423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-14 19:06:26 +08:00
Code Issues Packages Projects Releases Wiki Activity

support fsdp + qlora

Browse Source 
Former-commit-id: 8408225162
This commit is contained in:
hiyouga
2024-03-21 00:36:06 +08:00
parent 6646e18c02
commit 7999836fb6
19 changed files with 87 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
examples/full_multi_gpu/ds_z2_config.json
View File

@@ -1,28 +0,0 @@
{
"train_batch_size": "auto",
"train_micro_batch_size_per_gpu": "auto",
"gradient_accumulation_steps": "auto",
"gradient_clipping": "auto",
"zero_allow_untested_optimizer": true,
"fp16": {
"enabled": "auto",
"loss_scale": 0,
"loss_scale_window": 1000,
"initial_scale_power": 16,
"hysteresis": 2,
"min_loss_scale": 1
},
"bf16": {
"enabled": "auto"
},
"zero_optimization": {
"stage": 2,
"allgather_partitions": true,
"allgather_bucket_size": 5e8,
"overlap_comm": true,
"reduce_scatter": true,
"reduce_bucket_size": 5e8,
"contiguous_gradients": true,
"round_robin_gradients": true
}
}

32
examples/full_multi_gpu/ds_z2_offload_config.json
View File

@@ -1,32 +0,0 @@
{
"train_batch_size": "auto",
"train_micro_batch_size_per_gpu": "auto",
"gradient_accumulation_steps": "auto",
"gradient_clipping": "auto",
"zero_allow_untested_optimizer": true,
"fp16": {
"enabled": "auto",
"loss_scale": 0,
"loss_scale_window": 1000,
"initial_scale_power": 16,
"hysteresis": 2,
"min_loss_scale": 1
},
"bf16": {
"enabled": "auto"
},
"zero_optimization": {
"stage": 2,
"offload_optimizer": {
"device": "cpu",
"pin_memory": true
},
"allgather_partitions": true,
"allgather_bucket_size": 5e8,
"overlap_comm": true,
"reduce_scatter": true,
"reduce_bucket_size": 5e8,
"contiguous_gradients": true,
"round_robin_gradients": true
}
}

30
examples/full_multi_gpu/ds_z3_config.json
View File

@@ -1,30 +0,0 @@
{
"train_batch_size": "auto",
"train_micro_batch_size_per_gpu": "auto",
"gradient_accumulation_steps": "auto",
"gradient_clipping": "auto",
"zero_allow_untested_optimizer": true,
"fp16": {
"enabled": "auto",
"loss_scale": 0,
"loss_scale_window": 1000,
"initial_scale_power": 16,
"hysteresis": 2,
"min_loss_scale": 1
},
"bf16": {
"enabled": "auto"
},
"zero_optimization": {
"stage": 3,
"overlap_comm": true,
"contiguous_gradients": true,
"sub_group_size": 1e9,
"reduce_bucket_size": "auto",
"stage3_prefetch_bucket_size": "auto",
"stage3_param_persistence_threshold": "auto",
"stage3_max_live_parameters": 1e9,
"stage3_max_reuse_distance": 1e9,
"stage3_gather_16bit_weights_on_model_save": true
}
}

38
examples/full_multi_gpu/ds_z3_offload_config.json
View File

@@ -1,38 +0,0 @@
{
"train_batch_size": "auto",
"train_micro_batch_size_per_gpu": "auto",
"gradient_accumulation_steps": "auto",
"gradient_clipping": "auto",
"zero_allow_untested_optimizer": true,
"fp16": {
"enabled": "auto",
"loss_scale": 0,
"loss_scale_window": 1000,
"initial_scale_power": 16,
"hysteresis": 2,
"min_loss_scale": 1
},
"bf16": {
"enabled": "auto"
},
"zero_optimization": {
"stage": 3,
"offload_optimizer": {
"device": "cpu",
"pin_memory": true
},
"offload_param": {
"device": "cpu",
"pin_memory": true
},
"overlap_comm": true,
"contiguous_gradients": true,
"sub_group_size": 1e9,
"reduce_bucket_size": "auto",
"stage3_prefetch_bucket_size": "auto",
"stage3_param_persistence_threshold": "auto",
"stage3_max_live_parameters": 1e9,
"stage3_max_reuse_distance": 1e9,
"stage3_gather_16bit_weights_on_model_save": true
}
}

4
examples/full_multi_gpu/multi_node.sh
View File

@@ -7,11 +7,11 @@ python -m torch.distributed.run \
--master_addr $MASTER_ADDR \
--master_port $MASTER_PORT \
../../src/train_bash.py \
--deepspeed ds_z3_config.json \
--deepspeed ../deepspeed/ds_z3_config.json \
--stage sft \
--do_train \
--model_name_or_path meta-llama/Llama-2-7b-hf \
--dataset alpaca_gpt4_en \
--dataset alpaca_gpt4_en,glaive_toolcall \
--dataset_dir ../../data \
--template default \
--finetuning_type full \

4
examples/full_multi_gpu/single_node.sh
View File

@@ -1,11 +1,11 @@
#!/bin/bash
deepspeed --num_gpus 4 ../../src/train_bash.py \
--deepspeed ds_z3_config.json \
--deepspeed ../deepspeed/ds_z3_config.json \
--stage sft \
--do_train \
--model_name_or_path meta-llama/Llama-2-7b-hf \
--dataset alpaca_gpt4_en \
--dataset alpaca_gpt4_en,glaive_toolcall \
--dataset_dir ../../data \
--template default \
--finetuning_type full \