# recipes/qwen3.6-27b/b200/attune-tp1-pp8-split7-rb2.yaml # An OpenRecipes recipe; the fields are documented in schema/recipe.schema.yaml. schema_version: 1 model: Qwen3.6-27B gpu: type: B200 memory_gb: 180 count: 8 per_node: 8 instance: AWS p6-b200.48xlarge author: name: ATTUNE url: https://arxiv.org/abs/2610.17528 status: verified stack: framework: 'slime 0.3.0: THUDM/slime a897e1f plus 8 commits that do not change the training step' megatron_lm: commit 1dcf0daf (megatron-core 0.16.0rc0), with a backported upstream fix for the Qwen3.5 attention output gate at TP above the query-group count transformer_engine: 2.10.0 pytorch: 2.11.0 (CUDA 12.9 build) cuda: '12.9' nccl: 2.28.9 (CUDA 12.9 build; printed by the measured run) deepep: 1.2.1 python: 3.12.3 fla: 0.4.1 (flash-linear-attention) triton: 3.6.0 network: 8 GPUs per node, EFA, 3.2 Tb/s RDMA per node runtime_fixes: patches/fla_blackwell_guard.py, imported in every training process before FLA source: 'in-pod version check of the measured run''s training image (same digest) on 2026-10-18: installed TE, PyTorch, DeepEP, megatron-core, FLA and Triton, Megatron-LM git commit, Python; NCCL from the measured run''s log' workload: row_tokens: 8192 rows: SFT conversations global_batch: 128 precision: bf16 dataset: name: Tulu 3 SFT mixture source: allenai/tulu-3-sft-mixture revision: b14afda60f1bbebe55d5d2fa1e4df5042f97f8be license: ODC-By 1.0 rows_sha256: 7a837641dcde9ebcac9d3ed5ed84ddd941fc9bfc173606cc701e9a3f51c307f1 build: name: qwen35-tulu3-fixed8192-v1 tokenizer: Qwen3.5 (the Qwen3.5-122B-A10B tokenizer; tokenizer.json sha256 5f9e4d4901a92b997e463c1f46055088b6cca5ca61a6522d1b9f64c4bb81cb42) loss_mask: qwen3_5 rows: 6400 row_tokens: 8192 padding_tokens: 0 seed: 42 packing: source conversations in a seeded global permutation, concatenated with an EOS token between them and cut into consecutive 8,192-token rows; rows without supervised targets are skipped supervised_token_ratio: 0.676 parquet_sha256: f215bdc6aba60d6d0dff4471eaef805765cb393e40b95762190535376491ab01 parquet_bytes: 76127393 builder: data/generate_fixed_sft.py (data/README.md) layout: label: TP1 PP8 summary: TP 1, PP 8 (7 layers on the first stage, 9 on each of the six middle stages, 3 on the last), CP 1, DP 1; one 8,192-token row per microbatch; full recompute of the first 2 layers of each stage (block); precision-aware distributed optimizer; no offload; dense model, no experts highlights: 7/9/9/9/9/9/9/3 layer split, block recompute of 2 layers, no offload tp: 1 pp: 8 vpp: 1 cp: 1 ep: 1 etp: 1 dp: 1 sequence_parallel: false layer_split: first: 7 last: 3 max_tokens_per_gpu: 8192 rows_per_microbatch: 1 recompute: granularity: full method: block num_layers: 2 offload: null dispatcher: null result: tokens_per_s_per_gpu: 4330.87 usd_per_million_tokens: 0.91 step_time_cv_percent: 1.08 mfu: 0.2959 mfu_note: null active_params: count: 25623920640 source: https://huggingface.co/Qwen/Qwen3.6-27B revision: 6a9e13bd6fc8f0983b9b99948120bc37f49c13e9 measured: 2026-10-18 steps_per_run: 20 window: - 5 - 19 window_tokens: 15728640 runs: 1 profiler: false gpu_hours: 3.13 gpu_hours_note: GPUs x (scheduler admission to finish), including startup; the same run is the candidate arm of the convergence check correctness: result: pass summary: passed; 1 component test passed on this image reference: 'The layout of slime''s `scripts/run-qwen3.5-27B.sh` on one 8-GPU node: TP 4 with sequence parallelism, PP 2 (first stage 34 layers, last 30), CP 1 in place of its CP 4, full recompute and CPU optimizer offload, with this recipe''s optimizer settings.' configuration: pass component_tests: - test: precision-aware Adam on the GPU and CPU optimizer offload (fraction 1.0, overlapped copies), each against an FP64 reference shape: 'one Qwen3.5-397B-A17B MoE layer: hidden 4096, 512 experts, top-10, expert FFN 1024, EP 8, ETP 1, 4,096 tokens per rank' gpu: B200 date: 2026-10-14 result: pass detail: worst tensor (update) relative L2 