Fix checkpoint error

.github/workflows/lint.yml

@@ -5,7 +5,7 @@ on: [push, pull_request]
 jobs:
   Lint:
     name: Lint
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-24.04
     steps:
       - uses: actions/checkout@v3
         with:

.github/workflows/pipelines.yml

@@ -12,7 +12,7 @@ on:
 jobs:
   Pipelines-Test:
     name: Pipelines-Test
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-24.04
     steps:
       - uses: actions/checkout@v3
       - uses: actions/setup-python@v4

.github/workflows/release.yml

@@ -16,7 +16,7 @@ on:
 jobs:
   Release:
     name: Release
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-24.04
     steps:
       - uses: actions/checkout@v2
       - uses: actions/setup-python@v1

.github/workflows/stale.yml

@@ -13,7 +13,7 @@ permissions:
 
 jobs:
   stale:
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-24.04
 
     permissions:
       issues: write

.github/workflows/tests.yml

@@ -5,7 +5,7 @@ on: [push, pull_request]
 jobs:
   Test:
     name: Test
-    runs-on: ubuntu-20.04
+    runs-on: ubuntu-24.04
     permissions:
       pull-requests: write
       contents: read

docs/.readthedocs.yaml

@@ -5,7 +5,7 @@
 # Required
 version: 2
 build:
-  os: "ubuntu-20.04"
+  os: "ubuntu-24.04"
   tools:
     python: "3.10"
 

paddlenlp/trainer/trainer_utils.py

@@ -130,8 +130,9 @@ def _get_distributed_seeds(seed: int = 1234, topo: Topology = None):
         pp_rank, pp_size = 0, 1
         dp_rank, dp_size = 0, 1
         sharding_rank, _ = 0, 1
+    # in mpu/random.py , local_seed = seed + 1 + mp_rank * pp_size + pp_rank , so here add a offset to avoid same with local_seed
+    seed_offset = seed + 2 + mp_size * pp_size + pp_size
 
-    seed_offset = seed
     global_seed = (
         seed_offset
         + sep_rank * (mp_size)