update outputs path so that we can mount workspace to /workspace/data (#1623)

* update outputs path so that we can mount workspace to /workspace/data

* fix ln order

examples/cerebras/btlm-ft.yml

@@ -38,7 +38,7 @@ wandb_watch:
 wandb_name:
 wandb_log_model:
 
-output_dir: btlm-out
+output_dir: ./outputs/btlm-out
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 1

examples/cerebras/qlora.yml

@@ -25,7 +25,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 batch_size: 4
 micro_batch_size: 4
 num_epochs: 2

examples/code-llama/13b/lora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/code-llama/13b/qlora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/code-llama/34b/lora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/code-llama/34b/qlora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/code-llama/7b/lora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/code-llama/7b/qlora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/colab-notebooks/colab-axolotl-example.ipynb

@@ -84,7 +84,7 @@
         "    type: alpaca\n",
         "dataset_prepared_path:\n",
         "val_set_size: 0.05\n",
-        "output_dir: ./qlora-out\n",
+        "output_dir: ./outputs/qlora-out\n",
         "\n",
         "adapter: qlora\n",
         "lora_model_dir:\n",

examples/dbrx/16bit-lora.yaml

@@ -10,7 +10,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.0
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 512
 sample_packing: false

examples/dbrx/8bit-lora.yaml

@@ -10,7 +10,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.0
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 512
 sample_packing: false

examples/dbrx/fft-ds-zero3.yaml

@@ -10,7 +10,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.0
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 512
 sample_packing: false

examples/falcon/config-7b-lora.yml

@@ -28,7 +28,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./falcon-7b
+output_dir: ./outputs/falcon-7b
 batch_size: 2
 micro_batch_size: 1
 num_epochs: 4

examples/falcon/config-7b-qlora.yml

@@ -42,7 +42,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 # QLoRA paper Table 9
 # - 16 for 7b & 13b

examples/falcon/config-7b.yml

@@ -28,7 +28,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./falcon-7b
+output_dir: ./outputs/falcon-7b
 batch_size: 2
 micro_batch_size: 1
 num_epochs: 4

examples/gemma/qlora.yml

@@ -12,7 +12,7 @@ datasets:
   - path: mhenrichsen/alpaca_2k_test
     type: alpaca
 val_set_size: 0.1
-output_dir: ./out
+output_dir: ./outputs/out
 
 adapter: qlora
 lora_r: 32

examples/gptj/qlora.yml

@@ -23,7 +23,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 gradient_accumulation_steps: 2
 micro_batch_size: 2
 num_epochs: 2

examples/jamba/qlora.yaml

@@ -10,7 +10,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.0
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 4096
 sample_packing: false

examples/jamba/qlora_deepspeed.yaml

@@ -10,7 +10,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.0
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 4096
 sample_packing: false

examples/jeopardy-bot/config.yml

@@ -21,7 +21,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./jeopardy-bot-7b
+output_dir: ./outputs/jeopardy-bot-7b
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 4

examples/llama-2/fft_optimized.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.05
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 4096
 sample_packing: true

examples/llama-2/gptq-lora.yml

@@ -33,7 +33,7 @@ wandb_project:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./model-out
+output_dir: ./outputs/model-out
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 4

examples/llama-2/lisa.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.05
-output_dir: ./lisa-out
+output_dir: ./outputs/lisa-out
 
 sequence_len: 4096
 sample_packing: true

examples/llama-2/loftq.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/llama-2/lora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/llama-2/qlora-fsdp.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.05
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/llama-2/qlora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/llama-2/relora.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./relora-out
+output_dir: ./outputs/relora-out
 
 adapter: qlora
 lora_model_dir:

examples/llama-3/fft-8b.yaml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.05
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 8192
 sample_packing: true

examples/llama-3/lora-8b.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/llama-3/qlora-fsdp-70b.yaml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.05
-output_dir: ./out/qlora-llama3-70b
+output_dir: ./outputs/out/qlora-llama3-70b
 
 adapter: qlora
 lora_model_dir:

examples/llama-3/qlora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/mamba/config.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.0
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 2048
 sample_packing: false

examples/mistral/bigstral-ds-zero3.yaml

@@ -23,7 +23,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.05
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 2048
 sample_packing: true

examples/mistral/config.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 8192
 sample_packing: true

examples/mistral/lora-mps.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 eval_sample_packing: false
 
 adapter: lora

examples/mistral/lora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.1
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 adapter: lora
 lora_model_dir:

examples/mistral/mistral-qlora-fsdp.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.02
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 model_config:
   output_router_logits: true

examples/mistral/mistral-qlora-orpo.yml

@@ -16,7 +16,7 @@ datasets:
     type: chat_template.argilla
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.1
-output_dir: ./mistral-qlora-orpo-out
+output_dir: ./outputs/mistral-qlora-orpo-out
 
