[Model] Pipeline parallel support for Qwen2 (#6924)

2024-08-01 09:49:51 +08:00
parent 7ecee34321
commit 1d2e7fb73f
3 changed files with 101 additions and 27 deletions
--- a/vllm/config.py
+++ b/vllm/config.py
@@ -40,6 +40,8 @@ _PP_SUPPORTED_MODELS = [
    "GPT2LMHeadModel",
    "MixtralForCausalLM",
    "NemotronForCausalLM",
+    "Qwen2ForCausalLM",
+    "Qwen2MoeForCausalLM",
 ]