Anurag1734/cuda-error-resolution-analysis
07
1[2 {3 "post_stream": {4 "posts": [5 {6 "id": 271405,7 "name": "Shrutishrestha",8 "username": "shrutishrestha",9 "avatar_template": "/user_avatar/discuss.pytorch.org/shrutishrestha/{size}/51287_2.png",10 "created_at": "2021-03-19T06:20:23.250Z",11 "cooked": "<p>When I use model.eval() with torch.no_grad() during my evaluation, I get 36% accuracy. But, if I do not use model.eval() and only use torch.no_grad() then my accuracy is 87%. Why is this hapenning?<br>\nwith_model_eval<br>\n<div class=\"lightbox-wrapper\"><a class=\"lightbox\" href=\"https://discuss.pytorch.org/uploads/default/original/3X/6/8/68a243508a3a8033d838f0e23c2f6d05c131f959.jpeg\" data-download-href=\"https://discuss.pytorch.org/uploads/default/68a243508a3a8033d838f0e23c2f6d05c131f959\" title=\"with_model_eval\"><img src=\"https://discuss.pytorch.org/uploads/default/optimized/3X/6/8/68a243508a3a8033d838f0e23c2f6d05c131f959_2_345x97.jpeg\" alt=\"with_model_eval\" data-base62-sha1=\"eVDiPnxMxybQEAmz4NgZClm7DtL\" width=\"345\" height=\"97\" srcset=\"https://discuss.pytorch.org/uploads/default/optimized/3X/6/8/68a243508a3a8033d838f0e23c2f6d05c131f959_2_345x97.jpeg, https://discuss.pytorch.org/uploads/default/optimized/3X/6/8/68a243508a3a8033d838f0e23c2f6d05c131f959_2_517x145.jpeg 1.5x, https://discuss.pytorch.org/uploads/default/optimized/3X/6/8/68a243508a3a8033d838f0e23c2f6d05c131f959_2_690x194.jpeg 2x\" data-dominant-color=\"3A281F\"><div class=\"meta\"><svg class=\"fa d-icon d-icon-far-image svg-icon\" aria-hidden=\"true\"><use href=\"#far-image\"></use></svg><span class=\"filename\">with_model_eval</span><span class=\"informations\">1883×530 53.4 KB</span><svg class=\"fa d-icon d-icon-discourse-expand svg-icon\" aria-hidden=\"true\"><use href=\"#discourse-expand\"></use></svg></div></a></div></p>\n<p>without_model_eval<br>\n<div class=\"lightbox-wrapper\"><a class=\"lightbox\" href=\"https://discuss.pytorch.org/uploads/default/original/3X/1/2/126aa1d7ec1b019ab68813446e0da6a31aef7ac0.jpeg\" data-download-href=\"https://discuss.pytorch.org/uploads/default/126aa1d7ec1b019ab68813446e0da6a31aef7ac0\" title=\"without_model_eval\"><img src=\"https://discuss.pytorch.org/uploads/default/optimized/3X/1/2/126aa1d7ec1b019ab68813446e0da6a31aef7ac0_2_345x97.jpeg\" alt=\"without_model_eval\" data-base62-sha1=\"2CV3hmK9kpdOd5ggOMdA3QS1ujm\" width=\"345\" height=\"97\" srcset=\"https://discuss.pytorch.org/uploads/default/optimized/3X/1/2/126aa1d7ec1b019ab68813446e0da6a31aef7ac0_2_345x97.jpeg, https://discuss.pytorch.org/uploads/default/optimized/3X/1/2/126aa1d7ec1b019ab68813446e0da6a31aef7ac0_2_517x145.jpeg 1.5x, https://discuss.pytorch.org/uploads/default/optimized/3X/1/2/126aa1d7ec1b019ab68813446e0da6a31aef7ac0_2_690x194.jpeg 2x\" data-dominant-color=\"342118\"><div class=\"meta\"><svg class=\"fa d-icon d-icon-far-image svg-icon\" aria-hidden=\"true\"><use href=\"#far-image\"></use></svg><span class=\"filename\">without_model_eval</span><span class=\"informations\">1883×530 47.2 KB</span><svg class=\"fa d-icon d-icon-discourse-expand svg-icon\" aria-hidden=\"true\"><use href=\"#discourse-expand\"></use></svg></div></a></div></p>",12 "post_number": 1,13 "post_type": 1,14 "posts_count": 2,15 "updated_at": "2021-03-19T06:20:54.828Z",16 "reply_count": 0,17 "reply_to_post_number": null,18 "quote_count": 0,19 "incoming_link_count": 354,20 "reads": 10,21 "readers_count": 9,22 "score": 1782.0,23 "yours": false,24 "topic_id": 115308,25 "topic_slug": "using-and-not-using-model-eval",26 "display_username": "Shrutishrestha",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": 1,35 "can_edit": false,36 "can_delete": false,37 "can_recover": false,38 "can_see_hidden_post": false,39 "can_wiki": false,40 "link_counts": [41 {42 "url": "https://discuss.pytorch.org/uploads/default/original/3X/6/8/68a243508a3a8033d838f0e23c2f6d05c131f959.jpeg",43 "internal": true,44 "reflection": false,45 "clicks": 046 },47 {48 "url": "https://discuss.pytorch.org/uploads/default/original/3X/1/2/126aa1d7ec1b019ab68813446e0da6a31aef7ac0.jpeg",49 "internal": true,50 "reflection": false,51 "clicks": 052 }53 ],54 "read": true,55 "user_title": null,56 "bookmarked": false,57 "actions_summary": [58 {59 "id": 2,60 "count": 161 }62 ],63 "moderator": false,64 "admin": false,65 "staff": false,66 "user_id": 34483,67 "hidden": false,68 "trust_level": 2,69 "deleted_at": null,70 "user_deleted": false,71 "edit_reason": null,72 "can_view_edit_history": true,73 "wiki": false,74 "post_url": "/t/using-and-not-using-model-eval/115308/1",75 "can_accept_answer": false,76 "can_unaccept_answer": false,77 "accepted_answer": false,78 "topic_accepted_answer": null,79 "can_vote": false80 },81 {82 "id": 271408,83 "name": "",84 "username": "ptrblck",85 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",86 "created_at": "2021-03-19T06:28:37.200Z",87 "cooked": "<p>This might be happening if you are using e.g. batchnorm layers, with “bad” running stats for the mean and var. If you use the model in training mode during evaluation, the batch statistics will be used and the running stats updated (which could be seen as a data leak for future <code>model.eval()</code> validation runs).<br>\nYou could play around with the <code>momentum</code> to smooth the update of the running stats and check other posts in this forum, as other users are also facing this issue.</p>",88 "post_number": 2,89 "post_type": 1,90 "posts_count": 2,91 "updated_at": "2021-03-19T06:28:37.200Z",92 "reply_count": 0,93 "reply_to_post_number": null,94 "quote_count": 0,95 "incoming_link_count": 6,96 "reads": 9,97 "readers_count": 8,98 "score": 31.8,99 "yours": false,100 "topic_id": 115308,101 "topic_slug": "using-and-not-using-model-eval",102 "display_username": "",103 "primary_group_name": null,104 "flair_name": null,105 "flair_url": null,106 "flair_bg_color": null,107 "flair_color": null,108 "flair_group_id": null,109 "badges_granted": [],110 "version": 1,111 "can_edit": false,112 "can_delete": false,113 "can_recover": false,114 "can_see_hidden_post": false,115 "can_wiki": false,116 "read": true,117 "user_title": "",118 "bookmarked": false,119 "actions_summary": [],120 "moderator": true,121 "admin": true,122 "staff": true,123 "user_id": 3534,124 "hidden": false,125 "trust_level": 2,126 "deleted_at": null,127 "user_deleted": false,128 "edit_reason": null,129 "can_view_edit_history": true,130 "wiki": false,131 "post_url": "/t/using-and-not-using-model-eval/115308/2",132 "can_accept_answer": false,133 "can_unaccept_answer": false,134 "accepted_answer": false,135 "topic_accepted_answer": null136 }137 ],138 "stream": [139 271405,140 271408141 ]142 },143 "timeline_lookup": [144 [145 1,146 1682147 ]148 ],149 "suggested_topics": [150 {151 "fancy_title": "About the uniqueness of eigen decomposition with torch.eigh",152 "id": 212468,153 "title": "About the uniqueness of eigen decomposition with torch.eigh",154 "slug": "about-the-uniqueness-of-eigen-decomposition-with-torch-eigh",155 "posts_count": 2,156 "reply_count": 0,157 "highest_post_number": 2,158 "image_url": "https://discuss.pytorch.org/uploads/default/original/3X/2/1/21f5e2cb3119b8de74bce7c3b6be1a76554ed46d.png",159 "created_at": "2024-11-03T08:41:03.397Z",160 "last_posted_at": "2024-11-04T21:39:04.342Z",161 "bumped": true,162 "bumped_at": "2024-11-04T21:39:04.342Z",163 "archetype": "regular",164 "unseen": false,165 "pinned": false,166 "unpinned": null,167 "visible": true,168 "closed": false,169 "archived": false,170 "bookmarked": null,171 "liked": null,172 "tags_descriptions": {},173 "like_count": 1,174 "views": 91,175 "category_id": 1,176 "featured_link": null,177 "has_accepted_answer": true,178 "posters": [179 {180 "extras": null,181 "description": "Original Poster",182 "user": {183 "id": 80508,184 "username": "wasabi_linguist",185 "name": "",186 "avatar_template": "/user_avatar/discuss.pytorch.org/wasabi_linguist/{size}/73592_2.png",187 "trust_level": 1188 }189 },190 {191 "extras": "latest",192 "description": "Most Recent Poster, Accepted Answer",193 "user": {194 "id": 18088,195 "username": "KFrank",196 "name": "K. Frank",197 "avatar_template": "/letter_avatar_proxy/v4/letter/k/ecb155/{size}.png",198 "trust_level": 2199 }200 }201 ]202 },203 {204 "fancy_title": "Computing averages",205 "id": 213152,206 "title": "Computing averages",207 "slug": "computing-averages",208 "posts_count": 6,209 "reply_count": 2,210 "highest_post_number": 6,211 "image_url": null,212 "created_at": "2024-11-19T11:26:31.040Z",213 "last_posted_at": "2024-11-27T07:16:01.463Z",214 "bumped": true,215 "bumped_at": "2024-11-27T07:16:01.463Z",216 "archetype": "regular",217 "unseen": false,218 "pinned": false,219 "unpinned": null,220 "visible": true,221 "closed": false,222 "archived": false,223 "bookmarked": null,224 "liked": null,225 "tags_descriptions": {},226 "like_count": 1,227 "views": 93,228 "category_id": 1,229 "featured_link": null,230 "has_accepted_answer": true,231 "posters": [232 {233 "extras": "latest",234 "description": "Original Poster, Most Recent Poster",235 "user": {236 "id": 71200,237 "username": "KukumavMozolo",238 "name": "Mozolo",239 "avatar_template": "/user_avatar/discuss.pytorch.org/kukumavmozolo/{size}/65719_2.png",240 "trust_level": 1241 }242 },243 {244 "extras": null,245 "description": "Frequent Poster, Accepted Answer",246 "user": {247 "id": 3534,248 "username": "ptrblck",249 "name": "",250 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",251 "admin": true,252 "moderator": true,253 "trust_level": 2254 }255 }256 ]257 },258 {259 "fancy_title": "Saving predicted results consumes too much memory",260 "id": 214215,261 "title": "Saving predicted results consumes too much memory",262 "slug": "saving-predicted-results-consumes-too-much-memory",263 "posts_count": 2,264 "reply_count": 0,265 "highest_post_number": 2,266 "image_url": null,267 "created_at": "2024-12-14T04:51:01.618Z",268 "last_posted_at": "2024-12-14T20:31:01.239Z",269 "bumped": true,270 "bumped_at": "2024-12-14T20:31:01.239Z",271 "archetype": "regular",272 "unseen": false,273 "pinned": false,274 "unpinned": null,275 "visible": true,276 "closed": false,277 "archived": false,278 "bookmarked": null,279 "liked": null,280 "tags_descriptions": {},281 "like_count": 0,282 "views": 32,283 "category_id": 1,284 "featured_link": null,285 "has_accepted_answer": false,286 "posters": [287 {288 "extras": null,289 "description": "Original Poster",290 "user": {291 "id": 81499,292 "username": "linuxonly801",293 "name": "Linuxonly801",294 "avatar_template": "/user_avatar/discuss.pytorch.org/linuxonly801/{size}/74511_2.png",295 "trust_level": 1296 }297 },298 {299 "extras": "latest",300 "description": "Most Recent Poster",301 "user": {302 "id": 3534,303 "username": "ptrblck",304 "name": "",305 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",306 "admin": true,307 "moderator": true,308 "trust_level": 2309 }310 }311 ]312 },313 {314 "fancy_title": "StyleGAN2 Training Illegal Instruction on cache clear instruction",315 "id": 214804,316 "title": "StyleGAN2 Training Illegal Instruction on cache clear instruction",317 "slug": "stylegan2-training-illegal-instruction-on-cache-clear-instruction",318 "posts_count": 1,319 "reply_count": 0,320 "highest_post_number": 1,321 "image_url": null,322 "created_at": "2024-12-30T22:41:09.300Z",323 "last_posted_at": "2024-12-30T22:41:09.340Z",324 "bumped": true,325 "bumped_at": "2024-12-30T22:41:09.340Z",326 "archetype": "regular",327 "unseen": false,328 "pinned": false,329 "unpinned": null,330 "visible": true,331 "closed": false,332 "archived": false,333 "bookmarked": null,334 "liked": null,335 "tags_descriptions": {},336 "like_count": 