lovesenko
/

DeepSeek-V4-Flash-DSpark-Abliterated

8-bit precision

Model card Files Files and versions

lovesenko commited on Jul 3

Commit

235f754

·

verified ·

1 Parent(s): 5186e33

Update README.md

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -144,7 +144,7 @@ This abliterated checkpoint is a drop-in replacement for the original weights
 For inference guidance specific to the **NVIDIA RTX PRO 6000 Blackwell** (TP2/TP4, the `lucifer-default` / `lucifer-cutlass` / `b12x` backends, and the native DSpark `method=dspark` speculative-decoding path with `num_speculative_tokens=5`), see the community guide:
-👉 **https://github.com/local-inference-lab/rtx6kpro/blob/master/models/ds4dspark-v8.md**
 That page documents the validated Docker image, launch helpers, and full TP2/TP4 throughput sweep (decode + prefill) for this exact checkpoint on RTX PRO 6000, including DSpark speculative decoding which gives a ~41% single-stream throughput speedup on this abliterated release.

 For inference guidance specific to the **NVIDIA RTX PRO 6000 Blackwell** (TP2/TP4, the `lucifer-default` / `lucifer-cutlass` / `b12x` backends, and the native DSpark `method=dspark` speculative-decoding path with `num_speculative_tokens=5`), see the community guide:
+👉 **https://github.com/local-inference-lab/rtx6kpro/blob/master/models/ds4dspark-v9.md**
 That page documents the validated Docker image, launch helpers, and full TP2/TP4 throughput sweep (decode + prefill) for this exact checkpoint on RTX PRO 6000, including DSpark speculative decoding which gives a ~41% single-stream throughput speedup on this abliterated release.