Initial commit

swift-main/examples/pytorch/llm/scripts/torchacc/yi_34b_chat/swift_lora_sft.sh

+# Experimental environment: 4 * A100
+# 80GB GPU memory
+# Note: TorchAcc is currently only available internally.
+
+# MASTER_ADDR=127.0.0.1 \
+
+NPROC_PER_NODE=2 \
+CUDA_VISIBLE_DEVICES=0,1,2,3 \
+swift sft \
+  --model_type yi-34b-chat \
+  --dataset codefuse-python-en \
+  --sft_type lora \
+  --dtype AUTO \
+  --output_dir output \
+  --num_train_epochs 1 \
+  --max_length 2048 \
+  --batch_size  1 \
+  --use_flash_attn true \
+  --gradient_accumulation_steps 1 \
+  --dataset_test_ratio 0 \
+  --save_strategy no \
+  --eval_steps 2000000 \
+  --save_steps 2000000 \
+  --logging_steps 100 \
+  --preprocess_num_proc 1 \
+  --metric_warmup_step 0.1 \
+  --report_to 'none'

swift-main/examples/pytorch/llm/scripts/xverse_13b/qlora/infer.sh

+# Experimental environment: 3090
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/xverse-13b/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --max_new_tokens 2048 \
+    --temperature 0.7 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/xverse_13b/qlora/sft.sh

+# Experimental environment: 3090
+# 12GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_sft.py \
+    --model_id_or_path xverse/XVERSE-13B \
+    --model_revision master \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type default-generation \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset advertise-gen-zh \
+    --train_dataset_sample 20000 \
+    --num_train_epochs 1 \
+    --max_length 2048 \
+    --check_dataset_strategy warning \
+    --quantization_bit 4 \
+    --bnb_4bit_comp_dtype AUTO \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --lora_dropout_p 0.05 \
+    --lora_target_modules ALL \
+    --gradient_checkpointing true \
+    --batch_size 1 \
+    --weight_decay 0.1 \
+    --learning_rate 1e-4 \
+    --gradient_accumulation_steps 16 \
+    --max_grad_norm 0.5 \
+    --warmup_ratio 0.03 \
+    --eval_steps 100 \
+    --save_steps 100 \
+    --save_total_limit 2 \
+    --logging_steps 10 \

swift-main/examples/pytorch/llm/scripts/xverse_13b_256k/infer.sh

+# Experimental environment: A100
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+    --ckpt_dir "output/xverse-13b-256k/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --max_new_tokens 2048 \
+    --temperature 0.7 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/xverse_13b_256k/sft.sh

+# Experimental environment: A100
+# 40GB GPU memory
+CUDA_VISIBLE_DEVICES=0 \
+swift sft \
+    --model_type xverse-13b-256k \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type default-generation \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset advertise-gen-zh \
+    --train_dataset_sample 20000 \
+    --num_train_epochs 1 \
+    --max_length 2048 \
+    --check_dataset_strategy warning \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --lora_dropout_p 0.05 \
+    --lora_target_modules ALL \
+    --gradient_checkpointing true \
+    --batch_size 1 \
+    --weight_decay 0.1 \
+    --learning_rate 1e-4 \
+    --gradient_accumulation_steps 16 \
+    --max_grad_norm 0.5 \
+    --warmup_ratio 0.03 \
+    --eval_steps 100 \
+    --save_steps 100 \
+    --save_total_limit 2 \
+    --logging_steps 10 \

swift-main/examples/pytorch/llm/scripts/xverse_65b/qlora_mp/infer.sh

+# Experimental environment: A100
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/xverse-65b/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --max_new_tokens 2048 \
+    --temperature 0.7 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/xverse_65b/qlora_mp/sft.sh

