From 398b98d9ef14c881786bbc2e695185434f030ea3 Mon Sep 17 00:00:00 2001 From: "github-actions[bot]" Date: Tue, 28 Oct 2025 08:34:24 +0000 Subject: [PATCH 1/3] Add metadata for Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8 --- .../metadata.yaml | 48 +++++++++++++++++++ 1 file changed, 48 insertions(+) create mode 100644 models/qwen/qwen3-coder-480b-a35b-instruct-fp8-/metadata.yaml diff --git a/models/qwen/qwen3-coder-480b-a35b-instruct-fp8-/metadata.yaml b/models/qwen/qwen3-coder-480b-a35b-instruct-fp8-/metadata.yaml new file mode 100644 index 0000000..5783a1d --- /dev/null +++ b/models/qwen/qwen3-coder-480b-a35b-instruct-fp8-/metadata.yaml @@ -0,0 +1,48 @@ +apiVersion: model.hydra.io/v1alpha1 +kind: ModelSpec +metadata: + name: qwen3-coder-480b-a35b-instruct-fp8 +spec: + deployments: + - customRuntimeArgs: [] + resourceRequirements: + cpu: 8 + gpuCount: 8 + gpuType: nvidia-vgpu + memory: 640 + perGPUMemoryGB: 80 + runtime: vllm + versionRequired: '>=0.8.5' + descriptor: + description: + enUS: 'Today, we''re announcing Qwen3-Coder, our most agentic code model to + date. Qwen3-Coder is available in multiple sizes, but we''re excited to introduce + its most powerful variant first: Qwen3-Coder-480B-A35B-Instruct. featuring + the following key enhancements: - Significant Performance among open models + on Agentic Coding, Agentic Browser-Use, and other foundational coding tasks, + achieving results comparable to Claude Sonnet.' + zhCN: 'Today, we''re announcing Qwen3-Coder, our most agentic code model to + date. Qwen3-Coder is available in multiple sizes, but we''re excited to introduce + its most powerful variant first: Qwen3-Coder-480B-A35B-Instruct. featuring + the following key enhancements: - Significant Performance among open models + on Agentic Coding, Agentic Browser-Use, and other foundational coding tasks, + achieving results comparable to Claude Sonnet.' + display: Qwen3-Coder-480B-A35B-Instruct-FP8 + icon: + src: https://public-resources.d.run/models/logos/qwen-model-logo.svg + type: image/svg + links: + - description: About + url: https://github.com/QwenLM + provider: + id: alibaba + name: + enUS: Alibaba + zhCN: 通义千问 + tags: + - TEXT_GENERATION + source: + huggingface: + name: Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8 + modelscope: + name: Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8 From 78fc8a1891cc680a325dfdfb4c3df3fc6d26d58f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Nicole=20LiHui=20=F0=9F=A5=9C?= Date: Tue, 28 Oct 2025 09:37:14 +0000 Subject: [PATCH 2/3] update qwen3-coder-480b-a35b-instruct-fp8 metadata --- .../metadata.yaml | 48 +++++++++++++++++++ .../metadata.yaml | 48 ------------------- 2 files changed, 48 insertions(+), 48 deletions(-) create mode 100644 models/alibaba/qwen3-coder-480b-a35b-instruct-fp8/metadata.yaml delete mode 100644 models/qwen/qwen3-coder-480b-a35b-instruct-fp8-/metadata.yaml diff --git a/models/alibaba/qwen3-coder-480b-a35b-instruct-fp8/metadata.yaml b/models/alibaba/qwen3-coder-480b-a35b-instruct-fp8/metadata.yaml new file mode 100644 index 0000000..3d72766 --- /dev/null +++ b/models/alibaba/qwen3-coder-480b-a35b-instruct-fp8/metadata.yaml @@ -0,0 +1,48 @@ +apiVersion: model.hydra.io/v1alpha1 +kind: ModelSpec +metadata: + name: qwen3-coder-480b-a35b-instruct-fp8 +spec: + config: + maxTokens: 262144 + deployments: + - customRuntimeArgs: + - --max-model-len=131072 + - --enable-auto-tool-choice + - --tool-call-parser + - qwen3_coder + resourceRequirements: + cpu: 16 + gpuCount: 8 + gpuType: vgpu + memory: 640 + perGPUMemoryGB: 80 + runtime: vllm + versionRequired: '>=0.8.5' + descriptor: + description: + enUS: 'The Qwen3-Coder-480B-A35B-Instruct model supports 256K-token context (extendable to 1M via Yarn), + excels in Agentic Coding, Browser-Use, and code benchmarks, and offers robust function-call + support on platforms like Qwen Code and CLINE.' + zhCN: 'Qwen3-Coder-480B-A35B-Instruct 支持 256K Token 上下文 (可通过 Yarn 扩展至 1M), + 在智能体编程、浏览任务及代码基准测试中表现出色，并在 Qwen Code、CLINE 等平台提供完善的函数调用支持.' + display: Qwen3-Coder-480B-A35B-Instruct-FP8 + icon: + src: https://public-resources.d.run/models/logos/qwen-model-logo.svg + type: image/svg + links: + - description: About + url: https://github.com/QwenLM + provider: + id: alibaba + name: + enUS: Alibaba + zhCN: 通义千问 + tags: + - TEXT_GENERATION + - TOOLS + source: + huggingface: + name: Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8 + modelscope: + name: Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8 diff --git a/models/qwen/qwen3-coder-480b-a35b-instruct-fp8-/metadata.yaml b/models/qwen/qwen3-coder-480b-a35b-instruct-fp8-/metadata.yaml deleted file mode 100644 index 5783a1d..0000000 --- a/models/qwen/qwen3-coder-480b-a35b-instruct-fp8-/metadata.yaml +++ /dev/null @@ -1,48 +0,0 @@ -apiVersion: model.hydra.io/v1alpha1 -kind: ModelSpec -metadata: - name: qwen3-coder-480b-a35b-instruct-fp8 -spec: - deployments: - - customRuntimeArgs: [] - resourceRequirements: - cpu: 8 - gpuCount: 8 - gpuType: nvidia-vgpu - memory: 640 - perGPUMemoryGB: 80 - runtime: vllm - versionRequired: '>=0.8.5' - descriptor: - description: - enUS: 'Today, we''re announcing Qwen3-Coder, our most agentic code model to - date. Qwen3-Coder is available in multiple sizes, but we''re excited to introduce - its most powerful variant first: Qwen3-Coder-480B-A35B-Instruct. featuring - the following key enhancements: - Significant Performance among open models - on Agentic Coding, Agentic Browser-Use, and other foundational coding tasks, - achieving results comparable to Claude Sonnet.' - zhCN: 'Today, we''re announcing Qwen3-Coder, our most agentic code model to - date. Qwen3-Coder is available in multiple sizes, but we''re excited to introduce - its most powerful variant first: Qwen3-Coder-480B-A35B-Instruct. featuring - the following key enhancements: - Significant Performance among open models - on Agentic Coding, Agentic Browser-Use, and other foundational coding tasks, - achieving results comparable to Claude Sonnet.' - display: Qwen3-Coder-480B-A35B-Instruct-FP8 - icon: - src: https://public-resources.d.run/models/logos/qwen-model-logo.svg - type: image/svg - links: - - description: About - url: https://github.com/QwenLM - provider: - id: alibaba - name: - enUS: Alibaba - zhCN: 通义千问 - tags: - - TEXT_GENERATION - source: - huggingface: - name: Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8 - modelscope: - name: Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8 From 8f391edd8bf16fafc1f5c6c687008dc0bf6e49b7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Nicole=20LiHui=20=F0=9F=A5=9C?= Date: Tue, 28 Oct 2025 10:18:12 +0000 Subject: [PATCH 3/3] rm qwen 480 maxtoken args --- models/alibaba/qwen3-coder-480b-a35b-instruct-fp8/metadata.yaml | 1 - 1 file changed, 1 deletion(-) diff --git a/models/alibaba/qwen3-coder-480b-a35b-instruct-fp8/metadata.yaml b/models/alibaba/qwen3-coder-480b-a35b-instruct-fp8/metadata.yaml index 3d72766..7ad8d1d 100644 --- a/models/alibaba/qwen3-coder-480b-a35b-instruct-fp8/metadata.yaml +++ b/models/alibaba/qwen3-coder-480b-a35b-instruct-fp8/metadata.yaml @@ -7,7 +7,6 @@ spec: maxTokens: 262144 deployments: - customRuntimeArgs: - - --max-model-len=131072 - --enable-auto-tool-choice - --tool-call-parser - qwen3_coder