Search Results for "training" - Page 6
Sort By:
Showing 881 open source projects for "training"
View related business solutions-
Gemini 3 and 200+ AI Models on One PlatformBuild generative AI apps with Vertex AI. Switch between models without switching platforms.
-
Try Google Cloud Risk-Free With $300 in CreditUse your credit across every product. Compute, storage, AI, analytics. When it runs out, 20+ products stay free. You only pay when you choose to.
-
1
DI-engine
OpenDILab Decision AI Engine
DI-engine is a unified reinforcement learning (RL) platform for reproducible and scalable RL research. It offers modular pipelines for various RL algorithms, with an emphasis on production-level training and evaluation. -
2
CLIP
CLIP, Predict the most relevant text snippet given an image
...It was trained on large sets of (image, caption) pairs using a contrastive objective: images and their matching text are pulled together in embedding space, while mismatches are pushed apart. Once trained, you can give it any text labels and ask it to pick which label best matches a given image—even without explicit training for that classification task. The repository provides code for model architecture, preprocessing transforms, evaluation pipelines, and example inference scripts. Because it generalizes to arbitrary labels via text prompts, CLIP is a powerful tool for tasks that involve interpreting images in terms of descriptive language. -
3
LLM Action
Technical principles related to large models
LLM-Action is a knowledge/tutorial/repository that shares principles, techniques, and real-world experience related to large language models (LLMs), focusing on LLM engineering, deployment, optimization, inference, compression, and tooling. It organizes content in domains like training, inference, compression, alignment, evaluation, pipelines, and applications. Sections covering infrastructure, engineering, and deployment. Repository templates, sample code, and resource links. Articles/code on LLM compression (quantization, pruning). -
4
OpenRLHF
An Easy-to-use, Scalable and High-performance RLHF Framework
OpenRLHF is an easy-to-use, scalable, and high-performance framework for Reinforcement Learning with Human Feedback (RLHF). It supports various training techniques and model architectures. -
Forever Free Full-Stack Observability | Grafana CloudBuilt on open standards like Prometheus and OpenTelemetry, Grafana Cloud includes Kubernetes Monitoring, Application Observability, Incident Response, plus the AI-powered Grafana Assistant. Get started with our generous free tier today.
-
5
Transformer Engine
A library for accelerating Transformer models on NVIDIA GPUs
...As the number of parameters in Transformer models continues to grow, training and inference for architectures such as BERT, GPT, and T5 become very memory and compute-intensive. Most deep learning frameworks train with FP32 by default. This is not essential, however, to achieve full accuracy for many deep learning models. -
6
Step1X-3D
High-Fidelity and Controllable Generation of Textured 3D Assets
...To achieve this, the project includes a massive curated dataset: among more than 5 million candidate 3D assets, it filters and standardizes to produce a high-quality 2 million–asset subset suitable for training. -
7
TextWorld
TextWorld is a sandbox learning environment for the training
TextWorld is a learning environment designed to train reinforcement learning agents to play text-based games, where actions and observations are entirely in natural language. Developed by Microsoft Research, TextWorld focuses on language understanding, planning, and interaction in complex, narrative-driven environments. It generates games procedurally, enabling scalable testing of agents’ natural language processing and decision-making abilities. -
8
Whisper
Robust Speech Recognition via Large-Scale Weak Supervision
...These tasks are jointly represented as a sequence of tokens to be predicted by the decoder, allowing a single model to replace many stages of a traditional speech-processing pipeline. The multitask training format uses a set of special tokens that serve as task specifiers or classification targets. -
9
Ultravox
Fast multimodal LLM for real-time voice interaction and AI apps
...Ultravox works by combining text prompts with encoded audio inputs, allowing it to understand spoken language alongside written instructions in a unified pipeline. Internally, it leverages pretrained language models and speech encoders, with a multimodal adapter that integrates both modalities for inference and training. Ultravox is optimized for low latency, achieving fast response times suitable for interactive voice agents and real-time applications. It supports use cases such as conversational AI agents, speech-to-speech translation, and analysis of spoken audio content. Ultravox also includes tooling and configuration systems for training, evaluation, and dataset integration. -
Custom VMs From 1 to 96 vCPUs With 99.95% UptimeLive migration and automatic failover keep workloads online through maintenance. One free e2-micro VM every month.
