Change the bucket for different object (deepjavalibrary#691)

Lokiiiiii · May 5, 2023 · 98897f9 · 98897f9
1 parent 72fcac8
commit 98897f9
Show file tree

Hide file tree

Showing 2 changed files with 8 additions and 8 deletions.
diff --git a/.github/workflows/llm_integration.yml b/.github/workflows/llm_integration.yml
@@ -463,7 +463,7 @@ jobs:
           train
 
           # checking if checkpoint files are generated.
-          sudo /opt/djl/bin/s5cmd --retry-count 1 sync s3://djl-llm/t5-small-tp4/* $PWD/models/partition-test
+          sudo /opt/djl/bin/s5cmd --retry-count 1 sync s3://djl-llm/t5-small-tp4/ft-aot/* $PWD/models/partition-test
           if grep -q t5-small-fp32-4-1 $PWD/models/partition-test/*-gpu/verify ; then echo "checkpoint files generated"; else exit 1;  fi
       - name: Test t5-small-inference
         working-directory: tests/integration
@@ -590,7 +590,7 @@ jobs:
           train
           
           # checking if checkpoint files are generated.
-          /opt/djl/bin/s5cmd --retry-count 1 sync s3://djl-llm/t5-small-tp4/* $PWD/models/partition-test
+          /opt/djl/bin/s5cmd --retry-count 1 sync s3://djl-llm/t5-small-tp4/ft-aot-handler/* $PWD/models/partition-test
           if grep -q t5-small-fp32-4-1 $PWD/models/partition-test/*-gpu/verify ; then echo "checkpoint files generated"; else exit 1;  fi
       - name: Test t5-small-inference
         working-directory: tests/integration
@@ -727,7 +727,7 @@ jobs:
           
           # checking if pt files are generated.
           mkdir $PWD/models/partition-test
-          /opt/djl/bin/s5cmd --retry-count 1 sync s3://djl-llm/bloom-7b1-tp4/* $PWD/models/partition-test
+          /opt/djl/bin/s5cmd --retry-count 1 sync s3://djl-llm/bloom-7b1-tp4/ds-aot/* $PWD/models/partition-test
           if ls $PWD/models/partition-test/*.pt &>/dev/null ; then echo "checkpoint files generated"; else exit 1;  fi
           if ls $PWD/models/partition-test/ds_inference_config.json &>/dev/null ; \
           then echo "ds_inference_config.json generated"; else exit 1;  fi
@@ -818,7 +818,7 @@ jobs:
 
           # checking if pt files are generated.
           # downloads the uploaded partitioned checkpoints from s3url.
-          /opt/djl/bin/s5cmd --retry-count 1 sync s3://djl-llm/bloom-7b1-tp4/* $PWD/models/partition-test
+          /opt/djl/bin/s5cmd --retry-count 1 sync s3://djl-llm/bloom-7b1-tp4/ds-aot-handler/* $PWD/models/partition-test
           if ls $PWD/models/partition-test/*.pt &>/dev/null ; then echo "checkpoint files generated"; else exit 1;  fi
           if ls $PWD/models/partition-test/ds_inference_config.json &>/dev/null ; \
           then echo "ds_inference_config.json generated"; else exit 1;  fi

diff --git a/tests/integration/llm/prepare.py b/tests/integration/llm/prepare.py
@@ -33,7 +33,7 @@
         "option.tensor_parallel_degree": 4,
         "option.task": "text-generation",
         "option.dtype": "float16",
-        "option.save_mp_checkpoint_path": "s3://djl-llm/bloom-7b1-tp4"
+        "option.save_mp_checkpoint_path": "s3://djl-llm/bloom-7b1-tp4/ds-aot/"
     }
 }
 
@@ -50,7 +50,7 @@
         "option.tensor_parallel_degree": 4,
         "option.task": "text-generation",
         "option.dtype": "fp16",
-        "option.save_mp_checkpoint_path": "s3://djl-llm/bloom-7b1-tp4"
+        "option.save_mp_checkpoint_path": "s3://djl-llm/bloom-7b1-tp4/ds-aot-handler/"
     }
 }
 
@@ -388,7 +388,7 @@ def build_ft_raw_aot_model(model):
     options = ft_model_list[model]
     options["engine"] = "FasterTransformer"
     if model == 't5-small':
-        options["option.save_mp_checkpoint_path"] = "s3://djl-llm/t5-small-tp4"
+        options["option.save_mp_checkpoint_path"] = "s3://djl-llm/t5-small-tp4/ft-aot/"
     else:
         options["option.save_mp_checkpoint_path"] = "/opt/ml/input/data/training/partition-test"
     write_properties(options)
@@ -404,7 +404,7 @@ def builder_ft_handler_aot_model(model):
     options["engine"] = "FasterTransformer"
     # options["entryPoint"] = "djl_python.fastertransformer"
     if model == 't5-small':
-        options["option.save_mp_checkpoint_path"] = "s3://djl-llm/t5-small-tp4"
+        options["option.save_mp_checkpoint_path"] = "s3://djl-llm/t5-small-tp4/ft-aot-handler/"
     else:
         options["option.save_mp_checkpoint_path"] = "/opt/ml/input/data/training/partition-test"
     write_properties(options)