A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 18:30 UTC
[V1] Remove V0 fallback for mistral-tokenizer (#14873)
Signed-off-by: Roger Wang <ywang@roblox.com>
Roger Wang committed
Mar 15, 2025 at 20:21 UTC
b30c75dda4f6c5e0d8b3d2b39134da38b72ea96e
1 file changed
-7
vllm/engine/arg_utils.py
-7
@@ -1487,13 +1487,6 @@ class EngineArgs:
1487
recommend_to_remove=False)
1488
return False
1489
1490
- # No MistralTokenizer support so far (not compatible
1491
- # with xgrammar)
1492
- if model_config.tokenizer_mode == "mistral":
1493
- _raise_or_fallback(feature_name="--tokenizer-mode mistral",
1494
- recommend_to_remove=False)
1495
- return False
1496
-
1490
# No CPU offloading yet.
1491
if self.cpu_offload_gb != EngineArgs.cpu_offload_gb:
1492
_raise_or_fallback(feature_name="--cpu-offload-gb",