error 5.12e-7 on the GPU and 7.99e-7 offloaded against a bound of 1.0e-3 after one step, and 4.99e-5 and 7.73e-5 against 1.1e-3 after three steps at the recipe's hyperparameters (bound 2 x the default GPU optimizer's error + 1e-3); reconstructed FP32 masters bitwise identical to an arm with explicit FP32 masters, the TE remainder self-test passes and the negative controls fail note: run on B200 on this recipe's training image; the optimizer step is elementwise, so the result holds for every model on the image convergence: steps: 20 result: pass correlation: 0.9999729 final_loss_difference: 7.7e-04 pointwise_failures: 0 note: 'candidate arm: the measured run; reference arm: a 20-step run of the reference recipe on the same image and rows; the parity probe off in both; both with patches/fla_blackwell_guard.py' rules: argument classes of ATTUNE's skill/scripts/correctness/rules/numerics_list.yaml (amazon-science/attune f2cb450), with CPU optimizer offload admitted by the B200 component test above; 20-step check by Megatron-Bridge validate_convergence with its pipeline thresholds env: CUDA_DEVICE_MAX_CONNECTIONS: '1' PYTORCH_CUDA_ALLOC_CONF: expandable_segments:True TORCHINDUCTOR_COMPILE_THREADS: '1' NCCL_NVLS_ENABLE: $([ "$NVLINK_COUNT" -gt 0 ] && echo 1 || echo 0) patches: - patches/fla_blackwell_guard.py run: | export CUDA_DEVICE_MAX_CONNECTIONS=1 export PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True export TORCHINDUCTOR_COMPILE_THREADS=1 export NCCL_NVLS_ENABLE=$([ "$NVLINK_COUNT" -gt 0 ] && echo 1 || echo 0) python3 -m --async \ --actor-num-nodes 1 \ --actor-num-gpus-per-node 8 \ ${MODEL_ARGS[@]} \ --hf-checkpoint "$HF_CKPT" \ --ref-load "$TORCH_DIST" \ --rollout-function-path pretokenized_sft_rollout.generate_rollout \ --prompt-data "$SFT_DATA" \ --input-key messages \ --metadata-key metadata \ --num-rollout 20 \ --rollout-batch-size 128 \ --global-batch-size 128 \ --loss-type sft_loss \ --loss-mask-type qwen3_5 \ --calculate-per-token-loss \ --disable-compute-advantages-and-returns \ --debug-train-only \ --tensor-model-parallel-size 1 \ --pipeline-model-parallel-size 8 \ --decoder-first-pipeline-num-layers 7 \ --decoder-last-pipeline-num-layers 3 \ --context-parallel-size 1 \ --expert-model-parallel-size 1 \ --expert-tensor-parallel-size 1 \ --sequence-parallel \ --recompute-granularity full \ --recompute-method block \ --recompute-num-layers 2 \ --use-dynamic-batch-size \ --max-tokens-per-gpu 8192 \ --optimizer adam --lr 1e-5 --lr-decay-style cosine --min-lr 1e-6 \ --lr-warmup-fraction 0.1 --weight-decay 0.1 --adam-beta1 0.9 --adam-beta2 0.98 \ --use-distributed-optimizer \ --use-precision-aware-optimizer \ --attention-dropout 0.0 --hidden-dropout 0.0 \ --accumulate-allreduce-grads-in-fp32 --attention-softmax-in-fp32 \ --attention-backend flash \ --distributed-timeout-minutes 30 \ --tensorboard-dir "$PROFILE_DIR" notes: | - Starting point: slime's Qwen3.5-27B script, `scripts/run-qwen3.5-27B.sh` with `scripts/models/qwen3.5-27B.sh` at slime 0.3.0. slime ships it as a 4-node GRPO script and publishes no tokens/s figure for it. Ported to this workload on one 8-GPU node with its layout (TP 4 with sequence parallel, PP 2 with 34/30 layers, CP 1 in place of its CP 4, full recompute, Adam on the CPU), it measured 1,434.4 tokens/s/GPU on 8 B200; ATTUNE tuned it to 4,330.9 (+202%). - Recompute is at the memory limit: the tuning run with block recompute of 1 layer ran out of memory on the second stage (164.1 GiB allocated of 178 GiB). - Step 0 of the measured run took 718 s, against about 30 s for each later step. - B200 runs of this model need `patches/fla_blackwell_guard.py`: with the image's Triton 3.6, FLA 0.4.1's autotuner picks gated-delta-net kernel configs that race on Blackwell and give a different, wrong result on every call, and the patch keeps those kernels on the configs FLA main uses there. Every training process imports it before FLA (patches/README.md) and prints a `[fla-guard]` line.