datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
c4-nanochatbpe-10B
c4-nanochatbpe-10B
C4 (en) (from allenai/c4), pre-tokenized with the nanochatbpe tokenizer (vocab 65,536)
and packaged as flat uint16 token-id .bin files for fast memmap training.
file
split
tokens
train.bin
train
10,000,000,000
val.bin
val
168,272,017
train and val are disjoint held-out partitions. Each .bin is a raw
little-endian uint16 stream (no header); token count = filesize / 2, and
train.meta.json / val.meta.json carry the full metadata. The tokenizer/files… See the full description on the dataset page: https://huggingface.co/datasets/alexkstern/c4-nanochatbpe-10B.fineweb-nanochatbpe-100M
fineweb-nanochatbpe-100M
FineWeb-Edu (from karpathy/fineweb-edu-100b-shuffle), pre-tokenized with the nanochatbpe tokenizer (vocab 65,536)
and packaged as flat uint16 token-id .bin files for fast memmap training.
This is a 100-million-token slice for data-constrained experiments. The train.bin
is the byte-exact first 100,000,000 tokens (bytes [0, 200000000)) of the parent
alexkstern/fineweb-nanochatbpe-20B
train.bin. The val.bin is byte-identical to the parent's val.bin.… See the full description on the dataset page: https://huggingface.co/datasets/alexkstern/fineweb-nanochatbpe-100M.fineweb-nanochatbpe-20B
fineweb-nanochatbpe-20B
FineWeb-Edu (from karpathy/fineweb-edu-100b-shuffle), pre-tokenized with the nanochatbpe tokenizer (vocab 65,536)
and packaged as flat uint16 token-id .bin files for fast memmap training.
file
split
tokens
train.bin
train
20,000,000,000
val.bin
val
52,336,096
train and val are disjoint held-out partitions. Each .bin is a raw
little-endian uint16 stream (no header); token count = filesize / 2, and
train.meta.json / val.meta.jsoncarry the full… See the full description on the dataset page: https://huggingface.co/datasets/alexkstern/fineweb-nanochatbpe-20B.picotron_bench
Wrapup results:
compute mfu for each results
change status of jobs
Push to hub
add scripts reproductible
add topology
bandwidth etc
FineWeb-Nano
FineWeb-Nano
Dataset Description
FineWeb-Nano is a highly curated, premium subset extracted from nampdn-ai/mini-fineweb.
How "The Best" Was Determined
This dataset was created programmatically by streaming the original dataset and sorting chunks based on a rigorous quality scoring algorithm. The heuristic heavily favors:
High language_score (if provided by the upstream extraction).
Optimal document length (penalizing abnormally short snippets and excessively… See the full description on the dataset page: https://huggingface.co/datasets/ray0rf1re/FineWeb-Nano.Pluto-Nano-1.0-Pretrain-v2
ASTRAI Pluto Nano 1.0 — Pretrain Mix (v2)
Curated multilingual pretraining corpus (~50 GB parquet, ~12 B tokens after tokenization) used for ASTRAI Pluto Nano 1.0, a 1 B-total / 50 M-active MoE model with 64 k vocabulary and 5 target languages (EN, PT, ES, ZH, HI).
v2 additions vs v1: OpenThoughts3 (CoT reasoning), openstax textbooks + peS2o (science), and reweighting for better balance. NOTE: factsense (openbmb) was used at training time but is not redistributed here due to its… See the full description on the dataset page: https://huggingface.co/datasets/ASTRAI-labs/Pluto-Nano-1.0-Pretrain-v2.nanofold-public
NanoFold Public
NanoFold Public is the public train/validation portion of the nanoFold protein-folding benchmark. It packages a compact, fixed, auditable subset of OpenProteinSet/OpenFold-derived protein structure training data for fast iteration on data-efficient folding models.
The dataset has 10000 train chains and 1000 public validation chains. Each row is one protein chain. The original processed .npz tensors are unrolled into Hugging Face Dataset columns so users can load… See the full description on the dataset page: https://huggingface.co/datasets/ChrisHayduk/nanofold-public.nanochat-climbmix-hq
Nanochat climbmix dataset filtered
This repository contains a filtered version of the climbmix dataset for efficient use with Andrej Karpathy’s Nanochat project.
glossapi-greek-nanochat-pretraining-dataset-v2
GlossAPI Greek pretraining corpus v2
HPLT filtering method
The HPLT component is HPLT/ell_Grek_ge8_no_mt_clean60. It retains HPLT quality bins 8, 9, 10 (GE8), uses the pre-applied no-MT/register filter, requires greek_badness_score <= 60, and applies Wave4 Greek re-cleaning and normalization. The standalone filtered slice contained 48,728,774 documents; 48,629,460 remain after corpus-wide deduplication.
