Anurag1734/cuda-error-resolution-analysis
07
1[2 {3 "post_stream": {4 "posts": [5 {6 "id": 268890,7 "name": "Jack Rolph",8 "username": "Jack_Rolph",9 "avatar_template": "/user_avatar/discuss.pytorch.org/jack_rolph/{size}/20306_2.png",10 "created_at": "2021-03-09T13:03:15.775Z",11 "cooked": "<p>I have a tensor, <code>A</code> of shape <code>[X,Y,Z]</code> and an indexing tensor, <code>B</code>, of <code>[X,Y,1]</code>. I would like to index <code>A</code> with <code>B</code>, and replace those elements of <code>A</code> with 0.</p>\n<p><code>B</code> is the output of a neural network, with the desired result that it is either true or false for each <code>Y</code>.</p>\n<pre><code class=\"lang-auto\">r1 = -10\nr2 = 10 #some ranges\n\nX = 20\nY=100\nZ=5\n\nactivation = nn.Tanh() #an activation function\n \n#A is an arbitrary random tensor [0,1]\nA = torch.FloatTensor(X,Y,Z).uniform_(0, 1)\n\n#B is an arbitrary random tensor [-1,1]\nB = activation(torch.FloatTensor(X,Y,1).uniform_(r1, r2))\n\n#idx are where B is less than 0\nidx = (B<0)\n\n#Some function to mask A with B such that where B is True, A=0\nA = foo(A,B)\n</code></pre>\n<p>What is the correct function, <code>foo</code>, required such that <code>A</code> is both masked with zero and differentiable?</p>",12 "post_number": 1,13 "post_type": 1,14 "posts_count": 1,15 "updated_at": "2021-03-09T14:20:18.743Z",16 "reply_count": 0,17 "reply_to_post_number": null,18 "quote_count": 0,19 "incoming_link_count": 151,20 "reads": 12,21 "readers_count": 11,22 "score": 757.4,23 "yours": false,24 "topic_id": 114210,25 "topic_slug": "masking-rows-of-tensor",26 "display_username": "Jack Rolph",27 "primary_group_name": null,28 "flair_name": null,29 "flair_url": null,30 "flair_bg_color": null,31 "flair_color": null,32 "flair_group_id": null,33 "badges_granted": [],34 "version": 3,35 "can_edit": false,36 "can_delete": false,37 "can_recover": false,38 "can_see_hidden_post": false,39 "can_wiki": false,40 "read": true,41 "user_title": null,42 "bookmarked": false,43 "actions_summary": [],44 "moderator": false,45 "admin": false,46 "staff": false,47 "user_id": 27329,48 "hidden": false,49 "trust_level": 1,50 "deleted_at": null,51 "user_deleted": false,52 "edit_reason": null,53 "can_view_edit_history": true,54 "wiki": false,55 "post_url": "/t/masking-rows-of-tensor/114210/1",56 "can_accept_answer": false,57 "can_unaccept_answer": false,58 "accepted_answer": false,59 "topic_accepted_answer": null,60 "can_vote": false61 }62 ],63 "stream": [64 26889065 ]66 },67 "timeline_lookup": [68 [69 1,70 169171 ]72 ],73 "suggested_topics": [74 {75 "fancy_title": "Pt_main_thread occasionally gets stuck, leading to training time increased",76 "id": 213404,77 "title": "Pt_main_thread occasionally gets stuck, leading to training time increased",78 "slug": "pt-main-thread-occasionally-gets-stuck-leading-to-training-time-increased",79 "posts_count": 1,80 "reply_count": 0,81 "highest_post_number": 1,82 "image_url": "https://discuss.pytorch.org/uploads/default/optimized/3X/2/a/2afc851e8d5e13a3b7ec514e48189417e5182bb9_2_1024x236.png",83 "created_at": "2024-11-25T08:32:58.086Z",84 "last_posted_at": "2024-11-25T08:32:58.136Z",85 "bumped": true,86 "bumped_at": "2024-11-25T08:32:58.136Z",87 "archetype": "regular",88 "unseen": false,89 "pinned": false,90 "unpinned": null,91 "visible": true,92 "closed": false,93 "archived": false,94 "bookmarked": null,95 "liked": null,96 "tags_descriptions": {},97 "like_count": 0,98 "views": 312,99 "category_id": 1,100 "featured_link": null,101 "has_accepted_answer": false,102 "posters": [103 {104 "extras": "latest single",105 "description": "Original Poster, Most Recent Poster",106 "user": {107 "id": 72712,108 "username": "xgbj",109 "name": "Xgbj",110 "avatar_template": "/user_avatar/discuss.pytorch.org/xgbj/{size}/67179_2.png",111 "trust_level": 1112 }113 }114 ]115 },116 {117 "fancy_title": "`num_features` parameter of `nn.InstanceNorm2d` does not change results",118 "id": 214610,119 "title": "`num_features` parameter of `nn.InstanceNorm2d` does not change results",120 "slug": "num-features-parameter-of-nn-instancenorm2d-does-not-change-results",121 "posts_count": 3,122 "reply_count": 1,123 "highest_post_number": 3,124 "image_url": null,125 "created_at": "2024-12-24T23:59:04.629Z",126 "last_posted_at": "2024-12-31T04:48:24.867Z",127 "bumped": true,128 "bumped_at": "2024-12-31T04:48:24.867Z",129 "archetype": "regular",130 "unseen": false,131 "pinned": false,132 "unpinned": null,133 "visible": true,134 "closed": false,135 "archived": false,136 "bookmarked": null,137 "liked": null,138 "tags_descriptions": {},139 "like_count": 0,140 "views": 46,141 "category_id": 1,142 "featured_link": null,143 "has_accepted_answer": true,144 "posters": [145 {146 "extras": "latest",147 "description": "Original Poster, Most Recent Poster",148 "user": {149 "id": 68684,150 "username": "aoot",151 "name": "",152 "avatar_template": "/letter_avatar_proxy/v4/letter/a/df788c/{size}.png",153 "trust_level": 2154 }155 },156 {157 "extras": null,158 "description": "Frequent Poster, Accepted Answer",159 "user": {160 "id": 3534,161 "username": "ptrblck",162 "name": "",163 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",164 "admin": true,165 "moderator": true,166 "trust_level": 2167 }168 }169 ]170 },171 {172 "fancy_title": "How does PyTorch implement batch normalization?",173 "id": 215939,174 "title": "How does PyTorch implement batch normalization?",175 "slug": "how-does-pytorch-implement-batch-normalization",176 "posts_count": 5,177 "reply_count": 3,178 "highest_post_number": 6,179 "image_url": null,180 "created_at": "2025-01-27T15:01:56.590Z",181 "last_posted_at": "2025-01-28T13:44:17.202Z",182 "bumped": true,183 "bumped_at": "2025-01-28T13:44:17.202Z",184 "archetype": "regular",185 "unseen": false,186 "pinned": false,187 "unpinned": null,188 "visible": true,189 "closed": false,190 "archived": false,191 "bookmarked": null,192 "liked": null,193 "tags_descriptions": {},194 "like_count": 1,195 "views": 59,196 "category_id": 1,197 "featured_link": null,198 "has_accepted_answer": false,199 "posters": [200 {201 "extras": null,202 "description": "Original Poster",203 "user": {204 "id": 32812,205 "username": "Bjorn_Lindqvist",206 "name": "Björn Lindqvist",207 "avatar_template": "/user_avatar/discuss.pytorch.org/bjorn_lindqvist/{size}/25326_2.png",208 "trust_level": 2209 }210 },211 {212 "extras": "latest",213 "description": "Most Recent Poster",214 "user": {215 "id": 3534,216 "username": "ptrblck",217 "name": "",218 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",219 "admin": true,220 "moderator": true,221 "trust_level": 2222 }223 }224 ]225 },226 {227 "fancy_title": "RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation torch version 2.0.0",228 "id": 216978,229 "title": "RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation torch version 2.0.0",230 "slug": "runtimeerror-one-of-the-variables-needed-for-gradient-computation-has-been-modified-by-an-inplace-operation-torch-version-2-0-0",231 "posts_count": 2,232 "reply_count": 0,233 "highest_post_number": 2,234 "image_url": null,235 "created_at": "2025-02-21T07:01:27.534Z",236 "last_posted_at": "2025-02-21T17:57:30.658Z",237 "bumped": true,238 "bumped_at": "2025-02-21T17:57:30.658Z",239 "archetype": "regular",240 "unseen": false,241 "pinned": false,242 "unpinned": null,243 "visible": true,244 "closed": false,245 "archived": false,246 "bookmarked": null,247 "liked": null,248 "tags_descriptions": {},249 "like_count": 0,250 "views": 23,251 "category_id": 1,252 "featured_link": null,253 "has_accepted_answer": false,254 "posters": [255 {256 "extras": null,257 "description": "Original Poster",258 "user": {259 "id": 82832,260 "username": "hj_feng",261 "name": "hj feng",262 "avatar_template": "/user_avatar/discuss.pytorch.org/hj_feng/{size}/75792_2.png",263 "trust_level": 0264 }265 },266 {267 "extras": "latest",268 "description": "Most Recent Poster",269 "user": {270 "id": 18088,271 "username": "KFrank",272 "name": "K. Frank",273 "avatar_template": "/letter_avatar_proxy/v4/letter/k/ecb155/{size}.png",274 "trust_level": 2275 }276 }277 ]278 },279 {280 "fancy_title": "nn.BatchNorm is slow",281 "id": 212195,282 "title": "nn.BatchNorm is slow",283 "slug": "nn-batchnorm-is-slow",284 "posts_count": 1,285 "reply_count": 0,286 "highest_post_number": 1,287 "image_url": null,288 "created_at": "2024-10-28T08:24:38.399Z",289 "last_posted_at": "2024-10-28T08:24:38.458Z",290 "bumped": true,291 "bumped_at": "2024-10-28T08:24:38.458Z",292 "archetype": "regular",293 "unseen": false,294 "pinned": false,295 "unpinned": null,296 "visible": true,297 "closed": false,298 "archived": false,299 "bookmarked": null,300 "liked": null,301 "tags_descriptions": {},302 "like_count": 0,303 "views": 131,304 "category_id": 1,305 "featured_link": null,306 "has_accepted_answer": false,307 "posters": [308 {309 "extras": "latest single",310 "description": "Original Poster, Most Recent Poster",311 "user": {312 "id": 80459,313 "username": "dailaila",314 "name": null,315 "avatar_template": "/letter_avatar_proxy/v4/letter/d/50afbb/{size}.png",316 "trust_level": 0317 }318 }319 ]320 }321 ],322 "tags_descriptions": {},323 "fancy_title": "Masking Rows of Tensor",324 "id": 114210,325 "title": "Masking Rows of Tensor",326 "posts_count": 1,327 "created_at": "2021-03-09T13:03:15.700Z",328 "views": 515,329 "reply_count": 0,330 "like_count": 0,331 "last_posted_at": "2021-03-09T13:03:15.775Z",332 "visible": true,333 "closed": false,334 "archived": false,335 "has_summary": false,336 "archetype": "regular",337 "slug": "masking-rows-of-tensor",338 "category_id": 1,339 "word_count": 157,340 "deleted_at": null,341 "user_id": 27329,342 "featured_link": null,343 "pinned_globally": false,344 "pinned_at": null,345 "pinned_until": null,346 "image_url": null,347 "slow_mode_seconds": 0,348 "draft": null,349 "draft_key": "topic_114210",350 "draft_sequence": null,351 "unpinned": null,352 "pinned": false,353 "current_post_number": 1,354 "highest_post_number": 1,355 "deleted_by": null,356 "actions_summary": [357 {358 "id": 4,359 "count": 0,360 "hidden": false,361 "can_act": false362 },363 {364 "id": 8,365 "count": 0,366 "hidden": false,367 "can_act": false368 },369 {370 "id": 10,371 "count": 0,372 "hidden": false,373 "can_act": false374 },375 {376 "id": 7,377 "count": 0,378 "hidden": false,379 "can_act": false380 }381 ],382 "chunk_size": 20,383 "bookmarked": false,384 "topic_timer": null,385 "message_bus_last_id": 0,386 "participant_count": 1,387 "show_read_indicator": false,388 "thumbnails": null,389 "slow_mode_enabled_until": null,390 "can_vote": false,391 "vote_count": 0,392 "user_voted": false,393 "discourse_zendesk_plugin_zendesk_id": null,394 "discourse_zendesk_plugin_zendesk_url": "https://your-url.zendesk.com/agent/tickets/",395 "details": {396 "can_edit": false,397 "notification_level": 1,398 "participants": [399 {400 "id": 27329,401 "username": "Jack_Rolph",402 "name": "Jack Rolph",403 "avatar_template": "/user_avatar/discuss.pytorch.org/jack_rolph/{size}/20306_2.png",404 "post_count": 1,405 "primary_group_name": null,406 "flair_name": null,407 "flair_url": null,408 "flair_color": null,409 "flair_bg_color": null,410 "flair_group_id": null,411 "trust_level": 1412 }413 ],414 "created_by": {415 "id": 27329,416 "username": "Jack_Rolph",417 "name": "Jack Rolph",418 "avatar_template": "/user_avatar/discuss.pytorch.org/jack_rolph/{size}/20306_2.png"419 },420 "last_poster": {421 "id": 27329,422 "username": "Jack_Rolph",423 "name": "Jack Rolph",424 "avatar_template": "/user_avatar/discuss.pytorch.org/jack_rolph/{size}/20306_2.png"425 }426 },427 "bookmarks": []428 },429 {430 "post_stream": {431 "posts": [432 {433 "id": 268003,434 "name": "Deeply",435 "username": "Deeply",436 "avatar_template": "/user_avatar/discuss.pytorch.org/deeply/{size}/3840_2.png",437 "created_at": "2021-03-05T17:26:50.650Z",438 "cooked": "<p>I would like to try some new regularization methods and thought of using some <code>torchvision</code> models.</p>\n<p>I only want to work on one model; say for example <code>googlenet</code>, what is the best way to do this:<br>\n1- Download <code>torchvision</code> from <a href=\"https://github.com/pytorch/vision/tree/master/torchvision\" rel=\"noopener nofollow ugc\">source</a>, or<br>\n2 - Download the whole vision from <a href=\"https://github.com/pytorch/vision\" rel=\"noopener nofollow ugc\">source</a>, or<br>\n3- Only <a href=\"https://github.com/pytorch/vision/blob/master/torchvision/models/googlenet.py\" rel=\"noopener nofollow ugc\">download googlenet</a>, which I noticed won’t work as it has some dependencies (have already tried this option).</p>\n<p>Before spending time on this, I appreciate any thoughts on the best way to do this?