-
10
deepjazz
Deep learning driven jazz generation using Keras & Theano
...The project was originally created during a hackathon and was designed to show how neural networks can emulate creative tasks traditionally associated with human musicians. The repository includes preprocessing scripts for preparing MIDI data, training scripts for building the neural network model, and code for generating new compositions. -
11
seq2seq-couplet
Play couplet with seq2seq model
...The repository includes the code needed to train the model, configure file paths and hyperparameters, and evaluate progress through loss and BLEU score tracking. It also supports serving the trained model through a web service, allowing users to interact with the system after training is complete. In addition to local execution, the project includes Docker files, which make it easier to package and deploy the application in a more reproducible way. The repository also points users to an external dataset source and documents vocabulary formatting requirements for custom datasets, showing that it is meant for both experimentation and extension. -
12
verl-agent
Designed for training LLM/VLM agents via RL
verl-agent is an open-source reinforcement learning framework designed to train large language model agents and vision-language model agents for complex interactive environments. Built as an extension of the veRL reinforcement learning infrastructure, the project focuses on enabling scalable training for agents that perform multi-step reasoning and decision-making tasks. The framework supports multi-turn interactions between agents and their environments, allowing the system to receive feedback after each step and adjust its strategy accordingly. This step-wise interaction model makes it possible to train agents to operate in long-horizon scenarios where decisions depend on cumulative context and previous outcomes. ... -
13
PKU Beaver
Constrained Value Alignment via Safe Reinforcement Learning
...By incorporating constraint-based optimization methods, Safe-RLHF trains models that balance reward objectives with safety requirements, ensuring that harmful outputs are penalized during training. -
14
PRIME
Scalable RL solution for advanced reasoning of language models
...The system introduces the concept of process reinforcement through implicit rewards, allowing models to receive feedback on intermediate reasoning steps instead of evaluating only the final answer. This approach helps models learn better reasoning strategies and encourages them to generate more reliable multi-step solutions to complex tasks. PRIME provides training pipelines, datasets, and experimental infrastructure that allow researchers to train models with reinforcement learning tailored for reasoning improvement. The framework also includes data preprocessing utilities and example datasets such as mathematical reasoning tasks that are well suited for process-based reward signals. -
15
LongWriter
Unleashing 10,000+ Word Generation from Long Context LLMs
LongWriter is an open-source framework and set of large language models designed to enable ultra-long text generation that can exceed 10,000 words while maintaining coherence and structure. Traditional large language models can process large inputs but often struggle to generate long outputs due to limitations in training data and alignment strategies. LongWriter addresses this challenge by introducing a specialized dataset and training approach that encourages models to produce longer responses. The system uses an agent-based pipeline called AgentWrite that decomposes large writing tasks into smaller subtasks, allowing the model to produce long documents section by section. ... -
16
MatMul-Free LM
Implementation for MatMul-free LM
...Since matrix multiplication is one of the most computationally expensive components of modern language models, the project explores alternative computational strategies that reduce hardware requirements while maintaining comparable performance. The architecture relies on quantization-aware training and lightweight operations to replace conventional dense matrix multiplications with more efficient alternatives. These optimizations can significantly reduce memory consumption and potentially improve computational efficiency during both training and inference. The repository provides implementations of models at several parameter scales and includes tools for experimenting with the architecture using modern machine learning frameworks. -
17
Agent Lightning
The absolute trainer to light up AI agents
...It’s designed to be compatible with a wide range of agent architectures and frameworks — from LangChain and OpenAI Agent SDKs to AutoGen and custom Python agents — making it broadly applicable across different agent tooling ecosystems. Agent-Lightning introduces a lightweight training pipeline that observes agents’ execution traces, converts them into structured data, and feeds them into training algorithms, enabling users to improve agent behaviors systematically. The project emphasizes minimalist integration, so you can drop this into existing systems without extensive rewrites, focusing instead on iterative performance improvement. -
