EngineX-Ascend/enginex-ascend-910-vllm
11
0
Fork 1
Code Issues Pull Requests Actions Projects Releases Wiki Activity
Files
v0.23.0
enginex-ascend-910-vllm/tests/e2e/coverage.md
Sun Ruoxi 7f8a1b1f7a init v0.23.0 
Signed-off-by: Sun Ruoxi <sunruoxi@4paradigm.com>
2026-08-27 15:11:51 +08:00

116 KiB
Raw Permalink Blame History

The coverage of e2e is as follows:

1-Card Tests

Test file Test method Model 310P Dense MoE Embedding Classification Reranker Mamba/SSM Multimodal Reasoning TP PP EP PCP DCP Context Parallel EPLB Dynamic EPLB Multistream MoE Full Graph Full Decode Only Graph Default FULL_AND_PIECEWISE Graph Piecewise Graph Eager Mode PD disaggregation W8A8 W4A8 FP16 LoRA Multi-LoRA Runtime LoRA updating Fully sharded LoRA parameterization Spec Decode MTP Eagle-3 SFA/DSA DSA CP Pooling runner Score API Classification API Distributed executor mp Flash Attention 3 FIA comparison Chunked Prefill Prefix Caching CPU/KV offloading KV transfer/events Sleep/Wake memory Xlite Graph CP KV Interleave Long Sequence FlashComm1 env Skipped Conditional skip Logprobs Batch inference Mixed lengths
_310p/test_classification_310p.py test_qwen_pooling_classify_correctness Howeee/Qwen2.5-1.5B-apeach ✅ ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_dense_model_310p.py test_qwen3_5_dense_tp1_fp16 Qwen/Qwen3.5-4B ✅ ✅ ✅ ✅ ✅
_310p/test_dense_model_310p.py test_qwen3_5_dense_tp1_fp16_aclgraph Qwen/Qwen3.5-4B ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_dense_model_310p.py test_qwen3_dense_tp1_fp16 Qwen/Qwen3-8B ✅ ✅ ✅ ✅ ✅
_310p/test_dense_model_310p.py test_qwen3_dense_tp1_fp16_aclgraph Qwen/Qwen3-8B ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_dense_model_310p.py test_qwen3_dense_tp1_w8a8 vllm-ascend/Qwen3-8B-W8A8 ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_embedding_310p.py test_bge_m3_correctness BAAI/bge-m3 ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_embedding_310p.py test_embed_models_correctness Qwen/Qwen3-Embedding-0.6B
intfloat/multilingual-e5-small		 ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_scoring_310p.py test_cross_encoder_score_1_to_1 BAAI/bge-reranker-v2-m3 ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_scoring_310p.py test_cross_encoder_score_1_to_N BAAI/bge-reranker-v2-m3 ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_scoring_310p.py test_cross_encoder_score_N_to_N BAAI/bge-reranker-v2-m3 ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_vl_model_310p.py test_qwen3_vl_8b_tp1_fp16 Qwen/Qwen3-VL-8B-Instruct ✅ ✅ ✅ ✅
aclgraph/test_aclgraph_accuracy.py test_default_full_and_piecewise_res_consistency - ✅
aclgraph/test_aclgraph_accuracy.py test_full_decode_only_res_consistency - ✅
aclgraph/test_aclgraph_accuracy.py test_full_res_consistency - ✅
aclgraph/test_aclgraph_accuracy.py test_npugraph_ex_res_consistency - ✅
aclgraph/test_aclgraph_accuracy.py test_npugraph_ex_with_static_kernel - ✅
aclgraph/test_aclgraph_batch_invariant.py test_aclgraph_logprobs_bitwise_batch_invariance_bs1_vs_bsN - ✅ ✅ ✅ ✅ ✅
aclgraph/test_aclgraph_batch_invariant.py test_aclgraph_logprobs_without_batch_invariance_should_fail - ✅ ✅ ✅ ✅ ✅
aclgraph/test_aclgraph_batch_invariant.py test_aclgraph_simple_generation - ✅ ✅ ✅ ✅
aclgraph/test_aclgraph_batch_invariant.py test_aclgraph_v1_generation_is_deterministic_across_batch_sizes_with_needle - ✅ ✅ ✅ ✅
aclgraph/test_aclgraph_mem.py test_aclgraph_mem_use vllm-ascend/DeepSeek-V2-Lite-W8A8 ✅ ✅ ✅ ✅
compile/test_graphex_norm_quant_fusion.py test_rmsnorm_quant_fusion - ✅
compile/test_graphex_qknorm_rope_fusion.py test_rmsnorm_quant_fusion - ✅
compile/test_norm_quant_fusion.py test_rmsnorm_quant_fusion - ✅
lora/test_ilama_lora.py test_ilama_lora - ✅ ✅ ✅ ✅
lora/test_llama32_lora.py test_llama_lora - ✅ ✅ ✅
lora/test_lora_with_spec_decode.py test_batch_inference_correctness - ✅ ✅ ✅ ✅
lora/test_qwen35_densemodel_lora.py test_qwen35_text_lora - ✅
lora/test_qwen3_multi_loras.py test_multi_loras_with_tp_sync - ✅ ✅ ✅ ✅
lora/test_qwen3_reranker_lora.py test_reranker_models_lora - ✅ ✅ ✅
model_runner_v2/test_basic.py test_egale_spec_decoding Qwen/Qwen3-0.6B
vllm-ascend/DeepSeek-V2-Lite-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅ ✅
model_runner_v2/test_basic.py test_qwen3_dense_eager_mode Qwen/Qwen3-0.6B
vllm-ascend/DeepSeek-V2-Lite-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅ ✅
model_runner_v2/test_basic.py test_qwen3_dense_graph_mode Qwen/Qwen3-0.6B ✅ ✅ ✅ ✅
pooling/test_classification.py test_qwen_pooling_classify_correctness Howeee/Qwen2.5-1.5B-apeach ✅ ✅ ✅ ✅ ✅
pooling/test_embedding.py test_bge_m3_correctness BAAI/bge-m3 ✅ ✅ ✅ ✅
pooling/test_embedding.py test_causal_embed_models_using_prefix_caching_correctness Qwen/Qwen3-Embedding-0.6B ✅ ✅ ✅ ✅ ✅
pooling/test_embedding.py test_embed_models_correctness Qwen/Qwen3-Embedding-0.6B
intfloat/multilingual-e5-small		 ✅ ✅ ✅ ✅
pooling/test_scoring.py test_cross_encoder_score_1_to_1 BAAI/bge-reranker-v2-m3
dengcao/ms-marco-MiniLM-L6-v2
sentence-transformers/all-MiniLM-L12-v2		 ✅ ✅ ✅ ✅ ✅
pooling/test_scoring.py test_cross_encoder_score_1_to_N BAAI/bge-reranker-v2-m3
dengcao/ms-marco-MiniLM-L6-v2
sentence-transformers/all-MiniLM-L12-v2		 ✅ ✅ ✅ ✅ ✅
pooling/test_scoring.py test_cross_encoder_score_N_to_N BAAI/bge-reranker-v2-m3
dengcao/ms-marco-MiniLM-L6-v2
sentence-transformers/all-MiniLM-L12-v2		 ✅ ✅ ✅ ✅ ✅
pooling/test_scoring.py test_embedding_score_1_to_1 BAAI/bge-reranker-v2-m3
dengcao/ms-marco-MiniLM-L6-v2
sentence-transformers/all-MiniLM-L12-v2		 ✅ ✅ ✅ ✅ ✅
pooling/test_scoring.py test_embedding_score_1_to_N BAAI/bge-reranker-v2-m3
dengcao/ms-marco-MiniLM-L6-v2
