Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -72,14 +72,17 @@ vllm serve <target-model> \
|
|
| 72 |
|
| 73 |
6,903 rows generated on-policy from Arete-27B-Uncensored (loop rows
|
| 74 |
quarantined), hidden states extracted from layers 6/20/34/48/62/64 of the
|
| 75 |
-
target.
|
|
|
|
|
|
|
|
|
|
| 76 |
|
| 77 |
## Provenance
|
| 78 |
|
| 79 |
- Z-Lab drafter base: [z-lab/Qwen3.8-27B-DFlash2](https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2)
|
| 80 |
- incoai finetuned variant: [incoai/Qwen3.8-27B-DFlash2](https://huggingface.co/incoai/Qwen3.8-27B-DFlash2)
|
| 81 |
- Target: [SargeDev/Arete-27B-Uncensored-V2](https://huggingface.co/SargeDev/Arete-27B-Uncensored-V2) (our uncensored build of [NousResearch/Arete-27B](https://huggingface.co/NousResearch/Arete-27B))
|
| 82 |
-
- Training corpus: `SargeDev/jev-distill-corpus-v3`
|
| 83 |
- Trainer: [vllm-project/speculators](https://github.com/vllm-project/speculators) 0.9.0a0, DFlash2 (dflash2) speculator type
|
| 84 |
|
| 85 |
## License
|
|
|
|
| 72 |
|
| 73 |
6,903 rows generated on-policy from Arete-27B-Uncensored (loop rows
|
| 74 |
quarantined), hidden states extracted from layers 6/20/34/48/62/64 of the
|
| 75 |
+
target. The exact training rows are published:
|
| 76 |
+
**[SargeDev/arete-dflash2-trainset](https://huggingface.co/datasets/SargeDev/arete-dflash2-trainset)**
|
| 77 |
+
(97.6 MB, jsonl: `id`, `primary_id`, `source`, `input_ids`, `loss_mask`,
|
| 78 |
+
`text`, `metadata`). Corpus lineage: `SargeDev/jev-distill-corpus-v3`.
|
| 79 |
|
| 80 |
## Provenance
|
| 81 |
|
| 82 |
- Z-Lab drafter base: [z-lab/Qwen3.8-27B-DFlash2](https://huggingface.co/z-lab/Qwen3.8-27B-DFlash2)
|
| 83 |
- incoai finetuned variant: [incoai/Qwen3.8-27B-DFlash2](https://huggingface.co/incoai/Qwen3.8-27B-DFlash2)
|
| 84 |
- Target: [SargeDev/Arete-27B-Uncensored-V2](https://huggingface.co/SargeDev/Arete-27B-Uncensored-V2) (our uncensored build of [NousResearch/Arete-27B](https://huggingface.co/NousResearch/Arete-27B))
|
| 85 |
+
- Training corpus: [SargeDev/arete-dflash2-trainset](https://huggingface.co/datasets/SargeDev/arete-dflash2-trainset) (`SargeDev/jev-distill-corpus-v3` lineage)
|
| 86 |
- Trainer: [vllm-project/speculators](https://github.com/vllm-project/speculators) 0.9.0a0, DFlash2 (dflash2) speculator type
|
| 87 |
|
| 88 |
## License
|