slurmray 3.3.0__py3-none-any.whl → 3.3.2__py3-none-any.whl

slurmray/RayLauncher.py

@@ -6,6 +6,7 @@ import os
 import dill
 import paramiko
 from getpass import getpass
+import re
 
 dill.settings["recurse"] = True
 
@@ -18,6 +19,7 @@ class RayLauncher:
         project_name: str = None,
         func: Callable = None,
         args: dict = None,
+        files: List[str] = [],
         modules: List[str] = [],
         node_nbr: int = 1,
         use_gpu: bool = False,
@@ -33,6 +35,7 @@ class RayLauncher:
             project_name (str, optional): Name of the project. Defaults to None.
             func (Callable, optional): Function to execute. This function should not be remote but can use ray ressources. Defaults to None.
             args (dict, optional): Arguments of the function. Defaults to None.
+            files (List[str], optional): List of files to push to the cluster. This path must be **relative** to the project directory. Defaults to [].
             modules (List[str], optional): List of modules to load on the curnagl Cluster. Use `module spider` to see available modules. Defaults to None.
             node_nbr (int, optional): Number of nodes to use. Defaults to 1.
             use_gpu (bool, optional): Use GPU or not. Defaults to False.
@@ -46,6 +49,7 @@ class RayLauncher:
         self.project_name = project_name
         self.func = func
         self.args = args
+        self.files = files
         self.node_nbr = node_nbr
         self.use_gpu = use_gpu
         self.memory = memory
@@ -58,7 +62,7 @@ class RayLauncher:
             mod for mod in modules if mod not in ["gcc", "python/3.9.13"]
         ]
         if self.use_gpu is True and "cuda" not in self.modules:
-            self.modules += ["cuda/11.8.0", "cudnn"]
+            self.modules += ["cuda", "cudnn"]
 
         # Check if this code is running on a cluster
         self.cluster = os.path.exists("/usr/bin/sbatch")
@@ -116,6 +120,25 @@ class RayLauncher:
             result = dill.load(f)
 
         return result
+    
+    def __push_file(self, file_path: str, sftp: paramiko.SFTPClient, ssh_client: paramiko.SSHClient):
+        """Push a file to the cluster
+
+        Args:
+            file_path (str): Path to the file to push. This path must be **relative** to the project directory.
+        """
+        print(f"Pushing file {os.path.basename(file_path)} to the cluster...")
+        
+        # Determine the path to the file
+        local_path = file_path
+        local_path_from_pwd = os.path.relpath(local_path, self.pwd_path)
+        cluster_path = os.path.join("/users", self.server_username, "slurmray-server", ".slogs", "server", local_path_from_pwd)
+        
+        # Create the directory if not exists
+        ssh_client.exec_command(f"mkdir -p '{os.path.dirname(cluster_path)}'")
+        
+        # Copy the file to the server
+        sftp.put(file_path, cluster_path)
 
     def serialize_func_and_args(self, func: Callable = None, args: list = None):
         """Serialize the function and the arguments
@@ -382,16 +405,33 @@ class RayLauncher:
         subprocess.run(
             [f"pip freeze > {self.project_path}/requirements.txt"], shell=True
         )
-        # Add slurmray --pre
-        with open(f"{self.project_path}/requirements.txt", "r") as file:
-            requirements  = file.read()
-            requirements += "\nslurmray --pre"
+        
+        with open(f"{self.project_path}/requirements.txt", 'r') as file:
+            lines = file.readlines()
+            # Add slurmray --pre
+            lines.append("slurmray --pre")
+            # Adapt torch version
+            lines = [re.sub(r'\ntorch==.*', 'torch', line) for line in lines]
+            lines = [re.sub(r'\ntorchvision==.*', 'torchvision', line) for line in lines]
+            lines = [re.sub(r'\ntorchaudio==.*', 'torchaudio', line) for line in lines]
+            lines = [re.sub(r'\nbitsandbytes==.*', 'bitsandbytes', line) for line in lines]
             
+        with open(f"{self.project_path}/requirements.txt", 'w') as file:
+            file.writelines(lines)
 
         # Copy files from the project to the server
         for file in os.listdir(self.project_path):
             if file.endswith(".py") or file.endswith(".pkl") or file.endswith(".sh"):
                 sftp.put(os.path.join(self.project_path, file), file)
+                
+        
+        # Create the server directory and remove old files
+        ssh_client.exec_command(
+            "mkdir -p slurmray-server/.slogs/server && rm -rf slurmray-server/.slogs/server/*"
+        )
+        # Copy user files to the server
+        for file in self.files:
+            self.__push_file(file, sftp, ssh_client)
         # Copy the requirements.txt to the server
         sftp.put(
             os.path.join(self.project_path, "requirements.txt"), "requirements.txt"
@@ -456,18 +496,20 @@ class RayLauncher:
 # ---------------------------------------------------------------------------- #
 if __name__ == "__main__":
     import ray
-    import torch
 