</p>",439 "post_number": 1,440 "post_type": 1,441 "posts_count": 7,442 "updated_at": "2021-03-05T17:26:50.650Z",443 "reply_count": 0,444 "reply_to_post_number": null,445 "quote_count": 0,446 "incoming_link_count": 358,447 "reads": 13,448 "readers_count": 12,449 "score": 1787.6,450 "yours": false,451 "topic_id": 113849,452 "topic_slug": "what-is-the-best-way-to-do-changes-to-a-torchvision-model",453 "display_username": "Deeply",454 "primary_group_name": null,455 "flair_name": null,456 "flair_url": null,457 "flair_bg_color": null,458 "flair_color": null,459 "flair_group_id": null,460 "badges_granted": [],461 "version": 1,462 "can_edit": false,463 "can_delete": false,464 "can_recover": false,465 "can_see_hidden_post": false,466 "can_wiki": false,467 "link_counts": [468 {469 "url": "https://github.com/pytorch/vision/tree/master/torchvision",470 "internal": false,471 "reflection": false,472 "title": "vision/torchvision at master · pytorch/vision · GitHub",473 "clicks": 2474 },475 {476 "url": "https://github.com/pytorch/vision/blob/master/torchvision/models/googlenet.py",477 "internal": false,478 "reflection": false,479 "title": "vision/googlenet.py at master · pytorch/vision · GitHub",480 "clicks": 0481 },482 {483 "url": "https://github.com/pytorch/vision",484 "internal": false,485 "reflection": false,486 "title": "GitHub - pytorch/vision: Datasets, Transforms and Models specific to Computer Vision",487 "clicks": 0488 }489 ],490 "read": true,491 "user_title": null,492 "bookmarked": false,493 "actions_summary": [],494 "moderator": false,495 "admin": false,496 "staff": false,497 "user_id": 6888,498 "hidden": false,499 "trust_level": 2,500 "deleted_at": null,501 "user_deleted": false,502 "edit_reason": null,503 "can_view_edit_history": true,504 "wiki": false,505 "post_url": "/t/what-is-the-best-way-to-do-changes-to-a-torchvision-model/113849/1",506 "can_accept_answer": false,507 "can_unaccept_answer": false,508 "accepted_answer": false,509 "topic_accepted_answer": true,510 "can_vote": false511 },512 {513 "id": 268090,514 "name": "",515 "username": "ptrblck",516 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",517 "created_at": "2021-03-06T10:20:09.523Z",518 "cooked": "<p>Downloading the source and building it after your manipulation might work, but the easier way would probably be to create a custom <code>nn.Module</code> and use the <code>torchvision</code> model as the base class.<br>\nInside your custom model you could then redefine the internal parameters and methods.</p>",519 "post_number": 2,520 "post_type": 1,521 "posts_count": 7,522 "updated_at": "2021-03-06T10:20:09.523Z",523 "reply_count": 1,524 "reply_to_post_number": null,525 "quote_count": 0,526 "incoming_link_count": 3,527 "reads": 13,528 "readers_count": 12,529 "score": 52.6,530 "yours": false,531 "topic_id": 113849,532 "topic_slug": "what-is-the-best-way-to-do-changes-to-a-torchvision-model",533 "display_username": "",534 "primary_group_name": null,535 "flair_name": null,536 "flair_url": null,537 "flair_bg_color": null,538 "flair_color": null,539 "flair_group_id": null,540 "badges_granted": [],541 "version": 1,542 "can_edit": false,543 "can_delete": false,544 "can_recover": false,545 "can_see_hidden_post": false,546 "can_wiki": false,547 "read": true,548 "user_title": "",549 "bookmarked": false,550 "actions_summary": [551 {552 "id": 2,553 "count": 2554 }555 ],556 "moderator": true,557 "admin": true,558 "staff": true,559 "user_id": 3534,560 "hidden": false,561 "trust_level": 2,562 "deleted_at": null,563 "user_deleted": false,564 "edit_reason": null,565 "can_view_edit_history": true,566 "wiki": false,567 "post_url": "/t/what-is-the-best-way-to-do-changes-to-a-torchvision-model/113849/2",568 "can_accept_answer": false,569 "can_unaccept_answer": false,570 "accepted_answer": false,571 "topic_accepted_answer": true572 },573 {574 "id": 268101,575 "name": "Deeply",576 "username": "Deeply",577 "avatar_template": "/user_avatar/discuss.pytorch.org/deeply/{size}/3840_2.png",578 "created_at": "2021-03-06T11:50:26.437Z",579 "cooked": "<aside class=\"quote no-group\" data-username=\"ptrblck\" data-post=\"2\" data-topic=\"113849\">\n<div class=\"title\">\n<div class=\"quote-controls\"></div>\n<img loading=\"lazy\" alt=\"\" width=\"24\" height=\"24\" src=\"https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/48/1823_2.png\" class=\"avatar\"> ptrblck:</div>\n<blockquote>\n<p>model as the base class.<br>\nInside your custom model you could then redefine the internal parameters and methods</p>\n</blockquote>\n</aside>\n<p>not sure if this would work. For example, I would like to remove <code>droupout</code> from <code>googlenet</code> and use another regularization method.