Search Results for "stage"
Sort By:
Showing 16 open source projects for "stage"
View related business solutions-
MongoDB Atlas runs apps anywhereMongoDB Atlas gives you the freedom to build and run modern applications anywhere—across AWS, Azure, and Google Cloud. With global availability in over 115 regions, Atlas lets you deploy close to your users, meet compliance needs, and scale with confidence across any geography.
-
Cut Data Warehouse Costs by 54%BigQuery delivers 54% lower TCO with exabyte scale and flexible pricing. Free migration tools handle the SQL translation automatically.
-
1
MedicalGPT
MedicalGPT: Training Your Own Medical GPT Model with ChatGPT Training
MedicalGPT training medical GPT model with ChatGPT training pipeline, implementation of Pretraining, Supervised Finetuning, Reward Modeling and Reinforcement Learning. MedicalGPT trains large medical models, including secondary pre-training, supervised fine-tuning, reward modeling, and reinforcement learning training. -
2
LLM Council
LLM Council works together to answer your hardest questions
...Instead of relying on one provider, this application sends your query simultaneously to several LLMs supported via OpenRouter, collects each model’s independent response, and then orchestrates a multi-stage evaluation where the models critique and rank each other’s outputs anonymously. After this peer-review process, a designated “Chairman” model synthesizes a final consolidated answer drawing on the strengths and insights of all participants. The interface looks like a familiar chat app but under the hood it implements this ensemble and consensus workflow to reduce bias and leverage diverse reasoning styles. -
3
SAG
SQL-Driven RAG Engine
...These vectors allow the system to identify relationships between concepts and construct a graph representation of knowledge at runtime. The architecture also includes a three-stage retrieval pipeline consisting of recall, expansion, and reranking steps to improve search accuracy. The engine integrates semantic vector similarity with traditional full-text search to improve both recall and precision. Because the knowledge graph is generated dynamically, the system can adapt to new information without requiring manual graph maintenance. -
4
WhisperJAV
Uses Qwen3-ASR, local LLM, Whisper, TEN-VAD
WhisperJAV is an open-source speech transcription pipeline designed specifically for generating subtitles for Japanese adult video content. The project addresses challenges that standard speech recognition models face when transcribing this type of audio, which often includes low signal-to-noise ratios and large numbers of non-verbal vocalizations. Traditional automatic speech recognition systems can misinterpret these sounds as words, leading to inaccurate transcripts. WhisperJAV introduces... -
5
omp (Oh My Pi)
AI Coding agent for the terminal
...Instead of functioning as a simple prompt-based assistant, the system includes an agent architecture that can inspect Git repositories, analyze changes, and perform development actions with fine-grained control. The platform also supports tool-based workflows where the agent can run shell commands, read files, modify code, and stage changes during development tasks. It includes infrastructure for integrating different AI providers and models through a unified API layer, allowing developers to switch between models while keeping the same agent interface. -
6
HuixiangDou
Overcoming Group Chat Scenarios with LLM-based Technical Assistance
...The project addresses a common problem in developer communities where discussion channels become overwhelmed by repeated or irrelevant questions. To solve this issue, HuixiangDou implements a multi-stage pipeline that analyzes incoming messages, filters irrelevant conversations, and selectively generates responses when the assistant determines it can provide useful information. This design allows the system to participate in group discussions without flooding the chat with unnecessary messages. The assistant uses retrieval and ranking methods along with language model reasoning to produce accurate answers for technical topics such as computer vision and machine learning projects. ... -
7
VibeThinker
Diversity-driven optimization and large-model reasoning ability
...It contains about 1.5 billion parameters, far smaller than many “frontier” models, yet it is explicitly optimized for reasoning, mathematics, and code generation tasks rather than general open-domain chat. The innovation lies in its training methodology: the team uses what they call the Spectrum-to-Signal Principle (SSP), where a first stage emphasizes diversity of reasoning paths (the “spectrum” phase) and a second stage uses reinforcement techniques (the “signal” phase) to refine toward correctness and strong reasoning. The result is a model that outpaces many much larger models on domain-specific benchmarks, demonstrating that smaller models, if trained carefully and with the right objectives, can achieve high performance in reasoning-centric tasks. -
8
Agentless
An agentless approach to automatically solve software development
...When solving a problem, the system first performs localization to determine which files, functions, or code segments are most likely responsible for the issue. It then generates multiple candidate patches for the identified locations using language model reasoning and diff-style edits. In the final stage, the framework validates potential patches by running regression tests and additional reproduction tests to confirm whether the fix resolves the original error. Based on these results, the system ranks the candidate patches and selects the most reliable solution to submit. -
9
Parallax
Parallax is a distributed model serving framework
...Instead of relying on centralized GPU clusters in data centers, the system allows multiple heterogeneous machines to collaborate in serving AI inference workloads. Parallax divides model layers across different nodes and dynamically coordinates them to form a complete inference pipeline. A two-stage scheduling architecture determines how model layers are allocated to available hardware and how requests are routed across nodes during execution. This scheduling system optimizes latency, throughput, and hardware utilization even when nodes have different computational capabilities. The platform also supports model sharding and pipeline parallelism, allowing very large models to run across distributed resources. -
Demo Series - Small Business Backup By VeeamWatch this on-demand demo series and learn how to protect your Microsoft 365 data with clear, simple, actionable steps that are easy to implement for businesses of all sizes.
