|
1157 | 1157 | "cell_type": "markdown",
|
1158 | 1158 | "metadata": {},
|
1159 | 1159 | "source": [
|
1160 |
| - "**Answer:** I have tried difference sequence_lengths (10, 15, 8, 12) based on the medium sentence length, in the end it seems 12 is a good value, even if I got better result with sequence_length=10. As for hidden_dim and n_layers I have based my choice on the lessons here and I did some research. I have tried both n_layers=2 and n_layers=3, and n_layers=2 seems giving better results on this dataset. In my opinion in this specific case, as explained in the lessons, the exponential decay on the learning rate could help because of the loss increasing a bit, but I haven't tried." |
| 1160 | + "**Answer:** I have tried difference sequence_lengths (10, 15, 8, 12) I got better result with sequence_length=10. As for hidden_dim and n_layers I have based my choice on the lessons here and I did some research. I have tried both n_layers=2 and n_layers=3, and n_layers=2 seems giving better results on this dataset. In my opinion in this specific case, as explained in the lessons, the exponential decay on the learning rate could help because of the loss increasing a bit, but I haven't tried." |
1161 | 1161 | ]
|
1162 | 1162 | },
|
1163 | 1163 | {
|
|
1172 | 1172 | },
|
1173 | 1173 | {
|
1174 | 1174 | "cell_type": "code",
|
1175 |
| - "execution_count": 38, |
| 1175 | + "execution_count": 61, |
1176 | 1176 | "metadata": {},
|
1177 | 1177 | "outputs": [],
|
1178 | 1178 | "source": [
|
|
1200 | 1200 | },
|
1201 | 1201 | {
|
1202 | 1202 | "cell_type": "code",
|
1203 |
| - "execution_count": 39, |
| 1203 | + "execution_count": 62, |
1204 | 1204 | "metadata": {},
|
1205 | 1205 | "outputs": [],
|
1206 | 1206 | "source": [
|
|
1294 | 1294 | },
|
1295 | 1295 | {
|
1296 | 1296 | "cell_type": "code",
|
1297 |
| - "execution_count": 40, |
| 1297 | + "execution_count": 63, |
1298 | 1298 | "metadata": {},
|
1299 | 1299 | "outputs": [
|
1300 | 1300 | {
|
1301 | 1301 | "name": "stdout",
|
1302 | 1302 | "output_type": "stream",
|
1303 | 1303 | "text": [
|
1304 |
| - "jerry: us george: they're yeah i'm(.. to the ha it's don't good we statue,.. put) a material george: for? that up i night at? the hello your jerry: surprising? see well like good we different a he of,.. to my? the three tell we about look man:,.. it) a one george: right? just? the hello over! he's george jerry: alright, i on that the when you so like! seen,.. and my? the his the with! little back likes,.. elaine: yeah? the jerry out jerry: coffee in do our,.. elaine: that up i you.. to my? the jerry on,.. and i'm,,,.. to) listening george: i on that you i on that you.. to my? i on,,,.. and i'm? i'm, the on that well good he, well good we hitting was i,.. to) a running george: that at jerry: shmoopy judy: you.. to my? the not ha in of,.. elaine: yeah,,, i'm? my that at jerry: strained this alright,.. it that you.. to my? the three so like out! groom,.. elaine:) let's george just? what wa just was! cmon body, george:.. it right? i on that? the jerry gonna i a he no! putting, tell there i house,.. it yeah? i'm(i'm(.. elaine: that you.. to that listen you.. elaine:) a can george: the from, the when? the jerry on that the really at jerry: better them,.. elaine: that you that up i hi you.. and my? the don't good they look a jerry: woah,.. to) let's something george: i jerry gonna no a they,.. elaine: the on,.. to) stand george: the when? jerry: better them- jerry: who's them at kramer:,.. to the jerry\n" |
| 1304 | + "jerry: cartoons published.(george laughs)\n", |
| 1305 | + "\n", |
| 1306 | + "kramer: oh, yeah, well, i think i have to go. i don't know, i know, but i think i would be able to get a good idea.\n", |
| 1307 | + "\n", |
| 1308 | + "jerry:(to george) hey, what happened to the door?(to elaine) what is it?\n", |
| 1309 | + "\n", |
| 1310 | + "elaine: well, i was thinking of donating something to me. i think i'm going to do something like this. i'm sorry about this guy.(george starts to look at the table)\n", |
| 1311 | + "\n", |
| 1312 | + "elaine:(still shaking) oh, hi.\n", |
| 1313 | + "\n", |
| 1314 | + "elaine: hi, mr. pitt, i'm gonna get back to the office.\n", |
| 1315 | + "\n", |
| 1316 | + "george: what is this?\n", |
| 1317 | + "\n", |
| 1318 | + "kramer: i just don't have the money.\n", |
| 1319 | + "\n", |
| 1320 | + "elaine: i don't understand. i just don't know what i'm doing.\n", |
| 1321 | + "\n", |
| 1322 | + "jerry: oh, i don't know.\n", |
| 1323 | + "\n", |
| 1324 | + "elaine: oh, i think i was just a little too much fun. i think i may be in a really long room, but i'm not gonna go back to the game, and i got it for the cat, and then i have to go to the bathroom.\n", |
| 1325 | + "\n", |
| 1326 | + "helen:(looking at the door) oh, no!\n", |
| 1327 | + "\n", |
| 1328 | + "jerry: i can't.\n", |
| 1329 | + "\n", |
| 1330 | + "george:(to jerry) you got to go?(he leaves, jerry answers the window) hey...\n", |
| 1331 | + "\n", |
| 1332 | + "jerry: what?\n", |
| 1333 | + "\n", |
| 1334 | + "jerry: i don't know.\n", |
| 1335 | + "\n", |
| 1336 | + "kramer: i know what i did!\n", |
| 1337 | + "\n", |
| 1338 | + "george: you want a car?\n", |
| 1339 | + "\n", |
| 1340 | + "elaine: yeah.\n", |
| 1341 | + "\n", |
| 1342 | + "jerry:(to jerry) you know, i was wondering, i think i could go to a bathroom and i know...\n", |
| 1343 | + "\n", |
| 1344 | + "kramer: oh, no. no, not really.\n", |
| 1345 | + "\n", |
| 1346 | + "george:(to george) i told you that i was in my office, and she didn't know what it is.\n", |
| 1347 | + "\n", |
| 1348 | + "jerry: oh, yeah, that's good.(he leaves)\n", |
| 1349 | + "\n", |
| 1350 | + "jerry:(pause) oh, yeah.\n", |
| 1351 | + "\n", |
| 1352 | + "kramer: yeah\n" |
1305 | 1353 | ]
|
1306 | 1354 | }
|
1307 | 1355 | ],
|
|
1329 | 1377 | },
|
1330 | 1378 | {
|
1331 | 1379 | "cell_type": "code",
|
1332 |
| - "execution_count": 22, |
| 1380 | + "execution_count": 64, |
1333 | 1381 | "metadata": {},
|
1334 | 1382 | "outputs": [],
|
1335 | 1383 | "source": [
|
|
0 commit comments