MoS-27B-Checkpoints is an open-weight model from Ryan Wang. Its published files total 261.4 GB.
Epoch checkpoints of the Qwen3.8-27B DFlash2 MoS line (project MoS, branch ryan/mos-improve, experiments/dflash2/qwen3.8-27b-100k/).
Parameters—
Context—
Weights261.4 GB
License—
AccessOpen weights
Monthly Downloads—
Model Card
Epoch checkpoints of the Qwen3.8-27B DFlash2 MoS line (project MoS, branch ryan/mos-improve, experiments/dflash2/qwen3.8-27b-100k/). Started 2026-09-22, when the general archive repo ryan-0608/MoS-Aurora-Experiment-Archive reached the Hugging Face 20,000-file limit. Serving protocol: sglang main f5866545, TP2, DFLASH 16 draft tokens, fixed-500 prompts, max 128 new tokens, greedy, concurrency 1, 500/500 completed. Same protocol as the dense and K4-jitter references. Reference arms (model-only, no optimizer) are in ryan-0608/MoS-Aurora-Experiment-Archive under dflash227btraj20260915/main-800k/: dense/epoch1, dense/epoch2, mos/epoch1 (K4 shared+1% jitter, expert LR 1e-4; serving 5.1501…
Excerpt from the card by Ryan Wang.
Identity and Version
- Repository
- ryan-0608/MoS-27B-Checkpoints
- Publisher
- Ryan Wang
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- f1a6c984c005f80a1c23dc65b6b5091d134bcf72
- First published
- 2026-09-22
- Last updated
- 2026-09-29
Files and Weights
98 files, 261.4 GB in total. The weights are 24 files totalling 261.4 GB in pt, safetensors.
Weights24 files · 261.4 GB
Configuration40 files · 60.7 KB
Documentation1 file · 3.2 KB
Other32 files · 3.4 MB
Repository1 file · 1.5 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/model.safetensors | Weights | 8.9 GB | 38b6b70d240c |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/optimizer_state_dict.pt | Weights | 7.7 GB | b632c6dc421b |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/scheduler_state_dict.pt | Weights | 1.5 KB | d5eccd889c0b |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/model.safetensors | Weights | 8.9 GB | cbdbf3566f51 |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/optimizer_state_dict.pt | Weights | 7.7 GB | e56baa6f3748 |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/scheduler_state_dict.pt | Weights | 1.5 KB | 30d04c45e4c1 |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/model.safetensors | Weights | 8.9 GB | 096037dc4956 |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/optimizer_state_dict.pt | Weights | 7.7 GB | 0731fb3416fd |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/scheduler_state_dict.pt | Weights | 1.5 KB | 7cd998966e8b |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/model.safetensors | Weights | 8.9 GB | 127d0aa4b14e |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/optimizer_state_dict.pt | Weights | 7.7 GB | 267e9db56c25 |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/scheduler_state_dict.pt | Weights | 1.5 KB | 44c44665ef56 |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/model.safetensors | Weights | 19.6 GB | 24a76de2efe9 |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/optimizer_state_dict.pt | Weights | 29.1 GB | 2d9e23b4fd38 |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/scheduler_state_dict.pt | Weights | 1.5 KB | ac4675afb609 |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/model.safetensors | Weights | 19.6 GB | 26d9b2f5aa1b |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/optimizer_state_dict.pt | Weights | 29.1 GB | 5cd19123fe49 |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/scheduler_state_dict.pt | Weights | 1.5 KB | 5dcfe0d69072 |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/model.safetensors | Weights | 19.6 GB | de212f234bc5 |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/optimizer_state_dict.pt | Weights | 29.1 GB | c33dfef97f9b |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/scheduler_state_dict.pt | Weights | 1.5 KB | 51b372e3d295 |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/model.safetensors | Weights | 19.6 GB | 0b1e7ecaae46 |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/optimizer_state_dict.pt | Weights | 29.1 GB | e51bc99da507 |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/scheduler_state_dict.pt | Weights | 1.5 KB | 05404bb912e6 |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/config.json | Configuration | 2.3 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/config.py | Configuration | 1.4 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/serving/acceptance_summary.json | Configuration | 1.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/training_state.json | Configuration | 51 B | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/val_metrics.json | Configuration | 2.4 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/config.json | Configuration | 2.3 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/config.py | Configuration | 1.4 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/serving/acceptance_summary.json | Configuration | 1.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/training_state.json | Configuration | 52 B | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/val_metrics.json | Configuration | 2.4 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/config.json | Configuration | 2.3 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/config.py | Configuration | 1.4 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/serving/acceptance_summary.json | Configuration | 1.