From c004209052a3d1f269c19fb26d0ffd14f2ec17cc Mon Sep 17 00:00:00 2001 From: Jialong Wu Date: Fri, 20 Dec 2024 08:11:48 +0000 Subject: [PATCH] add bair finetune scripts --- scripts/finetune/bair-64-act-cond.sh | 21 +++++++++++++++++++++ scripts/finetune/bair-64-act-free.sh | 21 +++++++++++++++++++++ 2 files changed, 42 insertions(+) create mode 100644 scripts/finetune/bair-64-act-cond.sh create mode 100644 scripts/finetune/bair-64-act-free.sh diff --git a/scripts/finetune/bair-64-act-cond.sh b/scripts/finetune/bair-64-act-cond.sh new file mode 100644 index 0000000..bafe2b1 --- /dev/null +++ b/scripts/finetune/bair-64-act-cond.sh @@ -0,0 +1,21 @@ +accelerate launch train_tokenizer.py \ + --exp_name bair_tokenizer_ft --output_dir log_vqgan --seed 0 --mixed_precision bf16 \ + --model_type ctx_vqgan \ + --train_batch_size 16 --gradient_accumulation_steps 1 --disc_start 1000005 \ + --oxe_data_mixes_type bair --resolution 64 --dataloader_num_workers 16 \ + --rand_select --video_stepsize 1 --segment_horizon 16 --segment_length 8 --context_length 1 \ + --pretrained_model_name_or_path pretrained_models/ivideogpt-oxe-64-act-free/tokenizer + + +accelerate launch train_gpt.py \ + --exp_name bair_llama_ft --output_dir log_trm --seed 0 --mixed_precision bf16 \ + --vqgan_type ctx_vqgan \ + --pretrained_model_name_or_path {log directory of finetuned tokenizer}/unwrapped_model \ + --config_name configs/llama/config.json --load_internal_llm --action_conditioned --action_dim 4 \ + --pretrained_transformer_path pretrained_models/ivideogpt-oxe-64-act-free/transformer \ + --per_device_train_batch_size 16 --gradient_accumulation_steps 1 \ + --learning_rate 1e-4 --lr_scheduler_type cosine \ + --oxe_data_mixes_type bair --resolution 64 --dataloader_num_workers 16 \ + --video_stepsize 1 --segment_length 16 --context_length 1 \ + --use_eval_dataset --use_fvd --use_frame_metrics \ + --weight_decay 0.01 --llama_attn_drop 0.1 --embed_no_wd \ No newline at end of file diff --git a/scripts/finetune/bair-64-act-free.sh b/scripts/finetune/bair-64-act-free.sh new file mode 100644 index 0000000..7a2f27b --- /dev/null +++ b/scripts/finetune/bair-64-act-free.sh @@ -0,0 +1,21 @@ +accelerate launch train_tokenizer.py \ + --exp_name bair_tokenizer_ft --output_dir log_vqgan --seed 0 --mixed_precision bf16 \ + --model_type ctx_vqgan \ + --train_batch_size 16 --gradient_accumulation_steps 1 --disc_start 1000005 \ + --oxe_data_mixes_type bair --resolution 64 --dataloader_num_workers 16 \ + --rand_select --video_stepsize 1 --segment_horizon 16 --segment_length 8 --context_length 1 \ + --pretrained_model_name_or_path pretrained_models/ivideogpt-oxe-64-act-free/tokenizer + + +accelerate launch train_gpt.py \ + --exp_name bair_llama_ft --output_dir log_trm --seed 0 --mixed_precision bf16 \ + --vqgan_type ctx_vqgan \ + --pretrained_model_name_or_path {log directory of finetuned tokenizer}/unwrapped_model \ + --config_name configs/llama/config.json \ + --pretrained_transformer_path pretrained_models/ivideogpt-oxe-64-act-free/transformer \ + --per_device_train_batch_size 16 --gradient_accumulation_steps 1 \ + --learning_rate 1e-4 --lr_scheduler_type cosine \ + --oxe_data_mixes_type bair --resolution 64 --dataloader_num_workers 16 \ + --video_stepsize 1 --segment_length 16 --context_length 1 \ + --use_eval_dataset --use_fvd --use_frame_metrics \ + --weight_decay 0.01 --llama_attn_drop 0.1 --embed_no_wd \ No newline at end of file