From 225ff08ab5bd037c7c38863d53240215571df833 Mon Sep 17 00:00:00 2001 From: "young.zheng" Date: Thu, 19 Feb 2026 13:33:16 -0800 Subject: [PATCH 1/3] adds glm4.7 --- z-ai/glm-4-7-fp8/config.yaml | 0 1 file changed, 0 insertions(+), 0 deletions(-) create mode 100644 z-ai/glm-4-7-fp8/config.yaml diff --git a/z-ai/glm-4-7-fp8/config.yaml b/z-ai/glm-4-7-fp8/config.yaml new file mode 100644 index 00000000..e69de29b From cee134e00621b63931b73132f28744191c8f6d8f Mon Sep 17 00:00:00 2001 From: "young.zheng" Date: Thu, 19 Feb 2026 13:36:22 -0800 Subject: [PATCH 2/3] adds in content --- z-ai/glm-4-7-fp8/config.yaml | 32 ++++++++++++++++++++++++++++++++ 1 file changed, 32 insertions(+) diff --git a/z-ai/glm-4-7-fp8/config.yaml b/z-ai/glm-4-7-fp8/config.yaml index e69de29b..a5e4f0e0 100644 --- a/z-ai/glm-4-7-fp8/config.yaml +++ b/z-ai/glm-4-7-fp8/config.yaml @@ -0,0 +1,32 @@ +model_metadata: + example_model_input: + messages: + - role: system + content: "You are a helpful assistant." + - role: user + content: "Write FizzBuzz in Python" + stream: true + model: "glm-4.7-fp8" + max_tokens: 131072 + temperature: 1.0 + top_p: 0.95 + tags: + - openai-compatible +model_name: GLM-4.7-FP8 +base_image: + image: lmsysorg/sglang@sha256:cedd5a9d449444c2f1460e549efe7cded85e2e46246efb8d00c1eba1f917e5b0 +docker_server: + start_command: sh -c "GPU_COUNT=$(nvidia-smi --list-gpus | wc -l); python3 -m sglang.launch_server --model-path zai-org/GLM-4.7-FP8 --tp-size $GPU_COUNT --tool-call-parser glm47 --reasoning-parser glm45 --speculative-algorithm EAGLE --speculative-num-steps 3 --speculative-eagle-topk 1 --speculative-num-draft-tokens 4 --mem-fraction-static 0.8 --served-model-name glm-4.7-fp8 --host 0.0.0.0 --port 8000" + readiness_endpoint: /health_generate + liveness_endpoint: /health_generate + predict_endpoint: /v1/chat/completions + server_port: 8000 +resources: + accelerator: H100:8 + use_gpu: true +runtime: + predict_concurrency: 256 + health_checks: + restart_check_delay_seconds: 1500 # Waits 15 minutes after deployment before starting health checks + restart_threshold_seconds: 30 # Triggers a restart if health checks fail for 30 seconds + stop_traffic_threshold_seconds: 30 # Stops traffic if health checks fail for 30 seconds \ No newline at end of file From 53133ae4aa327976aba169ebee8dbbaeda8d1f16 Mon Sep 17 00:00:00 2001 From: "young.zheng" Date: Thu, 19 Feb 2026 13:40:10 -0800 Subject: [PATCH 3/3] precommit linting --- z-ai/glm-4-7-fp8/config.yaml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/z-ai/glm-4-7-fp8/config.yaml b/z-ai/glm-4-7-fp8/config.yaml index a5e4f0e0..c228d502 100644 --- a/z-ai/glm-4-7-fp8/config.yaml +++ b/z-ai/glm-4-7-fp8/config.yaml @@ -29,4 +29,4 @@ runtime: health_checks: restart_check_delay_seconds: 1500 # Waits 15 minutes after deployment before starting health checks restart_threshold_seconds: 30 # Triggers a restart if health checks fail for 30 seconds - stop_traffic_threshold_seconds: 30 # Stops traffic if health checks fail for 30 seconds \ No newline at end of file + stop_traffic_threshold_seconds: 30 # Stops traffic if health checks fail for 30 seconds