</p>",580 "post_number": 3,581 "post_type": 1,582 "posts_count": 7,583 "updated_at": "2021-03-06T11:50:26.437Z",584 "reply_count": 1,585 "reply_to_post_number": 2,586 "quote_count": 1,587 "incoming_link_count": 0,588 "reads": 12,589 "readers_count": 11,590 "score": 7.4,591 "yours": false,592 "topic_id": 113849,593 "topic_slug": "what-is-the-best-way-to-do-changes-to-a-torchvision-model",594 "display_username": "Deeply",595 "primary_group_name": null,596 "flair_name": null,597 "flair_url": null,598 "flair_bg_color": null,599 "flair_color": null,600 "flair_group_id": null,601 "badges_granted": [],602 "version": 1,603 "can_edit": false,604 "can_delete": false,605 "can_recover": false,606 "can_see_hidden_post": false,607 "can_wiki": false,608 "read": true,609 "user_title": null,610 "bookmarked": false,611 "actions_summary": [],612 "moderator": false,613 "admin": false,614 "staff": false,615 "user_id": 6888,616 "hidden": false,617 "trust_level": 2,618 "deleted_at": null,619 "user_deleted": false,620 "edit_reason": null,621 "can_view_edit_history": true,622 "wiki": false,623 "post_url": "/t/what-is-the-best-way-to-do-changes-to-a-torchvision-model/113849/3",624 "can_accept_answer": false,625 "can_unaccept_answer": false,626 "accepted_answer": false,627 "topic_accepted_answer": true628 },629 {630 "id": 268159,631 "name": "",632 "username": "ptrblck",633 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",634 "created_at": "2021-03-06T21:41:45.454Z",635 "cooked": "<p>The dropout is used as a module, you could replace it with <code>nn.Identity</code>. Alternatively, if it’s used via the functional API, you could override the <code>forward</code> method and reimplement it.<br>\nI still think this might be the faster and cleaner approach than to manipulate the original model and rebuild <code>torchvision</code>, but you might prefer it.</p>",636 "post_number": 4,637 "post_type": 1,638 "posts_count": 7,639 "updated_at": "2021-03-09T13:15:53.846Z",640 "reply_count": 1,641 "reply_to_post_number": 3,642 "quote_count": 0,643 "incoming_link_count": 1,644 "reads": 12,645 "readers_count": 11,646 "score": 27.4,647 "yours": false,648 "topic_id": 113849,649 "topic_slug": "what-is-the-best-way-to-do-changes-to-a-torchvision-model",650 "display_username": "",651 "primary_group_name": null,652 "flair_name": null,653 "flair_url": null,654 "flair_bg_color": null,655 "flair_color": null,656 "flair_group_id": null,657 "badges_granted": [],658 "version": 1,659 "can_edit": false,660 "can_delete": false,661 "can_recover": false,662 "can_see_hidden_post": false,663 "can_wiki": false,664 "read": true,665 "user_title": "",666 "reply_to_user": {667 "id": 6888,668 "username": "Deeply",669 "name": "Deeply",670 "avatar_template": "/user_avatar/discuss.pytorch.org/deeply/{size}/3840_2.png"671 },672 "bookmarked": false,673 "actions_summary": [674 {675 "id": 2,676 "count": 1677 }678 ],679 "moderator": true,680 "admin": true,681 "staff": true,682 "user_id": 3534,683 "hidden": false,684 "trust_level": 2,685 "deleted_at": null,686 "user_deleted": false,687 "edit_reason": null,688 "can_view_edit_history": true,689 "wiki": false,690 "post_url": "/t/what-is-the-best-way-to-do-changes-to-a-torchvision-model/113849/4",691 "can_accept_answer": false,692 "can_unaccept_answer": false,693 "accepted_answer": true,694 "topic_accepted_answer": true695 },696 {697 "id": 268566,698 "name": "Deeply",699 "username": "Deeply",700 "avatar_template": "/user_avatar/discuss.pytorch.org/deeply/{size}/3840_2.png",701 "created_at": "2021-03-08T14:25:23.980Z",702 "cooked": "<p>Thank you! I agree that overriding the forward function would be the best way to do it. To do this, do you mean to build a new <code>googlenet_like class</code> inherited from from <code>googlenet class</code>, then, to override the forward function?</p>\n<p>NB. Currently, <code>googlenet</code> in tochvision/googlenet.py is working out-of-the-box without any extra dependencies, except the need to complement it with utils.py (<code>from utils import load_state_dict_from_url</code>); although this has little impact for my case as the pretrained model has dropout in its structure. Got acc1 of 20% after 9 epochs on CIFAR-100, trained from scratch.</p>\n<p>Sounds I had the problem with the logits thing, and have resolved it already.