What An Input File Looks Like

Raw sample input

Here is the visible test_data/test0.txt input file with the same numbers and line breaks as the real file. The colours are added here only to make the sections easier to spot:

3 2 2 9
the subject foundations of algorithms is the best subject
1 1 0
1 0
0 1
1 1
1 1
0.5 -0.5
1 0
0 1
1 0
0 1
1 0
0 1

If you only want the big picture, read it like this:

• 3 2 2 9 means n = 3, d = 2, g = 2, and text_len = 9
• the next line is the token list used only for Stage 1 embedding creation
• 1 1 0 is the prompt mask
• the next n = 3 lines are prompt embeddings
• the next g = 2 lines are generated embeddings
• the last 3d = 6 lines are the matrices Wq, Wk, and Wv, each with d = 2 rows

Line-by-line breakdown

Each part of the file means:

Two details matter:

• the actual input file does not contain labels like mask or Wq; only this page adds labels and colours to make the structure easier to read
• input is whitespace-separated, so scanf does not care whether values are separated by spaces or newlines

If you want a more human-friendly example generated from real text instead of a tiny numeric test case, use the Toy demo.