sentence-transformers/all-MiniLM-L12-v2		 ✅ ✅ ✅ ✅ ✅
pooling/test_scoring.py test_embedding_score_N_to_N BAAI/bge-reranker-v2-m3
dengcao/ms-marco-MiniLM-L6-v2
sentence-transformers/all-MiniLM-L12-v2		 ✅ ✅ ✅ ✅ ✅
spec_decode/test_dflash.py test_dflash_acceptance - ✅ ✅ ✅ ✅ ✅
spec_decode/test_draft_parallel.py test_parallel_drafting_acceptance - ✅ ✅ ✅ ✅ ✅
spec_decode/test_eagle.py test_qwen3_vl_eagle - ✅
spec_decode/test_eagle.py test_qwen_eagle3_acceptance - ✅ ✅ ✅ ✅
spec_decode/test_extract_hidden_states.py test_extract_hidden_states_aclgraph_mode - ✅ ✅ ✅ ✅ ✅
spec_decode/test_extract_hidden_states.py test_extract_hidden_states_eager_mode - ✅ ✅ ✅ ✅ ✅
spec_decode/test_mtp_eagle_correctness.py test_deepseek_mtp wemaster/deepseek_mtp_main_random_bf16 ✅ ✅ ✅ ✅ ✅ ✅ ✅
spec_decode/test_ngram.py test_ngram - ✅ ✅
spec_decode/test_ngram_npu.py test_ngram_npu_async_acceptance - ✅ ✅ ✅ ✅
spec_decode/test_suffix.py test_suffix_acceptance - ✅ ✅
test_attention_fa3.py test_fa3_vs_fia_logprobs - ✅ ✅
test_attention_fa3.py test_fa3_vs_fia_mixed_lengths - ✅ ✅ ✅ ✅
test_attention_fa3.py test_fa3_vs_fia_single_prompt - ✅ ✅
test_attention_fa3.py test_fa3_vs_fia_with_chunkprefill - ✅ ✅
test_batch_invariant.py test_logprobs_bitwise_batch_invariance_bs1_vs_bsN - ✅ ✅ ✅ ✅ ✅
test_batch_invariant.py test_logprobs_without_batch_invariance_should_fail - ✅ ✅ ✅ ✅ ✅
test_batch_invariant.py test_simple_generation - ✅ ✅ ✅ ✅
test_batch_invariant.py test_v1_generation_is_deterministic_across_batch_sizes_with_needle - ✅ ✅ ✅ ✅
test_camem.py test_end_to_end Qwen/Qwen3-0.6B ✅ ✅ ✅
test_completion_with_prompt_embeds.py test_mixed_prompt_embeds_and_text Qwen/Qwen3-0.6B ✅ ✅ ✅
test_cpu_offloading.py test_cpu_offloading Qwen/Qwen3-0.6B ✅ ✅ ✅
test_guided_decoding.py test_guided_json_completion - ✅
test_guided_decoding.py test_guided_regex - ✅
test_minicpm.py test_minicpm OpenBMB/MiniCPM4-0.5B
openbmb/MiniCPM-2B-sft-bf16		 ✅ ✅
test_multi_instance.py test_two_instances_on_single_card - ✅
test_multistream_overlap_shared_expert.py test_models_with_multistream_overlap_shared_expert vllm-ascend/DeepSeek-V2-Lite-W8A8 ✅ ✅ ✅ ✅ ✅
test_qwen3_0_6b.py test_dense_default_full_and_piecewise_graph Qwen/Qwen3-0.6B ✅ ✅
test_qwen3_5_0_8b.py test_mamba_ssm_multimodal_reasoning_mtp_full_decode_only Qwen/Qwen3.5-0.8B ✅ ✅ ✅ ✅ ✅
test_qwen3_8b_w8a8.py test_dense_w8a8_eagle3_full_graph RedHatAI/Qwen3-8B-speculator.eagle3
vllm-ascend/Qwen3-8B-W8A8		 ✅ ✅ ✅ ✅ ✅
test_qwen3_embedding_0_6b.py test_embedding_full_decode_only Qwen/Qwen3-Embedding-0.6B ✅ ✅ ✅ ✅
test_sampler.py test_qwen3_exponential_overlap Qwen/Qwen3-0.6B ✅ ✅
test_sampler.py test_qwen3_prompt_logprobs Qwen/Qwen3-0.6B ✅ ✅ ✅
test_sampler.py test_qwen3_topk Qwen/Qwen3-0.6B ✅ ✅
test_vlm.py test_multimodal_audio Qwen/Qwen2-Audio-7B-Instruct ✅ ✅ ✅
test_vlm.py test_multimodal_vl openai-mirror/whisper-large-v3-turbo ✅ ✅ ✅
test_vlm.py test_multimodal_vl_language_model_only Qwen/Qwen3-VL-8B-Instruct ✅ ✅ ✅
test_vlm.py test_whisper openai-mirror/whisper-large-v3-turbo ✅ ✅ ✅
test_xlite.py test_models_with_xlite_decode_only - ✅ ✅
test_xlite.py test_models_with_xlite_full_mode - ✅

2-Card Tests

Test file Test method Model 310P Dense MoE Embedding Classification Reranker Mamba/SSM Multimodal Reasoning TP PP EP PCP DCP Context Parallel EPLB Dynamic EPLB Multistream MoE Full Graph Full Decode Only Graph Default FULL_AND_PIECEWISE Graph Piecewise Graph Eager Mode PD disaggregation W8A8 W4A8 FP16 LoRA Multi-LoRA Runtime LoRA updating Fully sharded LoRA parameterization Spec Decode MTP Eagle-3 SFA/DSA DSA CP Pooling runner Score API Classification API Distributed executor mp Flash Attention 3 FIA comparison Chunked Prefill Prefix Caching CPU/KV offloading KV transfer/events Sleep/Wake memory Xlite Graph CP KV Interleave Long Sequence FlashComm1 env Skipped Conditional skip Logprobs Batch inference Mixed lengths
aclgraph/test_aclgraph_capture_replay.py test_models_aclgraph_capture_replay_metrics_dp2 vllm-ascend/DeepSeek-V2-Lite-W8A8 ✅ ✅ ✅ ✅
aclgraph/test_full_graph_mode.py test_qwen3_moe_full_decode_only_tp2 Qwen/Qwen3-30B-A3B ✅ ✅ ✅ ✅
aclgraph/test_full_graph_mode.py test_qwen3_moe_full_graph_tp2 Qwen/Qwen3-30B-A3B ✅ ✅ ✅ ✅
lora/test_ilama_lora_tp2.py test_ilama_lora_tp2 - ✅ ✅ ✅ ✅ ✅
lora/test_llama32_lora_tp2.py test_llama_lora_tp2 - ✅ ✅ ✅ ✅
spec_decode/test_spec_decode.py test_eagle3_sp_acceptance - ✅ ✅ ✅ ✅ ✅ ✅
spec_decode/test_spec_decode.py test_p_eagle_acceptance - ✅ ✅ ✅ ✅ ✅ ✅
spec_decode/test_spec_decode.py test_qwen3_eagle3_pcp2_tp1 - ✅ ✅ ✅ ✅ ✅ ✅
test_data_parallel.py test_qwen3_inference_dp2 Qwen/Qwen3-30B-A3B
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅
test_deepseek_multistream_moe.py test_deepseek_multistream_moe_tp2 vllm-ascend/DeepSeek-V3-Pruning ✅ ✅ ✅ ✅ ✅
test_external_launcher.py test_qwen3_external_launcher Qwen/Qwen3-0.6B ✅ ✅ ✅
test_external_launcher.py test_qwen3_external_launcher_with_matmul_allreduce Qwen/Qwen3-0.6B ✅ ✅ ✅ ✅
test_external_launcher.py test_qwen3_external_launcher_with_sleepmode Qwen/Qwen3-8B ✅ ✅ ✅
test_external_launcher.py test_qwen3_external_launcher_with_sleepmode_level2 Qwen/Qwen3-8B ✅ ✅ ✅
test_external_launcher.py test_qwen3_moe_external_launcher_ep_tp2 Qwen/Qwen3-0.6B ✅ ✅ ✅
test_flashcomm_distributed.py test_deepseek_v2_lite_fc1_tp2 vllm-ascend/DeepSeek-V2-Lite-W8A8 ✅ ✅ ✅ ✅ ✅ ✅
test_flashcomm_distributed.py test_qwen3_dense_fc1_tp2 - ✅ ✅
test_flashcomm_distributed.py test_qwen3_dense_prefetch_mlp_weight_tp2 - ✅ ✅
test_flashcomm_distributed.py test_qwen3_moe_fc2_oshard_tp2 Qwen/Qwen3-30B-A3B ✅ ✅ ✅ ✅ ✅ ✅
test_gpt_oss_distributed.py test_gpt_oss_distributed_tp2 - ✅ ✅
test_moe_routing_replay.py test_qwen3_moe_routing_replay Qwen/Qwen3-30B-A3B
