Upload qwen3.6-35b-a3b.env with huggingface_hub

qwen3.6-35b-a3b.env

@@ -0,0 +1,29 @@
+# Qwen3.6 35B-A3B - optimized GB10 default
+# Source: huggingface.co/Qwen/Qwen3.6-35B-A3B
+# MoE VLM, 35B total / 3B active, 128k text-only context, Qwen3 reasoning parser, tool calls, MTP 2.
+# Selected from llama-benchy short-suite sweep: best balanced c1/c2/c4 profile.
+MODEL=Qwen/Qwen3.6-35B-A3B
+SERVED_NAME=Qwen/Qwen3.6-35B-A3B
+IMAGE=vllm/vllm-openai:v0.20.0
+GPU_MEM_UTIL=0.85
+# vllm/vllm-openai entrypoint already runs `vllm serve`.
+CMD_PREFIX=()
+
+EXTRA_DOCKER_ARGS=()
+
+EXTRA_VLLM_ARGS=(
+  --dtype bfloat16
+  --tensor-parallel-size 1
+  --pipeline-parallel-size 1
+  --data-parallel-size 1
+  --max-model-len 131072
+  --max-num-seqs 16
+  --max-num-batched-tokens 8192
+  --language-model-only
+  --enable-prefix-caching
+  --reasoning-parser qwen3
+  --enable-auto-tool-choice
+  --tool-call-parser qwen3_coder
+  --speculative-config '{"method":"qwen3_next_mtp","num_speculative_tokens":2}'
+)
+