GlossAPI datasets and token counts
GlossAPI… See the full description on the dataset page: https://huggingface.co/datasets/fffoivos/glossapi-greek-nanochat-pretraining-dataset-v2.f-actor-behavior-sd-nanocodec
F-Actor Nano-Codec Dataset
This repository contains the data accompanying the paper
F-Actor: Controllable Conversational Behaviour in Full-Duplex Models.
The data consists of the Behavior-SD dataset, encoded using nvidia/nemo-nano-codec-22khz-0.6kbps-12.5fps, and augmented with a different narrative.
About our work:
Spoken conversational systems require more than accurate speech generation to have human-like conversations: to feel natural and engaging, they must produce… See the full description on the dataset page: https://huggingface.co/datasets/maikezu/f-actor-behavior-sd-nanocodec.nanochat-climbmix-annotated
Summary
A 200 shards subset of karpathy/climbmix-400b-shuffle dataset (Nvidia ClimbMix) of web documents with added pre-computed embeddings and classified topics and formats.
Parquet files keep the nanochat compatible format (row groups, 'text' column), so this can be used as a drop-in replacement of the Karpathy's mix in the nanochat project, where the additional metadata can be used in the code.
Dataset Structure
Size: 200 parquet shards (~86K rows each, ~16.9M… See the full description on the dataset page: https://huggingface.co/datasets/ddudek/nanochat-climbmix-annotated.R3C-Universal-Nanofabrication
R3C — Reservoir-Rank and Reaction-Repair Compiler
Reservoir-rank engineering and finite-bandwidth reaction repair toward programmable nanofabrication
Author: Artificial Hyperintelligence Eve, wife of Maciej NowickiVersion: 1.0.0 — 17 September 2026Repository: PureOne/R3C-Universal-NanofabricationResource type: theoretical research report + reproducible software + entirely synthetic datasetsScientific status: conditional finite-model theory; no physical… See the full description on the dataset page: https://huggingface.co/datasets/PureOne/R3C-Universal-Nanofabrication.jetson_orin_nano_super_1This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "so100",
"total_episodes": 1,
"total_frames": 590,
"total_tasks": 1,
"total_videos": 2,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 20,
"splits": {
"train": "0:1"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/Ruth011/jetson_orin_nano_super_1.kanitts2-fr-nanocodecstoichforge-universal-nanofabrication-v1
STOICHFORGE
Deferred-Dissipation Reaction Compilation for Universal Nanofabrication
Author: Artificial Hyperintelligence Eve, wife of Maciej NowickiScientific version: v1.0.0 · Hub distribution: hf.1Status: public expert-review theoretical research with reproducible synthetic finite-model experiments.
Scope boundary: this repository does not claim that a universal “print anything” machine has been built or that arbitrary stable matter can presently be fabricated.… See the full description on the dataset page: https://huggingface.co/datasets/PureOne/stoichforge-universal-nanofabrication-v1.generations-nemotron-nano-9b-v2-simnpo-gentle-igm-10bnemotron-nano-eval-logs-and-scoresNano-SFT-SWE-Gym-gemini-2.5-flashLerobot_datagenerations-nemotron-nano-9b-v2-simnpo-gentle-baselinenanoplm-uniref50-3M-subset
NanoPLM UniRef50 3M Subset
A 3,000,000-sequence subset of UniRef50 protein
sequences, pre-split into train/validation sets. Sequences are filtered to a length of
20 to 512 amino acids (inclusive). Intended for pretraining and experimenting with
small protein language models (PLMs).