Qwen/Qwen3.5-35B-A3B		 ✅ ✅ ✅ ✅ ✅ ✅
test_offline_weight_load.py test_qwen3_offline_load_and_sleepmode_tp2 Qwen/Qwen3-30B-A3B ✅ ✅ ✅
test_prefix_caching.py test_models_prefix_cache_tp2 Qwen/Qwen3-8B
deepseek-ai/DeepSeek-V2-Lite-Chat		 ✅ ✅ ✅ ✅ ✅
test_qwen3_30b_a3b.py test_moe_tp_ep_eplb_full_decode_only Qwen/Qwen3-30B-A3B ✅ ✅ ✅
test_qwen3_6_27b_fia.py test_qwen3_6_27b_multimodel_fia_eager Qwen/Qwen3.6-27B/ ✅ ✅ ✅ ✅ ✅
test_qwen3_6_27b_fia.py test_qwen3_6_27b_multimodel_fia_acl_graph Qwen/Qwen3.6-27B/ ✅ ✅ ✅ ✅ ✅
test_qwen3_vl_30b_a3b_instruct.py test_multimodal_reasoning_pp_full_decode_only Qwen/Qwen3-VL-30B-A3B-Instruct ✅ ✅ ✅ ✅ ✅
test_sequence_parallelism_moe.py test_sequence_parallelism_moe_patterns - ✅
test_shared_expert_dp.py test_deepseek_v2_lite_enable_shared_expert_dp_tp2 deepseek-ai/DeepSeek-V2-Lite ✅ ✅
test_sp_pass.py test_qwen3_vl_sp_tp2 Qwen/Qwen3-VL-2B-Instruct ✅ ✅ ✅ ✅

4-Card Tests

Test file Test method Model 310P Dense MoE Embedding Classification Reranker Mamba/SSM Multimodal Reasoning TP PP EP PCP DCP Context Parallel EPLB Dynamic EPLB Multistream MoE Full Graph Full Decode Only Graph Default FULL_AND_PIECEWISE Graph Piecewise Graph Eager Mode PD disaggregation W8A8 W4A8 FP16 LoRA Multi-LoRA Runtime LoRA updating Fully sharded LoRA parameterization Spec Decode MTP Eagle-3 SFA/DSA DSA CP Pooling runner Score API Classification API Distributed executor mp Flash Attention 3 FIA comparison Chunked Prefill Prefix Caching CPU/KV offloading KV transfer/events Sleep/Wake memory Xlite Graph CP KV Interleave Long Sequence FlashComm1 env Skipped Conditional skip Logprobs Batch inference Mixed lengths
_310p/test_dense_model_310p.py test_qwen3_dense_tp2_fp16 Qwen/Qwen3-8B ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_dense_model_310p.py test_qwen3_dense_tp4_w8a8 vllm-ascend/Qwen3-32B-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_moe_model_310p.py test_qwen3_5_moe_tp4_fp16 Qwen/Qwen3.5-35B-A3B ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_moe_model_310p.py test_qwen3_moe_tp2_w8a8 vllm-ascend/Qwen3-30B-A3B-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_moe_model_310p.py test_qwen3_moe_tp4_fp16 Qwen/Qwen3-30B-A3B ✅ ✅ ✅ ✅ ✅ ✅ ✅
_310p/test_vl_model_310p.py test_qwen3_vl_8b_tp2_fp16 Qwen/Qwen3-VL-8B-Instruct ✅ ✅ ✅ ✅
context_parallel/test_accuracy.py test_accuracy_dcp_only_eager Qwen/Qwen3-8B
vllm-ascend/DeepSeek-V2-Lite-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_accuracy.py test_accuracy_dcp_only_graph Qwen/Qwen3-8B
vllm-ascend/DeepSeek-V2-Lite-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_accuracy.py test_accuracy_pcp_only Qwen/Qwen3-8B
vllm-ascend/DeepSeek-V2-Lite-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_accuracy.py test_models_long_sequence_cp_kv_interleave_size_output_between_tp_and_cp vllm-ascend/DeepSeek-V2-Lite-W8A8 ✅ ✅ ✅ ✅
context_parallel/test_accuracy.py test_models_long_sequence_output_between_tp_and_cp vllm-ascend/DeepSeek-V2-Lite-W8A8 ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_dcp_basic deepseek-ai/DeepSeek-V2-Lite-Chat ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_dcp_full_graph deepseek-ai/DeepSeek-V2-Lite-Chat ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_dcp_piece_wise deepseek-ai/DeepSeek-V2-Lite-Chat ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_deepseek_v4_w4a8_dsa_cp_basic_greedy gdydems/DeepSeek-V4-Flash-w4a8-mtp ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_models_pcp_dcp_basic Qwen/Qwen3-Next-80B-A3B-Instruct
deepseek-ai/DeepSeek-V2-Lite-Chat
vllm-ascend/DeepSeek-V3.2-W8A8-Pruning
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_models_pcp_dcp_full_graph deepseek-ai/DeepSeek-V2-Lite-Chat
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_models_pcp_dcp_piece_wise deepseek-ai/DeepSeek-V2-Lite-Chat
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_pcp_basic deepseek-ai/DeepSeek-V2-Lite-Chat ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_pcp_full_graph deepseek-ai/DeepSeek-V2-Lite-Chat ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_pcp_piece_wise deepseek-ai/DeepSeek-V2-Lite-Chat ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_qwen3_5_4b_multimodal_single_and_multi_image Qwen/Qwen3.5-4B ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_basic.py test_qwen3_vl_8b_multimodal_single_and_multi_image Qwen/Qwen3-VL-8B-Instruct ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_chunked_prefill_cp.py test_models_chunked_prefill_mixed_length_prompts_including_1_token vllm-ascend/DeepSeek-V2-Lite-W8A8
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_chunked_prefill_cp.py test_models_chunked_prefill_with_cp_basic vllm-ascend/DeepSeek-V2-Lite-W8A8
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_chunked_prefill_cp.py test_models_chunked_prefill_with_cp_default_full_and_piecewise vllm-ascend/DeepSeek-V2-Lite-W8A8
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_chunked_prefill_cp.py test_models_chunked_prefill_with_cp_full_graph vllm-ascend/DeepSeek-V2-Lite-W8A8
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_chunked_prefill_cp.py test_models_chunked_prefill_with_empty_kvcache vllm-ascend/DeepSeek-V2-Lite-W8A8
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_mtp.py test_dcp_mtp3_full_graph - ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_mtp.py test_pcp_dcp_mtp1_eager - ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_mtp.py test_pcp_dcp_mtp3_eager - ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_mtp.py test_pcp_dcp_mtp3_full_graph - ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_mtp.py test_pcp_dcp_mtp3_piecewise_graph - ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_mtp.py test_pcp_eagle3_eager - ✅ ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_prefix_caching_cp.py test_models_prefix_cache_with_cp_basic vllm-ascend/DeepSeek-V2-Lite-W8A8
