datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Cot-Drop
LuminaSFT
LuminaSFT is a synthetic SFT dataset suite specifically designed to improve both general-purpose and task-specific SLMs. LuminaSFT consists of multiple curated splits that target diverse capabilities:
UltraChat200K-DeepSeek - A regenerated base SFT dataset for broad instruction following.
InstructGPT-NaturalQA and InstructGPT-TriviaQA - Factual question answering datasets to strengthen knowledge recall and answer accuracy.
CoT-Drop - A reading comprehension dataset with… See the full description on the dataset page: https://huggingface.co/datasets/amd/Cot-Drop.SAND-MATH
SAND-Math: A Synthetic Dataset of Difficult Problems to Elevate LLM Math Performance
📃 Paper | 🤗 Dataset
SAND-Math (Synthetic Augmented Novel and Difficult Mathematics) is a high-quality, high-difficulty dataset of mathematics problems and solutions. It is generated using a novel pipeline that addresses the critical bottleneck of scarce, high-difficulty training data for mathematical Large Language Models (LLMs).
Key Features
Novel Problem Generation: Problems are… See the full description on the dataset page: https://huggingface.co/datasets/amd/SAND-MATH.SAND-Post-Training-Dataset
SAND-Post-Training-Dataset: High-Quality Synthetic Reasoning Dataset Built with AMD GPUs
Dataset Summary
We introduce the SAND-Post-Training-Dataset, a high-quality synthetic reasoning dataset for mathematics and science built entirely using a synthetic data pipeline running on the AMD ROCm™ stack and AMD Instinct™ MI325 GPUs.
This dataset prioritizes difficulty and novelty over volume, demonstrating that high-difficulty synthetic data can elevate… See the full description on the dataset page: https://huggingface.co/datasets/amd/SAND-Post-Training-Dataset.Instella-GSM8K-synthetic
Instella-GSM8K-synthetic
The Instella-GSM8K-synthetic dataset was used in the second stage pre-training of Instella-3B model, which was trained on top of the Instella-3B-Stage1 model.
This synthetic dataset was generated using the training set of GSM8k dataset, where we first used Qwen2.5-72B-Instruct to
Abstract numerical values as function parameters and generate a Python program to solve the math question.
Identify and replace numerical values in the existing question with… See the full description on the dataset page: https://huggingface.co/datasets/amd/Instella-GSM8K-synthetic.InstructGpt-educational
LuminaSFT
LuminaSFT is a synthetic SFT dataset suite specifically designed to improve both general-purpose and task-specific SLMs. LuminaSFT consists of multiple curated splits that target diverse capabilities:
UltraChat200K-DeepSeek - A regenerated base SFT dataset for broad instruction following.
InstructGPT-NaturalQA and InstructGPT-TriviaQA - Factual question answering datasets to strengthen knowledge recall and answer accuracy.
CoT-Drop - A reading comprehension dataset with… See the full description on the dataset page: https://huggingface.co/datasets/amd/InstructGpt-educational.TTT-Bench
TTT-Bench: A Benchmark for Evaluating Reasoning Ability with Simple and Novel Tic-Tac-Toe-style Games
📃 Paper | 🤗 Dataset | 🌐 Website
We introduce TTT-Bench, a new benchmark specifically created to evaluate the reasoning capability of LRMs through a suite of simple and novel two-player Tic-Tac-Toe-style games.
Although trivial for humans, these games require basic strategic reasoning, including predicting an opponent's intentions and understanding spatial configurations.… See the full description on the dataset page: https://huggingface.co/datasets/amd/TTT-Bench.InstructGpt-NaturalQa
LuminaSFT
LuminaSFT is a synthetic SFT dataset suite specifically designed to improve both general-purpose and task-specific SLMs. LuminaSFT consists of multiple curated splits that target diverse capabilities:
UltraChat200K-DeepSeek - A regenerated base SFT dataset for broad instruction following.
InstructGPT-NaturalQA and InstructGPT-TriviaQA - Factual question answering datasets to strengthen knowledge recall and answer accuracy.
CoT-Drop - A reading comprehension dataset with… See the full description on the dataset page: https://huggingface.co/datasets/amd/InstructGpt-NaturalQa.UltraChat200K-regenerated
LuminaSFT
LuminaSFT is a synthetic SFT dataset suite specifically designed to improve both general-purpose and task-specific SLMs. LuminaSFT consists of multiple curated splits that target diverse capabilities:
UltraChat200K-DeepSeek - A regenerated base SFT dataset for broad instruction following.
InstructGPT-NaturalQA and InstructGPT-TriviaQA - Factual question answering datasets to strengthen knowledge recall and answer accuracy.
CoT-Drop - A reading comprehension dataset with… See the full description on the dataset page: https://huggingface.co/datasets/amd/UltraChat200K-regenerated.InstructGpt-TriviaQa
LuminaSFT
LuminaSFT is a synthetic SFT dataset suite specifically designed to improve both general-purpose and task-specific SLMs. LuminaSFT consists of multiple curated splits that target diverse capabilities:
UltraChat200K-DeepSeek - A regenerated base SFT dataset for broad instruction following.
InstructGPT-NaturalQA and InstructGPT-TriviaQA - Factual question answering datasets to strengthen knowledge recall and answer accuracy.
