From 46a1ef0382ed851bfc5dcfe75fc55f401d5f1d00 Mon Sep 17 00:00:00 2001 From: ebsmothers Date: Thu, 19 Dec 2024 14:07:20 -0800 Subject: [PATCH] Set teacher ckptr output_dir to match student in KD configs (#2185) --- recipes/configs/llama3_2/8B_to_1B_KD_lora_distributed.yaml | 2 +- recipes/configs/llama3_2/8B_to_1B_KD_lora_single_device.yaml | 2 +- recipes/configs/qwen2/1.5_to_0.5B_KD_lora_distributed.yaml | 2 +- recipes/configs/qwen2/1.5_to_0.5B_KD_lora_single_device.yaml | 2 +- 4 files changed, 4 insertions(+), 4 deletions(-) diff --git a/recipes/configs/llama3_2/8B_to_1B_KD_lora_distributed.yaml b/recipes/configs/llama3_2/8B_to_1B_KD_lora_distributed.yaml index 3b5d34f13f..9ab51fc0e3 100644 --- a/recipes/configs/llama3_2/8B_to_1B_KD_lora_distributed.yaml +++ b/recipes/configs/llama3_2/8B_to_1B_KD_lora_distributed.yaml @@ -59,7 +59,7 @@ teacher_checkpointer: model-00004-of-00004.safetensors ] recipe_checkpoint: null - output_dir: /tmp/Meta-Llama-3.1-8B-Instruct/ + output_dir: ${output_dir} model_type: LLAMA3 # Dataset and Sampler diff --git a/recipes/configs/llama3_2/8B_to_1B_KD_lora_single_device.yaml b/recipes/configs/llama3_2/8B_to_1B_KD_lora_single_device.yaml index 7ab0f23bc2..0a2dfea9f5 100644 --- a/recipes/configs/llama3_2/8B_to_1B_KD_lora_single_device.yaml +++ b/recipes/configs/llama3_2/8B_to_1B_KD_lora_single_device.yaml @@ -59,7 +59,7 @@ teacher_checkpointer: model-00004-of-00004.safetensors ] recipe_checkpoint: null - output_dir: /tmp/Meta-Llama-3.1-8B-Instruct/ + output_dir: ${output_dir} model_type: LLAMA3 # Dataset and Sampler diff --git a/recipes/configs/qwen2/1.5_to_0.5B_KD_lora_distributed.yaml b/recipes/configs/qwen2/1.5_to_0.5B_KD_lora_distributed.yaml index faae6cd3b8..b68704cdc2 100644 --- a/recipes/configs/qwen2/1.5_to_0.5B_KD_lora_distributed.yaml +++ b/recipes/configs/qwen2/1.5_to_0.5B_KD_lora_distributed.yaml @@ -51,7 +51,7 @@ teacher_checkpointer: hf_model_0001_0.pt ] recipe_checkpoint: null - output_dir: /tmp/Qwen2-1.5B-Instruct-lora-finetune + output_dir: ${output_dir} model_type: QWEN2 resume_from_checkpoint: False diff --git a/recipes/configs/qwen2/1.5_to_0.5B_KD_lora_single_device.yaml b/recipes/configs/qwen2/1.5_to_0.5B_KD_lora_single_device.yaml index 48dbe4b117..385c1d453a 100644 --- a/recipes/configs/qwen2/1.5_to_0.5B_KD_lora_single_device.yaml +++ b/recipes/configs/qwen2/1.5_to_0.5B_KD_lora_single_device.yaml @@ -51,7 +51,7 @@ teacher_checkpointer: model.safetensors ] recipe_checkpoint: null - output_dir: /tmp/Qwen2-1.5B-Instruct + output_dir: ${output_dir} model_type: QWEN2 resume_from_checkpoint: False