</p>",703 "post_number": 5,704 "post_type": 1,705 "posts_count": 7,706 "updated_at": "2021-03-08T14:25:23.980Z",707 "reply_count": 1,708 "reply_to_post_number": 4,709 "quote_count": 0,710 "incoming_link_count": 4,711 "reads": 11,712 "readers_count": 10,713 "score": 27.2,714 "yours": false,715 "topic_id": 113849,716 "topic_slug": "what-is-the-best-way-to-do-changes-to-a-torchvision-model",717 "display_username": "Deeply",718 "primary_group_name": null,719 "flair_name": null,720 "flair_url": null,721 "flair_bg_color": null,722 "flair_color": null,723 "flair_group_id": null,724 "badges_granted": [],725 "version": 1,726 "can_edit": false,727 "can_delete": false,728 "can_recover": false,729 "can_see_hidden_post": false,730 "can_wiki": false,731 "read": true,732 "user_title": null,733 "reply_to_user": {734 "id": 3534,735 "username": "ptrblck",736 "name": "",737 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png"738 },739 "bookmarked": false,740 "actions_summary": [],741 "moderator": false,742 "admin": false,743 "staff": false,744 "user_id": 6888,745 "hidden": false,746 "trust_level": 2,747 "deleted_at": null,748 "user_deleted": false,749 "edit_reason": null,750 "can_view_edit_history": true,751 "wiki": false,752 "post_url": "/t/what-is-the-best-way-to-do-changes-to-a-torchvision-model/113849/5",753 "can_accept_answer": false,754 "can_unaccept_answer": false,755 "accepted_answer": false,756 "topic_accepted_answer": true757 },758 {759 "id": 268765,760 "name": "",761 "username": "ptrblck",762 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",763 "created_at": "2021-03-09T05:39:42.756Z",764 "cooked": "<p>Yes, that what I was thinking. However, if simply copy-pasting the model code (without the unused <code>utils</code>) works fine, it could be easier to just to this. In any way, I would try to avoid changing the code directly in <code>torchvision</code> and rebuilding it, as these changes would live in you branch of <code>torchvision</code> and (if I’m understanding your use case correctly) you don’t want to commit these changes, as they are a standalone experiment.</p>",765 "post_number": 6,766 "post_type": 1,767 "posts_count": 7,768 "updated_at": "2021-03-09T05:39:42.756Z",769 "reply_count": 1,770 "reply_to_post_number": 5,771 "quote_count": 0,772 "incoming_link_count": 1,773 "reads": 10,774 "readers_count": 9,775 "score": 27.0,776 "yours": false,777 "topic_id": 113849,778 "topic_slug": "what-is-the-best-way-to-do-changes-to-a-torchvision-model",779 "display_username": "",780 "primary_group_name": null,781 "flair_name": null,782 "flair_url": null,783 "flair_bg_color": null,784 "flair_color": null,785 "flair_group_id": null,786 "badges_granted": [],787 "version": 1,788 "can_edit": false,789 "can_delete": false,790 "can_recover": false,791 "can_see_hidden_post": false,792 "can_wiki": false,793 "read": true,794 "user_title": "",795 "reply_to_user": {796 "id": 6888,797 "username": "Deeply",798 "name": "Deeply",799 "avatar_template": "/user_avatar/discuss.pytorch.org/deeply/{size}/3840_2.png"800 },801 "bookmarked": false,802 "actions_summary": [803 {804 "id": 2,805 "count": 1806 }807 ],808 "moderator": true,809 "admin": true,810 "staff": true,811 "user_id": 3534,812 "hidden": false,813 "trust_level": 2,814 "deleted_at": null,815 "user_deleted": false,816 "edit_reason": null,817 "can_view_edit_history": true,818 "wiki": false,819 "post_url": "/t/what-is-the-best-way-to-do-changes-to-a-torchvision-model/113849/6",820 "can_accept_answer": false,821 "can_unaccept_answer": false,822 "accepted_answer": false,823 "topic_accepted_answer": true824 },825 {826 "id": 268891,827 "name": "Deeply",828 "username": "Deeply",829 "avatar_template": "/user_avatar/discuss.pytorch.org/deeply/{size}/3840_2.png",830 "created_at": "2021-03-09T13:16:26.180Z",831 "cooked": "<aside class=\"quote no-group\" data-username=\"ptrblck\" data-post=\"6\" data-topic=\"113849\">\n<div class=\"title\">\n<div class=\"quote-controls\"></div>\n<img loading=\"lazy\" alt=\"\" width=\"24\" height=\"24\" src=\"https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/48/1823_2.png\" class=\"avatar\"> ptrblck:</div>\n<blockquote>\n<p>I would try to avoid changing the code directly in <code>torchvision</code> and rebuilding it</p>\n</blockquote>\n</aside>\n<p>Absolutely, it is a separate project.