Remove grok model arch from vllm (#46706)

Signed-off-by: Xianbao QIAN <xianbao.qian@gmail.com>
This commit is contained in:
Tiezhen WANG
2026-06-25 23:02:10 -07:00
committed by GitHub
parent 35a49fcfc2
commit c7645bce04
13 changed files with 3 additions and 1401 deletions
+1 -1
View File
@@ -297,7 +297,7 @@ The `fastokens` Python package (>= 0.2.0) must be installed; if it isn't,
vLLM raises a clear `ImportError` at tokenizer load. The override applies to
any `--tokenizer-mode` that ends up loading an HF fast tokenizer (`hf`,
`deepseek_v32`, `deepseek_v4`, …). Models that don't use the HF
fast tokenizer (`mistral`, `grok2`, `kimi_audio`) ignore the flag.
fast tokenizer (`mistral`, `kimi_audio`) ignore the flag.
Tokenizer-bound workloads — long shared prefixes, bursty short prompts,
batch detokenization — see the largest wins. If your bottleneck is GPU