_parse_slurm_node_list leads to incorrect node names if no brackets are in node name

[Leonngm]

I realized that sometimes the number of nodes does not match the number of requested nodes. This is due to the function _parse_slurm_node_list failing, if the nodename does not contain any brackets.

Code to reproduce:

from dinov2.distributed import _parse_slurm_node_list

input = "node1g[0010,0011],node2g0005,node3g0007"
output = _parse_slurm_node_list(input)

# output: ['node1g0010', 'node1g0011', 'node2g0005,node3g007']
# len(output): 3
# expected output:  ['node1g0010', 'node1g0011', 'node2g0005', 'node3g007']
# len(expected output): 4

Can be fixed by adding:

    flattened_nodes = []
    for node in nodes:
        flattened_nodes.extend(node.split(","))
        
    return flattened_nodes

Fix using regular expressions thanks to @NikolasSchmitz:

def parse_slurm_node_list(s: str) -> List[str]:
    nodes = []
    p = re.compile(r"([^\[,]+)(?:\[([^\]]+)\])?,?")
    for m in p.finditer(s):
        prefix = m.group(1)
        suffixes = m.group(2)
        if suffixes:
            for suffix in suffixes.split(","):
                span = suffix.split("-")
                if len(span) == 1:
                    nodes.append(prefix + suffix)
                else:
                    width = len(span[0])
                    start, end = int(span[0]), int(span[1]) + 1
                    nodes.extend([prefix + f"{i:0{width}}" for i in range(start, end)])
        else:
            nodes.append(prefix)  # If no brackets, add the prefix as is
    return nodes