Update DPO Max Seq Len (#2176)

Author: pbontrager

recipes/configs/llama2/7B_lora_dpo.yaml

@@ -32,7 +32,7 @@ model:
 tokenizer:
   _component_: torchtune.models.llama2.llama2_tokenizer
   path: /tmp/Llama-2-7b-hf/tokenizer.model
-  max_seq_len: 1024
+  max_seq_len: 1024 # higher increases memory
 
 checkpointer:
   _component_: torchtune.training.FullModelHFCheckpointer

recipes/configs/llama2/7B_lora_dpo_single_device.yaml

@@ -31,7 +31,7 @@ model:
 tokenizer:
   _component_: torchtune.models.llama2.llama2_tokenizer
   path: /tmp/Llama-2-7b-hf/tokenizer.model
-  max_seq_len: 1024
+  max_seq_len: 1024 # higher increases memory
 
 checkpointer:
   _component_: torchtune.training.FullModelHFCheckpointer

recipes/configs/llama3_1/8B_lora_dpo.yaml

@@ -32,7 +32,7 @@ model:
 tokenizer:
   _component_: torchtune.models.llama3.llama3_tokenizer
   path: /tmp/Meta-Llama-3.1-8B-Instruct/original/tokenizer.model
-  max_seq_len: null
+  max_seq_len: 1024 # higher increases memory
 
 checkpointer:
   _component_: torchtune.training.FullModelHFCheckpointer

recipes/configs/llama3_1/8B_lora_dpo_single_device.yaml

@@ -31,7 +31,7 @@ model:
 tokenizer:
   _component_: torchtune.models.llama3.llama3_tokenizer
   path: /tmp/Meta-Llama-3.1-8B-Instruct/original/tokenizer.model
-  max_seq_len: null
+  max_seq_len: 1024 # higher increases memory
 
 checkpointer:
   _component_: torchtune.training.FullModelHFCheckpointer