Initial commit

swift-main/examples/pytorch/llm/scripts/qwen1half_32b_chat/lora_mp/infer.sh

+# Experimental environment: V100, A10, 3090
+# 66GB GPU memory
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+    --ckpt_dir "output/qwen1half-32b-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/qwen1half_32b_chat/lora_mp/sft.sh

+# Experimental environment: A100
+# 2*40GB GPU memory
+CUDA_VISIBLE_DEVICES=0,1 \
+swift sft \
+    --model_type qwen1half-32b-chat \
+    --sft_type lora \
+    --tuner_backend peft \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset alpaca-zh alpaca-en \
+    --train_dataset_sample 5000 \
+    --num_train_epochs 2 \
+    --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/qwen1half_72b_chat_int4/qlora/infer.sh

+# Experimental environment: A100
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+    --ckpt_dir "output/qwen1half-72b-chat-int4/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/qwen1half_72b_chat_int4/qlora/sft.sh

+# Experimental environment: A100
+# 50GB GPU memory
+
+CUDA_VISIBLE_DEVICES=0 \
+swift sft \
+    --model_type qwen1half-72b-chat-int4 \
+    --sft_type lora \
+    --output_dir output \
+    --dataset codefuse-python-en \
+    --train_dataset_sample -1 \
+    --num_train_epochs 3 \
+    --max_length 2048 \
+    --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/qwen1half_7b_chat/full/infer.sh

+# Experimental environment: A100
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+    --ckpt_dir "output/qwen1half-7b-chat/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn true \

swift-main/examples/pytorch/llm/scripts/qwen1half_7b_chat/full/sft.sh

+# Experimental environment: A100
+# 80GB GPU memory
+CUDA_VISIBLE_DEVICES=0 \
+swift sft \
+    --model_type qwen1half-7b-chat \
+    --sft_type full \
+    --train_dataset_sample -1 \
+    --eval_steps 1000 \
+    --output_dir output \
+    --num_train_epochs 1 \
+    --max_length 4096 \
+    --learning_rate 1e-5 \
+    --use_flash_attn true \
+    --save_only_model true \
+    --dataset codefuse-evol-instruction-zh \
+    --preprocess_num_proc 4 \

swift-main/examples/pytorch/llm/scripts/qwen1half_7b_chat/galore/infer.sh

+# Experimental environment: A100
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+    --ckpt_dir "output/qwen1half-7b-chat/vx-xxx/checkpoint-xxx" \
+    --load_dataset_config true \
+    --use_flash_attn true \

swift-main/examples/pytorch/llm/scripts/qwen1half_7b_chat/galore/sft.sh

+# Experimental environment: A100
+# 40GB GPU memory
+CUDA_VISIBLE_DEVICES=0 \
+swift sft \
+    --model_type qwen1half-7b-chat \
+    --sft_type full \
+    --use_galore true \
+    --galore_update_proj_gap 400 \
+    --train_dataset_sample -1 \
+    --eval_steps 1000 \
+    --output_dir output \
+    --num_train_epochs 1 \
+    --max_length 4096 \
+    --learning_rate 1e-5 \
+    --use_flash_attn true \
+    --save_only_model true \
+    --dataset codefuse-evol-instruction-zh \
+    --preprocess_num_proc 4 \

swift-main/examples/pytorch/llm/scripts/qwen1half_7b_chat/lora/infer.sh

+# Experimental environment: V100, A10, 3090
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/qwen1half-7b-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/qwen1half_7b_chat/lora/sft.sh

+# Experimental environment: A100
+# 30GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_sft.py \
+    --model_type qwen1half-7b-chat \
+    --sft_type lora \
+    --tuner_backend peft \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset alpaca-zh alpaca-en \
+    --train_dataset_sample 5000 \
+    --num_train_epochs 2 \
+    --max_length 1024 \
+    --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 false \
+    --self_cognition_sample 1000 \
+    --model_name 卡卡罗特 \
+    --model_author 陶白白 \

swift-main/examples/pytorch/llm/scripts/qwen1half_7b_chat_awq/lora/infer.sh

+# Experiment env: A10, RTX3090/4090, A100
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/qwen1half-7b-chat-awq/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 \
+    --stream false \
+    --merge_lora false \

swift-main/examples/pytorch/llm/scripts/qwen1half_7b_chat_awq/lora/sft.sh

+# Experiment env: A10, RTX3090/4090, A100
+# 1 * 17G GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_sft.py \
+  --model_type qwen1half-7b-chat-awq \
+  --dataset ms-agent \
+  --train_dataset_mix_ratio 3 \
+  --batch_size 4 \
+  --max_length 1024 \
+  --use_loss_scale true \
+  --gradient_accumulation_steps 2 \
+  --learning_rate 5e-5 \
+  --use_flash_attn true \
+  --eval_steps 2000 \
+  --save_steps 2000 \
+  --train_dataset_sample -1 \
+  --num_train_epochs 1 \
+  --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 ALL \
+  --lora_rank 8 \
+  --lora_alpha 32

swift-main/examples/pytorch/llm/scripts/qwen1half_7b_chat_int8/qlora/infer.sh

+# Experimental environment: V100, A10, 3090
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/qwen1half-7b-chat-int8/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/qwen1half_7b_chat_int8/qlora/sft.sh

+# Experimental environment: V100, A10, 3090
+# 20GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_sft.py \
+    --model_type qwen1half-7b-chat-int8 \
+    --sft_type lora \
+    --tuner_backend peft \
+    --dtype fp16 \
+    --output_dir output \
+    --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 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/qwen1half_moe_a2_7b/lora/infer.sh

+# Experimental environment: A100
+# 36GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/qwen1half-moe-a2_7b/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/qwen1half_moe_a2_7b/lora/sft.sh

+# Experimental environment: A100
+# 42GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_sft.py \
+    --model_type qwen1half-moe-a2_7b \
+    --sft_type lora \
+    --tuner_backend peft \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset dureader-robust-zh \
+    --train_dataset_sample -1 \
+    --num_train_epochs 1 \
+    --max_length 1024 \
+    --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/qwen1half_moe_a2_7b_chat/lora/infer.sh

+# Experimental environment: A100
+# 36GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_infer.py \
+    --ckpt_dir "output/qwen1half-moe-a2_7b-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/qwen1half_moe_a2_7b_chat/lora/sft.sh

+# Experimental environment: A100
+# 42GB GPU memory
+PYTHONPATH=../../.. \
+CUDA_VISIBLE_DEVICES=0 \
+python llm_sft.py \
+    --model_type qwen1half-moe-a2_7b-chat \
+    --sft_type lora \
+    --tuner_backend peft \
+    --dtype AUTO \
+    --output_dir output \
+    --dataset blossom-math-zh \
+    --train_dataset_sample -1 \
+    --num_train_epochs 1 \
+    --max_length 1024 \
+    --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/qwen1half_moe_a2_7b_chat_int4/qlora/infer.sh

+# Experimental environment: A100
+CUDA_VISIBLE_DEVICES=0 \
+swift infer \
+    --ckpt_dir "output/qwen1half-moe-a2_7b-chat-int4/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/qwen1half_moe_a2_7b_chat_int4/qlora/sft.sh

+# Experimental environment: A100
+# 17GB GPU memory
+
+CUDA_VISIBLE_DEVICES=0 \
+swift sft \
+    --model_type qwen1half-moe-a2_7b-chat-int4 \
+    --sft_type lora \
+    --output_dir output \
+    --dataset blossom-math-zh \
+    --train_dataset_sample -1 \
+    --num_train_epochs 3 \
+    --max_length 2048 \
+    --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 \