 adapter: qlora
 lora_model_dir:

examples/mistral/mixtral-8x22b-qlora-fsdp.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.02
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 model_config:
   output_router_logits: true

examples/mistral/mixtral-qlora-fsdp.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.02
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 model_config:
   output_router_logits: true

examples/mistral/mixtral.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.0
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 ## You can optionally freeze the entire model and unfreeze a subset of parameters
 unfrozen_parameters:

examples/mistral/mixtral_22.yml

@@ -21,7 +21,7 @@ model_config:
 datasets:
   - path: yahma/alpaca-cleaned
     type: alpaca
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 8000
 sample_packing: true

examples/mistral/qlora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.1
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/mpt-7b/config.yml

@@ -23,7 +23,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./mpt-alpaca-7b
+output_dir: ./outputs/mpt-alpaca-7b
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 4

examples/openllama-3b/config.yml

@@ -25,7 +25,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./openllama-out
+output_dir: ./outputs/openllama-out
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 4

examples/openllama-3b/lora.yml

@@ -31,7 +31,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 gradient_accumulation_steps: 1
 micro_batch_size: 2
 num_epochs: 4

examples/openllama-3b/qlora.yml

@@ -25,7 +25,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 gradient_accumulation_steps: 1
 micro_batch_size: 2
 num_epochs: 4

examples/phi/phi-ft.yml

@@ -12,7 +12,7 @@ datasets:
 
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./phi-sft-out
+output_dir: ./outputs/phi-sft-out
 
 sequence_len: 2048
 sample_packing: true

examples/phi/phi-qlora.yml

@@ -12,7 +12,7 @@ datasets:
 
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./phi-sft-out
+output_dir: ./outputs/phi-sft-out
 
 sequence_len: 2048
 sample_packing: true

examples/phi/phi2-ft.yml

@@ -12,7 +12,7 @@ datasets:
 
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./phi-sft-out
+output_dir: ./outputs/phi-sft-out
 
 sequence_len: 2048
 sample_packing: true

examples/pythia-12b/config.yml

@@ -26,7 +26,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./pythia-12b
+output_dir: ./outputs/pythia-12b
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 5

examples/pythia/lora.yml

@@ -20,7 +20,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./lora-alpaca-pythia
+output_dir: ./outputs/lora-alpaca-pythia
 gradient_accumulation_steps: 1
 micro_batch_size: 4
 num_epochs: 4

examples/qwen/lora.yml

@@ -13,7 +13,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 2048  # supports up to 8192
 sample_packing: false

examples/qwen/qlora.yml

@@ -13,7 +13,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 2048  # supports up to 8192
 sample_packing: false

examples/qwen/qwen2-moe-lora.yaml

@@ -10,7 +10,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 1024  # supports up to 32k
 sample_packing: false

examples/qwen/qwen2-moe-qlora.yaml

@@ -10,7 +10,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 1024  # supports up to 32k
 sample_packing: false

examples/redpajama/config-3b.yml

@@ -24,7 +24,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./redpajama-alpaca-3b
+output_dir: ./outputs/redpajama-alpaca-3b
 batch_size: 4
 micro_batch_size: 1
 num_epochs: 4

examples/replit-3b/config-lora.yml

@@ -23,7 +23,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./lora-replit
+output_dir: ./outputs/lora-replit
 batch_size: 8
 micro_batch_size: 1
 num_epochs: 4

examples/stablelm-2/1.6b/fft.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.05
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 4096
 sample_packing: true

examples/stablelm-2/1.6b/lora.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/starcoder2/qlora.yml

@@ -11,7 +11,7 @@ datasets:
 
 dataset_prepared_path:
 val_set_size: 0.2
-output_dir: ./qlora
+output_dir: ./outputs/qlora
 
 adapter: qlora
 lora_model_dir:

examples/tiny-llama/lora-mps.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/tiny-llama/lora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/tiny-llama/pretrain.yml

@@ -14,7 +14,7 @@ pretraining_dataset:
   type: pretrain
 dataset_prepared_path:
 val_set_size: 0.0
-output_dir: ./model-out
+output_dir: ./outputs/model-out
 
 sequence_len: 2048
 sample_packing: true

examples/tiny-llama/qlora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/xgen-7b/xgen-7b-8k-qlora.yml

@@ -40,7 +40,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 # QLoRA paper Table 9
 # - 16 for 7b & 13b

examples/yi-34B-chat/qlora.yml

@@ -33,7 +33,7 @@ eval_sample_packing: false
 eval_batch_size: 1
 
 # LoRA
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 adapter: qlora
 lora_model_dir:
 lora_r: 32