minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[MLX] Window-bounded SWA KV storage and in-graph sampling (#34166)

Browse Source 
Co-authored-by: Siming Deng <siming_deng_stat@163.com>
Co-authored-by: R0CKSTAR <yeahdongcn@gmail.com>
Co-authored-by: Jiminator <Jiminator@users.noreply.github.com>
Co-authored-by: damahua <damahua@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Alex Nails
2026-08-09 21:28:51 -07:00
committed by GitHub
co-authored by Siming Deng R0CKSTAR Jiminator damahua Claude Opus 5
parent 449f0da78f
commit 2969ab3d41
27 changed files with 3510 additions and 526 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/docs/advanced_features/server_arguments.mdx
+6
View File
@@ -673,6 +673,12 @@ Please consult the documentation below and [server_args.py](https://github.com/s
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>The random seed.</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}><code>None</code></td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>Type: int</td>
</tr>
<tr>
<td style={{padding: "9px 12px", fontWeight: 500, backgroundColor: "rgba(255,255,255,0.02)"}}>`--mlx-enable-sampling`</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>MLX backend only: sample decode tokens (temperature / top-k / top-p / min-p) instead of greedy argmax. Sampling runs inside the lazy MLX graph, so it works with the overlap scheduler; first tokens from prefill/extend are sampled too. Greedy requests keep exact argmax behavior. Also enables on the MLX path: grammar vocab masks and custom logit processors (these break decode chaining per step; custom processors run on pure-decode steps only), logit_bias, output logprobs (sampled token / top-k / token_ids; prompt input logprobs are not computed), NaN sanitization (SGLANG_SANITIZE_NAN_LOGITS), and per-request sampling_seed under --enable-deterministic-inference (deterministic within MLX only). Penalties are not applied.</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.02)"}}>`False`</td>
<td style={{padding: "9px 12px", backgroundColor: "rgba(255,255,255,0.05)"}}>bool flag (set to enable)</td>
</tr>
<tr>
<td style={{padding: "9px 12px", fontWeight: 500, backgroundColor: "rgba(255,255,255,0.02)"}}>`--constrained-json-whitespace-pattern`</td>