Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
67 changes: 49 additions & 18 deletions python-ff/runner-scripts/setup_multinode.py
Original file line number Diff line number Diff line change
@@ -1,22 +1,53 @@
import os
import json
from pathlib import Path

if 'BACKENDAI_CLUSTER_HOST' in os.environ: # Start mutli-instance setup.
env = {}
env['cluster'] = {}
env['cluster']['worker'] = []
env['cluster']['chief'] = [] # For compatibility
env['cluster']['chief'].append('main1:2220')
for container in os.environ['BACKENDAI_CLUSTER_HOSTS'].split(","):
if container != 'main1':
env['cluster']['worker'].append(container + ":2220")
env['task'] = {}
if os.environ['BACKENDAI_CLUSTER_ROLE'] == 'main':
env['task']['type'] = "chief" # For compatibility. Recent TF choose first worker as chief.
env['task']["index"] = str(int(os.environ['BACKENDAI_CLUSTER_IDX']) - 1) # Index starts from 0

def get_ssh_port(mapping, cluster_host):
if mapping is None:
return 2220
try:
_, port = mapping[cluster_host] # host-ip and port
return port
except KeyError:
return 2220


if __name__ == "__main__":
if "BACKENDAI_CLUSTER_HOST" in os.environ: # Start mutli-instance setup.
try:
ssh_port_mapping = json.loads(Path("/home/config/ssh/port-mapping.json").read_bytes())
except FileNotFoundError:
ssh_port_mapping = None

env = {
"cluster": {
"chief": ["main1:" + str(get_ssh_port(ssh_port_mapping, "main1"))],
"worker": [],
},
}
for cluster_host in os.environ["BACKENDAI_CLUSTER_HOSTS"].split(","):
if cluster_host != "main1":
env["cluster"]["worker"].append(
cluster_host + ":" + str(get_ssh_port(ssh_port_mapping, cluster_host))
)

# TF's worker index starts from 0.
cluster_role_idx = str(int(os.environ["BACKENDAI_CLUSTER_IDX"]) - 1)
if os.environ["BACKENDAI_CLUSTER_ROLE"] == "main":
task = {
"type": "chief",
"index": cluster_role_idx,
}
else:
task = {
"type": "worker",
"index": cluster_role_idx,
}
env["task"] = task
print(json.dumps(env))
else:
env['task']['type'] = "worker"
env['task']["index"] = str(int(os.environ['BACKENDAI_CLUSTER_IDX']) - 1) # Index starts from 0
print(json.dumps(env))
else:
print("")
print("")


# vim: et sts=4 sw=4 tw=105
67 changes: 49 additions & 18 deletions python-pytorch/runner-scripts/setup_multinode.py
Original file line number Diff line number Diff line change
@@ -1,22 +1,53 @@
import os
import json
from pathlib import Path

if 'BACKENDAI_CLUSTER_HOST' in os.environ: # Start mutli-instance setup.
env = {}
env['cluster'] = {}
env['cluster']['worker'] = []
env['cluster']['chief'] = [] # For compatibility
env['cluster']['chief'].append('main1:2220')
for container in os.environ['BACKENDAI_CLUSTER_HOSTS'].split(","):
if container != 'main1':
env['cluster']['worker'].append(container + ":2220")
env['task'] = {}
if os.environ['BACKENDAI_CLUSTER_ROLE'] == 'main':
env['task']['type'] = "chief" # For compatibility. Recent TF choose first worker as chief.
env['task']["index"] = str(int(os.environ['BACKENDAI_CLUSTER_IDX']) - 1) # Index starts from 0

def get_ssh_port(mapping, cluster_host):
if mapping is None:
return 2220
try:
_, port = mapping[cluster_host] # host-ip and port
return port
except KeyError:
return 2220


if __name__ == "__main__":
if "BACKENDAI_CLUSTER_HOST" in os.environ: # Start mutli-instance setup.
try:
ssh_port_mapping = json.loads(Path("/home/config/ssh/port-mapping.json").read_bytes())
except FileNotFoundError:
ssh_port_mapping = None

env = {
"cluster": {
"chief": ["main1:" + str(get_ssh_port(ssh_port_mapping, "main1"))],
"worker": [],
},
}
for cluster_host in os.environ["BACKENDAI_CLUSTER_HOSTS"].split(","):
if cluster_host != "main1":
env["cluster"]["worker"].append(
cluster_host + ":" + str(get_ssh_port(ssh_port_mapping, cluster_host))
)

