audio
audioduration (s) 0.53
8.25
| file
stringlengths 36
38
| instruction
stringclasses 14
values | label
stringlengths 6
53
|
|---|---|---|---|
clips/week7_1-clips/0043-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 接 下 來 輪 到 decoder 產 生 語 音 辨 識 的 結 果 |
|
clips/week7_1-clips/0044-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 decoder 怎 麼 產 生 這 個 語 音 辨 識 的 結 果 呢 |
|
clips/week7_1-clips/0080-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 你 都 可 以 把 它 用 一 個 one-hot 的 vector 來 表 示 |
|
clips/week7_1-clips/0083-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 所 以 begin 也 是 用 one-hot vector 來 表 示 |
|
clips/week7_1-clips/0132-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 助 教 說 他 是 用 subword 當 作 英 文 的 單 位 |
|
clips/week7_1-clips/0205-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 所 以 decoder 看 到 encoder 這 邊 的 輸 入 |
|
clips/week7_1-clips/0224-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 所 以 當 我 們 decoder 在 產 生 一 個 句 子 的 時 候 |
|
clips/week7_1-clips/0227-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 如 果 今 天 decoder 有 語 音 辨 識 的 錯 誤 |
|
clips/week7_1-clips/0230-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 接 下 來 decoder 就 會 看 到 錯 誤 的 辨 識 結 果 |
|
clips/week7_1-clips/0236-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 讓 decoder 看 到 自 己 產 生 出 來 的 錯 誤 的 輸 入 |
|
clips/week7_1-clips/0239-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 會 不 會 造 成 error propagation 的 問 題 呢 |
|
clips/week7_1-clips/0264-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 我 們 現 在 先 把 encoder 跟 decoder 放 在 一 起 |
|
clips/week7_1-clips/0297-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | multi-head attention 這 一 個 block 上 面 |
|
clips/week7_1-clips/0342-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 然 後 最 後 只 計 算 v1 跟 v2 的 weighted sum |
|
clips/week7_1-clips/0350-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 你 想 想 看 我 們 一 開 始 decoder 的 運 作 方 式 |
|
clips/week7_1-clips/0357-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | a1 跟 a4 是 一 次 整 個 輸 進 去 你 的 model 裡 面 的 |
|
clips/week7_1-clips/0368-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 所 以 你 根 本 就 沒 有 辦 法 把 a3 a4 考 慮 進 來 |
|
clips/week7_1-clips/0371-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | transformer 原 始 的 paper 特 別 跟 你 強 調 說 |
|
clips/week7_1-clips/0387-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 可 是 到 底 輸 出 的 sequence 的 長 度 應 該 是 多 少 呢 |
|
clips/week7_1-clips/0389-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 你 沒 有 辦 法 輕 易 的 從 輸 入 的 sequence 的 長 度 |
|
clips/week7_1-clips/0407-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 它 還 可 以 繼 續 重 複 一 模 一 樣 的 process |
|
clips/week7_1-clips/0477-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 就 decoder 看 到 encoder 輸 出 的 這 個 embedding |
|
clips/week7_1-clips/0502-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | non-autoregressive 的 model 是 怎 麼 運 作 的 呢 |
|
clips/week7_1-clips/0516-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 它 可 能 吃 的 是 一 整 排 的 begin 的 token |
|
clips/week7_1-clips/0517-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 你 就 把 一 堆 一 排 begin 的 token 都 丟 給 它 |
|
clips/week7_1-clips/0529-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 我 們 這 邊 怎 麼 知 道 begin 要 放 多 少 個 |
|
clips/week7_1-clips/0541-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 這 個 數 字 代 表 decoder 應 該 要 輸 出 的 長 度 |
|
clips/week7_1-clips/0572-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 這 是 另 外 一 種 處 理 nat 的 這 個 decoder |
|
clips/week7_1-clips/0593-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 有 這 種 self-attention 的 decoder 以 後 才 有 的 |
|
clips/week7_1-clips/0603-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 根 本 就 不 會 有 人 想 要 做 什 麼 nat 的 decoder |
|
clips/week7_1-clips/0607-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 nat 的 decoder 還 有 另 外 一 個 好 處 就 是 |
|
clips/week7_1-clips/0631-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 你 就 把 那 個 classifier 的 output 除 以 二 |
|
clips/week7_1-clips/0634-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 你 就 把 那 個 classifier 輸 出 的 那 個 長 度 啊 |
