From bce6e18f2b83cd13035c5ad6feef633f3bac3cb8 Mon Sep 17 00:00:00 2001 From: sirius0xdev Date: Sun, 10 May 2026 01:48:34 +0000 Subject: [PATCH] fix vllm and tini --- apps/base/customer1/hermes-agent/new-deployment.yaml | 3 ++- infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml | 2 +- 2 files changed, 3 insertions(+), 2 deletions(-) diff --git a/apps/base/customer1/hermes-agent/new-deployment.yaml b/apps/base/customer1/hermes-agent/new-deployment.yaml index 8b63d28..a7793cd 100644 --- a/apps/base/customer1/hermes-agent/new-deployment.yaml +++ b/apps/base/customer1/hermes-agent/new-deployment.yaml @@ -103,7 +103,8 @@ spec: secretKeyRef: name: telegram-webhook key: TELEGRAM_WEBHOOK_SECRET - + - name: TINI_SUBREAPER + value: "true" - name: TELEGRAM_WEBHOOK_PORT value: "8645" diff --git a/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml b/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml index dac27e4..9dd8f19 100644 --- a/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml +++ b/infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml @@ -42,7 +42,7 @@ spec: - --enable-auto-tool-choice - --kv-cache-dtype=fp8 - --max-num-batched-tokens=65536 - - --max-num-seqs=128 + - --max-num-seqs=64 - --enable-chunked-prefill - --trust-remote-code - --dtype=auto