# TF's worker index starts from 0.
cluster_role_idx = str(int(os.environ["BACKENDAI_CLUSTER_IDX"]) - 1)
if os.environ["BACKENDAI_CLUSTER_ROLE"] == "main":
task = {
"type": "chief",
"index": cluster_role_idx,
}
else:
task = {
"type": "worker",
"index": cluster_role_idx,
}
env["task"] = task
print(json.dumps(env))
else:
env['task']['type'] = "worker"
env['task']["index"] = str(int(os.environ['BACKENDAI_CLUSTER_IDX']) - 1) # Index starts from 0
print(json.dumps(env))
else:
print("")
print("")


# vim: et sts=4 sw=4 tw=105
64 changes: 49 additions & 15 deletions python-pytorchlightning/runner-scripts/setup_multinode.py
Original file line number Diff line number Diff line change
@@ -1,19 +1,53 @@
import os
import json
from pathlib import Path

if 'BACKENDAI_CLUSTER_HOST' in os.environ: # Start mutli-instance setup.
env = {}
env['cluster'] = {}
env['cluster']['worker'] = []
for container in os.environ['BACKENDAI_CLUSTER_HOSTS'].split(","):
env['cluster']['worker'].append(container + ":2220")
env['task'] = {}
if os.environ['BACKENDAI_CLUSTER_ROLE'] == 'main':
env['task']['type'] = "worker" # Was chief. but recent TF choose first worker as chief.
env['task']["index"] = str(int(os.environ['BACKENDAI_CLUSTER_IDX']) - 1) # Index starts from 0

def get_ssh_port(mapping, cluster_host):
if mapping is None:
return 2220
try:
_, port = mapping[cluster_host] # host-ip and port
return port
except KeyError:
return 2220


if __name__ == "__main__":
if "BACKENDAI_CLUSTER_HOST" in os.environ: # Start mutli-instance setup.
try:
ssh_port_mapping = json.loads(Path("/home/config/ssh/port-mapping.json").read_bytes())
except FileNotFoundError:
ssh_port_mapping = None

env = {
"cluster": {
"chief": ["main1:" + str(get_ssh_port(ssh_port_mapping, "main1"))],
"worker": [],
},
}
for cluster_host in os.environ["BACKENDAI_CLUSTER_HOSTS"].split(","):
if cluster_host != "main1":
env["cluster"]["worker"].append(
cluster_host + ":" + str(get_ssh_port(ssh_port_mapping, cluster_host))
)

# TF's worker index starts from 0.
cluster_role_idx = str(int(os.environ["BACKENDAI_CLUSTER_IDX"]) - 1)
if os.environ["BACKENDAI_CLUSTER_ROLE"] == "main":
task = {
"type": "chief",
"index": cluster_role_idx,
}
else:
task = {
"type": "worker",
"index": cluster_role_idx,
}
env["task"] = task
print(json.dumps(env))
else:
env['task']['type'] = "worker"
env['task']["index"] = os.environ['BACKENDAI_CLUSTER_IDX']
print(json.dumps(env))
else:
print("")
print("")


# vim: et sts=4 sw=4 tw=105
67 changes: 49 additions & 18 deletions python-tensorflow/runner-scripts/setup_multinode.py
Original file line number Diff line number Diff line change
@@ -1,22 +1,53 @@
import os
import json
from pathlib import Path

if 'BACKENDAI_CLUSTER_HOST' in os.environ: # Start mutli-instance setup.
env = {}
env['cluster'] = {}
env['cluster']['worker'] = []
env['cluster']['chief'] = [] # For compatibility
env['cluster']['chief'].append('main1:2220')
for container in os.environ['BACKENDAI_CLUSTER_HOSTS'].split(","):
if container != 'main1':
env['cluster']['worker'].append(container + ":2220")
env['task'] = {}
if os.environ['BACKENDAI_CLUSTER_ROLE'] == 'main':
env['task']['type'] = "chief" # For compatibility. Recent TF choose first worker as chief.
env['task']["index"] = str(int(os.environ['BACKENDAI_CLUSTER_IDX']) - 1) # Index starts from 0

