[Docs] Remove references to removed env vars (#59530)

Signed-off-by: Sundri Lai <laiyanting.neu@gmail.com>
This commit is contained in:
Sundri Lai
2026-10-01 08:26:14 +00:00
committed by GitHub
parent 24fb5fc08c
commit bcee730b1a
2 changed files with 4 additions and 8 deletions
+3 -6
View File
@@ -83,10 +83,7 @@ directly from the Triton kernel through unified virtual addressing (UVA),
prefetching on a side CUDA stream so transfers overlap with the compute of prefetching on a side CUDA stream so transfers overlap with the compute of
preceding layers. The freed GPU memory goes to the KV cache. preceding layers. The freed GPU memory goes to the KV cache.
`cpu_offload` defaults to `true` (on): it follows the `cpu_offload` defaults to `true` (on)
`VLLM_PLE_CPU_OFFLOAD` environment variable, which also defaults to `true`.
An explicit `--engram-config` value takes precedence over the environment
variable.
To keep the tables resident on GPU instead (e.g. when host memory is To keep the tables resident on GPU instead (e.g. when host memory is
scarce): scarce):
@@ -96,8 +93,8 @@ vllm serve deepseek-ai/DeepSeek-V4.1-Flash \
--engram-config.cpu_offload false --engram-config.cpu_offload false
``` ```
or equivalently `VLLM_PLE_CPU_OFFLOAD=0`. CPU offload requires a GPU with UVA CPU offload requires a GPU with UVA support;
support; vLLM fails fast when it is unavailable. vLLM fails fast when it is unavailable.
## Data-parallel topologies (DeepSeek V4.1) ## Data-parallel topologies (DeepSeek V4.1)
+1 -2
View File
@@ -51,8 +51,7 @@ and provision transport limits and memory accordingly.
For example, start a standard inference server with scale-out endpoints enabled: For example, start a standard inference server with scale-out endpoints enabled:
```bash ```bash
VLLM_ENABLE_SCALE_OUT_ENDPOINTS=1 \ vllm serve meta-llama/Llama-3.1-8B-Instruct --enable-scale-out
vllm serve meta-llama/Llama-3.1-8B-Instruct
``` ```
Send a Responses request and use `jq` to extract the `token_ids` field: Send a Responses request and use `jq` to extract the `token_ids` field: