chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

[Fix] Environment variable SGL_* is deprecated (#14943)

Browse Source
This commit is contained in:
Huang Lin
2025-12-14 11:55:43 +08:00
committed by GitHub
parent 20ce9938b5
commit 6abdf73f4d
6 changed files with 9 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
docs/references/environment_variables.md
View File

@@ -27,7 +27,7 @@ SGLang supports various environment variables that can be used to configure its
| `SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN` | Allows the scheduler to overwrite longer context length requests (often set to `1` in Docker builds) | `0` |
| `SGLANG_IS_FLASHINFER_AVAILABLE` | Control FlashInfer availability check | `true` |
| `SGLANG_SKIP_P2P_CHECK` | Skip P2P (peer-to-peer) access check | `false` |
| `SGL_CHUNKED_PREFIX_CACHE_THRESHOLD` | Sets the threshold for enabling chunked prefix caching | `8192` |
| `SGLANG_CHUNKED_PREFIX_CACHE_THRESHOLD` | Sets the threshold for enabling chunked prefix caching | `8192` |
| `SGLANG_FUSED_MLA_ENABLE_ROPE_FUSION` | Enable RoPE fusion in Fused Multi-Layer Attention | `1` |
| `SGLANG_DISABLE_CONSECUTIVE_PREFILL_OVERLAP` | Disable overlap schedule for consecutive prefill batches | `false` |
| `SGLANG_SCHEDULER_MAX_RECV_PER_POLL` | Set the maximum number of requests per poll, with a negative value indicating no limit | `-1` |
@@ -45,7 +45,7 @@ SGLang supports various environment variables that can be used to configure its
| `SGLANG_ENABLE_JIT_DEEPGEMM` | Enable Just-In-Time compilation of DeepGEMM kernels (enabled by default on NVIDIA Hopper (SM90) and Blackwell (SM100) GPUs when the DeepGEMM package is installed; set to `"0"` to disable) | `"true"` |
| `SGLANG_JIT_DEEPGEMM_PRECOMPILE` | Enable precompilation of DeepGEMM kernels | `"true"` |
| `SGLANG_JIT_DEEPGEMM_COMPILE_WORKERS` | Number of workers for parallel DeepGEMM kernel compilation | `4` |
| `SGL_IN_DEEPGEMM_PRECOMPILE_STAGE` | Indicator flag used during the DeepGEMM precompile script | `"false"` |
| `SGLANG_IN_DEEPGEMM_PRECOMPILE_STAGE` | Indicator flag used during the DeepGEMM precompile script | `"false"` |
| `SGLANG_DG_CACHE_DIR` | Directory for caching compiled DeepGEMM kernels | `~/.cache/deep_gemm` |
| `SGL_DG_USE_NVRTC` | Use NVRTC (instead of Triton) for JIT compilation (Experimental) | `"0"` |
| `SGL_USE_DEEPGEMM_BMM` | Use DeepGEMM for Batched Matrix Multiplication (BMM) operations | `"false"` |
@@ -89,7 +89,7 @@ SGLang supports various environment variables that can be used to configure its
| Environment Variable | Description | Default Value |
| --- | --- | --- |
| `SGLANG_BLOCK_NONZERO_RANK_CHILDREN` | Control blocking of non-zero rank children processes | `1` |
| `SGL_IS_FIRST_RANK_ON_NODE` | Indicates if the current process is the first rank on its node | `"true"` |
| `SGLANG_IS_FIRST_RANK_ON_NODE` | Indicates if the current process is the first rank on its node | `"true"` |
| `SGLANG_PP_LAYER_PARTITION` | Pipeline parallel layer partition specification | Not set |
| `SGLANG_ONE_VISIBLE_DEVICE_PER_PROCESS` | Set one visible device per process for distributed computing | `false` |

2
docs/references/multi_node_deployment/lws_pd/lws-examples/p.yaml
View File

@@ -226,7 +226,7 @@ spec:
value: "8"
- name: SGLANG_ENABLE_JIT_DEEPGEMM
value: "1"
- name: SGL_CHUNKED_PREFIX_CACHE_THRESHOLD
- name: SGLANG_CHUNKED_PREFIX_CACHE_THRESHOLD
value: "0"
- name: NCCL_IB_QPS_PER_CONNECTION
value: "8"

2
docs/references/multi_node_deployment/lws_pd/lws_pd_deploy.md
View File

@@ -259,7 +259,7 @@ spec:
value: "8"
- name: SGLANG_ENABLE_JIT_DEEPGEMM
value: "1"
- name: SGL_CHUNKED_PREFIX_CACHE_THRESHOLD
- name: SGLANG_CHUNKED_PREFIX_CACHE_THRESHOLD
value: "0"
- name: NCCL_IB_QPS_PER_CONNECTION
value: "8"

2
python/sglang/compile_deep_gemm.py
View File

@@ -33,7 +33,7 @@ envs.SGLANG_IN_DEEPGEMM_PRECOMPILE_STAGE.set(True)
# Force enable deep gemm
envs.SGLANG_ENABLE_JIT_DEEPGEMM.set(True)
# Force enable mha chunked kv for DeepSeek V3 to avoid missing kv_b_proj DeepGEMM case
os.environ["SGL_CHUNKED_PREFIX_CACHE_THRESHOLD"] = "0"
os.environ["SGLANG_CHUNKED_PREFIX_CACHE_THRESHOLD"] = "0"
@dataclasses.dataclass

4
python/sglang/srt/layers/deep_gemm_wrapper/compile_utils.py
View File

@@ -25,8 +25,8 @@ if ENABLE_JIT_DEEPGEMM:
_BUILTIN_M_LIST = list(range(1, 1024 * 16 + 1))
_ENABLE_JIT_DEEPGEMM_PRECOMPILE = envs.SGLANG_JIT_DEEPGEMM_PRECOMPILE.get()
_DO_COMPILE_ALL = True
_IS_FIRST_RANK_ON_NODE = get_bool_env_var("SGL_IS_FIRST_RANK_ON_NODE", "true")
_IN_PRECOMPILE_STAGE = get_bool_env_var("SGL_IN_DEEPGEMM_PRECOMPILE_STAGE", "false")
_IS_FIRST_RANK_ON_NODE = get_bool_env_var("SGLANG_IS_FIRST_RANK_ON_NODE", "true")
_IN_PRECOMPILE_STAGE = get_bool_env_var("SGLANG_IN_DEEPGEMM_PRECOMPILE_STAGE", "false")
# Force redirect deep_gemm cache_dir
os.environ["DG_JIT_CACHE_DIR"] = os.getenv(

2
python/sglang/srt/models/deepseek_v2.py
View File

@@ -1464,7 +1464,7 @@ class DeepseekV2AttentionMLA(nn.Module):
# TODO: Design a finer way to determine the threshold
self.chunked_prefix_cache_threshold = get_int_env_var(
"SGL_CHUNKED_PREFIX_CACHE_THRESHOLD", 8192
"SGLANG_CHUNKED_PREFIX_CACHE_THRESHOLD", 8192
)
# If we have self.fused_qkv_a_proj_with_mqa and we're running on CPU, we will choose the torch.ops.sgl_kernel.qkv_proj_with_rope_fused_weight kernel