[example] Add KTransformers Qwen3MoE example (#9511)

Co-authored-by: unknown <xiongchenhui@hisense.ad>
Co-authored-by: Kingsley <kingsleydodonow@gmail.com>

examples/inference/qwen3moe_lora_sft_kt.yaml

@@ -0,0 +1,10 @@
+model_name_or_path: Qwen/Qwen3-235B-A22B-Instruct-2507
+adapter_name_or_path: saves/Kllama_Qwen3MoE_235bA22b
+template: qwen3_nothink
+infer_backend: ktransformers  # choices: [huggingface, vllm, sglang, ktransformers]
+trust_remote_code: true
+
+use_kt: true # use KTransformers as LoRA sft backend to inference
+kt_optimize_rule: examples/kt_optimize_rules/Qwen3Moe-sft-amx.yaml
+cpu_infer: 32
+chunk_size: 8192

examples/kt_optimize_rules/Qwen3Moe-sft-amx.yaml

@@ -0,0 +1,80 @@
+- match:
+    class: ktransformers.models.modeling_qwen2_moe.Qwen2MoeRotaryEmbedding
+  replace:
+    class: ktransformers.operators.RoPE.RotaryEmbedding
+    kwargs:
+      generate_device: "cuda"
+      prefill_device: "cuda"
+
+- match:
+    name: "^lm_head$"  # regular expression 
+    class: torch.nn.Linear  # only match modules matching name and class simultaneously
+  replace:
+    class: ktransformers.operators.linear.KTransformersLinear  # optimized Kernel on quantized data types
+    kwargs:
+      generate_device: "cuda"
+      prefill_device: "cuda"
+      generate_op: "KLinearTorch"
+      prefill_op: "KLinearTorch"
+
+# - match:
+#     name: "^model\\.layers\\..*$"  # regular expression 
+#     class: torch.nn.Linear  # only match modules matching name and class simultaneously
+#   replace:
+#     class: ktransformers.operators.linear.KTransformersLinear  # optimized Kernel on quantized data types
+#     kwargs:
+#       generate_device: "cuda"
+#       prefill_device: "cuda"
+#       generate_op: "KLinearTorch"
+#       prefill_op: "KLinearTorch"
+- match:
+    name: "^model\\.layers\\.(?!.*mlp\\.shared_expert_gate).*$"  # regular expression 
+    class: torch.nn.Linear  # only match modules matching name and class simultaneously
+  replace:
+    class: ktransformers.operators.linear.KTransformersLinear  # optimized Kernel on quantized data types
+    kwargs:
+      generate_device: "cuda"
+      prefill_device: "cuda"
+      generate_op: "KLinearTorch"
+      prefill_op: "KLinearTorch"
+- match:
+    name: "^model\\.layers\\..*\\.mlp$"
+  replace:
+    class: ktransformers.operators.experts.KQwen3MoeSparseMoeBlock     # mlp module with custom forward function
+    kwargs:
+      generate_device: "cuda"
+      prefill_device: "cuda"
+
+- match:
+    name: "^model\\.layers\\..*\\.mlp\\.experts$"
+  replace:
+    class: ktransformers.operators.experts.KTransformersExperts     # custom MoE Kernel with expert paralleism
+    kwargs:
+      prefill_device: "cuda"
+      prefill_op: "KExpertsTorch"
+      generate_device: "cpu"
+      generate_op: "KSFTExpertsCPU"
+      out_device: "cuda"
+      backend: "AMXInt8" # or "AMXBF16" or "AMXInt8"
+  recursive: False # don't recursively inject submodules of this module
+- match:
+    name: "^model\\.layers\\..*\\.self_attn$"
+  replace:
+    class: ktransformers.operators.attention.KQwen3MoeAttention # optimized MLA implementation
+    kwargs:
+      generate_device: "cuda"
+      prefill_device: "cuda"
+- match:
+    name: "^model.embed_tokens"
+  replace:
+    class: "default"
+    kwargs:
+      generate_device: "cpu"
+      prefill_device: "cpu"
+
+- match:
+    name: "^model$"
+  replace:
+    class: "ktransformers.operators.models.KQwen3MoeModel"
+    kwargs:
+      per_layer_prefill_intput_threshold: 0

examples/train_lora/qwen3moe_lora_sft_kt.yaml

@@ -0,0 +1,52 @@
+### model
+model_name_or_path: Qwen/Qwen3-235B-A22B-Instruct-2507
+trust_remote_code: true
+
+### method
+stage: sft
+do_train: true
+finetuning_type: lora
+lora_rank: 8
+lora_target: all
+
+### dataset
+dataset: identity, alpaca_en_demo
+template: qwen3_nothink
+cutoff_len: 2048
+max_samples: 100000
+overwrite_cache: true
+preprocessing_num_workers: 16
+dataloader_num_workers: 4
+
+### output
+output_dir: saves/Kllama_Qwen3MoE_235bA22b
+logging_steps: 10
+save_steps: 200
+plot_loss: true
+overwrite_output_dir: true
+save_only_model: false
+report_to: none  # choices: [none, wandb, tensorboard, swanlab, mlflow]
+
+### train
+per_device_train_batch_size: 1
+gradient_accumulation_steps: 8
+learning_rate: 1.0e-4
+num_train_epochs: 3
+lr_scheduler_type: cosine
+warmup_ratio: 0.1
+bf16: true
+ddp_timeout: 180000000
+resume_from_checkpoint: null
+
+### ktransformers
+use_kt: true # use KTransformers as LoRA sft backend
+kt_optimize_rule: examples/kt_optimize_rules/Qwen3Moe-sft-amx.yaml
+cpu_infer: 32
+chunk_size: 8192
+
+### eval
+# eval_dataset: alpaca_en_demo
+# val_size: 0.1
+# per_device_eval_batch_size: 1
+# eval_strategy: steps
+# eval_steps: 500