def get_ssh_port(mapping, cluster_host):
if mapping is None:
return 2220
try:
_, port = mapping[cluster_host] # host-ip and port
return port
except KeyError:
return 2220


if __name__ == "__main__":
if "BACKENDAI_CLUSTER_HOST" in os.environ: # Start mutli-instance setup.
try:
ssh_port_mapping = json.loads(Path("/home/config/ssh/port-mapping.json").read_bytes())
except FileNotFoundError:
ssh_port_mapping = None

env = {
"cluster": {
"chief": ["main1:" + str(get_ssh_port(ssh_port_mapping, "main1"))],
"worker": [],
},
}
for cluster_host in os.environ["BACKENDAI_CLUSTER_HOSTS"].split(","):
if cluster_host != "main1":
env["cluster"]["worker"].append(
cluster_host + ":" + str(get_ssh_port(ssh_port_mapping, cluster_host))
)

# TF's worker index starts from 0.
cluster_role_idx = str(int(os.environ["BACKENDAI_CLUSTER_IDX"]) - 1)
if os.environ["BACKENDAI_CLUSTER_ROLE"] == "main":
task = {
"type": "chief",
"index": cluster_role_idx,
}
else:
task = {
"type": "worker",
"index": cluster_role_idx,
}
env["task"] = task
print(json.dumps(env))
else:
env['task']['type'] = "worker"
env['task']["index"] = str(int(os.environ['BACKENDAI_CLUSTER_IDX']) - 1) # Index starts from 0
print(json.dumps(env))
else:
print("")
print("")


# vim: et sts=4 sw=4 tw=105
64 changes: 49 additions & 15 deletions vendor/ngc-pytorch/runner-scripts/setup_multinode.py
Original file line number Diff line number Diff line change
@@ -1,19 +1,53 @@
import os
import json
from pathlib import Path

if 'BACKENDAI_CLUSTER_HOST' in os.environ: # Start mutli-instance setup.
env = {}
env['cluster'] = {}
env['cluster']['worker'] = []
for container in os.environ['BACKENDAI_CLUSTER_HOSTS'].split(","):
env['cluster']['worker'].append(container + ":2220")
env['task'] = {}
if os.environ['BACKENDAI_CLUSTER_ROLE'] == 'main':
env['task']['type'] = "worker" # Was chief. but recent TF choose first worker as chief.
env['task']["index"] = str(int(os.environ['BACKENDAI_CLUSTER_IDX']) - 1) # Index starts from 0

def get_ssh_port(mapping, cluster_host):
if mapping is None:
return 2220
try:
_, port = mapping[cluster_host] # host-ip and port
return port
except KeyError:
return 2220


if __name__ == "__main__":
if "BACKENDAI_CLUSTER_HOST" in os.environ: # Start mutli-instance setup.
try:
ssh_port_mapping = json.loads(Path("/home/config/ssh/port-mapping.json").read_bytes())
except FileNotFoundError:
ssh_port_mapping = None

env = {
"cluster": {
"chief": ["main1:" + str(get_ssh_port(ssh_port_mapping, "main1"))],
"worker": [],
},
}
for cluster_host in os.environ["BACKENDAI_CLUSTER_HOSTS"].split(","):
if cluster_host != "main1":
env["cluster"]["worker"].append(
cluster_host + ":" + str(get_ssh_port(ssh_port_mapping, cluster_host))
)

# TF's worker index starts from 0.
cluster_role_idx = str(int(os.environ["BACKENDAI_CLUSTER_IDX"]) - 1)
if os.environ["BACKENDAI_CLUSTER_ROLE"] == "main":
task = {
"type": "chief",
"index": cluster_role_idx,
}
else:
task = {
"type": "worker",
"index": cluster_role_idx,
}
env["task"] = task
print(json.dumps(env))
else:
env['task']['type'] = "worker"
env['task']["index"] = os.environ['BACKENDAI_CLUSTER_IDX']
print(json.dumps(env))
else:
print("")
print("")


# vim: et sts=4 sw=4 tw=105
Loading