[misc] refactor launch API and tensor constructor (#5666)

* [misc] remove config arg from initialize

* [misc] remove old tensor contrusctor

* [plugin] add npu support for ddp

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* [devops] fix doc test ci

* [test] fix test launch

* [doc] update launch doc

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Author: ver217

.github/workflows/doc_test_on_pr.yml

@@ -56,7 +56,7 @@ jobs:
     needs: detect-changed-doc
     runs-on: [self-hosted, gpu]
     container:
-      image: hpcaitech/pytorch-cuda:2.0.0-11.7.0
+      image: hpcaitech/pytorch-cuda:2.1.0-12.1.0
       options: --gpus all --rm
     timeout-minutes: 20
     defaults:

applications/Colossal-LLaMA/train.py

@@ -136,7 +136,7 @@ def main() -> None:
     # ==============================
     # Initialize Distributed Training
     # ==============================
-    colossalai.launch_from_torch({})
+    colossalai.launch_from_torch()
     accelerator = get_accelerator()
     coordinator = DistCoordinator()
 

applications/ColossalChat/benchmarks/benchmark_ppo.py

@@ -66,7 +66,7 @@ def benchmark_train(args):
     # ==============================
     # Initialize Distributed Training
     # ==============================
-    colossalai.launch_from_torch({})
+    colossalai.launch_from_torch()
     coordinator = DistCoordinator()
 
     # ======================================================

applications/ColossalChat/examples/training_scripts/train_dpo.py

@@ -37,7 +37,7 @@ def train(args):
     # ==============================
     # Initialize Distributed Training
     # ==============================
-    colossalai.launch_from_torch({})
+    colossalai.launch_from_torch()
     coordinator = DistCoordinator()
 
     # ==============================

applications/ColossalChat/examples/training_scripts/train_ppo.py

@@ -39,7 +39,7 @@ def train(args):
     # ==============================
     # Initialize Distributed Training
     # ==============================
-    colossalai.launch_from_torch({})
+    colossalai.launch_from_torch()
     coordinator = DistCoordinator()
 
     # ======================================================

applications/ColossalChat/examples/training_scripts/train_rm.py

@@ -34,7 +34,7 @@ def train(args):
     # ==============================
     # Initialize Distributed Training
     # ==============================
-    colossalai.launch_from_torch({})
+    colossalai.launch_from_torch()
     coordinator = DistCoordinator()
 
     # ======================================================

applications/ColossalChat/examples/training_scripts/train_sft.py

@@ -29,7 +29,7 @@ def train(args):
     # ==============================
     # Initialize Distributed Training
     # ==============================
-    colossalai.launch_from_torch({})
+    colossalai.launch_from_torch()
     coordinator = DistCoordinator()
 
     # ==============================

applications/ColossalEval/examples/dataset_evaluation/inference.py

@@ -81,7 +81,7 @@ def rm_and_merge(
 
 
 def main(args):
-    colossalai.launch_from_torch(config={}, seed=42)
+    colossalai.launch_from_torch(seed=42)
     accelerator = get_accelerator()
     world_size = dist.get_world_size()
 

applications/ColossalEval/examples/gpt_evaluation/inference.py

@@ -81,7 +81,7 @@ def rm_and_merge(
 
 
 def main(args):
-    colossalai.launch_from_torch(config={}, seed=42)
+    colossalai.launch_from_torch(seed=42)
     world_size = dist.get_world_size()
 
     rank = dist.get_rank()

applications/ColossalMoE/infer.py

@@ -57,7 +57,7 @@ def main():
     args = parse_args()
 
     # Launch ColossalAI
-    colossalai.launch_from_torch(config={}, seed=args.seed)
+    colossalai.launch_from_torch(seed=args.seed)
     coordinator = DistCoordinator()
 
     config = MixtralConfig.from_pretrained(args.model_name)
@@ -96,7 +96,11 @@ def main():
     if coordinator.rank == 0:
         text = ["Hello my name is"]
     else:
-        text = ["What's the largest country in the world?", "How many people live in China?", "帮我续写这首诗：离离原上草"]
+        text = [
+            "What's the largest country in the world?",
+            "How many people live in China?",
+            "帮我续写这首诗：离离原上草",
+        ]
     tokenizer.pad_token = tokenizer.unk_token
     inputs = tokenizer(text, return_tensors="pt", padding=True).to(torch.cuda.current_device())