How to read a large file block by block and judge by block header?

I don't know what you want to achieve exactly but maybe something like this:

with open(filename) as f:
    for line in f:
        if line.startswith('@'):
            print('header')
            # do something with header here
        else:
            print('regular line')
            # do something with the line here

Attached at bottom is a solution using a Python generator split_into_chunks(f) to extract each section (as list-of-string), squelch empty lines, detect missing @headers and EOF. The generator approach is really neat because it allows you to further wrap e.g. a CSV reader object which handles space-separated value (e.g. pandas read_csv):

with open('your.ssv') as f:
    for chunk in split_into_chunks(f):
        # Do stuff on chunk. Presumably, wrap a reader e.g. pandas read_csv
        # print(chunk)

Code is below. I also parameterized the value demarcator='@header' for you. Note that we have to iterate with line = inputstream.readline(), while line, instead of the usual iterating with for line in f, since if we see the @header of the next section, we need to pushback with seek/tell() ; see this and this for explanation why. And if you want to modify the generator to yield the chunk header and body separately (e.g. as a list of two items), that's trivial.

def split_into_chunks(inputstream, demarcator='@header'):
    """Utility generator to get sections from file, demarcated by '@header'"""

    while True:
        chunk = []

        line = inputstream.readline()
        # At EOF?
        if not line: break
        # Expect that each chunk starts with one header line
        if not line.startswith(demarcator):
            raise RuntimeError(f"Bad chunk, missing {demarcator}")

        chunk.append(line.rstrip('\n'))

        # Can't use `for line in inputstream:` since we may need to pushback
        while line:
            # Remember our file-pointer position in case we need to pushback a header row
            last_pos = inputstream.tell()
            line = inputstream.readline()

            # Saw next chunk's header line? Pushback the header line, then yield the current chunk
            if line.startswith(demarcator):
                inputstream.seek(last_pos)
                break

            # Ignore blank or whitespace-only lines
            #line = line.rstrip('\n')
            if line:
                chunk.append(line.rstrip('\n'))

        yield chunk


with open('your.ssv') as f:
    for chunk in split_into_chunks(f):
        # Do stuff on chunk. Presumably, wrap it with a reader which handles space-sparated value, e.g. pandas read_csv
        print(chunk)

I saw another post similar to this question and copied the idea here. I agree that SpghttCd is right although I have not tried that.

    with open(filename) as f:
        #find each line number that contains header
        for i,line in enumerate(f,1):
            if 'some_header' in line:
                index1=i
            elif 'another_header' in line:
                index2=i
            ...
    with open(filename) as f:
        #read the first block:
        for i in range(int(index1)):
            line=f.readline()
        for i in range('the block size'):
            'read, split and store'
        f.seek(0)
        #read the second block, third and ... 
        ...