chore: import upstream snapshot with attribution

examples/train/flash_attention_3/mcore.sh

@@ -0,0 +1,47 @@
+# test_env: 4 * H20
+# fa2: 4 * 49GiB; 17.2s/it
+# fa3: 4 * 49GiB; 15.2s/it
+# https://github.com/Dao-AILab/flash-attention/tree/main#flashattention-3-beta-release
+# pip install "transformers==4.53.*"
+
+PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
+NPROC_PER_NODE=4 \
+CUDA_VISIBLE_DEVICES=0,1,2,3 \
+megatron sft \
+    --model Qwen/Qwen3-30B-A3B-Base \
+    --save_safetensors true \
+    --merge_lora false \
+    --dataset 'swift/Chinese-Qwen3-235B-2507-Distill-data-110k-SFT' \
+    --load_from_cache_file true \
+    --tuner_type lora \
+    --lora_rank 32 \
+    --lora_alpha 64 \
+    --target_modules all-linear \
+    --split_dataset_ratio 0.01 \
+    --moe_permute_fusion true \
+    --expert_model_parallel_size 4 \
+    --moe_grouped_gemm true \
+    --moe_shared_expert_overlap true \
+    --moe_aux_loss_coeff 1e-3 \
+    --micro_batch_size 1 \
+    --global_batch_size 16 \
+    --recompute_granularity full \
+    --recompute_method uniform \
+    --recompute_num_layers 1 \
+    --num_train_epochs 3 \
+    --finetune true \
+    --cross_entropy_loss_fusion true \
+    --lr 1e-4 \
+    --lr_warmup_fraction 0.05 \
+    --min_lr 1e-5 \
+    --output_dir megatron_output/Qwen3-30B-A3B-Base \
+    --eval_steps 200 \
+    --save_steps 200 \
+    --packing true \
+    --max_length 8192 \
+    --dataloader_num_workers 8 \
+    --dataset_num_proc 8 \
+    --no_save_optim true \
+    --no_save_rng true \
+    --sequence_parallel true \
+    --attention_backend flash

examples/train/flash_attention_3/transformers.sh

@@ -0,0 +1,35 @@
+# test_env: 4 * H20
+# fa2: 4 * 43GiB; 35.5s/it
+# fa3: 4 * 43GiB; 32.4s/it
+# https://github.com/Dao-AILab/flash-attention/tree/main#flashattention-3-beta-release
+# pip install "transformers==4.53.*"
+
+PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
+NPROC_PER_NODE=4 \
+CUDA_VISIBLE_DEVICES=0,1,2,3 \
+swift sft \
+    --model Qwen/Qwen2.5-7B \
+    --tuner_type full \
+    --dataset 'AI-ModelScope/LongAlpaca-12k' \
+    --load_from_cache_file true \
+    --attn_impl flash_attention_3 \
+    --num_train_epochs 3 \
+    --split_dataset_ratio 0.01 \
+    --torch_dtype bfloat16 \
+    --per_device_train_batch_size 1 \
+    --per_device_eval_batch_size 1 \
+    --learning_rate 1e-5 \
+    --gradient_accumulation_steps 4 \
+    --packing true \
+    --eval_steps 200 \
+    --save_steps 200 \
+    --logging_steps 5 \
+    --max_length 16384 \
+    --warmup_ratio 0.05 \
+    --dataloader_num_workers 8 \
+    --dataset_num_proc 8 \
+    --save_total_limit 2 \
+    --save_only_model true \
+    --output_dir output/Qwen2.5-7B \
+    --deepspeed zero3 \
+    --use_liger_kernel true