chore: import upstream snapshot with attribution

examples/megatron/seq_cls/full.sh

@@ -0,0 +1,39 @@
+# 4 * 60GiB; 7.5s/it
+PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
+NPROC_PER_NODE=4 \
+CUDA_VISIBLE_DEVICES=0,1,2,3 \
+megatron sft \
+    --model Qwen/Qwen2.5-VL-7B-Instruct \
+    --save_safetensors true \
+    --dataset 'tany0699/garbage265#20000' \
+    --load_from_cache_file true \
+    --tensor_model_parallel_size 2 \
+    --sequence_parallel true \
+    --packing true \
+    --freeze_llm false \
+    --freeze_vit true \
+    --freeze_aligner true \
+    --split_dataset_ratio 0.01 \
+    --micro_batch_size 2 \
+    --global_batch_size 4 \
+    --model_kwargs '{"max_pixels": 1003520}' \
+    --recompute_granularity full \
+    --recompute_method uniform \
+    --recompute_num_layers 1 \
+    --finetune true \
+    --cross_entropy_loss_fusion true \
+    --lr 1e-5 \
+    --lr_warmup_fraction 0.05 \
+    --min_lr 1e-6 \
+    --num_train_epochs 1 \
+    --output_dir megatron_output/Qwen2.5-VL-7B-Instruct \
+    --save_steps 200 \
+    --eval_steps 200 \
+    --vit_gradient_checkpointing true \
+    --max_length 8192 \
+    --dataloader_num_workers 4 \
+    --no_save_optim true \
+    --no_save_rng true \
+    --dataset_num_proc 8 \
+    --num_labels 265 \
+    --task_type seq_cls \

examples/megatron/seq_cls/lora/infer.sh

@@ -0,0 +1,9 @@
+# acc: 0.9298597194388778
+# 60GiB
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+  --model megatron_output/Qwen3-30B-A3B-Instruct-2507/vx-xxx/checkpoint-xxx \
+  --load_data_args true \
+  --max_batch_size 16 \
+  --attn_impl flash_attn \
+  --metric acc

examples/megatron/seq_cls/lora/train.sh

@@ -0,0 +1,44 @@
+# 2 * 40GiB; 5s/it
+# eval_acc: 0.924
+PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
+NPROC_PER_NODE=2 \
+CUDA_VISIBLE_DEVICES=0,1 \
+megatron sft \
+    --model Qwen/Qwen3-30B-A3B-Instruct-2507 \
+    --save_safetensors true \
+    --merge_lora true \
+    --dataset 'DAMO_NLP/jd:cls' \
+    --load_from_cache_file true \
+    --split_dataset_ratio 0.01 \
+    --tuner_type lora \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --target_modules all-linear all-router \
+    --packing true \
+    --expert_model_parallel_size 2 \
+    --moe_permute_fusion true \
+    --moe_grouped_gemm true \
+    --moe_shared_expert_overlap true \
+    --moe_aux_loss_coeff 1e-6 \
+    --sequence_parallel true \
+    --micro_batch_size 1 \
+    --global_batch_size 2 \
+    --recompute_granularity full \
+    --recompute_method uniform \
+    --recompute_num_layers 1 \
+    --finetune true \
+    --cross_entropy_loss_fusion true \
+    --lr 1e-4 \
+    --lr_warmup_fraction 0.05 \
+    --min_lr 1e-5 \
+    --num_train_epochs 1 \
+    --output_dir megatron_output/Qwen3-30B-A3B-Instruct-2507 \
+    --eval_steps 200 \
+    --save_steps 200 \
+    --max_length 2048 \
+    --task_type seq_cls \
+    --num_labels 2 \
+    --dataloader_num_workers 4 \
+    --no_save_optim true \
+    --no_save_rng true \
+    --dataset_num_proc 4