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_prefix_caching_cp.py test_models_prefix_cache_with_cp_default_full_and_piecewise vllm-ascend/DeepSeek-V2-Lite-W8A8
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅
context_parallel/test_prefix_caching_cp.py test_models_prefix_cache_with_cp_full_graph vllm-ascend/DeepSeek-V2-Lite-W8A8
vllm-ascend/Qwen3-30B-A3B-W8A8		 ✅ ✅ ✅ ✅ ✅ ✅
spec_decode/test_mtp_qwen3_next.py test_qwen3_next_mtp_acceptance_tp4 Qwen/Qwen3-Next-80B-A3B-Instruct ✅ ✅ ✅ ✅ ✅ ✅ ✅
test_data_parallel_tp2.py test_qwen3_inference_dp2_tp2 Qwen/Qwen3-30B-A3B ✅ ✅ ✅
test_deepseek_v3_2_w8a8_pruning.py test_moe_w8a8_tp_pp_ep_full_decode_only vllm-ascend/DeepSeek-V3.2-W8A8-Pruning ✅ ✅ ✅ ✅ ✅ ✅
test_deepseek_v3_2_w8a8_pruning.py test_pd_disaggregation_w8a8_sfa_dsa_full_decode_only vllm-ascend/DeepSeek-V3.2-W8A8-Pruning ✅ ✅ ✅ ✅ ✅
test_deepseek_v4.py test_deepseek_v4_w4a8_tp4_basic_greedy gdydems/DeepSeek-V4-Flash-w4a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
test_deepseek_v4.py test_deepseek_v4_w4a8_tp4_index_cache_freq4 gdydems/DeepSeek-V4-Flash-w4a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
test_pipeline_parallel.py test_models_pp2_dp2 - ✅
test_pipeline_parallel.py test_models_pp2_tp2 - ✅
test_profiling_chunk_performance.py test_profiling_chunk_ttft_performance - ✅ ✅ ✅ ✅ ✅ ✅ ✅
test_qwen3_5.py test_qwen3_5_27b_distributed_mp_tp4 Qwen/Qwen3.5-27B ✅ ✅ ✅ ✅
test_qwen3_5.py test_qwen3_5_35b_distributed_mp_tp4 Qwen/Qwen3.5-35B-A3B ✅ ✅ ✅ ✅ ✅
test_qwen3_5.py test_qwen3_5_35b_distributed_mp_tp4_full_decode_only_mtp3 Qwen/Qwen3.5-35B-A3B ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
test_qwen3_5.py test_qwen3_5_35b_distributed_mp_tp4_full_decode_only_mtp3_flashcomm Qwen/Qwen3.5-35B-A3B ✅ ✅ ✅ ✅ ✅ ✅
test_qwen3_next.py test_qwen3_next_distributed_mp_flash_comm_tp4 Qwen/Qwen3-Next-80B-A3B-Instruct ✅ ✅ ✅ ✅ ✅
test_qwen3_next.py test_qwen3_next_distributed_mp_full_decode_only_tp4 Qwen/Qwen3-Next-80B-A3B-Instruct ✅ ✅ ✅ ✅
test_qwen3_next.py test_qwen3_next_distributed_mp_graph_mode_tp4 Qwen/Qwen3-Next-80B-A3B-Instruct ✅ ✅ ✅ ✅ ✅
test_qwen3_next.py test_qwen3_next_distributed_mp_tp4 Qwen/Qwen3-Next-80B-A3B-Instruct ✅ ✅ ✅ ✅
test_qwen3_next.py test_qwen3_next_w8a8dynamic_distributed_tp4_ep vllm-ascend/Qwen3-Next-80B-A3B-Instruct-W8A8 ✅ ✅ ✅ ✅ ✅

Nightly Tests

Test file Test method Model 310P Dense MoE Embedding Classification Reranker Mamba/SSM Multimodal Reasoning TP PP EP PCP DCP Context Parallel EPLB Dynamic EPLB Multistream MoE Full Graph Full Decode Only Graph Default FULL_AND_PIECEWISE Graph Piecewise Graph Eager Mode PD disaggregation W8A8 W4A8 FP16 LoRA Multi-LoRA Runtime LoRA updating Fully sharded LoRA parameterization Spec Decode MTP Eagle-3 SFA/DSA DSA CP Pooling runner Score API Classification API Distributed executor mp Flash Attention 3 FIA comparison Chunked Prefill Prefix Caching CPU/KV offloading KV transfer/events Sleep/Wake memory Xlite Graph CP KV Interleave Long Sequence FlashComm1 env Skipped Conditional skip Logprobs Batch inference Mixed lengths
310p/single_node/ops/singlecard_ops/test_recurrent_gated_delta_rule_v310.py test_recurrent_gated_delta_rule_v310 - ✅ ✅ ✅
multi_node/external_dp/config/GLM5_1-W8A8-EP-external.yaml multi-node-glm-5.1-w8a8-ep-external-dp Eco-Tech/GLM-5.1-w8a8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/external_dp/scripts/test_external_dp.py test_external_dp - ✅
multi_node/internal_dp/config/DeepSeek-R1-W8A8-EPLB.yaml test DeepSeek-R1-W8A8 disaggregated_prefill vllm-ascend/DeepSeek-R1-0528-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/DeepSeek-R1-W8A8-longseq.yaml test DeepSeek-R1-W8A8-longseq disaggregated_prefill vllm-ascend/DeepSeek-R1-0528-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/DeepSeek-R1-W8A8.yaml test DeepSeek-R1-W8A8 disaggregated_prefill vllm-ascend/DeepSeek-R1-0528-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/DeepSeek-V3.1-BF16.yaml test DeepSeek-V3.1-BF16 on A3 unsloth/DeepSeek-V3.1-BF16 ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/DeepSeek-V3_2-W8A8-A3-dual-nodes.yaml test DeepSeek-V3.2-W8A8 on A3 vllm-ascend/DeepSeek-V3.2-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/DeepSeek-V3_2-W8A8-EP.yaml test DeepSeek-V3.2-W8A8-EP disaggregated_prefill vllm-ascend/DeepSeek-V3.2-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/GLM5_1-W8A8-A2-dual-nodes.yaml multi-node-GLM-5.1-w8a8-A2 Eco-Tech/GLM-5.1-w8a8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/GLM5_1-W8A8-A3-dual-nodes.yaml multi-node-GLM-5.1-w8a8-A3 Eco-Tech/GLM-5.1-w8a8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/GLM5_1-W8A8-EP.yaml multi-node-GLM-5.1-w8a8-EP Eco-Tech/GLM-5.1-w8a8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/Kimi-K2_5-W4A8-A2-dual-nodes.yaml test Kimi-K2.5-W4A8 A2 dual nodes Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/Qwen3-235B-A22B-A2.yaml test Qwen3-235B-A22B multi-dp on A2 Qwen/Qwen3-235B-A22B ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/Qwen3-235B-A22B.yaml test Qwen3-235B-A22B multi-dp Qwen/Qwen3-235B-A22B ✅ ✅ ✅ ✅
