Instructions to use K-Iwa/time-anchor with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use K-Iwa/time-anchor with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("K-Iwa/time-anchor", device_map="auto") - Notebooks
- Google Colab
- Kaggle
| # Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved. | |
| # SPDX-License-Identifier: Apache-2.0 | |
| from typing import List | |
| import torch | |
| def left_pad_and_stack_1d(tensors: List[torch.Tensor]) -> torch.Tensor: | |
| """Left-pad variable-length 1D tensors with NaN and stack them into a batch.""" | |
| if not tensors: | |
| raise ValueError("At least one tensor is required.") | |
| max_len = max(len(c) for c in tensors) | |
| padded = [] | |
| for index, c in enumerate(tensors): | |
| if not isinstance(c, torch.Tensor): | |
| raise TypeError(f"Item {index} is not a torch.Tensor.") | |
| if c.ndim != 1: | |
| raise ValueError(f"Item {index} must be 1D; got shape {tuple(c.shape)}.") | |
| padding = torch.full( | |
| size=(max_len - len(c),), | |
| fill_value=torch.nan, | |
| dtype=c.dtype, | |
| device=c.device, | |
| ) | |
| padded.append(torch.concat((padding, c), dim=-1)) | |
| return torch.stack(padded) | |
| left_pad_and_stack_1D = left_pad_and_stack_1d | |