+# Experimental environment: 2 * A100
+# 2 * 23GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0,1 \
+python llm_sft.py \
+    --model_id_or_path xverse/XVERSE-65B \
+    --model_revision v1.0.0 \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type default-generation \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset dureader-robust-zh \
+    --train_dataset_sample -1 \
+    --num_train_epochs 1 \
+    --max_length 2048 \
+    --check_dataset_strategy warning \
+    --quantization_bit 4 \
+    --bnb_4bit_comp_dtype AUTO \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --lora_dropout_p 0.05 \
+    --lora_target_modules ALL \
+    --gradient_checkpointing true \
+    --batch_size 1 \
+    --weight_decay 0.1 \
+    --learning_rate 1e-4 \
+    --gradient_accumulation_steps 16 \
+    --max_grad_norm 0.5 \
+    --warmup_ratio 0.03 \
+    --eval_steps 100 \
+    --save_steps 100 \
+    --save_total_limit 2 \
+    --logging_steps 10 \

swift-main/examples/pytorch/llm/scripts/xverse_moe_a4_2b/lora/infer.sh

+# Experimental environment: A100
+# 60GB GPU memory
+
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+    --ckpt_dir "output/xverse-moe-a4_2b/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --max_new_tokens 2048 \
+    --temperature 0.7 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/xverse_moe_a4_2b/lora/sft.sh

+# Experimental environment: A100
+# 66GB GPU memory
+CUDA_VISIBLE_DEVICES=0 \
+swift sft \
+    --model_type xverse-moe-a4_2b \
+    --sft_type lora \
+    --tuner_backend peft \
+    --dtype fp16 \
+    --dataset dureader-robust-zh \
+    --train_dataset_sample -1 \
+    --num_train_epochs 1 \
+    --max_length 1024 \
+    --check_dataset_strategy warning \
+    --lora_dtype AUTO \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --lora_dropout_p 0.05 \
+    --lora_target_modules DEFAULT \
+    --gradient_checkpointing true \
+    --batch_size 1 \
+    --weight_decay 0.1 \
+    --learning_rate 1e-4 \
+    --gradient_accumulation_steps 16 \
+    --max_grad_norm 0.5 \
+    --warmup_ratio 0.03 \
+    --eval_steps 100 \
+    --save_steps 100 \
+    --save_total_limit 2 \
+    --logging_steps 10 \

swift-main/examples/pytorch/llm/scripts/yi_34b/lora_ddp_ds/infer.sh

+# Experimental environment: A100
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/yi-34b/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn true \
+    --max_new_tokens 2048 \
+    --temperature 0.7 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/yi_34b/lora_ddp_ds/sft.sh

+# Experimental environment: 2 * A100
+# 2 * 70GB GPU memory
+nproc_per_node=2
+
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0,1 \
+torchrun \
+    --nproc_per_node=$nproc_per_node \
+    --master_port 29500 \
+    llm_sft.py \
+    --model_id_or_path 01ai/Yi-34B \
+    --model_revision master \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type default-generation \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset dureader-robust-zh \
+    --train_dataset_sample -1 \
+    --num_train_epochs 1 \
+    --max_length 2048 \
+    --check_dataset_strategy warning \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --lora_dropout_p 0.05 \
+    --lora_target_modules DEFAULT \
+    --gradient_checkpointing true \
+    --batch_size 1 \
+    --weight_decay 0.1 \
+    --learning_rate 1e-4 \
+    --gradient_accumulation_steps $(expr 16 / $nproc_per_node) \
+    --max_grad_norm 0.5 \
+    --warmup_ratio 0.03 \
+    --eval_steps 100 \
+    --save_steps 100 \
+    --save_total_limit 2 \
+    --logging_steps 10 \
+    --use_flash_attn true \

swift-main/examples/pytorch/llm/scripts/yi_34b_chat/lora/infer.sh

+# Experimental environment: A100
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+    --ckpt_dir "output/yi-34b-chat/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn true \
+    --max_new_tokens 2048 \
+    --temperature 0.1 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/yi_34b_chat/lora/sft.sh

+# Experimental environment: A100
+# 70GB GPU memory
+
+CUDA_VISIBLE_DEVICES=0 \
+swift sft \
+    --model_type yi-34b-chat \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type AUTO \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset blossom-math-zh \
+    --train_dataset_sample -1 \
+    --num_train_epochs 3 \
+    --max_length 2048 \
+    --check_dataset_strategy warning \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --lora_dropout_p 0.05 \
+    --lora_target_modules ALL \
+    --gradient_checkpointing true \
+    --batch_size 1 \
+    --weight_decay 0.1 \
+    --learning_rate 1e-4 \
+    --gradient_accumulation_steps 16 \
+    --max_grad_norm 0.5 \
+    --warmup_ratio 0.03 \
+    --eval_steps 100 \
+    --save_steps 100 \
+    --save_total_limit 2 \
+    --logging_steps 10 \
+    --use_flash_attn true \

swift-main/examples/pytorch/llm/scripts/yi_34b_chat/lora_ddp_ds/infer.sh

+# Experimental environment: A100
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/yi-34b-chat/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn true \
+    --max_new_tokens 2048 \
+    --temperature 0.1 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/yi_34b_chat/lora_ddp_ds/sft.sh

