PEP-8

ClassNames should be BumpyWords; snake_case is used for things like method_names and variable_names. So you class should be:

class VisFileTools:
    ...

Unnecessary __init__ method

This method does nothing; it may safely be omitted.

    def __init__(self):
        pass

Public / Private

remove_jedec_line looks like a private method, used by the remove_jedecs_file method. If remove_jedec_line is not supposed to be used by actors outside of the class, it should be prefixed with a leading underscore:

    def _remove_jedec_line(self, inputline):
        ...

Garbage Collection

temp_data = ""    # Empties memory

This is unnecessary. The variable is about to go out of scope, which will naturally release the memory it is holding.

Stop Writing Classes

See the video Stop Writing Classes

Your class is unnecessary. It has no data members. It has one public method. It should just be a function.

Simplified code

Here is a function version of your code (untested):

def remove_jedecs_file(path_input_file: str, path_output_file: str) -> None:
    """
    Read `path_input_file`, copying each line to `path_output_file`.
    Once the start of components is reached (indicated by a line containing "COMP"),
    the last "word" (jedec) of each line is removed.
    """

    def remove_jedec_line(input_line: str) -> str:
        return " ".join(input_line.split(" ")[:-1]) + "\n"

    with open(path_input_file) as file, open(path_output_file, "w") as write_file:
        comp_line_reached = False

        for line in file:
            if comp_line_reached:
                line = remove_jedec_line(line)
            else:
                comp_line_reached = "COMP" in line

            write_file.write(line)

Notes:

• Type-hints (Python 3.6+) have been added to the function (: str and -> None)
• A """docstring""" has been added to describe the function.
• A nested function is used for remove_jedec_line(), making it "private".
• [:-1] returns all but the last element in a list, which is simpler than using del input_line_split[-1].
• Lines are not being accumulated in temp_data. Instead, lines are written out immediately after being read in, which is a much lighter load on memory, and avoid \$O(N^2)\$ string concatenation.
• "COMP" in line is a "slow" search operation, looking for that substring in the line string. When found, it sets a flag. Here, once the flag is set, we can skip the slow substring search. But I'm introducing a slight change in behaviour: if subsequent lines also contain "COMP", the OP version would not do the remove_jedec_line() transformation where as mine will. The assumption here is that "COMP" only appears once.