add dflash-32b-draft-v2test-phaseL-int4mlp (int4-MLP draft)
Browse files
README.md
CHANGED
|
@@ -24,7 +24,8 @@ tags:
|
|
| 24 |
| `dflash-7b-draft/` | DFlash draft for 7B target | SGLang 可部署 draft(speculative decoding) |
|
| 25 |
| `dflash-32b-draft/` | DFlash draft for **stage1-v2-32b** target (s5317) | SGLang 可部署 draft;對齊舊 32B deploy target |
|
| 26 |
| `dflash-32b-draft-v2test/` | DFlash draft for stage1-v2-32b-softdistill-v2test — **phase-1 warm-up(非 final)** | curriculum **phase-1 短 context warm-up**(step-10000 快照)。SWA512 / block_size 11 / 8L / GQA-8。**部署請用下方 `dflash-32b-draft-v2test-phaseL`。** |
|
| 27 |
-
| `dflash-32b-draft-v2test-phaseL/` | **DFlash draft for stage1-v2-32b-softdistill-v2test — phase-2 final(部署推薦)** | curriculum **phase-2 長 context 特化**(job 140680,warm-start 自 phase-1):train data = 真實長 proof 部署分佈(OPD 32B rollouts finish_reason=length filtered + dsflash-v2-test teacher proofs,micro 65536)、GAMMA 20。step_3000 完整收尾,acc 0.605 / greedy mean_prefix_len 4.90。SWA512 / block_size 11 / 8L / GQA-8。
|
|
|
|
| 28 |
|
| 29 |
DFlash draft 只含 `config.json` + `model.safetensors`(已 reshard 成多 shard),需搭配**對應的** target 模型使用(注意 v2test draft 對應 v2test target,與舊 `dflash-32b-draft` 的 target 不同)。
|
| 30 |
|
|
|
|
| 24 |
| `dflash-7b-draft/` | DFlash draft for 7B target | SGLang 可部署 draft(speculative decoding) |
|
| 25 |
| `dflash-32b-draft/` | DFlash draft for **stage1-v2-32b** target (s5317) | SGLang 可部署 draft;對齊舊 32B deploy target |
|
| 26 |
| `dflash-32b-draft-v2test/` | DFlash draft for stage1-v2-32b-softdistill-v2test — **phase-1 warm-up(非 final)** | curriculum **phase-1 短 context warm-up**(step-10000 快照)。SWA512 / block_size 11 / 8L / GQA-8。**部署請用下方 `dflash-32b-draft-v2test-phaseL`。** |
|
| 27 |
+
| `dflash-32b-draft-v2test-phaseL/` | **DFlash draft for stage1-v2-32b-softdistill-v2test — phase-2 final(部署推薦)** | curriculum **phase-2 長 context 特化**(job 140680,warm-start 自 phase-1):train data = 真實長 proof 部署分佈(OPD 32B rollouts finish_reason=length filtered + dsflash-v2-test teacher proofs,micro 65536)、GAMMA 20。step_3000 完整收尾,acc 0.605 / greedy mean_prefix_len 4.90。SWA512 / block_size 11 / 8L / GQA-8。serving accept ~3.1–4.1(單流,dev H200)。 |
|
| 28 |
+
| `dflash-32b-draft-v2test-phaseL-int4mlp/` | **phase-L draft 的 int4-MLP 量化版** | `dflash-32b-draft-v2test-phaseL` 的 **MLP(gate/up/down)量成 compressed-tensors int4**(RTN W4A16 g128),qkv/o + sink/fc/mask_embed 留 bf16(保 DFlash fused-KV)。**4.82→2.30GB(−55%)**。sglang DFLASH 部署實測:載成 int4(weight mem 2.16GB)、`fused KV materialization ENABLED`、accept 3.1–4.1(== bf16 draft)、單流比 bf16 draft **+2~15% tok/s**。serve 需 patched `dflash_sink.py`(thread quant_config→MLP)+ `--speculative-draft-model-quantization compressed-tensors`。詳見 `docs/quantization.md §13`。 |
|
| 29 |
|
| 30 |
DFlash draft 只含 `config.json` + `model.safetensors`(已 reshard 成多 shard),需搭配**對應的** target 模型使用(注意 v2test draft 對應 v2test target,與舊 `dflash-32b-draft` 的 target 不同)。
|
| 31 |
|