Fix ExaoneMoeMTP test that never ran in Transformers v4 (#36792)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
2026-03-11 17:10:23 +00:00
parent 196802dfa6
commit 5efa206a8c
4 changed files with 17 additions and 0 deletions
--- a/tests/models/multimodal/generation/vlm_utils/core.py
+++ b/tests/models/multimodal/generation/vlm_utils/core.py
@@ -74,6 +74,8 @@ def run_test(
    if model_info.require_embed_inputs:
        for k in ("skip_tokenizer_init", "enable_prompt_embeds", "enable_mm_embeds"):
            vllm_runner_kwargs_[k] = model_info.require_embed_inputs
+    if not model_info.enable_prefix_caching:
+        vllm_runner_kwargs_["enable_prefix_caching"] = False

    if vllm_runner_kwargs:
        vllm_runner_kwargs_.update(vllm_runner_kwargs)