aya train_claim_gpu_num: 3

yuiseki · yuiseki · commit e6e7b458c2c9 · 2024-03-26T12:52:48.000+09:00
diff --git a/recipes/A5000_24GB_x8/i18n-ar-wikipedia-aya.yaml b/recipes/A5000_24GB_x8/i18n-ar-wikipedia-aya.yaml
@@ -12,7 +12,7 @@ dataset_train_split_test_size: 0.2
 lora_r: 8
 lora_alpha: 16
 lora_dropout: 0.05
-train_claim_gpu_num: 4
+train_claim_gpu_num: 3
 train_per_device_train_batch_size: 8
 train_gradient_accumulation_steps: 4
 train_num_train_epochs: 4
diff --git a/recipes/A5000_24GB_x8/i18n-es-wikipedia-aya.yaml b/recipes/A5000_24GB_x8/i18n-es-wikipedia-aya.yaml
@@ -12,7 +12,7 @@ dataset_train_split_test_size: 0.2
 lora_r: 8
 lora_alpha: 16
 lora_dropout: 0.05
-train_claim_gpu_num: 4
+train_claim_gpu_num: 3
 train_per_device_train_batch_size: 8
 train_gradient_accumulation_steps: 4
 train_num_train_epochs: 4
diff --git a/recipes/A5000_24GB_x8/i18n-fr-wikipedia-aya.yaml b/recipes/A5000_24GB_x8/i18n-fr-wikipedia-aya.yaml
@@ -12,7 +12,7 @@ dataset_train_split_test_size: 0.2
 lora_r: 8
 lora_alpha: 16
 lora_dropout: 0.05
-train_claim_gpu_num: 4
+train_claim_gpu_num: 3
 train_per_device_train_batch_size: 8
 train_gradient_accumulation_steps: 4
 train_num_train_epochs: 4
diff --git a/recipes/A5000_24GB_x8/i18n-ja-wikipedia-aya.yaml b/recipes/A5000_24GB_x8/i18n-ja-wikipedia-aya.yaml
@@ -12,7 +12,7 @@ dataset_train_split_test_size: 0.2
 lora_r: 8
 lora_alpha: 16
 lora_dropout: 0.05
-train_claim_gpu_num: 4
+train_claim_gpu_num: 3
 train_per_device_train_batch_size: 8
 train_gradient_accumulation_steps: 4
 train_num_train_epochs: 4
diff --git a/recipes/A5000_24GB_x8/i18n-zh-wikipedia-aya.yaml b/recipes/A5000_24GB_x8/i18n-zh-wikipedia-aya.yaml
@@ -12,7 +12,7 @@ dataset_train_split_test_size: 0.2
 lora_r: 8
 lora_alpha: 16
 lora_dropout: 0.05
-train_claim_gpu_num: 4
+train_claim_gpu_num: 3
 train_per_device_train_batch_size: 8
 train_gradient_accumulation_steps: 4
 train_num_train_epochs: 4
