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

kv_cache must be (B,H,Nv,hd)={expected}, got k={tuple(k.shap

Error message

kv_cache must be (B,H,Nv,hd)={expected}, got k={tuple(k.shape)} v={tuple(v.shape)}

What it means

Error "kv_cache must be (B,H,Nv,hd)={expected}, got k={tuple(k.shape)} v={tuple(v.shape)}" thrown in rohitg00/ai-engineering-from-scratch.

Source

Thrown at phases/19-capstone-projects/61-cross-attention-fusion/code/main.py:124

                kv_cache: tuple[torch.Tensor, torch.Tensor] | None = None
                ) -> torch.Tensor:
        if x.dim() != 3:
            raise ValueError(f"expected (B, Nt, hidden), got {tuple(x.shape)}")
        if memory.shape[0] != x.shape[0]:
            raise ValueError(
                f"batch mismatch: text {x.shape[0]} vs memory {memory.shape[0]}"
            )
        b, nt, d = x.shape
        h, hd = self.cfg.heads, self.cfg.head_dim

        q = self.q_proj(x).reshape(b, nt, h, hd).transpose(1, 2)
        if kv_cache is None:
            k, v = self.project_memory(memory)
        else:
            k, v = kv_cache
            expected = (b, h, memory.shape[1], hd)
            if k.shape != expected or v.shape != expected:
                raise ValueError(
                    f"kv_cache must be (B,H,Nv,hd)={expected}, got "
                    f"k={tuple(k.shape)} v={tuple(v.shape)}"
                )

        scores = (q @ k.transpose(-2, -1)) * self.scale
        attn = F.softmax(scores, dim=-1)
        out = (attn @ v).transpose(1, 2).reshape(b, nt, d)
        return self.drop(self.out(out))


class FeedForward(nn.Module):
    def __init__(self, cfg: DecoderConfig) -> None:
        super().__init__()
        inner = int(cfg.hidden * cfg.mlp_ratio)
        self.fc1 = nn.Linear(cfg.hidden, inner)
        self.fc2 = nn.Linear(inner, cfg.hidden)

    def forward(self, x: torch.Tensor) -> torch.Tensor:

View on GitHub (pinned to 39ea8a1c6d)

When it happens

Trigger: Thrown at phases/19-capstone-projects/61-cross-attention-fusion/code/main.py:124 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/4fe0ccb57fb4232c. Report an issue: GitHub.