-
10
DATAGEN
AI-driven multi-agent research assistant automating hypothesis
...The system coordinates multiple specialized AI agents that collaborate to perform tasks such as hypothesis generation, data collection, analysis, visualization, and report creation. Instead of requiring users to manually orchestrate each stage of a research process, the platform allows these agents to coordinate automatically and handle the workflow end-to-end. The project integrates several modern AI frameworks including LangChain, LangGraph, and large language models to manage reasoning and data processing tasks. Through this architecture, the system can combine structured data analysis with natural language reasoning to generate insights and research outputs. ... -
11
Coconut
Training Large Language Model to Reason in a Continuous Latent Space
Coconut is the official PyTorch implementation of the research paper “Training Large Language Models to Reason in a Continuous Latent Space.” The framework introduces a novel method for enhancing large language models (LLMs) with continuous latent reasoning steps, enabling them to generate and refine reasoning chains within a learned latent space rather than relying solely on discrete symbolic reasoning. It supports training across multiple reasoning paradigms—including standard... -
12
Ring
Ring is a reasoning MoE LLM provided and open-sourced by InclusionAI
Ring is a reasoning Mixture-of-Experts (MoE) large language model (LLM) developed by inclusionAI. It is built from or derived from Ling. Its design emphasizes reasoning, efficiency, and modular expert activation. In its “flash” variant (Ring-flash-2.0), it optimizes inference by activating only a subset of experts. It applies reinforcement learning/reasoning optimization techniques. Its architectures and training approaches are tuned to enable efficient and capable reasoning performance.... -
13
The Alignment Handbook
Robust recipes to align language models with human and AI preferences
The Alignment Handbook is an open-source resource created to provide practical guidance for aligning large language models with human preferences and safety requirements. The project focuses on the post-training stage of model development, where models are refined after pre-training to behave more helpfully, safely, and reliably in real-world applications. It provides detailed training recipes that explain how to perform tasks such as supervised fine-tuning, preference modeling, and reinforcement learning from human feedback. The handbook also includes reproducible workflows for training instruction-following models and evaluating alignment quality across different datasets and benchmarks. ... -
14
kg-gen
Knowledge Graph Generation from Any Text
kg-gen is an open-source framework developed by the STAIR Lab that automatically generates knowledge graphs from unstructured text using large language models. The system is designed to transform plain text sources such as documents, articles, or conversation transcripts into structured graphs composed of entities and relationships. Instead of relying on traditional rule-based extraction techniques, KG-Gen uses language models to identify entities and their relationships, producing... -
15
VALL-E
PyTorch implementation of VALL-E (Zero-Shot Text-To-Speech)
...Specifically, we train a neural codec language model (called VALL-E) using discrete codes derived from an off-the-shelf neural audio codec model, and regard TTS as a conditional language modeling task rather than continuous signal regression as in previous work. During the pre-training stage, we scale up the TTS training data to 60K hours of English speech which is hundreds of times larger than existing systems. VALL-E emerges in-context learning capabilities and can be used to synthesize high-quality personalized speech with only a 3-second enrolled recording of an unseen speaker as an acoustic prompt. Experiment results show that VALL-E significantly outperforms the state-of-the-art zero-shot TTS system in terms of speech naturalness and speaker similarity. ... -
16
ZAYA1-8B
Efficient MoE reasoning model for coding and math workloads
...It introduces architectural innovations such as Compressed Convolutional Attention, a novel MLP-based expert router, and learned residual scaling to improve routing stability and inference efficiency. The model was trained entirely on AMD infrastructure and refined through supervised fine-tuning and multi-stage reinforcement learning focused on reasoning and coding.
- Previous
- You're on page 1
- Next