</p>",832 "post_number": 7,833 "post_type": 1,834 "posts_count": 7,835 "updated_at": "2021-03-09T13:16:26.180Z",836 "reply_count": 0,837 "reply_to_post_number": 6,838 "quote_count": 1,839 "incoming_link_count": 4,840 "reads": 8,841 "readers_count": 7,842 "score": 21.6,843 "yours": false,844 "topic_id": 113849,845 "topic_slug": "what-is-the-best-way-to-do-changes-to-a-torchvision-model",846 "display_username": "Deeply",847 "primary_group_name": null,848 "flair_name": null,849 "flair_url": null,850 "flair_bg_color": null,851 "flair_color": null,852 "flair_group_id": null,853 "badges_granted": [],854 "version": 1,855 "can_edit": false,856 "can_delete": false,857 "can_recover": false,858 "can_see_hidden_post": false,859 "can_wiki": false,860 "read": true,861 "user_title": null,862 "bookmarked": false,863 "actions_summary": [],864 "moderator": false,865 "admin": false,866 "staff": false,867 "user_id": 6888,868 "hidden": false,869 "trust_level": 2,870 "deleted_at": null,871 "user_deleted": false,872 "edit_reason": null,873 "can_view_edit_history": true,874 "wiki": false,875 "post_url": "/t/what-is-the-best-way-to-do-changes-to-a-torchvision-model/113849/7",876 "can_accept_answer": false,877 "can_unaccept_answer": false,878 "accepted_answer": false,879 "topic_accepted_answer": true880 }881 ],882 "stream": [883 268003,884 268090,885 268101,886 268159,887 268566,888 268765,889 268891890 ]891 },892 "timeline_lookup": [893 [894 1,895 1695896 ],897 [898 2,899 1694900 ],901 [902 5,903 1692904 ],905 [906 7,907 1691908 ]909 ],910 "suggested_topics": [911 {912 "fancy_title": "An error regarding the 5070TI not supporting SM_120",913 "id": 218072,914 "title": "An error regarding the 5070TI not supporting SM_120",915 "slug": "an-error-regarding-the-5070ti-not-supporting-sm-120",916 "posts_count": 3,917 "reply_count": 1,918 "highest_post_number": 3,919 "image_url": null,920 "created_at": "2025-03-20T12:44:55.261Z",921 "last_posted_at": "2025-09-15T06:56:12.228Z",922 "bumped": true,923 "bumped_at": "2025-09-15T06:56:12.228Z",924 "archetype": "regular",925 "unseen": false,926 "pinned": false,927 "unpinned": null,928 "visible": true,929 "closed": false,930 "archived": false,931 "bookmarked": null,932 "liked": null,933 "tags_descriptions": {},934 "like_count": 0,935 "views": 2905,936 "category_id": 1,937 "featured_link": null,938 "has_accepted_answer": false,939 "posters": [940 {941 "extras": null,942 "description": "Original Poster",943 "user": {944 "id": 83383,945 "username": "qiang",946 "name": "七月的天气",947 "avatar_template": "/letter_avatar_proxy/v4/letter/q/c68b51/{size}.png",948 "trust_level": 0949 }950 },951 {952 "extras": null,953 "description": "Frequent Poster",954 "user": {955 "id": 3534,956 "username": "ptrblck",957 "name": "",958 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",959 "admin": true,960 "moderator": true,961 "trust_level": 2962 }963 },964 {965 "extras": "latest",966 "description": "Most Recent Poster",967 "user": {968 "id": 85870,969 "username": "frank85729",970 "name": "Frank85729",971 "avatar_template": "/user_avatar/discuss.pytorch.org/frank85729/{size}/78231_2.png",972 "trust_level": 1973 }974 }975 ]976 },977 {978 "fancy_title": "Educational Framework Inspired by PyTorch",979 "id": 215288,980 "title": "Educational Framework Inspired by PyTorch",981 "slug": "educational-framework-inspired-by-pytorch",982 "posts_count": 1,983 "reply_count": 0,984 "highest_post_number": 1,985 "image_url": null,986 "created_at": "2025-01-12T00:56:41.683Z",987 "last_posted_at": "2025-01-12T00:56:41.721Z",988 "bumped": true,989 "bumped_at": "2025-01-12T00:56:41.721Z",990 "archetype": "regular",991 "unseen": false,992 "pinned": false,993 "unpinned": null,994 "visible": true,995 "closed": false,996 "archived": false,997 "bookmarked": null,998 "liked": null,999 "tags_descriptions": {},1000 "like_count": 1,1001 "views": 87,1002 "category_id": 1,1003 "featured_link": null,1004 "has_accepted_answer": false,1005 "posters": [1006 {1007 "extras": "latest single",1008 "description": "Original Poster, Most Recent Poster",1009 "user": {1010 "id": 82031,1011 "username": "joseortizm",1012 "name": "",1013 "avatar_template": "/user_avatar/discuss.pytorch.org/joseortizm/{size}/75060_2.png",1014 "trust_level": 01015 }1016 }1017 ]1018 },1019 {1020 "fancy_title": "Pytorch and torchvision for CUDA10.2",1021 "id": 215865,1022 "title": "Pytorch and torchvision for CUDA10.2",1023 "slug": "pytorch-and-torchvision-for-cuda10-2",1024 "posts_count": 4,1025 "reply_count": 2,1026 "highest_post_number": 4,1027 "image_url": null,1028 "created_at": "2025-01-25T17:12:17.383Z",1029 "last_posted_at": "2025-01-28T13:26:07.975Z",1030 "bumped": true,1031 "bumped_at": "2025-01-28T13:26:07.975Z",1032 "archetype": "regular",1033 "unseen": false,1034 "pinned": false,1035 "unpinned": null,1036 "visible": true,1037 "closed": false,1038 "archived": false,1039 "bookmarked": null,1040 "liked": null,1041 "tags_descriptions": {},1042 "like_count": 1,1043 "views": 335,1044 "category_id": 1,1045 "featured_link": null,1046 "has_accepted_answer": false,1047 "posters": [1048 {1049 "extras": null,1050 "description": "Original Poster",1051 "user": {1052 "id": 82311,1053 "username": "Bekzod",1054 "name": "",1055 "avatar_template": "/user_avatar/discuss.pytorch.org/bekzod/{size}/75336_2.png",1056 "trust_level": 01057 }1058 },1059 {1060 "extras": "latest",1061 "description": "Most Recent Poster",1062 "user": {1063 "id": 3534,1064 "username": "ptrblck",1065 "name": "",1066 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",1067 "admin": true,1068 "moderator": true,1069 "trust_level": 21070 }1071 }1072 ]1073 },1074 {1075 "fancy_title": "Conda Pytorch keeps downgrading itself from cuda -> cpu",1076 "id": 216525,1077 "title": "Conda Pytorch keeps downgrading itself from cuda -> cpu",1078 "slug": "conda-pytorch-keeps-downgrading-itself-from-cuda-cpu",1079 "posts_count": 4,1080 "reply_count": 1,1081 "highest_post_number": 4,1082 "image_url": null,1083 "created_at": "2025-02-11T10:46:19.656Z",1084 "last_posted_at": "2025-02-11T20:21:44.809Z",1085 "bumped": true,1086 "bumped_at": "2025-02-11T20:21:44.809Z",1087 "archetype": "regular",1088 "unseen": false,1089 "pinned": false,1090 "unpinned": null,1091 "visible": true,1092 "closed": false,1093 "archived": false,1094 "bookmarked": null,1095 "liked": null,1096 "tags_descriptions": {},1097 "like_count": 0,1098 "views": 251,1099 "category_id": 1,1100 "featured_link": null,1101 "has_accepted_answer": false,1102 "posters": [1103 {1104 "extras": null,1105 "description": "Original Poster",1106 "user": {1107 "id": 71072,1108 "username": "wilhelm",1109 "name": "Dave",1110 "avatar_template": "/user_avatar/discuss.pytorch.org/wilhelm/{size}/70280_2.png",1111 "trust_level": 21112 }1113 },1114 {1115 "extras": "latest",1116 "description": "Most Recent Poster",1117 "user": {1118 "id": 3534,1119 "username": "ptrblck",1120 "name": "",1121 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",1122 "admin": true,1123 "moderator": true,1124 "trust_level": 21125 }1126 }1127 ]1128 },1129 {1130 "fancy_title": "Profiling what happens in thread",1131 "id": 220919,1132 "title": "Profiling what happens in thread",1133 "slug": "profiling-what-happens-in-thread",1134 "posts_count": 3,1135 "reply_count": 0,1136 "highest_post_number": 3,1137 "image_url": "https://discuss.pytorch.org/uploads/default/optimized/3X/3/4/34b45504a210d31bebc9103a23a6383bf4e95724_2_1024x341.png",1138 "created_at": "2025-06-19T13:19:33.147Z",1139 "last_posted_at": "2025-06-23T08:27:42.581Z",1140 "bumped": true,1141 "bumped_at": "2025-06-23T08:27:42.581Z",1142 "archetype": "regular",1143 "unseen": false,1144 "pinned": false,1145 "unpinned": null,1146 "visible": true,1147 "closed": false,1148 "archived": false,1149 "bookmarked": null,1150 "liked": null,1151 "tags_descriptions": {},1152 "like_count": 0,1153 "views": 81,1154 "category_id": 1,1155 "featured_link": null,1156 "has_accepted_answer": true,1157 "posters": [1158 {1159 "extras": "latest single",1160 "description": "Original Poster, Most Recent Poster, Accepted Answer",1161 "user": {1162 "id": 12347,1163 "username": "Valentin_Thorey",1164 "name": "Valentin Thorey",1165 "avatar_template": "/user_avatar/discuss.pytorch.org/valentin_thorey/{size}/66795_2.png",1166 "trust_level": 11167 }1168 }1169 ]1170 }1171 ],1172 "tags_descriptions": {},1173 "fancy_title": "What is the best way to do changes to a torchvision model",1174 "id": 113849,1175 "title": "What is the best way to do changes to a torchvision model",1176 "posts_count": 7,1177 "created_at": "2021-03-05T17:26:50.592Z",1178 "views": 881,1179 "reply_count": 5,1180 "like_count": 4,1181 "last_posted_at": "2021-03-09T13:16:26.180Z",1182 "visible": true,1183 "closed": false,1184 "archived": false,1185 "has_summary": false,1186 "archetype": "regular",1187 "slug": "what-is-the-best-way-to-do-changes-to-a-torchvision-model",1188 "category_id": 1,1189 "word_count": 478,1190 "deleted_at": null,1191 "user_id": 6888,1192 "featured_link": null,1193 "pinned_globally": false,1194 "pinned_at": null,1195 "pinned_until": null,1196 "image_url": null,1197 "slow_mode_seconds": 0,1198 "draft": null,1199 "draft_key": "topic_113849",1200 "draft_sequence": null,