Improvements
- Add functionality for cluster kill process (#1588)
- Cluster factory refactor to be more robust (#1591)
- Fix bug with
run_bash
in Image setup (#1608) - Update
run_bash
output to match input type (#1639) - Bring up the cluster server after launching (#1645)
- Fix to Pytorch distributed (#1650)
- Detect and update stale ondemand ips in
restart_server
(#1633)
Deprecations
- Deprecate
accelerators
in favor ofgpus
for the cluster factory (#1648)
BC-Breaking
The following were deprecated in an earlier release, and are now no longer supported
config_for_rns
: Resource config is now just namedconfig
instead ofconfig_for_rns
(#1627)default_env
: Please userh.Image()
class to specify default setup for a Runhouse cluster, and pass it in theimage
field of the cluster factory. Links to: docs, examples (#1627)address
: Usecluster.head_ip
insteadcluster.address
(#1627)cluster.run()
: Please instead userun_bash
(to run over HTTP) orrun_bash_over_ssh
(to run over SSH)num_instances
in cluster factory: Please usenum_nodes
instead (#1634)