multi_node/internal_dp/config/Qwen3-235B-W8A8-EPLB.yaml test Qwen3-235B-A22B-W8A8 disaggregated_prefill vllm-ascend/Qwen3-235B-A22B-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/Qwen3-235B-W8A8-longseq.yaml test Qwen3-235B-A22B-W8A8-longseq disaggregated_prefill vllm-ascend/Qwen3-235B-A22B-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/Qwen3-235B-W8A8.yaml test Qwen3-235B-A22B-W8A8 disaggregated_prefill vllm-ascend/Qwen3-235B-A22B-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/Qwen3-235B-disagg-pd.yaml test Qwen3-235B-A22B disaggregated_prefill Qwen/Qwen3-235B-A22B ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/Qwen3-VL-235B-disagg-pd.yaml test Qwen3-VL-235B-A22B disaggregated_prefill Qwen/Qwen3-VL-235B-A22B-Instruct ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/DeepSeek-R1-0528-W8A8.yaml DeepSeek-R1-0528-W8A8-EPLB vllm-ascend/DeepSeek-R1-0528-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/DeepSeek-R1-0528-W8A8.yaml DeepSeek-R1-0528-W8A8-aclgraph vllm-ascend/DeepSeek-R1-0528-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/DeepSeek-R1-0528-W8A8.yaml DeepSeek-R1-0528-W8A8-single vllm-ascend/DeepSeek-R1-0528-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/DeepSeek-V3.2-W8A8.yaml DeepSeek-V3.2-W8A8-TP8-DP2 vllm-ascend/DeepSeek-V3.2-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/DeepSeek-V4-Flash-W8A8-A3.yaml DeepSeek-V4-Flash-W8A8-A3 Eco-Tech/DeepSeek-V4-Flash-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/GLM-4.7.yaml GLM-4.7-TP8-DP2-decodegraph Eco-Tech/GLM-4.7-W8A8-floatmtp ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Hy3-preview.yaml Hy3-preview-TP16-EP-MTP Tencent-Hunyuan/Hy3-preview ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Kimi-K2-Thinking.yaml Kimi-K2-Thinking-TP16-Case moonshotai/Kimi-K2-Thinking ✅ ✅ ✅ ✅
single_node/models/configs/Kimi-K2.5.yaml Kimi-K2.5-W4A8-Case Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/MTPX-DeepSeek-R1-0528-W8A8.yaml MTPX-DeepSeek-R1-0528-W8A8-mtp2 vllm-ascend/DeepSeek-R1-0528-W8A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/MTPX-DeepSeek-R1-0528-W8A8.yaml MTPX-DeepSeek-R1-0528-W8A8-mtp3 vllm-ascend/DeepSeek-R1-0528-W8A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/MiniMax-M2.5-w8a8-QuaRot-A2.yaml MiniMax-M2.5-w8a8 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Prefix-Cache-DeepSeek-R1-0528-W8A8.yaml prefix-cache-deepseek-r1-0528-w8a8 vllm-ascend/DeepSeek-R1-0528-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Prefix-Cache-Qwen3-32B-Int8.yaml prefix-cache-qwen3-32b-w8a8 vllm-ascend/Qwen3-32B-W8A8 ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-235B-A22B-W8A8.yaml Qwen3-235B-A22B-W8A8-EPLB vllm-ascend/Qwen3-235B-A22B-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-235B-A22B-W8A8.yaml Qwen3-235B-A22B-W8A8-full_graph vllm-ascend/Qwen3-235B-A22B-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-235B-A22B-W8A8.yaml Qwen3-235B-A22B-W8A8-piecewise vllm-ascend/Qwen3-235B-A22B-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-30B-A3B-W4A8-llm-compressor.yaml Qwen3-30B-A3B-W4A8-llm-compressor vllm-ascend/Qwen3-30B-A3B-Instruct-2507-quantized.w4a8 ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-30B-A3B-W8A8.yaml Qwen3-30B-A3B-W8A8-TP1 vllm-ascend/Qwen3-30B-A3B-W8A8 ✅ ✅ ✅
single_node/models/configs/Qwen3-30B-QuaRot-eagle3.yaml Qwen3-30B-QuaRot vllm-ascend/Qwen3-30B-A3B-W8A8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-32B-Int8-A2.yaml Qwen3-32B-W8A8-aclgraph-a2 vllm-ascend/Qwen3-32B-W8A8 ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-32B-Int8-A2.yaml Qwen3-32B-W8A8-single-a2 vllm-ascend/Qwen3-32B-W8A8 ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-32B-Int8.yaml Qwen3-32B-W8A8-aclgraph-a3 vllm-ascend/Qwen3-32B-W8A8 ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-32B-Int8.yaml Qwen3-32B-W8A8-single-a3 vllm-ascend/Qwen3-32B-W8A8 ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-32B-QuaRot-eagle3.yaml Qwen3-32B-QuaRot vllm-ascend/Qwen3-32B-W8A8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-VL-235B-A22B-Instruct-W8A8.yaml Qwen3-VL-235B-A22B-Instruct-W8A8 Eco-Tech/Qwen3-VL-235B-A22B-Instruct-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3-VL-32B-Instruct-W8A8.yaml Qwen3-VL-32B-Instruct-W8A8 Eco-Tech/Qwen3-VL-32B-Instruct-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3.5-122B-A10B-W8A8-A3.yaml Qwen3.5-122B-A10B-W8A8-A3 Eco-Tech/Qwen3.5-122B-A10B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3.5-27B-w8a8-A2.yaml Qwen3.5-27B-w8a8 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3.yaml Qwen3.5-397B-A17B-w8a8-mtp Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/models/configs/Qwen3.5-397B-A17B-w4a8-mtp-A2.yaml Qwen3.5-397B-A17B-w4a8-mtp Eco-Tech/Qwen3.5-397B-A17B-w4a8-mtp ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/ops/multicard_ops_a2/test_matmul_allreduce_add_rmsnorm.py test_matmul_allreduce_add_rmsnorm_kernel - ✅
single_node/ops/multicard_ops_a3/test_dispatch_ffn_combine.py test_dispatch_ffn_combine_kernel - ✅
single_node/ops/multicard_ops_a3/test_dispatch_ffn_combine_bf16.py test_dispatch_ffn_combine_kernel - ✅
single_node/ops/multicard_ops_a3/test_dispatch_ffn_combine_w4a8.py test_dispatch_ffn_combine_kernel - ✅
single_node/ops/multicard_ops_a3/test_dispatch_gmm_combine_decode.py test_dispatch_gmm_combine_decode_base - ✅
single_node/ops/multicard_ops_a3/test_dispatch_gmm_combine_decode.py test_dispatch_gmm_combine_decode_dynamic_eplb - ✅
single_node/ops/multicard_ops_a3/test_dispatch_gmm_combine_decode.py test_dispatch_gmm_combine_decode_with_mc2_mask - ✅
single_node/ops/singlecard_ops/test_add_rms_norm_bias.py test_quant_fpx_linear - ✅
single_node/ops/singlecard_ops/test_batch_matmul_transpose.py test_boundary_conditions - ✅
single_node/ops/singlecard_ops/test_batch_matmul_transpose.py test_random_shapes - ✅
single_node/ops/singlecard_ops/test_batch_matmul_transpose.py test_zero_values - ✅
