is a lot it’s doing under the hood to make this process easier (for example introducing automatic caching and resumability).
However, we also tried to simplify the API and have sensible defaults to make it usable for anyone / make ML research code cleaner :)
- patelajay285 1 year ago
- imjonse 1 year agoThe first paragraphs says RLHF can be used to align models, and the seconds say here's how to do it by using DPO. These two methods are not the same, and the latter is not an instance of the former.
- patelajay285 1 year agoFair, DPO is considered a fairly well established technique now that is far more stable in training than PPO, but also helps align LLMs from human feedback. The package also helps do PPO, so you can do traditional RLHF, but figured more people would be interested in seeing a DPO example, given how unstable PPO is.
- Der_Einzige 1 year agoThe latter is strictly superior to the former though. RlHF has been abandoned in the open source world.
- numeri 1 year agoI don't know about strictly superior. It's certainly strictly easier for people with a budget, who just need "good enough" results the first try. I don't have any evidence whatsoever, but I'd expect that enough tuning and retries can get squeeze a bit more performance out of RLHF than you can get out of DPO.
- patelajay285 1 year agoYep, DPO is not technically “RL” and implicitly uses the LLM itself as a reward model, but training with DPO is far more stable for that reason.
- espadrine 1 year agoDPO is as close to RL as RLHF. The latter also uses the LLM as a reward model.
I'm not a fan of the RL/SL dichotomy, because the line gets so foggy. If you squint, every loss is a negative reward, and every policy improvement a supervised target.
Still, what the code does isn't what is described in the paper that the page links to.
- patelajay285 1 year agoI tend to agree @espadrine, it's semantics for the most part
- espadrine 1 year ago
- imjonse 1 year agoI am just saying the intro paragraphs are confusing.
- patelajay285 1 year agoThanks, appreciate the feedback, will update when I get a chance!
- patelajay285 1 year ago
- numeri 1 year ago
- patelajay285 1 year ago
- proto-n 1 year agoYeah well in bash I can do it in one line: `python train.py`. I hate examples like this, the 50loc statement is totally useless (and so is the code example, as I can't learning anything from it).
- MrYellowP 1 year agoI don't prefer aligned models and I'm a human. It's not okay to claim that that's what humans prefer. There might be a subset of humans who can't handle words, but they're not even remotely in the majority.
Algined models are dumber, treat everyone like they're stupid immature idiots who can't handle words and they're a wannabe moral authority.
- theptip 1 year agoInterested if local RLHF is actually viable; can you get meaningful steering from 1k feedback points on a narrow task? I feel that annotation count is achievable with a single dedicated annotator making a few comments per minute (though tedious), 10k would be a week of work so achievable for a very dedicated hobbyist, and 100k seems out of reach for a hobby project.
Say for simple conversation usecases (eg customer support for a specific product, interactive fiction, things like that without deep technical knowledge).
I was also wondering if it’s possible to do such RLHF for SD running locally.
- aethelyon 1 year agoThis is cool, but the data collection is the hard part, right?
- patelajay285 1 year agoYes it is :), but the library is also a synthetic data generation library, so for example you can create the data for DPO fully synthetically, check out the self-rewarding LLMs example:
https://datadreamer.dev/docs/latest/pages/get_started/quick_...
- sillysaurusx 1 year agoI’m extremely skeptical of this approach. Until proven otherwise, with a model that users actually find useful, I don’t think this can work.
It would be nice. But I’ve seen too many nice ideas completely fall apart in practice to accept this without some justification. Even if there are papers on the topic, and those papers show that the models rank highly according to some eval metrics, the only metric that truly matters is "the user likes the model and it solves their problems."
By the way, on a separate topic, the 90/10 dataset split that you do in all of your examples turns out to be fraught with peril in practice. The issue is that the validation dataset quality turns out to be crucial, and randomly yeeting 10% of your data into the validation dataset without manual review is a recipe for problems.
- patelajay285 1 year agoIt's a demo snippet of how to setup the workflow, it's not meant to be a working production example a self-rewarding model or a faithful reproduction of the original paper. Whether self-rewarding LLMs are a good idea or not, it's a valuable and very active area of research in the literature today. This is a library for ML researchers who should actively research and study these avenues along with the pitfalls you're mentioning. But in order for them to do that, building these workflows have to be accessible to them, which is what this library is meant to do. It's not meant for the "hobbyist" ML-community, they should not be using synthetic data today in this way, it would likely lead to subpar results for any practical model or task.
- patelajay285 1 year ago
- sillysaurusx 1 year ago
- patelajay285 1 year ago
- bbstats 1 year agoI can abstract this to 2 lines
- v4dok 1 year agoI feel like the current meta on finetuning LLMs is random accounts at X/Twitter. Google results are littered with SEO garbage or some kind of guides that fail to work the moment you need something slightly different.
- rldjbpin 1 year agoit is very conflicting to see "do x in y LOC" in this field, especially when most of the workflow for different models are fragmented across non-overlapping frameworks/tooling.
to actually do something from scratch or using the author's code requires adopting something esoteric just for this purpose. for these scenarios it is nice to appreciate hf and their abstraction. but the reinventing the wheel situation is very frustrating to work with.
if you want to go beyond the demo, you have to deal with this painful reality. i hope there is more progress on this rather than making stacks of api.
- spdustin 1 year agoIt occurs to me that there must be a model that's been "aligned" opposite to the usual RLHF. Or has nobody done that?
- ilaksh 1 year agoHow do you normally do DPO? Is that built in to PyTorch or something?
Theoretically the hard part is collecting the examples with rejections etc.
- patelajay285 1 year agoCollecting data is hard, but the library is also a synthetic data generation library, so for example you can create the data for DPO fully synthetically, check out the self-rewarding LLMs example: https://datadreamer.dev/docs/latest/pages/get_started/quick_...
- patelajay285 1 year ago
- potatoman22 1 year agoThis seems useful, thanks!
- rrr_oh_man 1 year agoRLHF = Reinforcement Learning from Human Feedback
- cztomsik 1 year agoDPO is not RLHF.