CoT-Drop - A reading comprehension dataset with… See the full description on the dataset page: https://huggingface.co/datasets/amd/InstructGpt-TriviaQa.pepsi-2021-10k-8192-without-year-llama3.2-3b-with-trained-cartridge-amd-2021
Dataset: Phudish/pepsi-2021-10k-8192-without-year-llama3.2-3b-with-trained-cartridge-amd-2021
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/pepsi-2021-10k-8192-without-year-llama3.2-3b-with-trained-cartridge-amd-2021")
amd-2021-10k-8192-without-year-llama3.2-3b
Dataset: Phudish/amd-2021-10k-8192-without-year
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-10k-8192-without-year")
amd-2022-10k-8192-without-year-thinking-1-with-trained-cartridge
Dataset: Phudish/amd-2022-10k-8192-without-year-thinking-1-with-trained-cartridge
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2022-10k-8192-without-year-thinking-1-with-trained-cartridge")
amd-2021-10k_8192_1024_0.2_no_cartridge_llama-3.2-3b
Dataset: Phudish/amd-2021-10k_8192_1024_no_cartridge
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-10k_8192_1024_no_cartridge")
amd-2022-10k-8192-without-year-llama3.2-3b-with-trained-cartridge-amd-2021
Dataset: Phudish/amd-2022-10k-8192-without-year-llama3.2-3b-with-trained-cartridge-amd-2021
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2022-10k-8192-without-year-llama3.2-3b-with-trained-cartridge-amd-2021")
amd-2021-10k-64-without-year-new-prompt-qwen3-4b-0.2-thinking
Dataset: Phudish/amd-2021-10k-64-without-year-new-prompt-qwen3-4b
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-10k-64-without-year-new-prompt-qwen3-4b")
amd-2022-10k-8192-without-year-thinking-1-with-untrained-cartridge
Dataset: Phudish/amd-2022-10k-8192-without-year-thinking-1-with-untrained-cartridge
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2022-10k-8192-without-year-thinking-1-with-untrained-cartridge")
amd-2021-10k_8192_8192_1.0_no_cartridge_qwen3-4b
Dataset: Phudish/amd-2021-10k_8192_8192_1.0_no_cartridge_qwen3-4b
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-10k_8192_8192_1.0_no_cartridge_qwen3-4b")
AM-DeepSeek-R1-Distilled-1.4M_preprocess
AM-DeepSeek-R1-Distilled-1.4M_preprocess
Preprocessed from [a-m-team/AM-DeepSeek-R1-Distilled-1.4M] (configs: am_0.5M, am_0.9M).
Two columns:
question
answer: reasoning enclosed in <think>...</think> and final in #### answer.
Format:
<think> ...reasoning... </think>
#### answer ...final...
amd-2022-10k_8192_8192_0.2_no_cartridge_qwen3-4b
Dataset: Phudish/amd-2022-10k_8192_8192_0.2_no_cartridge_qwen3-4b
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2022-10k_8192_8192_0.2_no_cartridge_qwen3-4b")
amd-2021-llama-3.2-3b-n65536
Dataset: Phudish/amd-2021-llama-3.2-3b-n65536
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-llama-3.2-3b-n65536")
amd-2022-10k_8192_1024_1.0_no_cartridge_qwen3-4b
Dataset: Phudish/amd-2022-10k_8192_1024_1.0_no_cartridge_qwen3-4b
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2022-10k_8192_1024_1.0_no_cartridge_qwen3-4b")
amd-2021-10k-64-without-year-new-prompt-1-thinking-both-qwen3-4b
Dataset: Phudish/amd-2021-10k-64-without-year-new-prompt-1-thinking-both-qwen3-4b
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-10k-64-without-year-new-prompt-1-thinking-both-qwen3-4b")
amd-2021-10k-8192-without-year-thinking-0.2-1024
Dataset: Phudish/amd-2021-10k-8192-without-year-thinking-0.2-1024
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-10k-8192-without-year-thinking-0.2-1024")
amd-2022-10k_8192_1024_0.2_no_cartridge_llama-3.2-3b
Dataset: Phudish/amd-2022-10k_8192_1024_no_cartridge
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2022-10k_8192_1024_no_cartridge")
amd-2021-10k_8192_8192_0.2_no_cartridge_llama-3.2-3b
Dataset: Phudish/amd-2021-10k_8192_8192_0.2_no_cartridge_llama3.2-3b
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-10k_8192_8192_0.2_no_cartridge_llama3.2-3b")
amd-2021-10k_8192_8192_0.2_no_cartridge_qwen3-4b
Dataset: Phudish/amd-2021-10k_8192_8192_0.2_no_cartridge_qwen3-4b
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-10k_8192_8192_0.2_no_cartridge_qwen3-4b")
amd-2021-10k_8192_1024_1.0_no_cartridge_llama-3.2-3b
Dataset: Phudish/amd-2021-10k_8192_1024_1.0_no_cartridge_llama-3.2-3b
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-10k_8192_1024_1.0_no_cartridge_llama-3.2-3b")
amd-2021-qwen3-4b-n65536
Dataset: Phudish/amd-2021-qwen3-4b-n65536
Self-study synthetic QA data (N=65,536) used to train a Cartridge (compressed KV
memory) on AMD 2021 (FinanceBench 10-K), distilled from Qwen/Qwen3-4B-Instruct-2507.
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-qwen3-4b-n65536")
amd-2021-10k-64-without-year-new-prompt-llama-3.2-3b
Dataset: Phudish/amd-2021-10k-64-without-year-new-prompt-llama-3.2-3b
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2021-10k-64-without-year-new-prompt-llama-3.2-3b")
amd-2022-10k_8192_1024_0.2_no_cartridge_qwen3-4b
Dataset: Phudish/amd-2022-10k_8192_1024_no_cartridge_qwen3_4b
Usage
from datasets import load_dataset
ds = load_dataset("Phudish/amd-2022-10k_8192_1024_no_cartridge_qwen3_4b")