chore: import upstream snapshot with attribution

examples/megatron/lora/dense.sh

@@ -0,0 +1,39 @@
+# full: 2 * 70GiB 0.61s/it
+# lora: 2 * 14GiB 0.45s/it
+PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
+NPROC_PER_NODE=2 \
+CUDA_VISIBLE_DEVICES=0,1 \
+megatron sft \
+    --model Qwen/Qwen2.5-7B-Instruct \
+    --save_safetensors true \
+    --merge_lora false \
+    --dataset 'AI-ModelScope/alpaca-gpt4-data-zh#500' \
+              'AI-ModelScope/alpaca-gpt4-data-en#500' \
+              'swift/self-cognition#500' \
+    --tuner_type lora \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --target_modules all-linear \
+    --tensor_model_parallel_size 2 \
+    --sequence_parallel true \
+    --micro_batch_size 16 \
+    --global_batch_size 16 \
+    --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/Qwen2.5-7B-Instruct \
+    --save_steps 100 \
+    --max_length 2048 \
+    --system 'You are a helpful assistant.' \
+    --dataloader_num_workers 4 \
+    --no_save_optim true \
+    --no_save_rng true \
+    --dataset_num_proc 4 \
+    --model_author swift \
+    --model_name swift-robot

examples/megatron/lora/dpo.sh

@@ -0,0 +1,45 @@
+# 2 * 65GiB; 4.50s/it
+PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
+NPROC_PER_NODE=2 \
+CUDA_VISIBLE_DEVICES=0,1 \
+megatron rlhf \
+    --rlhf_type dpo \
+    --model Qwen/Qwen3-30B-A3B-Instruct-2507 \
+    --save_safetensors true \
+    --merge_lora false \
+    --dataset AI-ModelScope/orpo-dpo-mix-40k \
+    --load_from_cache_file true \
+    --tuner_type lora \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --target_modules all-linear \
+    --split_dataset_ratio 0.01 \
+    --expert_model_parallel_size 2 \
+    --moe_permute_fusion true \
+    --moe_grouped_gemm true \
+    --moe_shared_expert_overlap true \
+    --moe_aux_loss_coeff 1e-3 \
+    --micro_batch_size 8 \
+    --global_batch_size 16 \
+    --recompute_granularity full \
+    --recompute_method uniform \
+    --recompute_num_layers 1 \
+    --num_train_epochs 1 \
+    --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-Instruct-2507 \
+    --eval_steps 100 \
+    --save_steps 100 \
+    --max_length 2048 \
+    --dataloader_num_workers 8 \
+    --dataset_num_proc 8 \
+    --no_save_optim true \
+    --no_save_rng true \
+    --sequence_parallel true \
+    --attention_backend flash \
+    --rpo_alpha 0.1 \
+    --beta 0.1 \
+    --loss_type sigmoid

examples/megatron/lora/loss_scale.sh

@@ -0,0 +1,44 @@
+# 2 * 60GiB, 3.4s/it
+PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
+NPROC_PER_NODE=2 \
+CUDA_VISIBLE_DEVICES=0,1 \
+megatron sft \
+    --model Qwen/Qwen3-30B-A3B-Base \
+    --save_safetensors true \
+    --merge_lora false \
+    --tuner_type lora \
+    --dataset AI-ModelScope/function-calling-chatml#10000 \
+    --load_from_cache_file true \
+    --loss_scale hermes \
+    --agent_template hermes \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --target_modules all-linear \
+    --modules_to_save word_embeddings output_layer \
+    --split_dataset_ratio 0.01 \
+    --expert_model_parallel_size 2 \
+    --moe_permute_fusion true \
+    --moe_grouped_gemm true \
+    --moe_shared_expert_overlap true \
+    --moe_aux_loss_coeff 1e-3 \
+    --micro_batch_size 8 \
+    --global_batch_size 16 \
+    --recompute_granularity full \
+    --recompute_method uniform \
+    --recompute_num_layers 1 \
+    --num_train_epochs 1 \
+    --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 \
+    --max_length 2048 \
+    --dataloader_num_workers 8 \
+    --dataset_num_proc 8 \
+    --no_save_optim true \
+    --no_save_rng true \
+    --sequence_parallel true \
+    --attention_backend flash

examples/megatron/lora/moe.sh

@@ -0,0 +1,44 @@
+# 2 * 62GiB, 5.10s/it
+PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
+NPROC_PER_NODE=2 \
+CUDA_VISIBLE_DEVICES=0,1 \
+megatron sft \
+    --model Qwen/Qwen3-30B-A3B \
+    --save_safetensors true \
+    --merge_lora false \
+    --dataset 'swift/Qwen3-SFT-Mixin#2000' \
+              'swift/self-cognition:empty_think#600' \
+    --loss_scale ignore_empty_think \
+    --tuner_type lora \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --target_modules all-linear \
+    --split_dataset_ratio 0.01 \
+    --expert_model_parallel_size 2 \
+    --moe_permute_fusion true \
+    --moe_grouped_gemm true \
+    --moe_shared_expert_overlap true \
+    --moe_aux_loss_coeff 1e-3 \
+    --micro_batch_size 8 \
+    --global_batch_size 16 \
+    --recompute_granularity full \
+    --recompute_method uniform \
+    --recompute_num_layers 1 \
+    --num_train_epochs 1 \
+    --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 \
+    --eval_steps 200 \
+    --save_steps 200 \
+    --max_length 2048 \
+    --dataloader_num_workers 8 \
+    --dataset_num_proc 8 \
+    --no_save_optim true \
+    --no_save_rng true \
+    --sequence_parallel true \
+    --attention_backend flash \
+    --model_author swift \
+    --model_name swift-robot