18
Lingvo
Framework for building neural networks
Lingvo is a TensorFlow based framework focused on building and training sequence models, especially for language and speech tasks. It was originally developed for internal research and later open sourced to support reproducible experiments and shared model implementations. The framework provides a structured way to define models, input pipelines, and training configurations using a common interface for layers, which encourages reuse across different tasks. -
19
Flax
Flax is a neural network library for JAX
...Modules define parameterized computations, but initialization and application remain side-effect free, which pairs naturally with JAX’s staging and compilation model. Flax emphasizes composability: optimizers, training loops, and checkpointing are provided as examples or utilities rather than monolithic frameworks, encouraging research-friendly customization. The library is widely used in vision, language, and reinforcement learning, often serving as a thin layer atop NumPy-like JAX primitives. Tutorials and examples show patterns for multi-host training, mixed precision, and advanced input pipelines that scale from laptops to TPUs. -
20
MuJoCo Playground
An open source library for GPU-accelerated robot learning
...The project includes classic control benchmarks from dm_control, advanced quadruped and bipedal locomotion systems, and dexterous as well as non-prehensile manipulation setups. It also offers optional vision-based training capabilities through integration with Madrona-MJX, allowing researchers to train policies directly from image input on GPUs. MuJoCo Playground supports both the MJX JAX implementation and the Warp physics engine, enabling flexible use across research pipelines. The environments are designed for fast training, compatibility with reinforcement learning libraries, and real-time trajectory visualization using rscope. -
21
4M
4M: Massively Multimodal Masked Modeling
4M is a training framework for “any-to-any” vision foundation models that uses tokenization and masking to scale across many modalities and tasks. The same model family can classify, segment, detect, caption, and even generate images, with a single interface for both discriminative and generative use. The repository releases code and models for multiple variants (e.g., 4M-7 and 4M-21), emphasizing transfer to unseen tasks and modalities. -
22
DLRM
An implementation of a deep learning recommendation model (DLRM)
...The implementation is optimized for performance at scale, supporting multi-GPU and multi-node execution, quantization, embedding partitioning, and pipelined I/O to feed huge embeddings efficiently. It includes data loaders for standard benchmarks (like Criteo), training scripts, evaluation tools, and capabilities like mixed precision, gradient compression, and memory fusion to maximize throughput. -
23
MoCo (Momentum Contrast)
Self-supervised visual learning using momentum contrast in PyTorch
...The core idea of MoCo is to maintain a dynamic dictionary with a momentum-updated encoder, allowing efficient contrastive learning across large batches. The repository includes implementations for both MoCo v1 and MoCo v2, the latter improving training stability and performance through architectural and augmentation enhancements. Training is optimized for distributed multi-GPU environments, using DistributedDataParallel for speed and simplicity. -
24
Mistral Finetune
Memory-efficient and performant finetuning of Mistral's models
mistral-finetune is an official lightweight codebase designed for memory-efficient and performant finetuning of Mistral’s open models (e.g. 7B, instruct variants). It builds on techniques like LoRA (Low-Rank Adaptation) to allow customizing models without full parameter updates, which reduces GPU memory footprint and training cost. The repo includes utilities for data preprocessing (e.g. reformat_data.py), validation scripts, and example YAML configs for training variants like 7B base or instruct models. It supports function-calling style datasets (via "messages" keys) as well as plain text formats, with guidelines on formatting, tokenization, and vocabulary extension (e.g. extending vocab to 32768 for some models) before finetuning. ... -
25
NVIDIA Isaac GR00T
NVIDIA Isaac GR00T N1.5 is the world's first open foundation model
...It accepts multimodal inputs—such as language and images—and uses a diffusion transformer architecture built upon vision-language encoders, enabling adaptive robot behaviors across diverse environments. It is designed to be customizable via post-training with real or synthetic data. The vision-language model remains frozen during both pretraining and finetuning, preserving language understanding and improving generalization. Streamlined MLP connection between vision encoder and LLM with added layer normalization.