+# Experimental environment: 2 * A100
+# 2 * 72GB GPU memory
+nproc_per_node=2
+
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0,1 \
+torchrun \
+    --nproc_per_node=$nproc_per_node \
+    --master_port 29500 \
+    llm_sft.py \
+    --model_type yi-34b-chat \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type AUTO \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset blossom-math-zh \
+    --train_dataset_sample -1 \
+    --num_train_epochs 1 \
+    --max_length 2048 \
+    --check_dataset_strategy warning \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --lora_dropout_p 0.05 \
+    --lora_target_modules DEFAULT \
+    --gradient_checkpointing true \
+    --batch_size 1 \
+    --weight_decay 0.1 \
+    --learning_rate 1e-4 \
+    --gradient_accumulation_steps $(expr 16 / $nproc_per_node) \
+    --max_grad_norm 0.5 \
+    --warmup_ratio 0.03 \
+    --eval_steps 100 \
+    --save_steps 100 \
+    --save_total_limit 2 \
+    --logging_steps 10 \
+    --use_flash_attn true \

swift-main/examples/pytorch/llm/scripts/yi_34b_chat/qlora/infer.sh

+# Experimental environment: A10, 3090
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/yi-34b-chat/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn false \
+    --max_new_tokens 2048 \
+    --temperature 0.1 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/yi_34b_chat/qlora/sft.sh

+# Experimental environment: A10, 3090
+# 21GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_sft.py \
+    --model_type yi-34b-chat \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type AUTO \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset blossom-math-zh \
+    --train_dataset_sample -1 \
+    --num_train_epochs 1 \
+    --max_length 2048 \
+    --check_dataset_strategy warning \
+    --quantization_bit 4 \
+    --bnb_4bit_comp_dtype AUTO \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --lora_dropout_p 0.05 \
+    --lora_target_modules DEFAULT \
+    --gradient_checkpointing true \
+    --batch_size 1 \
+    --weight_decay 0.1 \
+    --learning_rate 1e-4 \
+    --gradient_accumulation_steps 16 \
+    --max_grad_norm 0.5 \
+    --warmup_ratio 0.03 \
+    --eval_steps 100 \
+    --save_steps 100 \
+    --save_total_limit 2 \
+    --logging_steps 10 \
+    --use_flash_attn false \

swift-main/examples/pytorch/llm/scripts/yi_6b/lora/infer.sh

+# Experimental environment: A10
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/yi-6b/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --max_new_tokens 2048 \
+    --temperature 0.7 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/yi_6b/lora/sft.sh

+# Experimental environment: A10
+# 15GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_sft.py \
+    --model_id_or_path 01ai/Yi-6B \
+    --model_revision master \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type default-generation \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset dureader-robust-zh \
+    --train_dataset_sample -1 \
+    --num_train_epochs 1 \
+    --max_length 2048 \
+    --check_dataset_strategy warning \
+    --lora_rank 8 \
+    --lora_alpha 32 \
+    --lora_dropout_p 0.05 \
+    --lora_target_modules DEFAULT \
+    --gradient_checkpointing true \
+    --batch_size 1 \
+    --weight_decay 0.1 \
+    --learning_rate 1e-4 \
+    --gradient_accumulation_steps 16 \
+    --max_grad_norm 0.5 \
+    --warmup_ratio 0.03 \
+    --eval_steps 100 \
+    --save_steps 100 \
+    --save_total_limit 2 \
+    --logging_steps 10 \

swift-main/examples/pytorch/llm/scripts/yi_6b_chat/llamapro/infer.sh

+# Experimental environment: A10
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/yi-6b-chat/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --max_new_tokens 2048 \
+    --temperature 0.7 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \