遇见数据集

arcee-ai/DeepSeek-DCLM-Logits-Packed-8192

收藏
Hugging Face2025-12-10 更新2026-01-03 收录
官方服务:

资源简介:

--- dataset_info: features: - name: input_ids sequence: uint64 - name: packed_indices sequence: sequence: uint64 - name: exact_values sequence: sequence: float32 - name: coeffs sequence: sequence: float32 - name: attention_mask sequence: int8 splits: - name: train num_bytes: 688429456092 num_examples: 158859 download_size: 570661614388 dataset_size: 688429456092 configs: - config_name: default data_files: - split: train path: data/train-* --- ```yaml sequence_length: 8192 dataset: train_dataset: repo_id: arcee-ai/DeepSeek-DCLM-Logits-Packed-8192 split: train prepacked: true teacher: kind: dataset legacy_logit_compression: exact_k: 32 invert_polynomial: true k: 128 polynomial_degree: 8 term_dtype: float32 vocab_size: 129280 with_sqrt_term: false ```

提供机构:
arcee-ai
二维码
社区交流群
二维码
科研交流群
商业服务