The dataset is currently empty. Upload or create new data files. Then, you will be able to explore them in the Dataset Viewer.
๐ MQ-Cognitive-Data // Domain-Adaptive Pretraining Corpus
This repository hosts the highly curated, domain-specific text corpora and instruction-tuning dataset splits engineered by the MAH Quantum Research Scholars cohort. This data matrix is explicitly designed for vocabulary expansion and domain adaptation of baseline models targeting accelerated quantum emulations and advanced cognitive logic structures.
๐ Dataset Structure & Token Allocation
Our processing pipelines filter, deduplicate, and tokenize heavy industrial computer science, classical tensor network mathematics, and hardware acceleration logs.
| Split Node | Purpose | Target Tokens / Pairs | Access Control |
|---|---|---|---|
train_pretrain |
Domain Vocabulary Adaptation | ~1.2 Billion Tokens | Internal Cluster Sync |
train_instruct |
Multi-turn Cognitive Instruction Splits | ~450,000 High-Quality Pairs | Active Evaluation |
validation_eval |
Zero-shot Architecture Benchmarks | ~50,000 Test Matrices | Public Verified |
๐ Verification & Compliance Registry
- Governance Frame: Ministry of MSME, Govt of India Registered Entity
- Infrastructure Pipeline: Managed via active NVIDIAยฎ NGC Enterprise Org ID
0963318590610147 - Data Curation Node: Verified distributed pipelines originating from
mahquantum.tech
For data contribution frameworks, compliance auditing, or remote university collaboration pipelines, contact our administrative layer at nirooph@mahquantum.tech.
- Downloads last month
- 13