Qwen3.6 - Miles

1. Model Introduction

Qwen3.6 is the next iteration of Alibaba’s Qwen3 line, focused on agentic-coding workflows and on preserving reasoning context across long sessions. The family ships two variants — a sparse MoE (Qwen3.6-35B-A3B) and a dense GDN-backbone model (Qwen3.6-27B) — both with native hybrid reasoning (thinking by default), built-in tool calling, and multimodal text / image / video input. Context windows reach 262 K and extend past 1 M. Weights are Apache 2.0, available in BF16 and FP8. The dense Qwen3.6-27B is the single-GPU-friendly variant. In miles it reuses the Qwen3.5 Megatron spec (miles_plugins.models.qwen3_5.get_qwen3_5_spec); architecturally it’s a wider, deeper Qwen3.5 with the gated-attention design preserved. Key highlights:

Dense GDN backbone: 27 B parameters, single-GPU friendly footprint.
Attention-output gate: shared with Qwen3.5, trained alongside attention weights.
Extended rotary base: --rotary-base 10000000, --rotary-percent 0.25.
Larger vocabulary: 248320 tokens.
Shape: hidden-size 5120, ffn-hidden-size 17408, 64 layers.
Long context: 262 K tokens, extensible past 1 M.

2. Supported Variants

Model	HF ID
Qwen3.6-27B	Qwen/Qwen3.6-27B

3. Environment Setup

3.1 Download model + datasets

hf download --repo-type dataset zhuzilin/dapo-math-17k --local-dir /root/dapo-math-17k
hf download --repo-type dataset zhuzilin/aime-2024     --local-dir /root/aime-2024

3.2 HF → Megatron `torch_dist` conversion

cd /root/miles
source scripts/models/qwen3.6-27B.sh
PYTHONPATH=/root/Megatron-LM python tools/convert_hf_to_torch_dist.py \
   ${MODEL_ARGS[@]} \
   --hf-checkpoint /root/Qwen3.6-27B \
   --save          /root/Qwen3.6-27B_torch_dist

4. Launch

4.1 Quick start

cd /root/miles
bash scripts/run-qwen3.6-27B.sh

The script targets 1 node × 8 GPU.

5. Recipe Configuration

5.1 Parallelism

TP	PP	CP	EP	`max_tokens_per_gpu`	SGLang `mem-fraction-static`	CPU Adam	GPUs
4	1	1	1	8192	0.5	✓	8 (1 × 8)

--sequence-parallel is enabled. Activation checkpointing is on (--recompute-granularity full --recompute-method uniform --recompute-num-layers 1).

5.2 Algorithm

GRPO with low-variance KL:

GRPO_ARGS=(
   --advantage-estimator grpo
   --use-kl-loss
   --kl-loss-coef 0.00
   --kl-loss-type low_var_kl
   --entropy-coef 0.00
   --eps-clip 0.2
   --eps-clip-high 0.28
)

5.3 Rollout & SGLang

SGLANG_ARGS=(
   --rollout-num-gpus-per-engine 1
   --sglang-mem-fraction-static 0.5
)

--rollout-num-gpus-per-engine 1 follows the Qwen3.5 line; SGLang TP > 1 has been problematic on this family. If your SGLang version carries the fix for sglang#21039, you can raise it.

5.4 Optimizer

CPU Adam is enabled (--optimizer-cpu-offload --overlap-cpu-optimizer-d2h-h2d --use-precision-aware-optimizer).

5.5 Notable quirks

From scripts/models/qwen3.6-27B.sh:

--spec miles_plugins.models.qwen3_5 get_qwen3_5_spec — Qwen3.6 reuses the Qwen3.5 spec (gated attention, FP32 A_log).
--rotary-base 10000000, --rotary-percent 0.25.
--vocab-size 248320.
--apply-layernorm-1p, --qk-layernorm, --group-query-attention.
--attention-output-gate.

See Backends Beyond Megatron for how miles preserves FP32 parameters like A_log through Megatron’s mixed-precision pipeline.

​1. Model Introduction

​2. Supported Variants

​3. Environment Setup

​3.1 Download model + datasets

​3.2 HF → Megatron torch_dist conversion

​4. Launch

​4.1 Quick start

​5. Recipe Configuration

​5.1 Parallelism

​5.2 Algorithm

​5.3 Rollout & SGLang

​5.4 Optimizer

​5.5 Notable quirks

​6. Pairs Well With