Replace the copy+merge_adapter pair in refresh_lora_merge_from_pristine with one torch.addmm(pristine, B, A, alpha=scaling, out=W_inf) per LoraLayer, then set merged_adapters directly so PEFT's forward short-circuits to base_layer(x). Previously each refresh did two passes per weight: a bf16 copy from pristine, then PEFT merge_adapter which materialises a full [out, in] fp32 delta via get_delta_weight and in-place adds it back. The fused path skips the transient delta allocation and runs one cuBLAS GEMM instead. cuBLAS accumulates the bf16 matmul in fp32 internally, so the numerical result stays within 1 bf16 ULP of PEFT's path (verified on the rank-32 Qwen3-4B adapter: max abs diff 1.22e-04). DoRA, fan_in_fan_out, and lora_bias=True layers fall back to PEFT's get_delta_weight/merge path via a single trailing merge_adapter call after restoring their base_layer.weight from pristine. rslora is not a fallback -- PEFT folds alpha/sqrt(r) into module.scaling[adapter], so the fused addmm picks it up transparently via alpha=. Drift verification still passes: base bit-identical across 10 perturb+refresh cycles, inference state deterministic after LoRA restore. |
||
|---|---|---|
| .. | ||
| results | ||
| cb_sync_driver.py | ||
| cb_vs_vllm_generation.py | ||
| compare_grpo_runs.py | ||
| flash_attn_fa4_shim.py | ||
| flex_autotune_replay.py | ||
| flex_paged_attention.py | ||
| make_lora_adapter.py | ||
| persistent_cb.py | ||
| qwen3_flex_inference.py | ||
| qwen3_grpo_naive.py | ||
| qwen3_grpo_notebook.py | ||
| qwen3_grpo_tpaged.py | ||
| qwen3_grpo_unified.py | ||
| qwen3_grpo_vllm.py | ||
| README.md | ||
| unsloth_grpo_common.py | ||
Qwen3-4B GRPO rollout engine benchmarks
This directory holds the reproducible scripts behind the experiment
documented in the accompanying PR: can Hugging Face transformers'
continuous-batching API (model.generate_batch, backed by
PagedAttentionCache) serve as a drop-in replacement for vLLM during GRPO
rollouts on the Qwen3-4B notebook?
The short answer on a single NVIDIA B200 with Unsloth Qwen3-4B-Base, LoRA
rank 32, bf16: transformers continuous batching is functionally correct and
integrates with TRL's use_transformers_paged=True path, but end-to-end
throughput lands at around 7 to 10 percent of vLLM colocated even after
wiring in Flash Attention 4 on Blackwell. Full numbers and per-step timings
are in the PR description.
Files
| File | Purpose |
|---|---|
unsloth_grpo_common.py |
Shared dataset loading, reward functions, and GRPO hyperparameters |
qwen3_grpo_vllm.py |
vLLM baseline training entry (fast_inference=True, use_vllm=True, vllm_mode="colocate") |
qwen3_grpo_naive.py |
Naive TRL path (vanilla HF model.generate, no vLLM, no CB) matching https://huggingface.co/docs/trl/grpo_trainer |
qwen3_grpo_tpaged.py |
Continuous-batching candidate (fast_inference=False, use_transformers_paged=True, vanilla HF + PEFT). Supports --persistent_cb |
cb_vs_vllm_generation.py |
Standalone generation microbenchmark across both engines; supports --attn_impl, --persistent_cb |
flash_attn_fa4_shim.py |
Installs two monkey-patches that let CB dispatch to FA4 when --attn_impl flash_attention_2 is selected |
persistent_cb.py |
Replaces model.generate_batch with a version that reuses a single ContinuousBatchingManager |
Flash Attention 4 on Blackwell (sm_100)
The CB code path has three attention implementations: eager_paged,
sdpa_paged, flash_attention_2. The last one requires the legacy
flash_attn Python package, which does not install cleanly on B200 today:
flash_attn==2.8.3+cu12torch2.8cxx11abiTRUE-cp313from the Dao-AILab releases hitsundefined symbol: _ZNK3c106SymInt6sym_neERKS0_on torch 2.9.1 (ABI drift between torch 2.8 and 2.9).flash_attn_3-3.0.0-cp39-abi3-manylinux_2_28_x86_64.whlfrom the PyTorch wheel index installs but was built for sm_80 and sm_90a only. B200 is sm_100. The kernel call fails with "no kernel image is available for execution on the device".flash-attn-4==4.0.0b9(pure Python CuTeDSL, Dao-AILab) works on B200. It exposesflash_attn.cute.flash_attn_varlen_func.
The recipe this repo uses:
uv pip install --no-deps flash-attn-4==4.0.0b9
plus a tiny site-packages shim that re-exports FA4 symbols under the
FA2 flash_attn namespace so transformers' is_flash_attn_2_available() and
_lazy_imports("flash_attention_2") succeed. The shim lives out of tree in
lib/python3.13/site-packages/flash_attn/__init__.py +
flash_attn/bert_padding.py + a flash_attn-2.8.3.dist-info/ directory
with enough metadata to satisfy importlib.metadata.version("flash_attn").