single_node/ops/singlecard_ops/test_bgmv_expand.py test_bgmv_expand - ✅
single_node/ops/singlecard_ops/test_bgmv_shrink.py test_bgmv_shrink - ✅
single_node/ops/singlecard_ops/test_causal_conv1d_310.py test_ascend_causal_conv1d_310_fn - ✅ ✅
single_node/ops/singlecard_ops/test_causal_conv1d_310.py test_causal_conv1d_310_update - ✅ ✅
single_node/ops/singlecard_ops/test_copy_and_expand_eagle_inputs.py test_copy_and_expand_eagle_inputs - ✅
single_node/ops/singlecard_ops/test_copy_and_expand_eagle_inputs.py test_large_tokens_per_request - ✅
single_node/ops/singlecard_ops/test_copy_and_expand_eagle_inputs.py test_large_tokens_shift_true - ✅
single_node/ops/singlecard_ops/test_copy_and_expand_eagle_inputs.py test_minimal_case - ✅
single_node/ops/singlecard_ops/test_copy_and_expand_eagle_inputs.py test_no_rejected_tokens - ✅
single_node/ops/singlecard_ops/test_dequant_swiglu_quant.py test_npu_dequant_swiglu_quant_with_limit - ✅
single_node/ops/singlecard_ops/test_fused_gdn_gating.py test_fused_gdn_gating_bulk_dma_alignment - ✅
single_node/ops/singlecard_ops/test_fused_gdn_gating.py test_fused_gdn_gating_extreme_large_batch - ✅ ✅
single_node/ops/singlecard_ops/test_fused_gdn_gating.py test_fused_gdn_gating_large_batch_multi_row - ✅ ✅
single_node/ops/singlecard_ops/test_fused_gdn_gating.py test_fused_gdn_gating_non_bulk_dma_fallback - ✅
single_node/ops/singlecard_ops/test_fused_gdn_gating.py test_fused_gdn_gating_non_default_params - ✅
single_node/ops/singlecard_ops/test_fused_gdn_gating.py test_fused_gdn_gating_output_shapes - ✅
single_node/ops/singlecard_ops/test_fused_gdn_gating.py test_fused_gdn_gating_small_batch_optimization - ✅ ✅
single_node/ops/singlecard_ops/test_fused_gdn_gating.py test_fused_gdn_gating_vs_reference - ✅
single_node/ops/singlecard_ops/test_fused_moe.py test_select_experts - ✅ ✅
single_node/ops/singlecard_ops/test_fused_moe.py test_select_experts_invalid_scoring_func - ✅ ✅
single_node/ops/singlecard_ops/test_fused_moe.py test_token_dispatcher_with_all_gather - ✅ ✅
single_node/ops/singlecard_ops/test_fused_moe.py test_token_dispatcher_with_all_gather_quant - ✅ ✅ ✅
single_node/ops/singlecard_ops/test_gating_top_k_softmax.py test_quant_fpx_linear - ✅
single_node/ops/singlecard_ops/test_gmm_swiglu_quant_weight_nz_tensor_list.py test_gmm_swiglu_quant_weight_nz_tensor_list - ✅
single_node/ops/singlecard_ops/test_grouped_matmul_swiglu_quant.py test_grouped_matmul_swiglu_quant_kernel - ✅
single_node/ops/singlecard_ops/test_hamming_dist_top_k.py test_hamming_dist_top_k - ✅
single_node/ops/singlecard_ops/test_hamming_dist_top_k.py test_hamming_dist_top_k_compare - ✅
single_node/ops/singlecard_ops/test_mla_preprocess.py test_mla_preprocess_kernel - ✅ ✅
single_node/ops/singlecard_ops/test_mla_preprocess_nq.py test_mla_preprocess_kernel - ✅ ✅
single_node/ops/singlecard_ops/test_mla_preprocess_qdown.py test_mla_preprocess_kernel - ✅ ✅
single_node/ops/singlecard_ops/test_moe_init_routing_custom.py test_moe_init_routing_custom - ✅
single_node/ops/singlecard_ops/test_ngram_spec_decode.py test_ngram_spec_decode_attrs - ✅
single_node/ops/singlecard_ops/test_ngram_spec_decode.py test_ngram_spec_decode_basic - ✅
single_node/ops/singlecard_ops/test_ngram_spec_decode.py test_ngram_spec_decode_decode - ✅
single_node/ops/singlecard_ops/test_ngram_spec_decode.py test_ngram_spec_decode_discard - ✅
single_node/ops/singlecard_ops/test_ngram_spec_decode.py test_ngram_spec_decode_exact_match - ✅
single_node/ops/singlecard_ops/test_ngram_spec_decode.py test_ngram_spec_decode_full_capacity - ✅
single_node/ops/singlecard_ops/test_ngram_spec_decode.py test_ngram_spec_decode_k1 - ✅
single_node/ops/singlecard_ops/test_ngram_spec_decode.py test_ngram_spec_decode_minimal - ✅
single_node/ops/singlecard_ops/test_ngram_spec_decode.py test_ngram_spec_decode_no_valid_sampled - ✅
single_node/ops/singlecard_ops/test_ngram_spec_decode.py test_ngram_spec_decode_padding - ✅
single_node/ops/singlecard_ops/test_ngram_spec_decode.py test_ngram_spec_decode_prefill - ✅
single_node/ops/singlecard_ops/test_npu_hc_pre.py test_npu_hc_pre_v1_v2_bf16_3d_input - ✅
single_node/ops/singlecard_ops/test_npu_hc_pre.py test_npu_hc_pre_v1_v2_bf16_4d_input - ✅
single_node/ops/singlecard_ops/test_npu_moe_gating_top_k.py test_npu_moe_gating_topk_compare - ✅
single_node/ops/singlecard_ops/test_recurrent_gated_delta_rule.py test_recurrent_gated_delta_rule - ✅
single_node/ops/singlecard_ops/test_recurrent_gated_delta_rule.py test_recurrent_gated_delta_rule_no_accepted - ✅
single_node/ops/singlecard_ops/test_recurrent_gated_delta_rule_310.py test_fused_recurrent_gated_delta_rule_310 - ✅ ✅
single_node/ops/singlecard_ops/test_reshape_and_cache_bnsd.py test_reshape_and_cache_bnsd_bf16_shape - ✅
single_node/ops/singlecard_ops/test_reshape_and_cache_bnsd.py test_reshape_and_cache_bnsd_compare - ✅
single_node/ops/singlecard_ops/test_reshape_and_cache_bnsd.py test_reshape_and_cache_bnsd_with_expected_output - ✅
single_node/ops/singlecard_ops/test_transpose_kv_cache_by_block.py test_transpose_kv_cache_by_block - ✅
single_node/ops/singlecard_ops/test_vocabparallelembedding.py test_get_masked_input_and_mask - ✅
single_node/ops/singlecard_ops/triton/test_apply_penalties_triton.py test_apply_all_penalties_v1_vs_ascend - ✅ ✅
single_node/ops/singlecard_ops/triton/test_bad_words.py test_apply_bad_words_different_shapes - ✅
single_node/ops/singlecard_ops/triton/test_bad_words.py test_apply_bad_words_edge_cases - ✅
single_node/ops/singlecard_ops/triton/test_bad_words.py test_apply_bad_words_no_bad_words - ✅
single_node/ops/singlecard_ops/triton/test_bad_words.py test_apply_bad_words_token_limit - ✅