-    def function_inside_function(x):
-        return ray.cluster_resources(), x + 1
+    def function_inside_function():
+        with open("slurmray/RayLauncher.py", "r") as f:
+            return f.read()[0:10]
 
     def example_func(x):
-        return function_inside_function(x), torch.cuda.is_available()
+        result = ray.cluster_resources(), x + 1, function_inside_function()
+        return result
 
     launcher = RayLauncher(
         project_name="example",
         func=example_func,
         args={"x": 1},
+        files=["slurmray/RayLauncher.py"],
         modules=[],
         node_nbr=1,
         use_gpu=True,

slurmray/assets/slurmray_server.sh

@@ -2,12 +2,6 @@
 
 echo "Installing slurmray server"
 
-# Create a folder if not exists
-mkdir -p slurmray-server/.slogs/server
-
-# Remove every old file in the server folder
-rm -rf slurmray-server/.slogs/server/*
-
 # Copy files
 mv -t slurmray-server requirements.txt slurmray_server.py
 mv -t slurmray-server/.slogs/server func.pkl args.pkl 

slurmray/assets/spython_template.py

@@ -1,9 +1,13 @@
 import ray
 import dill
 import os
+import sys
 
 PROJECT_PATH = {{PROJECT_PATH}}
 
+# Add the project path to the python path
+sys.path.append(PROJECT_PATH)
+
 # Start the ray cluster
 ray.init({{LOCAL_MODE}})
 

{slurmray-3.3.0.dist-info → slurmray-3.3.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: slurmray
-Version: 3.3.0
+Version: 3.3.2
 Summary: SlurmRay is a module for effortlessly distributing tasks on a Slurm cluster using the Ray library. 
 Home-page: https://henri-jamet.vercel.app/
 License: Apache License
@@ -13,7 +13,6 @@ Requires-Dist: dill (>=0.3.7,<0.4.0)
 Requires-Dist: paramiko (>=3.3.1,<4.0.0)
 Requires-Dist: pdoc3 (>=0.10.0,<0.11.0)
 Requires-Dist: ray[data,serve,train,tune] (>=2.7.1,<3.0.0)
-Requires-Dist: torch (>=2.1.1,<3.0.0)
 Project-URL: Documentation, https://henri-jamet.vercel.app/cards/documentation/slurm-ray/slurm-ray/
 Description-Content-Type: text/markdown
 

slurmray-3.3.2.dist-info/RECORD

@@ -0,0 +1,10 @@
+slurmray/RayLauncher.py,sha256=KqzDqAnKg_LmApYoqID_0bETMwFCEfPj_bQ-ErrvkM8,20251
+slurmray/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+slurmray/assets/sbatch_template.sh,sha256=c-7J4ItzrctDrbF5Znu8p1d_xIgayC9puhjX3nLMzsk,2273
+slurmray/assets/slurmray_server.sh,sha256=-Mt6CYjGdxui_wlD5NwKcv0wge23AjQqIscPbTA14mg,578
+slurmray/assets/slurmray_server_template.py,sha256=4VPJWq9sqW8BBarJB0XKvdQzbDomeHCIiqt0SBVnNIo,573
+slurmray/assets/spython_template.py,sha256=kRUvNQs9iCcg0wJLmm9LV0TnbUdlenZMYPr_bZPkXLg,597
+slurmray-3.3.2.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+slurmray-3.3.2.dist-info/METADATA,sha256=Aula9_eShrZcrtj_oNRkxPAl0_WyWnec3umnI7epAqk,2292
+slurmray-3.3.2.dist-info/WHEEL,sha256=d2fvjOD7sXsVzChCqf0Ty0JbHKBaLYwDbGQDwQTnJ50,88
+slurmray-3.3.2.dist-info/RECORD,,

slurmray-3.3.0.dist-info/RECORD

@@ -1,10 +0,0 @@
-slurmray/RayLauncher.py,sha256=UhE65i3_bEK4mD5KkGzbOLcszIY35PCRcDHtYewflUc,18230
-slurmray/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-slurmray/assets/sbatch_template.sh,sha256=c-7J4ItzrctDrbF5Znu8p1d_xIgayC9puhjX3nLMzsk,2273
-slurmray/assets/slurmray_server.sh,sha256=VVyeM8cdrPeCm24VTTcgrFX1inxUAhXPovjMl_sa1wU,735
-slurmray/assets/slurmray_server_template.py,sha256=4VPJWq9sqW8BBarJB0XKvdQzbDomeHCIiqt0SBVnNIo,573
-slurmray/assets/spython_template.py,sha256=JwOQxCvM-wW4CDUri_sgIKGtigvqGlVI0fBJDvBVmjY,513
-slurmray-3.3.0.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-slurmray-3.3.0.dist-info/METADATA,sha256=r39imnnXJ_nQqD37KM_6i4mZDgcXwLJZNlFO0QHLXq0,2330
-slurmray-3.3.0.dist-info/WHEEL,sha256=d2fvjOD7sXsVzChCqf0Ty0JbHKBaLYwDbGQDwQTnJ50,88
-slurmray-3.3.0.dist-info/RECORD,,