add more recipe

Author: yuiseki

recipes/A5000_24GB_x8/Mistral-7B-v0.1/coder-math-ja-wikipedia.yaml

@@ -0,0 +1,16 @@
+target_task: tasks/nlp/translation.md
+base_model_id: yuiseki/Mistral-7B-v0.1-ja-wikipedia-v0.1
+model_name: Mistral-7B-v0.1-ja-wikipedia-OpenMath-v0.1
+output_base_dir: /data/output
+dataset_id: kunishou/OpenMathInstruct-1-1.8m-ja
+dataset_input_field_name: question_ja
+dataset_output_field_name: generated_solution_ja
+dataset_train_split_seed: 42
+dataset_train_split_test_size: 0.2
+lora_r: 8
+lora_alpha: 16
+lora_dropout: 0.05
+train_claim_gpu_num: 4
+train_per_device_train_batch_size: 8
+train_gradient_accumulation_steps: 2
+train_num_train_epochs: 4

recipes/A5000_24GB_x8/Mistral-7B-v0.1/coder-python-ja-wikipedia-amenokaku.yaml

@@ -0,0 +1,17 @@
+target_task: tasks/text-generation/text2sql.md
+base_model_id: yuiseki/Mistral-7B-v0.1-ja-wikipedia-v0.1
+model_name: Mistral-7B-v0.1-ja-wikipedia-amenokaku-v0.1
+output_base_dir: /data/output
+dataset_id: kunishou/amenokaku-code-instruct
+dataset_input_field_name: instruction
+dataset_context_field_name: input
+dataset_output_field_name: output
+dataset_train_split_seed: 42
+dataset_train_split_test_size: 0.2
+lora_r: 8
+lora_alpha: 16
+lora_dropout: 0.05
+train_claim_gpu_num: 4
+train_per_device_train_batch_size: 8
+train_gradient_accumulation_steps: 4
+train_num_train_epochs: 4

recipes/A5000_24GB_x8/Mistral-7B-v0.1/i18n-ja-wikipedia-aya.yaml

@@ -13,6 +13,6 @@ lora_r: 8
 lora_alpha: 16
 lora_dropout: 0.05
 train_claim_gpu_num: 8
-train_per_device_train_batch_size: 2
-train_gradient_accumulation_steps: 8
+train_per_device_train_batch_size: 1
+train_gradient_accumulation_steps: 16
 train_num_train_epochs: 2

recipes/A5000_24GB_x8/Mistral-7B-v0.1/i18n-ja-wikipedia.yaml

@@ -11,6 +11,6 @@ lora_r: 8
 lora_alpha: 16
 lora_dropout: 0.05
 train_claim_gpu_num: 8
-train_per_device_train_batch_size: 2
-train_gradient_accumulation_steps: 8
+train_per_device_train_batch_size: 1
+train_gradient_accumulation_steps: 16
 train_num_train_epochs: 2