Add bundle README
Browse files
README.md
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
tags:
|
| 4 |
+
- proof-pilot
|
| 5 |
+
- olmo3
|
| 6 |
+
- dflash
|
| 7 |
+
- speculative-decoding
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# proof-pilot deploy bundle (private)
|
| 11 |
+
|
| 12 |
+
內部部署用模型打包，三個子資料夾：
|
| 13 |
+
|
| 14 |
+
| 子資料夾 | 模型 | 說明 |
|
| 15 |
+
|---|---|---|
|
| 16 |
+
| `soft-distill-7b-deploy/` | stage1-v2-7b soft-distill v2 (deploy) | off-policy soft distillation 完成版（bf16, `Olmo3SinkForCausalLM`），含 tokenizer + chat template |
|
| 17 |
+
| `dflash-7b-draft/` | DFlash draft for 7B target | SGLang 可部署 draft（speculative decoding），共用 target tokenizer |
|
| 18 |
+
| `dflash-32b-draft/` | DFlash draft for 32B target | SGLang 可部署 draft（speculative decoding），共用 target tokenizer |
|
| 19 |
+
|
| 20 |
+
DFlash draft 只含 `config.json` + `model.safetensors`，需搭配對應 target 模型使用。
|