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/training_state.json | Configuration | 52 B | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/val_metrics.json | Configuration | 1.8 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/config.json | Configuration | 2.3 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/config.py | Configuration | 1.4 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/serving/acceptance_summary.json | Configuration | 1.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/training_state.json | Configuration | 52 B | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/val_metrics.json | Configuration | 1.8 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/config.json | Configuration | 2.7 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/config.py | Configuration | 1.9 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/serving/acceptance_summary.json | Configuration | 1.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/training_state.json | Configuration | 51 B | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/val_metrics.json | Configuration | 2.4 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/config.json | Configuration | 2.7 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/config.py | Configuration | 1.9 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/serving/acceptance_summary.json | Configuration | 1.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/training_state.json | Configuration | 52 B | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/val_metrics.json | Configuration | 2.4 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/config.json | Configuration | 2.7 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/config.py | Configuration | 1.9 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/serving/acceptance_summary.json | Configuration | 1.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/training_state.json | Configuration | 52 B | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/val_metrics.json | Configuration | 2.4 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/config.json | Configuration | 2.7 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/config.py | Configuration | 1.9 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/serving/acceptance_summary.json | Configuration | 1.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/training_state.json | Configuration | 52 B | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/val_metrics.json | Configuration | 2.4 KB | — |
| README.md | Documentation | 3.2 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/serving/acceptance_trace.jsonl | Other | 134.8 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/serving/client.log | Other | 2.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/serving/server.log | Other | 291.8 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch1/train_command.txt | Other | 1.2 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/serving/acceptance_trace.jsonl | Other | 134.5 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/serving/client.log | Other | 2.0 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/serving/server.log | Other | 291.8 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch2/train_command.txt | Other | 1.2 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/serving/acceptance_trace.jsonl | Other | 134.7 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/serving/client.log | Other | 2.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/serving/server.log | Other | 291.6 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch3/train_command.txt | Other | 1.2 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/serving/acceptance_trace.jsonl | Other | 134.6 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/serving/client.log | Other | 2.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/serving/server.log | Other | 293.6 KB | — |
| dflash2_27b_traj_20260915/main-800k/dense4/epoch4/train_command.txt | Other | 1.2 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/serving/acceptance_trace.jsonl | Other | 134.8 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/serving/client.log | Other | 2.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/serving/server.log | Other | 289.9 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch1/train_command.txt | Other | 1.3 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/serving/acceptance_trace.jsonl | Other | 134.6 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/serving/client.log | Other | 2.0 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/serving/server.log | Other | 288.2 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch2/train_command.txt | Other | 1.3 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/serving/acceptance_trace.jsonl | Other | 134.3 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/serving/client.log | Other | 2.0 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/serving/server.log | Other | 287.6 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch3/train_command.txt | Other | 1.3 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/serving/acceptance_trace.jsonl | Other | 134.9 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/serving/client.log | Other | 2.1 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/serving/server.log | Other | 287.9 KB | — |
| dflash2_27b_traj_20260915/main-800k/x4rand/epoch4/train_command.txt | Other | 1.3 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 261.4 GB
Download from Ryan Wang
Released by Ryan Wang through its official repository on Hugging Face.
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 261.4 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.