examples/megatron/lora/mtp.sh

@@ -0,0 +1,61 @@
+# demo: thinking -> non-thinking
+# 4 * 70GiB; 40s/it
+PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
+NPROC_PER_NODE=4 \
+CUDA_VISIBLE_DEVICES=0,1,2,3 \
+megatron sft \
+    --model ZhipuAI/GLM-4.5-Air \
+    --save_safetensors true \
+    --merge_lora true \
+    --mtp_num_layers 1 \
+    --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 linear_qkv linear_proj \
+    --split_dataset_ratio 0.01 \
+    --moe_permute_fusion true \
+    --tensor_model_parallel_size 4 \
+    --expert_tensor_parallel_size 1 \
+    --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 2 \
+    --finetune true \
+    --cross_entropy_loss_fusion true \
+    --lr 1e-4 \
+    --lr_warmup_fraction 0.05 \
+    --min_lr 1e-5 \
+    --output_dir megatron_output/GLM-4.5-Air \
+    --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
+
+# If not using the MTP module, please remove the speculative-related parameters.
+# CUDA_VISIBLE_DEVICES=0,1,2,3 \
+# swift infer \
+#     --model megatron_output/GLM-4.5-Air/vx-xxx/checkpoint-xxx-merged \
+#     --sglang_tp_size 4 \
+#     --infer_backend sglang \
+#     --load_data_args true \
+#     --sglang_context_length 8192 \
+#     --max_new_tokens 2048 \
+#     --sglang_mem_fraction_static 0.7 \
+#     --sglang_speculative_algorithm EAGLE \
+#     --sglang_speculative_eagle_topk 1 \
+#     --sglang_speculative_num_steps 3 \
+#     --sglang_speculative_num_draft_tokens 4

examples/megatron/lora/new_special_tokens.sh

@@ -0,0 +1,45 @@
+# 2 * 60GiB, 2.7s/it
+# Note: The conversion script has no differences.
+# It will read the new_special_tokens parameter from args.json.
+
+PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
+NPROC_PER_NODE=2 \
+CUDA_VISIBLE_DEVICES=0,1 \
+megatron sft \
+    --model Qwen/Qwen3-30B-A3B \
+    --save_safetensors true \
+    --merge_lora false \
+    --dataset 'swift/new_special_tokens' \
+    --new_special_tokens 'examples/train/new_special_tokens/tokens.txt' \
+    --tuner_type lora \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --target_modules all-linear \
+    --modules_to_save word_embeddings output_layer \
+    --split_dataset_ratio 0.01 \
+    --expert_model_parallel_size 2 \
+    --moe_permute_fusion true \
+    --moe_grouped_gemm true \
+    --moe_shared_expert_overlap true \
+    --moe_aux_loss_coeff 1e-3 \
+    --micro_batch_size 32 \
+    --global_batch_size 64 \
+    --recompute_granularity full \
+    --recompute_method uniform \
+    --recompute_num_layers 1 \
+    --num_train_epochs 5 \
+    --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 \
+    --eval_steps 500 \
+    --save_steps 500 \
+    --max_length 2048 \
+    --dataloader_num_workers 8 \
+    --dataset_num_proc 8 \
+    --no_save_optim true \
+    --no_save_rng true \
+    --sequence_parallel true \
+    --attention_backend flash

examples/megatron/lora/qwen3_235b.sh

@@ -0,0 +1,48 @@
+# 8 * 80GiB, 3.2s/it
+# If you're doing full-parameter training, you'll need 64 × 80 GiB of GPU memory
+
+PYTORCH_CUDA_ALLOC_CONF='expandable_segments:True' \
+NPROC_PER_NODE=8 \
+CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
+megatron sft \
+    --model Qwen/Qwen3-235B-A22B-Instruct-2507 \
+    --dataset 'swift/Chinese-Qwen3-235B-2507-Distill-data-110k-SFT#2000' \
+              'swift/self-cognition#1000' \
+    --save_safetensors true \
+    --merge_lora false \
+    --load_from_cache_file true \
+    --tuner_type lora \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --target_modules all-linear \
+    --split_dataset_ratio 0.01 \
+    --moe_permute_fusion true \
+    --tensor_model_parallel_size 4 \
+    --expert_tensor_parallel_size 1 \
+    --expert_model_parallel_size 8 \
+    --moe_grouped_gemm true \
+    --moe_shared_expert_overlap true \
+    --moe_aux_loss_coeff 1e-3 \
+    --micro_batch_size 8 \
+    --global_batch_size 16 \
+    --recompute_granularity full \
+    --recompute_method uniform \
+    --recompute_num_layers 1 \
+    --num_train_epochs 1 \
+    --finetune true \
+    --cross_entropy_loss_fusion true \
+    --lr 1e-4 \
+    --lr_warmup_fraction 0.05 \
+    --min_lr 1e-5 \
+    --output_dir megatron_output/Qwen3-235B-A22B-Instruct-2507 \
+    --eval_steps 200 \
+    --save_steps 200 \
+    --max_length 2048 \
+    --dataloader_num_workers 8 \
+    --dataset_num_proc 8 \
+    --no_save_optim true \
+    --no_save_rng true \
+    --sequence_parallel true \
+    --attention_backend flash \
+    --model_author swift \
+    --model_name swift-robot