Implement a word embedding lookup step, which maps token IDs to dense vectors used in NLP models. Given a matrix of embeddings and a list of token IDs, return the corresponding sequence of embedding vectors using the lookup rule (E[t]).

$\mathbf{Y}_i = \mathbf{E}_{t_i}$

where $\mathbf{E} \in \mathbb{R}^{V \times D}$ is the embedding matrix and $t_i$ is the token ID at index $i$.

Requirements

Implement the function

Rules:

• Use the lookup rule (Y_j = E[\text{token_ids}_j]) for each position (j).
• Preserve the order of token_ids in the output.
• Return a np.ndarray.
• Do not use any prebuilt embedding layers (e.g., from PyTorch/TensorFlow).

Example

Output: