DFloat11: Lossless LLM Compression for Efficient GPU Inference
active 2025-04-18 → 2026-05-12 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 390 days from 2025-04-18 to 2026-05-12. Pushes: 6 total, peak 2 in a day. Pull requests: 2 total, peak 2 in a day. Issues: 28 total, peak 2 in a day. Comments: 61 total, peak 6 in a day. Stars: 328 total, peak 77 in a day.
- Pushes
- Pull requests
- Issues
- Comments
- Stars
Stars, PRs, issues and forks are under-captured in the later part of this window. GH Archive progressively stopped capturing non-push events during 2026 — −95% or worse by the end of the window. Every series here except Pushes fades for that reason, so a decline above reflects the archive, not this repository. Pushes stay reliable throughout, so read them, and the contributor counts derived from them, as the real signal. Data health has the measurements.
Top contributors
Pushes, PRs, issues, reviews and comments — stars and forks excluded, so this is contribution rather than popularity
| Contributor | Contributions | Pushes | PRs | Comments |
|---|---|---|---|---|
| tonyzhang617 | 34 | 6 | 2 | 22 |
| mingyi456 | 14 | 0 | 0 | 11 |
| Andy0422 | 3 | 0 | 0 | 2 |
| unquietwiki | 3 | 0 | 0 | 2 |
| pramodith | 3 | 0 | 0 | 1 |
| ehartford | 3 | 0 | 0 | 2 |
| Luwill6 | 2 | 0 | 0 | 1 |
| Bruhmement | 2 | 0 | 0 | 1 |
| kabachuha | 2 | 0 | 0 | 2 |
| whatever1983 | 2 | 0 | 0 | 2 |
| amosyou | 2 | 0 | 0 | 1 |
| Manni1000 | 2 | 0 | 0 | 1 |
| calanquee | 2 | 0 | 0 | 0 |
| sankexin | 2 | 0 | 0 | 1 |
| mitkox | 2 | 0 | 0 | 0 |
| MeiYi-dev | 1 | 0 | 0 | 0 |
| LatentSpacer | 1 | 0 | 0 | 1 |
| win10ogod | 1 | 0 | 0 | 0 |
| animemory | 1 | 0 | 0 | 0 |
| BigBIueWhale | 1 | 0 | 0 | 0 |
Recent activity
Latest issues, pull requests and releases
- Issue comment#35erusiamu2026-05-05 11:24Request update for Cuda13 compatibility
- Issue comment#37mingyi4562026-03-16 07:38Apply Bfloat11 to activation
- Issue comment#37mingyi4562026-03-16 06:58Apply Bfloat11 to activation
- Issue comment#37jhy20k2026-03-16 06:25Apply Bfloat11 to activation
- Issue#36calanquee2026-02-07 09:47[Performance] Significant Decompression Throughput Degradation on GPU 0 and when using CUDA_VISIBLE_DEVICES (H200 HGX)
- Issue#36calanquee2026-02-06 08:04[Performance] Significant Decompression Throughput Degradation on GPU 0 and when using CUDA_VISIBLE_DEVICES (H200 HGX)
- Issue#34matalama80td3l2026-01-16 10:39Request for DFloat11 Quantization Support for FLUX.2-klein-9B and FLUX.2-klein-base-9B Model
- Issue comment#15zji9962025-12-17 10:03Can it be compatible with other quantization models such as fp8, int4, and models like Gapeleon/bytedance-BAGEL-7B-MoT-INT8 to reduce their memory usage
- Issue comment#15mingyi4562025-12-16 08:34Can it be compatible with other quantization models such as fp8, int4, and models like Gapeleon/bytedance-BAGEL-7B-MoT-INT8 to reduce their memory usage
- Issue#33jaxiez2025-11-18 06:29Converted DFloat11 shards into safetensors for ComfyUI
- Issue#32mingyi4562025-11-03 17:34Feature request: support for decompressing `nn.Conv2d` (and similar) tensors
- Issue#30tonyzhang6172025-09-30 04:25DF11 for Qwen-Image-Edit-2509 (make it fit on 32 GB 🤝)
- Issue#30BigBIueWhale2025-09-23 15:17DF11 for Qwen-Image-Edit-2509 (make it fit on 32 GB 🤝)
- Issue comment#29mingyi4562025-09-22 12:36qwen-image-edit load error
- Issue comment#29Luwill62025-09-22 11:39qwen-image-edit load error
- Issue comment#29mingyi4562025-09-22 09:06qwen-image-edit load error
- Issue comment#29mingyi4562025-09-22 07:01qwen-image-edit load error
- Issue comment#22mingyi4562025-09-21 13:18Can we quantize the df11 model ourselves?
- Issue#29Luwill62025-09-19 02:51qwen-image-edit load error
- Issue comment#28mingyi4562025-09-18 17:46How to compress t5-v1_1-xxl and Gemma-2-2B?
- Issue comment#28mingyi4562025-09-18 04:55How to compress t5-v1_1-xxl?
- Issue comment#28mingyi4562025-09-17 17:24How to compress t5-v1_1-xxl?
- Issue#28mingyi4562025-09-14 12:03How to compress t5-v1_1-xxl?
- Issue#24tonyzhang6172025-08-28 07:49could you please support bagel?
- Issue comment#27tonyzhang6172025-08-28 07:48How do load DF11 models in `diffusers` without the original transformer weights?
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 328 stars here means stars gained during the window, not the repo's star count.