Add support for LoRA adapters trained with Rank-Stabilized scaling #299
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
arnavgarg1
changed the title
tgaddair
reviewed
Mar 4, 2024
server/lorax_server/models/model.py
Outdated
| lora_a, lora_a_name = module_map[weight_name]["lora_A"] | ||
| lora_a = lora_a.to(base_device, self.dtype) | ||
|
|
||
| lora_b, lora_b_name = module_map[weight_name]["lora_B"] | ||
| lora_b = lora_b.to(base_device, self.dtype) | ||
|
|
||
| scale = adapter_config.lora_alpha / adapter_config.r | ||
| scale: float = get_scaling_factor( |
There was a problem hiding this comment.
can we remove : float here?
tgaddair
approved these changes
Mar 4, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In PEFT 0.9, support was added for adapters trained using a new flag called
use_rslora.When set to True, Rank-Stabilized LoRA sets the adapter scaling factor to
lora_alpha/math.sqrt(r), since it was proven to work better. Otherwise, it will use the original default value oflora_alpha/r.In equation form:
W0X + (lora_alpha/r)(BAX)where W0 is the base model, BA are the lora weight matrices and X is the input from the embedding layer/previous transformer layer.In particular, this is useful when using larger ranks since it prevents the gradient from collapsing as rank increases, which may result in higher ranks actually leading to better performance (not true by default today and in the original LoRA paper). Paper: https://arxiv.org/pdf/2312.03732.pdf.