|
clips/week7_1-clips/0638-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 所 以 你 可 以 如 果 有 這 種 nat 的 decoder |
|
clips/week7_1-clips/0653-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | nat 的 decoder 的 performance 越 來 越 好 |
|
clips/week7_1-clips/0657-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 你 必 須 要 用 非 常 多 的 trick 才 能 夠 辦 到 |
|
clips/week7_1-clips/0661-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 為 什 麼 nat 的 decoder performance 不 好 呢 |
|
clips/week7_1-clips/0683-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 它 是 連 接 encoder 跟 decoder 之 間 的 橋 樑 |
|
clips/week7_1-clips/0700-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 begin 這 個 special 的 token 讀 進 來 以 後 |
|
clips/week7_1-clips/0704-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 就 是 self-attention 就 算 是 有 做 mask |
|
clips/week7_1-clips/0725-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 就 是 接 下 來 會 丟 到 fully-connected 的 |
|
clips/week7_1-clips/0733-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | fully-connected 的 network 的 input |
|
clips/week7_1-clips/0734-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 好 那 這 個 就 是 cross attention 運 作 的 過 程 |
|
clips/week7_1-clips/0741-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 這 個 向 量 一 樣 乘 上 一 個 linear 的 transform |
|
clips/week7_1-clips/0746-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 一 樣 跟 v1 v2 v3 做 weighted sum 做 加 權 |
|
clips/week7_1-clips/0748-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 交 給 接 下 來 fully-connected network 做 處 理 |
|
clips/week7_1-clips/0758-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 是 比 較 早 使 用 sequence to sequence model |
|
clips/week7_1-clips/0763-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 我 還 特 別 有 親 耳 聽 了 這 篇 paper 的 報 告 |
|
clips/week7_1-clips/0771-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 它 展 現 的 sequence to sequence model |
|
clips/week7_1-clips/0776-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | sequence to sequence 用 在 語 音 辨 識 上 |
|
clips/week7_1-clips/0779-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 時 候 的 encoder 跟 decoder 都 是 用 lstm |
|
clips/week7_1-clips/0780-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 不 過 那 個 時 候 就 已 經 有 cross attention |
|
clips/week7_1-clips/0783-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 其 實 就 已 經 有 cross attention 這 樣 的 機 制 |
|
clips/week7_1-clips/0788-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 如 果 你 是 用 sequence to sequence 的 model |
|
clips/week7_1-clips/0794-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 所 以 像 這 篇 是 listen, attend and spell |
|
clips/week7_1-clips/0796-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 然 後 做 attention 就 cross attention |
|
clips/week7_1-clips/0797-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | spell 就 是 把 它 聽 到 的 東 西 把 它 拼 出 來 |
|
clips/week7_1-clips/0803-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 聲 音 訊 號 輸 入 給 這 個 encoder 的 時 候 |
|
clips/week7_1-clips/0824-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | how much wood would a woodchuck chuck |
|
clips/week7_1-clips/0836-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 所 以 它 就 attend 在 這 個 地 方 然 後 產 生 h |
|
clips/week7_1-clips/0843-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 它 就 是 聽 到 這 個 地 方 有 h 的 聲 音 所 以 產 生 h |
|
clips/week7_1-clips/0849-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 你 會 看 到 說 這 個 attention 的 這 個 weight |
|
clips/week7_1-clips/0851-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 跟 你 想 像 attention 應 該 運 作 的 機 制 很 像 |
|
clips/week7_1-clips/0855-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 所 以 確 實 如 果 你 看 model attention 的 話 |
|
clips/week7_1-clips/0864-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 都 是 拿 encoder 的 最 後 一 層 的 輸 出 這 樣 對 嗎 |
|
clips/week7_1-clips/0871-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 也 有 人 嘗 試 不 同 的 cross attension 的 方 式 |
|
clips/week7_1-clips/0874-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 為 什 麼 decoder 這 邊 每 一 層 都 一 定 要 看 |
