updata to v0.9.1_stable

examples/extras/badam/llama3_full_sft.yaml

@@ -15,7 +15,7 @@ badam_verbose: 2
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/extras/fsdp_qlora/llama3_lora_sft.yaml

@@ -11,7 +11,7 @@ lora_target: all
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/extras/galore/llama3_full_sft.yaml

@@ -14,7 +14,7 @@ galore_scale: 2.0
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/extras/llama_pro/llama3_freeze_sft.yaml

@@ -12,7 +12,7 @@ use_llama_pro: true
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/extras/loraplus/llama3_lora_sft.yaml

@@ -11,7 +11,7 @@ loraplus_lr_ratio: 16.0
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/extras/mod/llama3_full_sft.yaml

@@ -10,7 +10,7 @@ mixture_of_depths: convert
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/extras/pissa/llama3_lora_sft.yaml

@@ -13,7 +13,7 @@ pissa_convert: true
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_full/llama3_full_predict.yaml

@@ -9,7 +9,7 @@ finetuning_type: full
 ### dataset
 eval_dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 50
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_full/llama3_full_sft_ds3.yaml

@@ -10,7 +10,7 @@ deepspeed: examples/deepspeed/ds_z3_config.json
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_full/qwen2vl_full_sft.yaml

@@ -10,7 +10,7 @@ deepspeed: examples/deepspeed/ds_z3_config.json
 ### dataset
 dataset: mllm_demo,identity
 template: qwen2_vl
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_lora/llama3_lora_dpo.yaml

@@ -12,7 +12,7 @@ pref_loss: sigmoid  # choices: [sigmoid (dpo), orpo, simpo]
 ### dataset
 dataset: dpo_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_lora/llama3_lora_kto.yaml

@@ -11,7 +11,7 @@ pref_beta: 0.1
 ### dataset
 dataset: kto_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_lora/llama3_lora_ppo.yaml

@@ -11,7 +11,7 @@ lora_target: all
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_lora/llama3_lora_predict.yaml

@@ -10,7 +10,7 @@ finetuning_type: lora
 ### dataset
 eval_dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 50
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_lora/llama3_lora_pretrain.yaml

@@ -9,7 +9,7 @@ lora_target: all
 
 ### dataset
 dataset: c4_demo
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_lora/llama3_lora_reward.yaml

@@ -10,7 +10,7 @@ lora_target: all
 ### dataset
 dataset: dpo_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_lora/llama3_lora_sft.yaml

@@ -10,7 +10,7 @@ lora_target: all
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_lora/llama3_lora_sft_ds0.yaml

@@ -11,7 +11,7 @@ deepspeed: examples/deepspeed/ds_z0_config.json
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_lora/llama3_lora_sft_ds3.yaml

@@ -11,7 +11,7 @@ deepspeed: examples/deepspeed/ds_z3_config.json
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16

examples/train_lora/llama3_preprocess.yaml

@@ -10,7 +10,7 @@ lora_target: all
 ### dataset
 dataset: identity,alpaca_en_demo
 template: llama3
-cutoff_len: 1024
+cutoff_len: 2048
 max_samples: 1000
 overwrite_cache: true
 preprocessing_num_workers: 16