From 9cb6cbc909525c2f92537428c9ad3ab7f3f9fed0 Mon Sep 17 00:00:00 2001 From: Marc Blatter Date: Wed, 15 Jul 2026 19:34:24 +0000 Subject: [PATCH] compose-vLLM.yml aktualisiert --- compose-vLLM.yml | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/compose-vLLM.yml b/compose-vLLM.yml index 088ddc4..afe8c17 100644 --- a/compose-vLLM.yml +++ b/compose-vLLM.yml @@ -38,8 +38,12 @@ services: - "0.90" - "--max-model-len" - "65536" + - "--rope-scaling" + - '{"rope_type":"yarn","factor":2.0,"original_max_position_embeddings":40960}' - "--max-num-seqs" - - "32" + - "16" + - "--kv-cache-dtype" + - "fp8" - "--enable-prefix-caching" - "--enable-auto-tool-choice" - "--tool-call-parser"