Founce 5dd78a0fe8 fix(chapter5): align labels/attention_mask semantics and add padding-aware batch generation (#170)

2026-02-26 15:34:10 +08:00

..

tokenizer_k

fix(tokenizer): 将add_prefix_space配置改为false

2025-06-08 09:27:21 +08:00

dataset.py

fix(chapter5): align labels/attention_mask semantics and add padding-aware batch generation (#170)

2026-02-26 15:34:10 +08:00

ddp_pretrain.py

fix(chapter5): align labels/attention_mask semantics and add padding-aware batch generation (#170)

2026-02-26 15:34:10 +08:00

ddp_sft_full.py

fix(chapter5): align labels/attention_mask semantics and add padding-aware batch generation (#170)

2026-02-26 15:34:10 +08:00

deal_dataset.py

fix: 为文件读取添加utf-8编码以避免潜在编码问题

2025-07-25 09:03:43 +08:00

download_dataset.sh

created windows_download_dataset.sh, deleted original changes in download_dataset.sh

2025-06-19 17:52:24 +08:00

export_model.py

fix(chapter5): align labels/attention_mask semantics and add padding-aware batch generation (#170)

2026-02-26 15:34:10 +08:00

k_model.py

fix(chapter5): align labels/attention_mask semantics and add padding-aware batch generation (#170)

2026-02-26 15:34:10 +08:00

model_sample.py

fix(chapter5): align labels/attention_mask semantics and add padding-aware batch generation (#170)

2026-02-26 15:34:10 +08:00

requirements.txt

docs(chapter5): 添加LLaMA2结构图并更新依赖

2025-06-09 22:14:01 +08:00

train_tokenizer.py

fix(tokenizer): 将add_prefix_space配置改为false

2025-06-08 09:27:21 +08:00

windows_download_dataset.sh

created windows_download_dataset.sh, deleted original changes in download_dataset.sh

2025-06-19 17:52:24 +08:00