[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>

examples/language/gpt/experiments/auto_offload/train_gpt_offload.py

@@ -94,8 +94,7 @@ def train_gpt(args):
 
 
 def run(rank, world_size, port, args):
-    config = {}
-    colossalai.launch(config=config, rank=rank, world_size=world_size, host="localhost", port=port, backend="nccl")
+    colossalai.launch(rank=rank, world_size=world_size, host="localhost", port=port, backend="nccl")
     train_gpt(args)
 
 

examples/language/gpt/experiments/auto_parallel/auto_parallel_with_gpt.py

@@ -47,7 +47,7 @@ def get_data(batch_size, seq_len, vocab_size):
 
 def main():
     disable_existing_loggers()
-    launch_from_torch(config={})
+    launch_from_torch()
     logger = get_dist_logger()
     config = transformers.GPT2Config(n_position=SEQ_LENGTH, n_layer=NUM_LAYERS, n_head=NUM_HEADS, n_embd=HIDDEN_DIM)
     if FP16:

examples/language/gpt/gemini/train_gpt_demo.py

@@ -132,7 +132,7 @@ def main():
     PROF_FLAG = False  # The flag of profiling, False by default
 
     disable_existing_loggers()
-    colossalai.launch_from_torch(config={})
+    colossalai.launch_from_torch()
 
     logger = get_dist_logger()
     logger.info(f"{args.model_type}, {args.distplan}, batch size {BATCH_SIZE}", ranks=[0])

examples/language/gpt/hybridparallelism/benchmark.py

@@ -67,7 +67,7 @@ def main():
     parser.add_argument("--cpu_offload", action="store_true", help="Use gradient checkpointing")
     args = parser.parse_args()
 
-    colossalai.launch_from_torch({})
+    colossalai.launch_from_torch()
     coordinator = DistCoordinator()
 
     def empty_init():

examples/language/gpt/hybridparallelism/finetune.py

@@ -196,7 +196,7 @@ def main():
     # ==============================
     # Launch Distributed Environment
     # ==============================
-    colossalai.launch_from_torch(config={}, seed=42)
+    colossalai.launch_from_torch(seed=42)
     coordinator = DistCoordinator()
 
     # local_batch_size = BATCH_SIZE // coordinator.world_size

examples/language/gpt/titans/train_gpt.py

@@ -36,9 +36,9 @@ def main():
     args = parser.parse_args()
     disable_existing_loggers()
     if args.from_torch:
-        colossalai.launch_from_torch(config=args.config)
+        colossalai.launch_from_torch()
     else:
-        colossalai.launch_from_slurm(config=args.config, host=args.host, port=29500, seed=42)
+        colossalai.launch_from_slurm(host=args.host, port=29500, seed=42)
     logger = get_dist_logger()
 
     data_path = None if args.use_dummy_dataset else os.environ["DATA"]