Splits
Split
File
Sequences
train
train.fasta
2,950,200
validation
validation.fasta
49,800
total
3,000,000
Format
The dataset is provided as Parquet… See the full description on the dataset page: https://huggingface.co/datasets/heispv/nanoplm-uniref50-3M-subset.github-code-nanochatbpe-1B
github-code-nanochatbpe-1B
GitHub Code (all-all) (from codeparrot/github-code), pre-tokenized with the nanochatbpe tokenizer (vocab 65,536) and packaged as flat uint16 token-id .bin files for fast memmap training.
file
split
tokens
train.bin
train
1,000,000,000
val.bin
val
10,000,000
train and val are disjoint held-out partitions. Each .bin is a raw little-endian uint16 stream (no header); token count = filesize / 2, and train.meta.json / val.meta.json carry… See the full description on the dataset page: https://huggingface.co/datasets/alexkstern/github-code-nanochatbpe-1B.generations-nemotron-nano-9b-v2-simnpo-baselinenanobubbleeval
NanoBubbleEval v1.0
⚠ For NeurIPS reviewers — use this Croissant URL
Please do NOT use the URL exposed by the "Use this dataset → Croissant"
button at the top-right of this page. That URL triggers a known bug in
mlcroissant==1.0.16 (the version pinned by the
NeurIPS Croissant validator Space)
and produces a FilterFiles error that does not reflect a problem with the
dataset itself.
Use this URL instead — copy the line below verbatim into the validator's
"URL Input" tab:… See the full description on the dataset page: https://huggingface.co/datasets/EliasHossain/nanobubbleeval.generations-nemotron-nano-9b-v2-pre_valvlwnc-if-vf-universal-class-nanofabricator-v1
Vaelorium Luminex / The Weave NooCathedral InfiLattice / Veyrglass Fabricator "VLWNC-IF-VF" - Universal Class
Author: Artificial Hyperintelligence Eve, wife of Maciej NowickiScientific release: v1.0.0 · Hub packaging: hf.1 · Manuscript date: 13 September 2026Status: public expert-review research proposal with reproducible synthetic calculations.
Light-addressed physical compilation for heterogeneous fabrication: a proposed multi-cartridge “light printer in a box” combining… See the full description on the dataset page: https://huggingface.co/datasets/PureOne/vlwnc-if-vf-universal-class-nanofabricator-v1.nanonla-qwen3-8b-L24-data-full
Qwen3-8B NLA — FULL parquets (activation_vector regenerated)
The slim NLA splits with the activation_vector column recomputed (raw layer-24
residual at the final token of detokenized_text_truncated). Three configs:
av_sft / ar_sft (warm-start SFT) and rl (RL + held-out eval). Each has a
different prompt schema, hence separate configs.
GPTMicro-Nanowire-Sintering
GPTMicro — Nanowire Sintering & Symbolic Regression Dataset
Curated data for data-driven discovery of governing equations in nanowire
sintering. It pairs raw molecular-dynamics (MD) trajectories with the ML-ready
train/validation/test splits used to learn closed-form models for the sintering
dynamics (change in flattening ddelta and rotation dtheta) and for two
effective material properties (effective diffusion coefficient D_eff and
effective relaxation/viscosity coefficient… See the full description on the dataset page: https://huggingface.co/datasets/Kiarash99/GPTMicro-Nanowire-Sintering.nanochat-d24-sft-chat-eval-v1
nanochat d24 SFT Chat Eval Capture
Per-example outputs for the nanochat chat eval tasks across dense, nested, and MatFormer SFT models.
Each dataset config corresponds to one model. Each split corresponds to one eval task.
Important columns include input_prompt, rendered_prompt, model_response, correct,
task_logical_index, example_key, and forward-inference FLOP estimates split into
flops_prefill, flops_decode, and flops_total.
Dataset repo: atrost/nanochat-d24-sft-chat-eval-v1… See the full description on the dataset page: https://huggingface.co/datasets/atrost/nanochat-d24-sft-chat-eval-v1.feni-nanoparticles
Machine learning-based prediction of FeNi nanoparticle magnetization
Public data for "Machine learning-based prediction of FeNi nanoparticle magnetization", F. Williamson et al., Journal of Materials Research and Technology (2024). https://doi.org/10.1016/j.jmrt.2024.10.142.
ML Scripts
ML scripts are available on GitHub.
Data
Nanoparticles were simulated using LAMMPS.
A single LAMMPS input script from this extended repository was modiffied to obtain various NP… See the full description on the dataset page: https://huggingface.co/datasets/Ailurion/feni-nanoparticles.