add transfer engine for training and inference. #569
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,2 @@ | ||
| # Import for backward compatibility | ||
|
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,31 @@ | ||
| import torch | ||
| import zmq | ||
|
|
||
| from transfer_engine import MooncakeTransferEngine | ||
| from new_tensor import create_tensor | ||
|
|
||
| class MooncakeInference: | ||
|
There was a problem hiding this comment. I think these class names do not precisely reveal what they are actually doing. Maybe the usage is for RL training and inference, but the names are very confusing. |
||
| def __init__(self, config: dict): | ||
| self.config = config | ||
| self._engine = MooncakeTransferEngine( | ||
| hostname = config["inference_ip"], | ||
| gpu_id = config["inference_gpu_id"], # Using GPU | ||
| ib_device = None # No specific IB device | ||
| ) | ||
| self._context = zmq.Context() | ||
| self._socket = self._context.socket(zmq.REQ) | ||
| self._socket.connect(f"tcp://{config['training_ip']}:{config['port']}") | ||
|
|
||
| def __del__(self): | ||
| self._socket.close() | ||
| self._context.destroy() | ||
|
|
||
| def recv_tensor(self) -> dict: | ||
| self._socket.send_json({"session_id": self._engine.get_session_id()}) | ||
| ret = self._socket.recv_json() # ["name", "shape", "dtype"] | ||
| tensor = create_tensor(ret[1], ret[2], self.config["inference_gpu_id"]) | ||
| self._engine.register(tensor.data_ptr(), tensor.numel() * tensor.element_size()) | ||
| self._socket.send_json({"ptr": tensor.data_ptr()}) | ||
| self._socket.recv_json() | ||
| self._engine.deregister(tensor.data_ptr()) | ||
| return {ret[0]: tensor} | ||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change | ||||
|---|---|---|---|---|---|---|
| @@ -0,0 +1,55 @@ | ||||||
| import torch | ||||||
| import zmq | ||||||
| import time | ||||||
| from threading import Thread | ||||||
| from queue import Queue | ||||||
| from new_tensor import get_dtype_str | ||||||
|
|
||||||
| from transfer_engine import MooncakeTransferEngine | ||||||
|
There was a problem hiding this comment. Use a relative import for the local transfer_engine module to avoid import errors (e.g.,
Suggested change
|
||||||
|
|
||||||
| class MooncakeTraining(Thread): | ||||||
| def __init__(self, config: dict): | ||||||
| super().__init__() | ||||||
| self._stop = False | ||||||
| self.config = config | ||||||
| self._queue = Queue(maxsize=config["send_bulk"]) | ||||||
| self.engine = MooncakeTransferEngine( | ||||||
| hostname = config["training_ip"], | ||||||
| gpu_id = config["training_gpu_id"], # Using GPU | ||||||
| ib_device = None # No specific IB device | ||||||
| ) | ||||||
| self._context = zmq.Context() | ||||||
| self._socket = self._context.socket(zmq.REP) | ||||||
| self._socket.bind(f"tcp://*:{config['port']}") | ||||||
| self._report = 0 | ||||||
| self.start() | ||||||
|
|
||||||
| def __del__(self): | ||||||
| self._socket.close() | ||||||
| self._context.destroy() | ||||||
|
|
||||||
|
|
||||||
| def reg_tensor(self, name: str, tensor: torch.Tensor): | ||||||
| self._queue.put((name, tensor)) | ||||||
|
|
||||||
| def run(self): | ||||||
| while True: | ||||||
| name, tensor = self._queue.get() | ||||||
| if name is None: | ||||||
| break | ||||||
| ret = self._socket.recv_json() # get req. | ||||||
| session_id = ret["session_id"] | ||||||
| self._socket.send_json([name, tensor.shape, get_dtype_str(tensor.dtype)]) | ||||||
| ptr, size = tensor.data_ptr(), tensor.numel() * tensor.element_size() | ||||||
| self.engine.register(ptr, size) | ||||||
| ret = self._socket.recv_json() | ||||||
| t1 = time.time() | ||||||
| self.engine.transfer_sync(session_id, ptr, ret["ptr"], size) | ||||||
| t2 = time.time() | ||||||
| self._report += t2 - t1 | ||||||
| self._socket.send_json(ret) | ||||||
| self.engine.deregister(ptr) | ||||||
|
|
||||||
| def report(self): | ||||||
| return self._report | ||||||
|
|
||||||
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Use a relative import for the local transfer_engine module to avoid import errors (e.g.,
from .transfer_engine import MooncakeTransferEngine).