On top of that, flash_attn_fa4_shim.py monkey-patches two rough edges in
the CB to FA integration that are unrelated to which FA version you use:
ContinuousBatchProcessor.return_attention_maskreturnsFalseforflash_attention_2/flash_attention_3so CB does not emit a 4-D paged attention mask that breaks_flash_attention_forward's_upad_inputbranch._flash_attention_forwardacceptsmax_seqlen_q/max_seqlen_kas aliases formax_length_q/max_length_k. Without this rename CB's model kwargs never bind and FA is called withmax_seqlen_q=None.
Reproduce
pip install unsloth "transformers>=4.57" "trl>=0.25" peft vllm
uv pip install --no-deps flash-attn-4==4.0.0b9
# Generation microbenchmark (32 prompts, 512 new tokens each)
CUDA_VISIBLE_DEVICES=2 python scripts/benchmarks/cb_vs_vllm_generation.py \
--backend vllm --stats_path logs/vllm_gen.json \
--n_prompts 32 --n_rounds 2 --max_new_tokens 512 \
--gpu_memory_utilization 0.6
# CB variants
CUDA_VISIBLE_DEVICES=6 python scripts/benchmarks/cb_vs_vllm_generation.py \
--backend tpaged --attn_impl sdpa \
--stats_path logs/cb_gen_sdpa.json \
--n_prompts 32 --n_rounds 2 --max_new_tokens 512
CUDA_VISIBLE_DEVICES=6 python scripts/benchmarks/cb_vs_vllm_generation.py \
--backend tpaged --attn_impl flash_attention_2 \
--stats_path logs/cb_gen_fa.json \
--n_prompts 32 --n_rounds 2 --max_new_tokens 512
# Full GRPO training (20 steps)
CUDA_VISIBLE_DEVICES=2 python scripts/benchmarks/qwen3_grpo_vllm.py \
--max_steps 20 --num_generations 2 --per_device_train_batch_size 2 \
--output_dir outputs/grpo_vllm --stats_path logs/vllm_stats.json \
--gpu_memory_utilization 0.6
CUDA_VISIBLE_DEVICES=7 python scripts/benchmarks/qwen3_grpo_naive.py \
--max_steps 20 --num_generations 2 --per_device_train_batch_size 2 \
--output_dir outputs/grpo_naive --stats_path logs/naive_stats.json
CUDA_VISIBLE_DEVICES=6 python scripts/benchmarks/qwen3_grpo_tpaged.py \
--max_steps 20 --num_generations 2 --per_device_train_batch_size 2 \
--attn_impl flash_attention_2 \
--output_dir outputs/grpo_tpaged_fa --stats_path logs/tpaged_stats_fa.json \
--max_batch_tokens 16384 --num_blocks 16384
Known integration notes for transformers continuous batching + TRL + Unsloth
These are the sharp edges you hit going down the continuous-batching path and
how qwen3_grpo_tpaged.py handles them:
-
top_k=-1is not a valid value for transformers. vLLM treats-1as "disabled", butTopKLogitsWarperraisesValueError: top_k has to be a strictly positive integer. The script rewritestop_k=Noneon the shared GRPOConfig before handing it toGRPOConfig(use_transformers_paged=True, ...). -
PagedAttentionCachedefault upper bounds are extremely conservative._upper_bound_max_batch_tokens=256and_upper_bound_num_blocks=4096choke decode throughput. The script passesgeneration_kwargs={"max_batch_tokens": 16384, "num_blocks": 16384}which TRL forwards toGenerationConfig, and the CB manager reads them when sizing the paged cache. -
Unsloth's
Qwen3Attention_fast_forwardbypasses the functional attention interface. Callingmodel.generate_batchon an Unsloth-patched Qwen3 model fails insideunsloth.utils.attention_dispatch.run_attentionbecause Unsloth routes through its own dispatcher rather than readingconfig._attn_implementation. The benchmark script works around this by loading a vanilla HF Qwen3 with PEFT LoRA for the tpaged and naive paths. This costs the Unsloth training kernels but keeps the comparison clean. A proper upstream fix is to detectconfig._attn_implementationbeingflash_attention_2/sdpa_paged/eager_pagedand delegate to the stock transformers forward. -
TRL imports
GuidedDecodingParamsfromvllm.sampling_params. Newer vLLM releases (>= 0.13) have moved or removed that symbol, sotrl.trainer.grpo_trainerfails to import on a fresh vLLM install even if you are not using vLLM.qwen3_grpo_tpaged.pyinstalls a minimal shim before importing TRL. -
UnslothGRPOTrainercallsmodel.for_training()/for_inference(). Importingunslothreplacestrl.GRPOTrainerwithUnslothGRPOTrainer, which assumes the model has these hooks. A vanilla HF model does not, soqwen3_grpo_tpaged.pydoes notimport unslothat all.
Why continuous batching is still slower than vLLM on this workload
ContinuousBatchingManagerdoes not yet implement CUDA graphs (use_cuda_graph=TrueraisesNotImplementedError). vLLM captures 100+ mixed prefill-decode and decode graphs during warmup.- CB re-allocates a fresh
PagedAttentionCacheon everygenerate_batchcall. For GRPO that is once per step.--persistent_cb(viapersistent_cb.py) keeps the cache warm across steps. - FA4 is a CuTeDSL package: first call per shape pays a one-time JIT-compile.
- vLLM uses its own colocated attention + FlashInfer / TRTLLM kernels tuned for decode, which currently outperform everything a generic CB path can do.
These are all upstream transformers issues, not Unsloth issues. The scripts in this directory are intentionally simple so they are easy to port into a future upstream fix.