<div style="line-height:1.7;color:#000000;font-size:14px;font-family:Arial"><div style="margin:0;"><span style="font-family: "Times New Roman"; font-size: large; text-align: justify;">Hi Adapters,</span></div><div style="margin:0;"><span style="font-family: "Times New Roman"; font-size: large; text-align: justify;"><br></span></div><div style="margin:0;"><span style="font-family: "Times New Roman"; font-size: large; text-align: justify;">We all know that training large transformer models on long sequence is expensive and may not be possible on a standard GPU card because of the self-attention mechanism that grows quadratically with sequence length.  In this seminar, I'll first give an analysis of memory occupation for standard transformer-based models (BART). Then, I'll talk about two solutions from the view of the model and the input data respectively to tackle the long document summarization problem.</span></div><div style="language:zh-CN;line-height:normal;margin-top:0pt;margin-bottom:0pt;
margin-left:0in;margin-right:0in;text-indent:0in;text-align:left;direction:
ltr;unicode-bidi:embed;mso-vertical-align-alt:auto;mso-line-break-override:
none;word-break:normal;punctuation-wrap:hanging"><br></div><div style="language:zh-CN;line-height:normal;margin-top:0pt;margin-bottom:0pt;
margin-left:0in;margin-right:0in;text-indent:0in;text-align:left;direction:
ltr;unicode-bidi:embed;mso-vertical-align-alt:auto;mso-line-break-override:
none;word-break:normal;punctuation-wrap:hanging"><div data-zone-id="0" data-line-index="1" style="font-family: Helvetica, "Microsoft Yahei", verdana; white-space: pre;"><font face="Times New Roman" size="4">See you there~</font></div><div style=""><p class="MsoNormal" style="font-family: Helvetica, "Microsoft Yahei", verdana; line-height: 23.8px; margin: 0cm; text-align: justify;"><font face="Times New Roman" size="4"><br></font></p><p class="MsoNormal" style="font-family: Helvetica, "Microsoft Yahei", verdana; line-height: 23.8px; margin: 0cm; text-align: justify;"><span lang="EN-US"><font face="Times New Roman" size="4">Time: Wed 4:00pm<o:p></o:p></font></span></p><p class="MsoNormal" style="font-family: Helvetica, "Microsoft Yahei", verdana; line-height: 23.8px; margin: 0cm; text-align: justify;"><span lang="EN-US"><font face="Times New Roman" size="4">Venue: SEIEE 3-414<o:p></o:p></font></span></p><p class="MsoNormal" style="font-family: Helvetica, "Microsoft Yahei", verdana; line-height: 23.8px; margin: 0cm; text-align: justify;"><span lang="EN-US"><font face="Times New Roman" size="4"> </font></span></p><p class="MsoNormal" style="font-family: Helvetica, "Microsoft Yahei", verdana; line-height: 23.8px; margin: 0cm; text-align: justify;"><span lang="EN-US"><font face="Times New Roman" size="4">Best Regards,<o:p></o:p></font></span></p><p class="MsoNormal" style="line-height: 23.8px; margin: 0cm; text-align: justify;"><font face="Times New Roman" size="4">Angel</font></p></div></div>

<!--EndFragment--></div><br><br><span title="neteasefooter"><p> </p></span>