From 8b9d30c48c1b904c356c2988e10a8e2c06857984 Mon Sep 17 00:00:00 2001
From: DNXie <dxie@meta.com>
Date: Thu, 16 Oct 2025 15:57:42 -0700
Subject: [PATCH] <Replace this line with a title. Use 1 line only, 67 chars or
 less>

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:
---
 apps/grpo/qwen3_1_7b.yaml | 5 +++--
 apps/grpo/qwen3_32b.yaml  | 5 +++--
 apps/grpo/qwen3_8b.yaml   | 5 +++--
 apps/sft/llama3_8b.yaml   | 5 +++--
 apps/sft/qwen3_8b.yaml    | 5 +++--
 5 files changed, 15 insertions(+), 10 deletions(-)

diff --git a/apps/grpo/qwen3_1_7b.yaml b/apps/grpo/qwen3_1_7b.yaml
index 14e4871cf..057ee2bcf 100644
--- a/apps/grpo/qwen3_1_7b.yaml
+++ b/apps/grpo/qwen3_1_7b.yaml
@@ -74,8 +74,9 @@ trainer:
     disable_loss_parallel: true
   checkpoint:
     enable: true
-    initial_load_path: hf://${model}
-    initial_load_in_hf: true
+    folder: ./checkpoint              # The folder to save checkpoints to.
+    initial_load_path: hf://${model}  # The path to load the initial checkpoint from. Ignored if `folder` exists.
+    initial_load_in_hf: true          # If true, interpret initial_load_path as a HuggingFace model repo
     last_save_in_hf: true
     interval: 500
     async_mode: "disabled"
diff --git a/apps/grpo/qwen3_32b.yaml b/apps/grpo/qwen3_32b.yaml
index e7a0cf509..c39260b45 100644
--- a/apps/grpo/qwen3_32b.yaml
+++ b/apps/grpo/qwen3_32b.yaml
@@ -77,8 +77,9 @@ trainer:
     disable_loss_parallel: true
   checkpoint:
     enable: true
-    initial_load_path: hf://${model}
-    initial_load_in_hf: true
+    folder: ./checkpoint              # The folder to save checkpoints to.
+    initial_load_path: hf://${model}  # The path to load the initial checkpoint from. Ignored if `folder` exists.
+    initial_load_in_hf: true          # If true, interpret initial_load_path as a HuggingFace model repo
     last_save_in_hf: true
     interval: 500
     async_mode: "disabled"
diff --git a/apps/grpo/qwen3_8b.yaml b/apps/grpo/qwen3_8b.yaml
index 534e5b92a..b0b4bf96e 100644
--- a/apps/grpo/qwen3_8b.yaml
+++ b/apps/grpo/qwen3_8b.yaml
@@ -70,8 +70,9 @@ trainer:
     disable_loss_parallel: true
   checkpoint:
     enable: true
-    initial_load_path: hf://${model}
-    initial_load_in_hf: true
+    folder: ./checkpoint              # The folder to save checkpoints to.
+    initial_load_path: hf://${model}  # The path to load the initial checkpoint from. Ignored if `folder` exists.
+    initial_load_in_hf: true          # If true, interpret initial_load_path as a HuggingFace model repo
     last_save_in_hf: true
     interval: 500
     async_mode: "disabled"
diff --git a/apps/sft/llama3_8b.yaml b/apps/sft/llama3_8b.yaml
index 43a690c1e..44e4485e4 100644
--- a/apps/sft/llama3_8b.yaml
+++ b/apps/sft/llama3_8b.yaml
@@ -45,8 +45,9 @@ parallelism:
 
 checkpoint:
   enable: true
-  initial_load_path: hf://${model_name}
-  initial_load_in_hf: true
+  folder: ./checkpoint              # The folder to save checkpoints to.
+  initial_load_path: hf://${model}  # The path to load the initial checkpoint from. Ignored if `folder` exists.
+  initial_load_in_hf: true          # If true, interpret initial_load_path as a HuggingFace model repo
   last_save_in_hf: true
   interval: 500
   async_mode: "disabled"
diff --git a/apps/sft/qwen3_8b.yaml b/apps/sft/qwen3_8b.yaml
index 2ab88bbd3..1c0d5bc8b 100644
--- a/apps/sft/qwen3_8b.yaml
+++ b/apps/sft/qwen3_8b.yaml
@@ -44,8 +44,9 @@ parallelism:
 
 checkpoint:
   enable: true
-  initial_load_path: hf://${model_name}
-  initial_load_in_hf: true
+  folder: ./checkpoint              # The folder to save checkpoints to.
+  initial_load_path: hf://${model}  # The path to load the initial checkpoint from. Ignored if `folder` exists.
+  initial_load_in_hf: true          # If true, interpret initial_load_path as a HuggingFace model repo
   last_save_in_hf: true
   interval: 500
   async_mode: "disabled"