single_node/ops/singlecard_ops/triton/test_batch_memcpy.py test_batch_memcpy - ✅ ✅
single_node/ops/singlecard_ops/triton/test_bincount.py test_bincount_kernel - ✅ ✅
single_node/ops/singlecard_ops/triton/test_causal_conv1d.py test_ascend_causal_conv1d - ✅
single_node/ops/singlecard_ops/triton/test_causal_conv1d.py test_causal_conv1d - ✅ ✅
single_node/ops/singlecard_ops/triton/test_causal_conv1d.py test_causal_conv1d_update_qwen3_next_shape - ✅ ✅
single_node/ops/singlecard_ops/triton/test_causal_conv1d.py test_causal_conv1d_update_with_batch_gather - ✅ ✅ ✅
single_node/ops/singlecard_ops/triton/test_chunk_gated_delta_rule.py test_chunk_gated_delta_rule_310_state_layout_matches_vllm - ✅
single_node/ops/singlecard_ops/triton/test_chunk_gated_delta_rule.py test_triton_fusion_ops - ✅
single_node/ops/singlecard_ops/triton/test_clear_ssm_states.py test_clear_ssm_states_ref_parity - ✅
single_node/ops/singlecard_ops/triton/test_compute_slot_mapping.py test_compute_slot_mapping_npu_kernel - ✅
single_node/ops/singlecard_ops/triton/test_compute_token_logprobs.py test_topk_log_softmax_deterministic - ✅ ✅
single_node/ops/singlecard_ops/triton/test_compute_token_logprobs.py test_topk_log_softmax_dtypes - ✅ ✅
single_node/ops/singlecard_ops/triton/test_compute_token_logprobs.py test_topk_log_softmax_edge_cases - ✅ ✅
single_node/ops/singlecard_ops/triton/test_compute_token_logprobs.py test_topk_log_softmax_kernel - ✅ ✅
single_node/ops/singlecard_ops/triton/test_compute_topk_logprobs.py test_compute_topk_logprobs - ✅
single_node/ops/singlecard_ops/triton/test_fused_gdn_gating.py test_fused_gdn_gating_310p_parity_precision - ✅
single_node/ops/singlecard_ops/triton/test_fused_qkvzba_split_reshape_cat.py test_fused_qkvzba_split_reshape_cat - ✅
single_node/ops/singlecard_ops/triton/test_fused_recurrent_gated_delta_rule.py test_fused_recurrent_gated_delta_rule_310_state_layout_matches_vllm - ✅
single_node/ops/singlecard_ops/triton/test_fused_recurrent_gated_delta_rule.py test_fused_recurrent_gated_delta_rule_310p_parity_precision - ✅ ✅
single_node/ops/singlecard_ops/triton/test_fused_sigmoid_gating_delta_rule.py test_triton_fusion_ops - ✅
single_node/ops/singlecard_ops/triton/test_l2norm.py test_l2norm - ✅
single_node/ops/singlecard_ops/triton/test_log_softmax.py test_topk_log_softmax_kernel - ✅
single_node/ops/singlecard_ops/triton/test_min_p.py test_apply_min_p_kernel - ✅
single_node/ops/singlecard_ops/triton/test_mrope.py test_mrotary_embedding_triton_kernel - ✅
single_node/ops/singlecard_ops/triton/test_muls_add.py test_muls_add_triton_correctness - ✅
single_node/ops/singlecard_ops/triton/test_penality.py test_apply_penalties - ✅ ✅
single_node/ops/singlecard_ops/triton/test_post_update.py test_post_update - ✅ ✅
single_node/ops/singlecard_ops/triton/test_prepare_inputs_padded.py test_prepare_inputs_padded - ✅
single_node/ops/singlecard_ops/triton/test_rejection_sample.py test_rejection_random_sample - ✅ ✅
single_node/ops/singlecard_ops/triton/test_rejection_sample.py test_rejection_sampler_block_verify_triton_kernel - ✅ ✅
single_node/ops/singlecard_ops/triton/test_rope.py test_rotary_embedding_triton_kernel - ✅
single_node/ops/singlecard_ops/triton/test_rope.py test_rotary_embedding_triton_kernel_siso - ✅
single_node/ops/singlecard_ops/triton/test_rope.py test_rotary_embedding_triton_kernel_with_cos_sin_cache - ✅
single_node/ops/singlecard_ops/triton/test_split_qkv_rmsnorm_mrope.py test_split_qkv_rmsnorm_mrope - ✅
single_node/ops/singlecard_ops/triton/test_split_qkv_rmsnorm_rope.py test_split_qkv_rmsnorm_rope - ✅
single_node/ops/singlecard_ops/triton/test_split_qkv_rmsnorm_rope.py test_split_qkv_rmsnorm_rope_with_bias - ✅
single_node/ops/singlecard_ops/triton/test_split_qkv_tp_rmsnorm_rope.py test_split_qkv_tp_rmsnorm_rope - ✅
single_node/ops/singlecard_ops/triton/test_temperature.py test_temperature_kernel - ✅

Weekly Tests

Test file Test method Model 310P Dense MoE Embedding Classification Reranker Mamba/SSM Multimodal Reasoning TP PP EP PCP DCP Context Parallel EPLB Dynamic EPLB Multistream MoE Full Graph Full Decode Only Graph Default FULL_AND_PIECEWISE Graph Piecewise Graph Eager Mode PD disaggregation W8A8 W4A8 FP16 LoRA Multi-LoRA Runtime LoRA updating Fully sharded LoRA parameterization Spec Decode MTP Eagle-3 SFA/DSA DSA CP Pooling runner Score API Classification API Distributed executor mp Flash Attention 3 FIA comparison Chunked Prefill Prefix Caching CPU/KV offloading KV transfer/events Sleep/Wake memory Xlite Graph CP KV Interleave Long Sequence FlashComm1 env Skipped Conditional skip Logprobs Batch inference Mixed lengths
multi_node/internal_dp/config/DeepSeek-V3.yaml test DeepSeek-V3 disaggregated_prefill vllm-ascend/DeepSeek-V3-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/DeepSeek-V3_2-W8A8-EP_weekly.yaml weekly test DeepSeek-V3.2-W8A8-EP disaggregated_prefill vllm-ascend/DeepSeek-V3.2-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
