rohitg00/ai-engineering-from-scratch · error · ValueError

table must be (L, D)

Error message

table must be (L, D)

What it means

Error "table must be (L, D)" thrown in rohitg00/ai-engineering-from-scratch.

Source

Thrown at phases/19-capstone-projects/32-token-positional-embeddings/code/main.py:175

        if ids.dim() != 2:
            raise ValueError(f"ids must be (B, T), got shape {tuple(ids.shape)}")
        seq_len = ids.shape[1]
        tok = self.token_embedding(ids)
        pos = self.positional_embedding(seq_len)
        return tok + pos.unsqueeze(0)


def count_parameters(module: nn.Module) -> int:
    return sum(p.numel() for p in module.parameters() if p.requires_grad)


def neighbour_cosine_curve(table: torch.Tensor, max_offset: int = 8) -> list[float]:
    """Average cosine similarity between row p and row p+k for k in 1..max_offset.

    Returns a list of length max_offset.
    """
    if table.dim() != 2:
        raise ValueError("table must be (L, D)")
    if max_offset < 1:
        raise ValueError(f"max_offset must be >= 1, got {max_offset}")
    if max_offset >= table.shape[0]:
        raise ValueError(
            f"max_offset {max_offset} must be < number of rows {table.shape[0]}"
        )
    rows = table.detach().to(torch.float32)
    norms = rows.norm(dim=1, keepdim=True).clamp(min=1e-8)
    unit = rows / norms
    result: list[float] = []
    for k in range(1, max_offset + 1):
        a = unit[:-k]
        b = unit[k:]
        dot = (a * b).sum(dim=1).mean().item()
        result.append(dot)
    return result

View on GitHub (pinned to 39ea8a1c6d)

When it happens

Trigger: Thrown at phases/19-capstone-projects/32-token-positional-embeddings/code/main.py:175 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of rohitg00/ai-engineering-from-scratch@39ea8a1c6d (2026-08-26). Data as JSON: /api/errors/a702e664e4a1e2fe. Report an issue: GitHub.