[diffusion] feat: let every layerwise component be configurable (#35688)

docs/docs/sglang-diffusion/api/cli.mdx

@@ -246,6 +246,22 @@ Values passed to the compatibility option `--layerwise-offload-components` must
 
 Layerwise tuning options such as `--dit-offload-prefetch-size`, `--dit-layerwise-resident-layers`, and `--dit-layerwise-residency-policy` continue to control the streamed layer working set. Prefer the smallest component set that solves the memory issue because layerwise offload can increase latency.
 
+Those three set the default for every streamed component. To give one component its own values, use the `component=value` forms, which also accept JSON:
+
+```bash
+sglang serve --model-path <MODEL> \
+  --layerwise-offload-components dit,text_encoder \
+  --layerwise-prefetch-size text_encoder=2 \
+  --layerwise-resident-layers text_encoder=4 \
+  --layerwise-residency-policy text_encoder=strided
+```
+
+- `--layerwise-prefetch-size`: how many layers to fetch ahead. Fractional values are a share of the stack, `>= 1` an absolute count. Deeper prefetch overlaps more of the transfer with compute, at the cost of staging buffers.
+- `--layerwise-resident-layers`: how many layers stay on the GPU instead of being streamed. Resident layers are transferred once at startup, so they are removed from every pass. Fractional values are a share of the stack.
+- `--layerwise-residency-policy`: `leading` keeps the first layers, `strided` spreads them across the stack so the transfers do not arrive as one burst.
+
+A component without an entry keeps the group default, so adding these options changes nothing until one is set. Both knobs trade VRAM for transfer, and the return differs by component: a DiT earns it back once per denoising step, a text encoder or VAE once per request. Measure before raising either.
+
 ## Serve
 
 `sglang serve` starts the HTTP server and keeps the model loaded for repeated requests.