multi_node/internal_dp/config/GLM-4.7-W8A8C8-Mooncake-Layerwise.yaml test GLM-4.7-W8A8C8 PD separation with mooncake layerwise connector vllm-ascend/GLM-4.7-W8A8C8 ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/DeepSeek-V3.2-W8A8_A3_weekly.yaml DeepSeek-V3.2-W8A8-weekly vllm-ascend/DeepSeek-V3.2-W8A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/GLM-5.yaml GLM-5-TP16-DP1-decodegraph Eco-Tech/GLM-5-w4a8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/GLM-5_1-W8A8_A3_weekly.yaml GLM-5_1-w8a8-High—Throughput-bs10 Eco-Tech/GLM-5.1-w8a8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/GLM-5_1-W8A8_A3_weekly.yaml GLM-5_1-w8a8-High—Throughput-bs20 Eco-Tech/GLM-5.1-w8a8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/GLM-5_1-W8A8_A3_weekly.yaml GLM-5_1-w8a8-High—Throughput-bs32 Eco-Tech/GLM-5.1-w8a8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/GLM-5_1-W8A8_A3_weekly.yaml GLM-5_1-w8a8-High—Throughput-bs8 Eco-Tech/GLM-5.1-w8a8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Kimi-K2.5-32k-512.yaml Kimi-K2.5-W4A8-TOPT20-32k-0.5k Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Kimi-K2.5-32k-512.yaml Kimi-K2.5-W4A8-TOPT20-32k-0.5k-pc90 Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Kimi-K2.5-32k-512.yaml Kimi-K2.5-W4A8-TOPT50-32k-0.5k Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Kimi-K2.5-32k-512.yaml Kimi-K2.5-W4A8-TOPT50-32k-0.5k-prefix-cache90 Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Kimi-K2.5.yaml Kimi-K2.5-W4A8-Case Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Kimi-K2.5.yaml Kimi-K2.5-W4A8-TOPT20-128k-1k-pc90 Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Kimi-K2.5.yaml Kimi-K2.5-W4A8-TOPT20-16k-1k Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Kimi-K2.5.yaml Kimi-K2.5-W4A8-TOPT20-64k-1k-pc90 Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Kimi-K2.5.yaml Kimi-K2.5-W4A8-TOPT50-128k-1k-pc90 Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Kimi-K2.5.yaml Kimi-K2.5-W4A8-TOPT50-16k-1k Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Kimi-K2.5.yaml Kimi-K2.5-W4A8-TOPT50-64k-1k-pc90 Eco-Tech/Kimi-K2.5-W4A8 ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-W8A8-A3.yaml MiniMax-M2.5-w8a8 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8-in128k-32-8 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8-in128k-4-1 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8-in128k-64-16 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8-in16k-120-30 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8-in16k-16-4 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8-in32k-36-9 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8-in32k-4-1 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8-in32k-4-1-90 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8-in32k-80-20 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8-in64k-4-1 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/MiniMax-M2.5-w8a8-QuaRot-A3.yaml MiniMax-M2.5-w8a8-in64k-72-18 Eco-Tech/MiniMax-M2.5-w8a8-QuaRot ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen2.5-VL-7B-Instruct-EPD.yaml Qwen2.5-VL-7B-Instruct-epd Qwen/Qwen2.5-VL-7B-Instruct ✅ ✅
single_node/configs/Qwen3-32B.yaml Qwen3-32B-TP4 Qwen/Qwen3-32B ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-122B-A10B-W8A8-A2.yaml Qwen3.5-122B-A10B-W8A8-single-A2 Eco-Tech/Qwen3.5-122B-A10B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-122B-A10B-W8A8-A3.yaml Qwen3.5-122B-A10B-W8A8-A3 Eco-Tech/Qwen3.5-122B-A10B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-122B-A10B-W8A8-A3.yaml Qwen3.5-122B-A10B-W8A8-TPOT20-16k-1k Eco-Tech/Qwen3.5-122B-A10B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-122B-A10B-W8A8-A3.yaml Qwen3.5-122B-A10B-W8A8-TPOT20-32k-0.5k Eco-Tech/Qwen3.5-122B-A10B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-122B-A10B-W8A8-A3.yaml Qwen3.5-122B-A10B-W8A8-TPOT20-64k-1k Eco-Tech/Qwen3.5-122B-A10B-w8a8-mtp ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-122B-A10B-W8A8-A3.yaml Qwen3.5-122B-A10B-W8A8-TPOT50-16k-1k Eco-Tech/Qwen3.5-122B-A10B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-122B-A10B-W8A8-A3.yaml Qwen3.5-122B-A10B-W8A8-TPOT50-32k-0.5k Eco-Tech/Qwen3.5-122B-A10B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-122B-A10B-W8A8-A3.yaml Qwen3.5-122B-A10B-W8A8-TPOT50-64k-1k Eco-Tech/Qwen3.5-122B-A10B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8-in128k-28-7 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8-in128k-4-1 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8-in16k-16-4 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8-in16k-56-14 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8-in32k-16-4 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8-in32k-56-14 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8-in32k-8-2 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8-in64k-16-4 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8-in64k-48-12 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8-in64k-8-2 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-27B-w8a8-A3.yaml Qwen3.5-27B-w8a8-in64k-8-2-90 Eco-Tech/Qwen3.5-27B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3.yaml Qwen3.5-397B-A17B-w8a8-mtp Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-High—Throughput-bs136 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-High—Throughput-bs144 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-High—Throughput-bs32_in65536 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-High—Throughput-bs48 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-High—Throughput-bs8 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-High—Throughput-bs80 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-Minimal-Delay-bs16 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-Minimal-Delay-bs160 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-Minimal-Delay-bs16_in65536 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-Minimal-Delay-bs2 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-Minimal-Delay-bs32 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅
single_node/configs/Qwen3.5-397B-A17B-W8A8-mtp-A3_weekly.yaml Qwen3-397B-A17B-w8a8-A3-Minimal-Delay-bs36 Eco-Tech/Qwen3.5-397B-A17B-w8a8-mtp ✅ ✅ ✅ ✅ ✅ ✅