Initial commit

swift-main/examples/pytorch/llm/scripts/mamba-1.4b/lora/infer.sh

+# Experiment env: A10, RTX3090/4090, A100
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0,1 \
+python llm_infer.py \
+    --ckpt_dir "output/mamba-1.4b/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn true \
+    --max_new_tokens 2048 \
+    --temperature 0.5 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/mamba-1.4b/lora/sft.sh

+# Experiment env: A10, RTX3090/4090, A100
+# 1 * 12GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_sft.py \
+  --model_type mamba-1.4b \
+  --dataset dureader-robust-zh \
+  --batch_size 4 \
+  --max_length 1024 \
+  --gradient_accumulation_steps 2 \
+  --learning_rate 5e-5 \
+  --use_flash_attn true \
+  --eval_steps 1000 \
+  --save_steps 1000 \
+  --train_dataset_sample -1 \
+  --num_train_epochs 2 \
+  --check_dataset_strategy none \
+  --gradient_checkpointing true \
+  --weight_decay 0.1 \
+  --max_grad_norm 1.0 \
+  --warmup_ratio 0.03 \
+  --save_total_limit 2 \
+  --logging_steps 10 \
+  --sft_type lora \
+  --lora_target_modules DEFAULT \
+  --lora_rank 8 \
+  --lora_alpha 32

swift-main/examples/pytorch/llm/scripts/mengzi3_13b_base/lora_ddp_ds/infer.sh

+# Experimental environment: A100
+# 30GB GPU memory
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+    --ckpt_dir "output/mengzi3-13b-base/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --temperature 0.1 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/mengzi3_13b_base/lora_ddp_ds/sft.sh

+# Experimental environment: 2 * A100
+# 2 * 36GB GPU memory
+nproc_per_node=2
+
+CUDA_VISIBLE_DEVICES=0,1 \
+NPROC_PER_NODE=$nproc_per_node \
+MASTER_PORT=29500 \
+swift sft \
+    --model_id_or_path langboat/Mengzi3-13B-Base \
+    --model_revision master \
+    --sft_type lora \
+    --tuner_backend peft \
+    --dtype bf16 \
+    --output_dir output \
+    --ddp_backend nccl \
+    --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 \
+    --deepspeed default-zero2 \

swift-main/examples/pytorch/llm/scripts/minicpm_2b_sft_chat/lora_ddp/infer.sh

+# Experimental environment: A10
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+    --ckpt_dir "output/minicpm-2b-sft-chat/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn false \
+    --max_new_tokens 2048 \
+    --do_sample false \

swift-main/examples/pytorch/llm/scripts/minicpm_2b_sft_chat/lora_ddp/sft.sh

+# Experimental environment: 2 * A10
+# 2 * 12GB GPU memory
+nproc_per_node=2
+
+CUDA_VISIBLE_DEVICES=0,1 \
+NPROC_PER_NODE=$nproc_per_node \
+MASTER_PORT=29500 \
+swift sft \
+    --model_id_or_path OpenBMB/MiniCPM-2B-sft-fp32 \
+    --model_revision master \
+    --sft_type lora \
+    --template_type AUTO \
+    --dtype AUTO \
+    --output_dir output \
+    --ddp_backend nccl \
+    --dataset jd-sentiment-zh \
+    --train_dataset_sample -1 \
+    --val_dataset_sample 1000 \
+    --num_train_epochs 1 \
+    --max_length 2048 \
+    --check_dataset_strategy warning \
+    --lora_target_modules ALL \
+    --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 false \

swift-main/examples/pytorch/llm/scripts/minicpm_moe_8x2b/lora_ddp/infer.sh

+# Experimental environment: A100
+# 30GB GPU memory
+
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+    --ckpt_dir "output/minicpm-moe-8x2b/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config 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/minicpm_moe_8x2b/lora_ddp/sft.sh

+# Experimental environment: 4 * A100
+# 4 * 41GB GPU memory
+nproc_per_node=4
+CUDA_VISIBLE_DEVICES=0,1,2,3 \
+NPROC_PER_NODE=$nproc_per_node \
+MASTER_PORT=29500 \
+swift sft \
+    --model_id_or_path OpenBMB/MiniCPM-MoE-8x2B \
+    --model_revision master \
+    --sft_type lora \
+    --tuner_backend peft \
+    --dtype bf16 \
+    --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 false \
+    --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 false \

swift-main/examples/pytorch/llm/scripts/mistral_7b_instruct/lora_ddp_ds/infer.sh

+# Experimental environment: A10
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/mistral-7b-instruct/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config 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/mistral_7b_instruct/lora_ddp_ds/sft.sh

