Add output dir to top of all configs (meta-pytorch#2183)

Author: ebsmothers

recipes/configs/eleuther_evaluation.yaml

@@ -3,6 +3,8 @@
 # To launch, run the following command from root torchtune directory:
 #    tune run eleuther_eval --config eleuther_evaluation tasks=["truthfulqa_mc2","hellaswag"]
 
+output_dir: ./ # Not needed
+
 # Model Arguments
 model:
   _component_: torchtune.models.llama2.llama2_7b
@@ -14,7 +16,7 @@ checkpointer:
     pytorch_model-00001-of-00002.bin,
     pytorch_model-00002-of-00002.bin,
   ]
-  output_dir: /tmp/Llama-2-7b-hf
+  output_dir: ${output_dir}
   model_type: LLAMA2
 
 # Tokenizer

recipes/configs/gemma/evaluation.yaml

@@ -3,6 +3,8 @@
 # To launch, run the following command:
 #    tune run eleuther_eval --config gemma/evaluation
 
+output_dir: ./ # Not needed
+
 # Model Arguments
 model:
   _component_: torchtune.models.gemma.gemma_2b
@@ -15,7 +17,7 @@ checkpointer:
     model-00001-of-00002.safetensors,
     model-00002-of-00002.safetensors,
   ]
-  output_dir: ./ # Not needed
+  output_dir: ${output_dir}
   model_type: GEMMA
 
 # Tokenizer

recipes/configs/generation.yaml

@@ -3,6 +3,8 @@
 # To launch, run the following command from root torchtune directory:
 #    tune run generate --config generation
 
+output_dir: ./ # Not needed
+
 # Model arguments
 model:
   _component_: torchtune.models.llama2.llama2_7b
@@ -14,7 +16,7 @@ checkpointer:
     pytorch_model-00001-of-00002.bin,
     pytorch_model-00002-of-00002.bin,
   ]
-  output_dir: /tmp/Llama-2-7b-hf/
+  output_dir: ${output_dir}
   model_type: LLAMA2
 
 device: cuda

recipes/configs/llama2/generation_v2.yaml

@@ -6,6 +6,8 @@
 # To launch, run the following command:
 #    tune run dev/generate_v2 --config llama2/generation_v2
 
+output_dir: ./ # Not needed
+
 # Model arguments
 model:
   _component_: torchtune.models.llama2.llama2_7b
@@ -24,7 +26,7 @@ checkpointer:
     pytorch_model-00001-of-00002.bin,
     pytorch_model-00002-of-00002.bin
   ]
-  output_dir: ./
+  output_dir: ${output_dir}
   model_type: LLAMA2
 
 # Device

recipes/configs/llama3_2_vision/11B_evaluation.yaml

@@ -9,6 +9,8 @@
 # To launch, run the following command from root torchtune directory:
 #    tune run eleuther_eval --config llama3_2_vision/11B_evaluation
 
+output_dir: ./ # Not needed
+
 # Model arguments
 model:
   _component_: torchtune.models.llama3_2_vision.llama3_2_vision_11b
@@ -26,7 +28,7 @@ checkpointer:
   checkpoint_files:
     filename_format: model-{}-of-{}.safetensors
     max_filename: "00005"
-  output_dir: ./
+  output_dir: ${output_dir}
   model_type: LLAMA3_VISION
 
 # Environment

recipes/configs/llama3_2_vision/11B_generation_v2.yaml

@@ -7,6 +7,8 @@
 # To launch, run the following command from root torchtune directory:
 #    tune run dev/generate_v2 --config llama3_2_vision/generation_v2
 
+output_dir: ./ # Not needed
+
 # Model arguments
 model:
   _component_: torchtune.models.llama3_2_vision.llama3_2_vision_11b
@@ -25,7 +27,7 @@ checkpointer:
   checkpoint_files:
     filename_format: model-{}-of-{}.safetensors
     max_filename: "00005"
-  output_dir: ./
+  output_dir: ${output_dir}
   model_type: LLAMA3_VISION
 
 # Device

recipes/configs/mistral/evaluation.yaml

@@ -3,6 +3,8 @@
 # To launch, run the following command:
 #    tune run eleuther_eval --config mistral/evaluation
 
+output_dir: ./ # Not needed
+
 # Model Arguments
 model:
   _component_: torchtune.models.mistral.mistral_7b
@@ -15,7 +17,7 @@ checkpointer:
     pytorch_model-00001-of-00002.bin,
     pytorch_model-00002-of-00002.bin
   ]
-  output_dir: /tmp/Mistral-7B-v0.1/
+  output_dir: ${output_dir}
   model_type: MISTRAL
 resume_from_checkpoint: False
 

recipes/configs/phi3/evaluation.yaml

@@ -3,6 +3,8 @@
 # To launch, run the following command:
 #    tune run eleuther_eval --config phi3/evaluation
 
+output_dir: ./ # Not needed
+
 # Model Arguments
 model:
   _component_: torchtune.models.phi3.phi3_mini
@@ -16,7 +18,7 @@ checkpointer:
     model-00002-of-00002.safetensors
   ]
   recipe_checkpoint: null
-  output_dir: /tmp/Phi-3-mini-4k-instruct
+  output_dir: ${output_dir}
   model_type: PHI3_MINI
 resume_from_checkpoint: False
 

recipes/configs/quantization.yaml

@@ -3,6 +3,8 @@
 # To launch, run the following command from root torchtune directory:
 #    tune run quantize --config quantization
 
+output_dir: /tmp/torchtune/llama2_7B/quantized # /tmp may be deleted by your system. Change it to your preference.
+
 #
 # Model arguments
 model:
@@ -16,7 +18,7 @@ checkpointer:
     pytorch_model-00002-of-00002.bin,
   ]
   recipe_checkpoint: null
-  output_dir: /tmp/Llama-2-7b-hf
+  output_dir: ${output_dir}
   model_type: LLAMA2
 
 device: cuda

recipes/configs/qwen2/evaluation.yaml

@@ -3,6 +3,8 @@
 # To launch, run the following command:
 #    tune run eleuther_eval --config qwen2/evaluation
 
+output_dir: ./ # Not needed
+
 # Model Arguments
 model:
   _component_: torchtune.models.qwen2.qwen2_7b
@@ -17,7 +19,7 @@ checkpointer:
     model-00003-of-00004.safetensors,
     model-00004-of-00004.safetensors
   ]
-  output_dir: ./ # Not needed
+  output_dir: ${output_dir}
   model_type: QWEN2
 
 # Tokenizer