|
clips/week7_1-clips/0877-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 這 完 全 可 以 當 做 一 個 研 究 的 問 題 來 study |
|
clips/week7_1-clips/0916-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 機 這 個 字 會 被 表 示 成 一 個 one-hot 的 vector |
|
clips/week7_1-clips/0927-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 跟 這 個 distribution 它 們 之 間 的 cross entropy |
|
clips/week7_1-clips/0928-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 然 後 我 們 希 望 這 個 cross entropy 的 值 |
|
clips/week7_1-clips/0935-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 每 一 次 decoder 在 產 生 一 個 中 文 字 的 時 候 |
|
clips/week7_1-clips/0948-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 跟 這 四 個 字 的 one-hot vector 越 接 近 越 好 |
|
clips/week7_1-clips/0950-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 每 一 個 輸 出 都 會 有 一 個 cross entropy |
|
clips/week7_1-clips/0952-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 跟 它 對 應 的 正 確 答 案 都 有 一 個 cross entropy |
|
clips/week7_1-clips/0954-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 我 們 要 希 望 所 有 的 cross entropy 的 總 和 最 小 |
|
clips/week7_1-clips/0958-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 它 總 合 起 來 的 cross entropy 越 小 越 好 |
|
clips/week7_1-clips/0964-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 你 要 decoder 輸 出 的 不 是 只 有 這 四 個 中 文 字 |
|
clips/week7_1-clips/0974-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 希 望 decoder 的 輸 出 跟 正 確 答 案 越 接 近 越 好 |
|
clips/week7_1-clips/0980-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 在 訓 練 的 時 候 我 們 會 給 decoder 看 正 確 答 案 |
|
clips/week7_1-clips/0985-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 有 begin 有 機 有 器 有 學 的 情 況 下 輸 出 習 |
|
clips/week7_1-clips/0992-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 其 實 我 不 太 確 定 為 什 麼 叫 teacher forcing |
|
clips/week7_1-clips/0995-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 但 是 這 個 技 術 呢 就 叫 做 teacher forcing |
|
clips/week7_1-clips/1014-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 訓 練 這 種 sequence to sequence model 的 |
|
clips/week7_1-clips/1021-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 也 許 decoder 沒 有 必 要 自 己 創 造 輸 出 出 來 |
|
clips/week7_1-clips/1063-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 人 的 輸 入 去 copy 一 些 詞 彙 當 做 它 的 輸 出 |
|
clips/week7_1-clips/1125-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 好 那 其 實 這 個 sequence-to-sequence model |
|
clips/week7_1-clips/1182-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 就 是 你 這 個 sequence-to-sequence model |
|
clips/week7_1-clips/1269-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 顯 然 這 樣 子 的 attention 是 有 問 題 的 |
|
clips/week7_1-clips/1272-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 所 以 guiding attention 要 做 的 事 情 就 是 |
|
clips/week7_1-clips/1334-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 那 像 這 樣 子 每 次 找 分 數 最 高 的 那 個 token |
|
clips/week7_1-clips/1413-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | the curious case of neural text degeneration |
|
clips/week7_1-clips/1463-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 舉 例 來 說 在 這 邊 的 sentence completion |
|
clips/week7_1-clips/1520-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 有 時 候 我 們 其 實 期 待 decorder 有 隨 機 性 |
|
clips/week7_1-clips/1525-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 對 於 tts 或 sentence completion 來 說 |
|
clips/week7_1-clips/1555-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 所 以 我 們 minimize cross entropy |
|
clips/week7_1-clips/1556-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 不 見 得 可 以 讓 bleu score 比 較 大 |
|
clips/week7_1-clips/1558-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 比 如 說 在 助 教 在 做 validation 的 時 候 |
|
clips/week7_1-clips/1559-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 並 不 是 拿 cross entropy 來 挑 最 好 的 model |
|
clips/week7_1-clips/1560-week7_1.wav | This is a recording of a machine learning course. The topic is transformer. Transcribe the speech. | 他 並 不 是 挑 cross entropy 最 低 的 那 個 model |
End of preview. Expand
in Dataset Viewer.
- Downloads last month
- 3