Set up a node! Bare boards that work with the app are like $50 and take a few clicks to flash and setup. The basic antenna with no amp makes contacts up to 50mi away if the conditions are right. I have one in a window and one in a backpack at all times.
By the way, some time ago when I checked there were two cool applications of LoRa: (1) a mesh, for (hopefully) truly decentralized and more difficult to disrupt communication, (2) a gateway, so that you could get data from your sensors in remote places via standard internet protocols.
Both are very cool, but I wonder if I missed something else?
> The paper describes the finding that LoRA and full fine-tuning, with equal performance on the fine-tuning task, can have solutions with very different generalization behaviors outside the fine-tuning task distribution. We found that LoRA and full fine-tuning yield models with significant differences spectral properties of their weight matrices: LoRA models often containing “intruder dimensions”, high-ranking singular vectors approximately orthogonal to the singular vectors of pre-trained weight matrices. The existence of intruder dimensions correlates with the fine-tuned model forgetting more of the pre-training distribution as well as forgetting more when trained on tasks sequentially in a continual learning setup.
I'm surprised they didn't cite this; it's a well known paper.
Efficient Estimation of Word Representations in Vector Space[1], one of the most influential papers in the space with tens of thousands of citations[2]? Or the RoBERTa[3] paper (dramatically improved upon BERT; RoBERTa and derived models currently have tens of millions of downloads on HF and still serve as a reliable industry workhorse)? Or the Mamba paper[4] (pretty much the only alternative to transformers that actually gets used)? Do you want me to keep going?
Honestly, I find that whether a paper gets rejected or not means diddly squat considering how broken the review system is, and through how much honestly terrible papers I have to wade through every time I'm looking through the conference submissions for anything good.
Standard LoRA (W_delta = B@A with standard inits) generally underperforms FT, primarily because of "intruder dimensions" (new high-ranking singular vectors which misalign with the singular vectors of the underlying weights) as outlined in the paper.
There are techniques like PiCa and SVFT which can mitigate much of the loss, though.
> LoRA works well when not capacity constrained, i.e., the number of trainable parameters exceeds the amount of information to be learned, which can be estimated in terms of dataset size
I’m shocked they didn’t look at progressive merging of LoRAs. Research shows that’s the best way of improving its ability to model higher level features.
Seems like a massive miss, not to mention there is other research that contradicts a lot of their findings. This feels a bit like a researchers first pass at learning LoRA
I'm not sure why progressive LoRa merging needs to be addressed here. They show there is a regime of problem where LoRa performs equivalently to FFT.
Progressive merging of LoRa is somewhere inbetween and categorically more complex than just LoRa so would be dominated by standard LoRa in that case.
While progressive merging could train faster as fewer params are trainable at any given time, it results in very larger adapter diffs OTO the size of the original model and doesn't retain the benefits of being able to deploy multiple adapters over the same base model idt.
Question for dudes building modern nn's... what's the thinking on estimating structural capacity for real world problem? How should I estimate how many parameters to choose for the model?
Can someone explain the bit counting argument in the reinforcement learning part?
I don’t get why a trajectory would provide only one bit of information.
Each step of the trajectory is at least giving information about what state transitions are possible.
An infinitely long trajectory can explore the whole state space if there are no absorbing states. Such a trajectory would provide a massive amount of information about the system, even if we ignored the final reward.
I believe it's because the way you measure things in RL, each episode only tells you whether it was good (say reward +1) or bad (say 0 or negative reward), it does not tell you anything about the trace that was produced to get the outcome. This reward is the only thing measured to produce your gradients. Hence why the amount of info in it is O(1).
This is in contrast to more "supervised" forms of learning where you could get a loss for each token produced (e.g. cross entropy loss), and where you'd get, as a consequence O(number of tokens) information into your gradients.
A fair amount of research has shown that RL doesn’t add knowledge to the base model it just optimizes paths that already exist.
Now ProRL from Nvidia showed there are ways of adding knowledge, mostly through progressive merging.
I’m still not fully convinced of the 1bit claim, they made other mistakes in the blog post
Readit News
Both are very cool, but I wonder if I missed something else?
Deleted Comment
Dead Comment
I think the literature is clear on that?
"LoRA vs Full Fine-tuning: An Illusion of Equivalence" -- https://arxiv.org/abs/2410.21228v1
Quoting from the conclusions:
> The paper describes the finding that LoRA and full fine-tuning, with equal performance on the fine-tuning task, can have solutions with very different generalization behaviors outside the fine-tuning task distribution. We found that LoRA and full fine-tuning yield models with significant differences spectral properties of their weight matrices: LoRA models often containing “intruder dimensions”, high-ranking singular vectors approximately orthogonal to the singular vectors of pre-trained weight matrices. The existence of intruder dimensions correlates with the fine-tuned model forgetting more of the pre-training distribution as well as forgetting more when trained on tasks sequentially in a continual learning setup.
I'm surprised they didn't cite this; it's a well known paper.
Oh, you mean rejected just like these papers?
Efficient Estimation of Word Representations in Vector Space[1], one of the most influential papers in the space with tens of thousands of citations[2]? Or the RoBERTa[3] paper (dramatically improved upon BERT; RoBERTa and derived models currently have tens of millions of downloads on HF and still serve as a reliable industry workhorse)? Or the Mamba paper[4] (pretty much the only alternative to transformers that actually gets used)? Do you want me to keep going?
Honestly, I find that whether a paper gets rejected or not means diddly squat considering how broken the review system is, and through how much honestly terrible papers I have to wade through every time I'm looking through the conference submissions for anything good.
[1] -- https://openreview.net/forum?id=idpCdOWtqXd60
[2] -- https://scholar.google.com/scholar?cites=7447715766504981253
[3] -- https://openreview.net/forum?id=SyxS0T4tvS
[4] -- https://openreview.net/forum?id=AL1fq05o7H
I'm surprised you copied and pasted all of that without explaining what it means.
Does LoRA perform worse, better or statistically insignificantly different to FullFT?
You aren't able to tell from what you pasted, are you?
There are techniques like PiCa and SVFT which can mitigate much of the loss, though.
I’m shocked they didn’t look at progressive merging of LoRAs. Research shows that’s the best way of improving its ability to model higher level features.
Seems like a massive miss, not to mention there is other research that contradicts a lot of their findings. This feels a bit like a researchers first pass at learning LoRA
Progressive merging of LoRa is somewhere inbetween and categorically more complex than just LoRa so would be dominated by standard LoRa in that case.
While progressive merging could train faster as fewer params are trainable at any given time, it results in very larger adapter diffs OTO the size of the original model and doesn't retain the benefits of being able to deploy multiple adapters over the same base model idt.
https://arxiv.org/abs/2410.22911
https://arxiv.org/abs/2409.16167
I don’t get why a trajectory would provide only one bit of information.
Each step of the trajectory is at least giving information about what state transitions are possible.
An infinitely long trajectory can explore the whole state space if there are no absorbing states. Such a trajectory would provide a massive amount of information about the system, even if we ignored the final reward.
This is in contrast to more "supervised" forms of learning where you could get a loss for each token produced (e.g. cross entropy loss), and where you'd get, as a consequence O(number of tokens) information into your gradients.
Deleted Comment
I’m still not fully convinced of the 1bit claim, they made other mistakes in the blog post