Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Library for sequence-to-sequence numeric prediction, applicable to any tokenizable input, and allows pretraining and fine-tuning over multiple tasks.
| Date | Stars |
|---|---|
| 2026-07-31 | 350 |
| 2026-08-01 | 350 |
| 2026-08-06 | 350 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# RegressLM: Easy Sequence-to-Sequence Numeric Prediction [](https://github.com/google-deepmind/regress-lm/actions?query=branch%3Amain) [**Overview**](#overview) | [**Setup**](#setup) | [**Colabs**](https://github.com/google-deepmind/regress-lm/tree/main/colabs) | [**Usage**](#usage) | [**Extended Usage**](#extended_usage) **Core Contributors**: Xingyou Song, Yash Akhauri, Jiyoun Ha, Bryan Lewandowski ## Overview <a name="overview"></a> RegressLM is a library for sequence-to-sequence numeric prediction, applicable to tokenizable inputs (e.g. strings, images) and allows pretraining and fine-tuning over multiple tasks. <p align="center"> <img src="https://raw.githubusercontent.com/akhauriyash/figures_placeholder/refs/heads/main/teaser_rlm_compressed.gif" alt="RegressLM decoding a numerical performance metric from text." width="100%"/> <br> <em><b><a href="https://research.google/blog/simulating-large-systems-with-regression-language-models/">Example Application</a>: Directly predicting performance metrics from unstructured, textually represented system states from Google's massive compute clusters.</b></em> </p> ## Setup <a name="setup"></a> Get started by installing the core libraries ([dependencies](https://github.com/google-deepmind/regress-lm/blob/main/pyproject.toml)): ``` pip install -e . ``` To run e.g. T5Gemma variants and LoRA fine-tuning, install additional libraries: ``` pip install ".[extras]" ``` Installation should take less than a minute. **Supported Platforms:** Requires Python 3.10+. Linux (e.g. Ubuntu) strongly preferred for deep learning. ## Colabs Example Colabs for getting started and demonstrating flagship results: * **Synthetic Density Training:** <a href="https://colab.research.google.com/github/google-deepmind/regress-lm/blob/main/colabs/synthetic_density_demo.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open in Colab" style="vertical-align: middle"></a> * **Triton GPU Kernel Latency Prediction:** <a href="https://colab.research.google.com/github/google-deepmind/regress-lm/blob/main/colabs/triton_demo.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open in Colab" style="vertical-align: middle"></a> * **Kaggle Experiment Outcome Prediction:** <a href="https://colab.research.google.com/github/google-deepmind/regress-lm/blob/main/colabs/kaggle_demo.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open in Colab" style="vertical-align: middle"></a> ## Usage <a name="usage"></a> There are two main stages: **inference** and **pretraining** (optional but recommended). ### Inference The intended use-case is to import a RegressLM class, which can decode floating-point predictions from a given input, and also fine-tune against new data. ```python from regress_lm import core from regress_lm import rlm # Create RegressLM from scratch. Optionally, use `from_t5gemma_encoder`. reg_lm = rlm.RegressLM.from_scratch(max_input_len=2048) # Example (x,y) pairs, which can be fine-tuned against. examples = [core.Example(x='hello', y=0.3), core.Example(x='world', y=-0.3)] reg_lm.fine_tune(examples) # Query inputs. query1, query2 = core.ExampleInput(x='hi'), core.ExampleInput(x='bye') samples1, samples2 = reg_lm.sample([query1, query2], num_samples=128) ``` ### Pretraining To produce better initial checkpoints for transfer learning, we recommend the user pretrains over large amounts of their own training data. Example pseudocode with PyTorch: ```python from regress_lm.pytorch import model as model_lib from regress_lm.pytorch import training model = model_lib.PyTorchModelConfig(...).make_model() trainer = training.Trainer(model, optimizer_factory, train_dataset, ...) for batch in trainer.train_dl: train_metrics = trainer.run_train_step(batch) ``` ## Boosting Performance and Extended Applications
Excerpt of 6,106 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:c3f13036be28135f, desc:fine-tuning, desc:fine tuning