datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
MTPqwen36-kquant-offload-mtp-swebench-lite100-results
Qwen3.6 K-Quant Offload MTP SWE-bench Lite 100 Results
This dataset contains the complete 5-model x 100-prompt runtime benchmark artifacts plus a detailed statistical analysis layer.
Primary conclusion: hot30/cold30 was the best decode-throughput run, while Q4_K_M had the best total wall clock. The ATX hot30/cold30 quantization significantly outperformed both Q4_K_M and Q3_K_XL on paired decode throughput, but Q4_K_M remains the elapsed-time control.
The ATX/K3 hot10, hot20, and… See the full description on the dataset page: https://huggingface.co/datasets/jakeatx/qwen36-kquant-offload-mtp-swebench-lite100-results.nvfp4-mtp-survey
Do Qwen3.8-27B NVFP4 repos actually ship a working MTP draft head?
A static survey of every NVFP4 quantization of Qwen3.8-27B and its finetunes that I could find
on the Hugging Face Hub, last run on 2026-08-24 (Rev 4) with
nvfp4_mtp_audit.py. Raw output: results.json.
I ran this to check a claim I had made in public, and the claim did not survive. The correction
is the first section, because it is the most important result here.
Revision history — read this, it is… See the full description on the dataset page: https://huggingface.co/datasets/windowsxp811203/nvfp4-mtp-survey.recap_dish_20260611This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "bi_so101_follower",
"total_episodes": 71,
"total_frames": 39618,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:71"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/mt-prox/recap_dish_20260611.recap_intervention0625This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "bi_so101_follower",
"total_episodes": 42,
"total_frames": 64742,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:42"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/mt-prox/recap_intervention0625.glm52-fidelity-exl3-tr3v4-3.5bpw-mtp78-brandonmusic-v1
fidelity--glm52.malaiwah.quant.exl3-tr3v4-3.5bpw-mtp78-brandonmusic
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from brandonmusic/GLM-5.2-EXL3-TR3v4-3.5bpw-MTP78.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm52-fidelity-exl3-tr3v4-3.5bpw-mtp78-brandonmusic-v1.datasets_for_magnetic_MTP_NatSR2024_training
Cite this dataset Kotykhov, A. S., Gubaev, K., Hodapp, M., Tantardini, C., Shapeev, A. V., and Novikov, I. S. datasets for magnetic MTP NatSR2024 training. ColabFit, 2024. https://doi.org/10.60732/9d635e27
This dataset has been curated and formatted for the ColabFit Exchange
This dataset is also available on the ColabFit Exchange:
https://materials.colabfit.org/id/DS_mf8sn11cn6wa_0
Visit the ColabFit Exchange to search additional datasets by author… See the full description on the dataset page: https://huggingface.co/datasets/colabfit/datasets_for_magnetic_MTP_NatSR2024_training.MT-prefdatasets_for_magnetic_MTP_NatSR2024_verification
Cite this dataset Kotykhov, A. S., Gubaev, K., Hodapp, M., Tantardini, C., Shapeev, A. V., and Novikov, I. S. datasets for magnetic MTP NatSR2024 verification. ColabFit, 2024. https://doi.org/10.60732/acd42be9
This dataset has been curated and formatted for the ColabFit Exchange
This dataset is also available on the ColabFit Exchange:
https://materials.colabfit.org/id/DS_wu6xd9i8cf7i_0
Visit the ColabFit Exchange to search additional datasets by… See the full description on the dataset page: https://huggingface.co/datasets/colabfit/datasets_for_magnetic_MTP_NatSR2024_verification.Zn_MTP_CMS2023
Cite this dataset Mei, H., Cheng, L., Chen, L., Wang, F., Li, J., and Kong, L. Zn MTP CMS2023. ColabFit, 2024. https://doi.org/10.60732/54902e18
This dataset has been curated and formatted for the ColabFit Exchange
This dataset is also available on the ColabFit Exchange:
https://materials.colabfit.org/id/DS_58y020ce6b6j_0
Visit the ColabFit Exchange to search additional datasets by author, description, element content and more.… See the full description on the dataset page: https://huggingface.co/datasets/colabfit/Zn_MTP_CMS2023.MTPu_2023
Cite this dataset Zongo, K., Sun, H., Ouellet-Plamondon, C., and Beland, L. K. MTPu 2023. ColabFit, 2024. https://doi.org/10.60732/41115bd2
This dataset has been curated and formatted for the ColabFit Exchange
This dataset is also available on the ColabFit Exchange:
https://materials.colabfit.org/id/DS_326i4urabisb_0
Visit the ColabFit Exchange to search additional datasets by author, description, element content and more.… See the full description on the dataset page: https://huggingface.co/datasets/colabfit/MTPu_2023.MT-pref-humantest-gemma4-mtp-uv
Generated Responses Dataset (Gemma 4 + MTP)
This dataset contains generated responses for prompts from davanstrien/haiku_dpo,
produced with Google Gemma 4 and vLLM's Multi-Token Prediction support.
Generation Details
Source Dataset: davanstrien/haiku_dpo
Source Split: train
Input Column: question (plain text prompts)
Model: google/gemma-4-26B-A4B-it
Rows Processed: 5
Batches: 3 (chunk size: 2)
Generation Date: 2026-05-06T13:42:18.500799
Script: gemma4-mtp.py… See the full description on the dataset page: https://huggingface.co/datasets/davanstrien/test-gemma4-mtp-uv.test-gemma4-mtp-OFF
Generated Responses Dataset (Gemma 4 + MTP)
This dataset contains generated responses for prompts from davanstrien/haiku_dpo,
produced with Google Gemma 4 and vLLM's Multi-Token Prediction support.
Generation Details
Source Dataset: davanstrien/haiku_dpo
Source Split: train
Input Column: question (plain text prompts)
Model: google/gemma-4-26B-A4B-it
Rows Processed: 30
Batches: 3 (chunk size: 10)
Generation Date: 2026-05-06T13:52:42.477251
Script: gemma4-mtp.py… See the full description on the dataset page: https://huggingface.co/datasets/davanstrien/test-gemma4-mtp-OFF.test-gemma4-mtp
Generated Responses Dataset (Gemma 4 + MTP)
This dataset contains generated responses for prompts from davanstrien/haiku_dpo,
produced with Google Gemma 4 and vLLM's Multi-Token Prediction support.
Generation Details
Source Dataset: davanstrien/haiku_dpo
Source Split: train
Input Column: question (plain text prompts)
Model: google/gemma-4-26B-A4B-it
Rows Processed: 5
Batches: 3 (chunk size: 2)
Generation Date: 2026-05-06T13:33:39.595844
Script: gemma4-mtp.py… See the full description on the dataset page: https://huggingface.co/datasets/davanstrien/test-gemma4-mtp.test-gemma4-mtp-ON
Generated Responses Dataset (Gemma 4 + MTP)
This dataset contains generated responses for prompts from davanstrien/haiku_dpo,
produced with Google Gemma 4 and vLLM's Multi-Token Prediction support.
Generation Details
Source Dataset: davanstrien/haiku_dpo
Source Split: train
Input Column: question (plain text prompts)
Model: google/gemma-4-26B-A4B-it
Rows Processed: 30
Batches: 3 (chunk size: 10)
Generation Date: 2026-05-06T13:52:41.595582
Script: gemma4-mtp.py… See the full description on the dataset page: https://huggingface.co/datasets/davanstrien/test-gemma4-mtp-ON.