A port of muP to JAX/Haiku

Overview

MUP for Haiku

This is a (very preliminary) port of Yang and Hu et al.'s μP repo to Haiku and JAX. It's not feature complete, and I'm very open to suggestions on improving the usability.

Installation

pip install haiku-mup

Learning rate demo

These plots show the evolution of the optimal learning rate for a 3-hidden-layer MLP on MNIST, trained for 10 epochs (5 trials per lr/width combination).

With standard parameterization, the learning rate optimum (w.r.t. training loss) continues changing as the width increases, but μP keeps it approximately fixed:

Here's the same kind of plot for 3 layer transformers on the Penn Treebank, this time showing Validation loss instead of training loss, scaling both the number of heads and the embedding dimension simultaneously:

Note that the optima have the same value for n_embd=80. That's because the other hyperparameters were tuned using an SP model with that width, so this shouldn't be biased in favor of μP.

Usage

from functools import partial

import jax
import jax.numpy as jnp
import haiku as hk
from optax import adam, chain

from haiku_mup import apply_mup, Mup, Readout

class MyModel(hk.Module):
    def __init__(self, width, n_classes=10):
        super().__init__(name='model')
        self.width = width
        self.n_classes = n_classes

    def __call__(self, x):
        x = hk.Linear(self.width)(x)
        x = jax.nn.relu(x)
        return Readout(2)(x) # 1. Replace output layer with Readout layer

def fn(x, width=100):
    with apply_mup(): # 2. Modify parameter creation with apply_mup()
        return MyModel(width)(x)

mup = Mup()

init_input = jnp.zeros(123)
base_model = hk.transform(partial(fn, width=1))

with mup.init_base(): # 3. Use this context manager when initializing the base model
    hk.init(fn, jax.random.PRNGKey(0), init_input) 

model = hk.transform(fn)

with mup.init_target(): # 4. Use this context manager when initializng the target model
    params = model.init(jax.random.PRNGKey(0), init_input)

model = mup.wrap_model(model) # 5. Modify your model with Mup

optimizer = optax.adam(3e-4)
optimizer = mup.wrap_optimizer(optimizer, adam=True) # 6. Use wrap_optimizer to get layer specific learning rates

# Now the model can be trained as normal

Summary

  1. Replace output layers with Readout layers
  2. Modify parameter creation with the apply_mup() context manager
  3. Initialize a base model inside a Mup.init_base() context
  4. Initialize the target model inside a Mup.init_target() context
  5. Wrap the model with Mup.wrap_model
  6. Wrap optimizer with Mup.wrap_optimizer

Shared Input/Output embeddings

If you want to use the input embedding matrix as the output layer's weight matrix make the following two replacements:

# old: embedding_layer = hk.Embed(*args, **kwargs)
# new:
embedding_layer = haiku_mup.SharedEmbed(*args, **kwargs)
input_embeds = embedding_layer(x)

#old: output = hk.Linear(n_classes)(x)
# new:
output = haiku_mup.SharedReadout()(embedding_layer.get_weights(), x) 
DeepSpamReview: Detection of Fake Reviews on Online Review Platforms using Deep Learning Architectures. Summer Internship project at CoreView Systems.

Detection of Fake Reviews on Online Review Platforms using Deep Learning Architectures Dataset: https://s3.amazonaws.com/fast-ai-nlp/yelp_review_polar

Ashish Salunkhe 37 Dec 17, 2022
null

DeformingThings4D dataset Video | Paper DeformingThings4D is an synthetic dataset containing 1,972 animation sequences spanning 31 categories of human

208 Jan 03, 2023
Rotation Robust Descriptors

RoRD Rotation-Robust Descriptors and Orthographic Views for Local Feature Matching Project Page | Paper link Evaluation and Datasets MMA : Training on

Udit Singh Parihar 25 Nov 15, 2022
A python library for self-supervised learning on images.