+# Experimental environment: 2 * A10
+# 2 * 21GB 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 AI-ModelScope/Mistral-7B-Instruct-v0.1 \
+    --model_revision master \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type AUTO \
+    --dtype AUTO \
+    --output_dir output \
+    --ddp_backend nccl \
+    --dataset leetcode-python-en \
+    --train_dataset_sample -1 \
+    --num_train_epochs 1 \
+    --max_length 4096 \
+    --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 $(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 \
+    --deepspeed default-zero2 \

swift-main/examples/pytorch/llm/scripts/mistral_7b_instruct/lora_mp_ddp/infer.sh

+# Experimental environment: 3090
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/mistral-7b-instruct/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config 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/mistral_7b_instruct/lora_mp_ddp/sft.sh

+# Experimental environment: 4 * 3090
+# 4 * 19GB GPU memory
+nproc_per_node=2
+
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0,1,2,3 \
+torchrun \
+    --nproc_per_node=$nproc_per_node \
+    --master_port 29500 \
+    llm_sft.py \
+    --model_id_or_path AI-ModelScope/Mistral-7B-Instruct-v0.1 \
+    --model_revision master \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type AUTO \
+    --dtype AUTO \
+    --output_dir output \
+    --ddp_backend nccl \
+    --dataset damo-agent-mini-zh \
+    --train_dataset_sample 20000 \
+    --num_train_epochs 1 \
+    --max_length 4096 \
+    --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 \

swift-main/examples/pytorch/llm/scripts/mistral_7b_v2/lora/infer.sh

+# Experimental environment: A100
+# 16GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/mistral-7b-v2/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn true \
+    --max_new_tokens 2048 \
+    --temperature 0.5 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/mistral_7b_v2/lora/sft.sh

+# Experimental environment: A100
+# 19GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_sft.py \
+    --model_id_or_path AI-ModelScope/Mistral-7B-v0.2-hf \
+    --model_revision master \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type AUTO \
+    --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 \
+    --use_flash_attn true \

swift-main/examples/pytorch/llm/scripts/mixtral_moe_7b/lora/infer.sh

+# Experimental environment: A100
+# 2 * 45GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0,1 \
+python llm_infer.py \
+    --ckpt_dir "output/mixtral-moe-7b/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn true \
+    --max_new_tokens 2048 \
+    --temperature 0.5 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/mixtral_moe_7b/lora/sft.sh

+# Experimental environment: 2 * A100
+# 2 * 60GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0,1 \
+python llm_sft.py \
+    --model_id_or_path AI-ModelScope/Mixtral-8x7B-v0.1 \
+    --model_revision master \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type AUTO \
+    --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 q_proj k_proj v_proj o_proj \
+    --gradient_checkpointing false \
+    --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/mixtral_moe_7b_instruct/lora/infer.sh

+# Experimental environment: A100
+# 2 * 45GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0,1 \
+python llm_infer.py \
+    --ckpt_dir "output/mixtral-moe-7b-instruct/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn true \
+    --max_new_tokens 2048 \
+    --temperature 0.5 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/mixtral_moe_7b_instruct/lora/sft.sh

+# Experimental environment: 2 * A100
+# 2 * 60GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0,1 \
+python llm_sft.py \
+    --model_id_or_path AI-ModelScope/Mixtral-8x7B-Instruct-v0.1 \
+    --model_revision master \
+    --sft_type lora \
+    --tuner_backend peft \
+    --template_type AUTO \
+    --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 q_proj k_proj v_proj o_proj \
+    --gradient_checkpointing false \
+    --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/mixtral_moe_8x22b_v1/lora_ddp_ds/infer.sh

+# Experimental environment: A100
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0,1,2,3 \
+python llm_infer.py \
+    --ckpt_dir "output/mixtral-8x22b-v0.1/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn true \
+    --max_new_tokens 2048 \
+    --temperature 0.5 \
+    --top_p 0.7 \
+    --repetition_penalty 1. \
+    --do_sample true \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/mixtral_moe_8x22b_v1/lora_ddp_ds/sft.sh

+# Experimental environment: 8 * A100
+# 4 * 41GB GPU memory
+nproc_per_node=8
+CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
+NPROC_PER_NODE=$nproc_per_node \
+MASTER_PORT=29500 \
+swift sft \
+    --model_id_or_path AI-ModelScope/Mixtral-8x22B-v0.1 \
+    --model_revision master \
+    --sft_type lora \
+    --tuner_backend peft \
+    --dtype bf16 \
+    --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 false \
+    --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 false \
+    --deepspeed default-zero3