Lightly is a computer vision framework for self-supervised learning. We, at Lightly, are passionate engineers who want to make deep learning more effi

Lightly 2k Jan 08, 2023
YOLOX-RMPOLY

本算法为适应robomaster比赛,而改动自矩形识别的yolox算法。 基于旷视科技YOLOX,实现对不规则四边形的目标检测 TODO 修改onnx推理模型 更改/添加标注: 1.yolox/models/yolox_polyhead.py: 1.1继承yolox/models/yolo_

3 Feb 25, 2022
Breaking the Dilemma of Medical Image-to-image Translation

Breaking the Dilemma of Medical Image-to-image Translation Supervised Pix2Pix and unsupervised Cycle-consistency are two modes that dominate the field

Kid Liet 86 Dec 21, 2022
The official PyTorch implementation of the paper: *Xili Dai, Xiaojun Yuan, Haigang Gong, Yi Ma. "Fully Convolutional Line Parsing." *.

F-Clip — Fully Convolutional Line Parsing This repository contains the official PyTorch implementation of the paper: *Xili Dai, Xiaojun Yuan, Haigang

Xili Dai 115 Dec 28, 2022
Fast and Simple Neural Vocoder, the Multiband RNNMS

Multiband RNN_MS Fast and Simple vocoder, Multiband RNN_MS. Demo Quick training How to Use System Details Results References Demo ToDO: Link super gre

tarepan 5 Jan 11, 2022
Rational Activation Functions - Replacing Padé Activation Units

Rational Activations - Learnable Rational Activation Functions First introduce as PAU in Padé Activation Units: End-to-end Learning of Activation Func

<a href=[email protected]"> 38 Nov 22, 2022
Generative Autoregressive, Normalized Flows, VAEs, Score-based models (GANVAS)

GANVAS-models This is an implementation of various generative models. It contains implementations of the following: Autoregressive Models: PixelCNN, G

MRSAIL (Mini Robotics, Software & AI Lab) 6 Nov 26, 2022
Implementation for Paper "Inverting Generative Adversarial Renderer for Face Reconstruction"

StyleGAR TODO: add arxiv link Implementation of Inverting Generative Adversarial Renderer for Face Reconstruction TODO: for test Currently, some model

155 Oct 27, 2022
This is the code of paper ``Contrastive Coding for Active Learning under Class Distribution Mismatch'' with python.

Contrastive Coding for Active Learning under Class Distribution Mismatch Official PyTorch implementation of ["Contrastive Coding for Active Learning u

21 Dec 22, 2022
Application of K-means algorithm on a music dataset after a dimensionality reduction with PCA

PCA for dimensionality reduction combined with Kmeans Goal The Goal of this notebook is to apply a dimensionality reduction on a big dataset in order

Arturo Ghinassi 0 Sep 17, 2022
"NAS-Bench-301 and the Case for Surrogate Benchmarks for Neural Architecture Search".

NAS-Bench-301 This repository containts code for the paper: "NAS-Bench-301 and the Case for Surrogate Benchmarks for Neural Architecture Search". The

AutoML-Freiburg-Hannover 57 Nov 30, 2022
Single-stage Keypoint-based Category-level Object Pose Estimation from an RGB Image

CenterPose Overview This repository is the official implementation of the paper "Single-stage Keypoint-based Category-level Object Pose Estimation fro

NVIDIA Research Projects 188 Dec 27, 2022
The implementation of "Bootstrapping Semantic Segmentation with Regional Contrast".

ReCo - Regional Contrast This repository contains the source code of ReCo and baselines from the paper, Bootstrapping Semantic Segmentation with Regio

Shikun Liu 128 Dec 30, 2022
Multi-Target Adversarial Frameworks for Domain Adaptation in Semantic Segmentation

Multi-Target Adversarial Frameworks for Domain Adaptation in Semantic Segmentation Paper Multi-Target Adversarial Frameworks for Domain Adaptation in

Valeo.ai 20 Jun 21, 2022
Image Restoration Toolbox (PyTorch). Training and testing codes for DPIR, USRNet, DnCNN, FFDNet, SRMD, DPSR, BSRGAN, SwinIR

Image Restoration Toolbox (PyTorch). Training and testing codes for DPIR, USRNet, DnCNN, FFDNet, SRMD, DPSR, BSRGAN, SwinIR

Kai Zhang 2k Dec 31, 2022
Code accompanying the paper "How Tight Can PAC-Bayes be in the Small Data Regime?"

How Tight Can PAC-Bayes be in the Small Data Regime? This is the code to reproduce all experiments for the following paper: @inproceedings{Foong:2021:

5 Dec 21, 2021
Implementation based on Paper - Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling

Implementation based on Paper - Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling

HamasKhan 3 Jul 08, 2022