0,337 "views": 31,338 "category_id": 1,339 "featured_link": null,340 "has_accepted_answer": false,341 "posters": [342 {343 "extras": "latest single",344 "description": "Original Poster, Most Recent Poster",345 "user": {346 "id": 76733,347 "username": "YM2132",348 "name": "",349 "avatar_template": "/user_avatar/discuss.pytorch.org/ym2132/{size}/70814_2.png",350 "trust_level": 1351 }352 }353 ]354 },355 {356 "fancy_title": "`torch.nn.functional.conv2d` 8 times slower in torch 2.5.1 compared to 2.3.1",357 "id": 216011,358 "title": "`torch.nn.functional.conv2d` 8 times slower in torch 2.5.1 compared to 2.3.1",359 "slug": "torch-nn-functional-conv2d-8-times-slower-in-torch-2-5-1-compared-to-2-3-1",360 "posts_count": 2,361 "reply_count": 0,362 "highest_post_number": 2,363 "image_url": null,364 "created_at": "2025-01-29T03:27:00.031Z",365 "last_posted_at": "2025-02-07T15:29:19.315Z",366 "bumped": true,367 "bumped_at": "2025-02-07T15:29:19.315Z",368 "archetype": "regular",369 "unseen": false,370 "pinned": false,371 "unpinned": null,372 "visible": true,373 "closed": false,374 "archived": false,375 "bookmarked": null,376 "liked": null,377 "tags_descriptions": {},378 "like_count": 0,379 "views": 156,380 "category_id": 1,381 "featured_link": null,382 "has_accepted_answer": true,383 "posters": [384 {385 "extras": "latest single",386 "description": "Original Poster, Most Recent Poster, Accepted Answer",387 "user": {388 "id": 82377,389 "username": "navidsam",390 "name": "Navid",391 "avatar_template": "/user_avatar/discuss.pytorch.org/navidsam/{size}/75363_2.png",392 "trust_level": 0393 }394 }395 ]396 }397 ],398 "tags_descriptions": {},399 "fancy_title": "Using and not using model.eval()",400 "id": 115308,401 "title": "Using and not using model.eval()",402 "posts_count": 2,403 "created_at": "2021-03-19T06:20:23.191Z",404 "views": 970,405 "reply_count": 0,406 "like_count": 1,407 "last_posted_at": "2021-03-19T06:28:37.200Z",408 "visible": true,409 "closed": false,410 "archived": false,411 "has_summary": false,412 "archetype": "regular",413 "slug": "using-and-not-using-model-eval",414 "category_id": 1,415 "word_count": 137,416 "deleted_at": null,417 "user_id": 34483,418 "featured_link": null,419 "pinned_globally": false,420 "pinned_at": null,421 "pinned_until": null,422 "image_url": "https://discuss.pytorch.org/uploads/default/optimized/3X/6/8/68a243508a3a8033d838f0e23c2f6d05c131f959_2_1023x288.jpeg",423 "slow_mode_seconds": 0,424 "draft": null,425 "draft_key": "topic_115308",426 "draft_sequence": null,427 "unpinned": null,428 "pinned": false,429 "current_post_number": 1,430 "highest_post_number": 2,431 "deleted_by": null,432 "actions_summary": [433 {434 "id": 4,435 "count": 0,436 "hidden": false,437 "can_act": false438 },439 {440 "id": 8,441 "count": 0,442 "hidden": false,443 "can_act": false444 },445 {446 "id": 10,447 "count": 0,448 "hidden": false,449 "can_act": false450 },451 {452 "id": 7,453 "count": 0,454 "hidden": false,455 "can_act": false456 }457 ],458 "chunk_size": 20,459 "bookmarked": false,460 "topic_timer": null,461 "message_bus_last_id": 0,462 "participant_count": 2,463 "show_read_indicator": false,464 "thumbnails": [465 {466 "max_width": null,467 "max_height": null,468 "width": 1883,469 "height": 530,470 "url": "https://discuss.pytorch.org/uploads/default/original/3X/6/8/68a243508a3a8033d838f0e23c2f6d05c131f959.jpeg"471 },472 {473 "max_width": 1024,474 "max_height": 1024,475 "width": 1023,476 "height": 288,477 "url": "https://discuss.pytorch.org/uploads/default/optimized/3X/6/8/68a243508a3a8033d838f0e23c2f6d05c131f959_2_1023x288.jpeg"478 }479 ],480 "slow_mode_enabled_until": null,481 "can_vote": false,482 "vote_count": 0,483 "user_voted": false,484 "discourse_zendesk_plugin_zendesk_id": null,485 "discourse_zendesk_plugin_zendesk_url": "https://your-url.zendesk.com/agent/tickets/",486 "details": {487 "can_edit": false,488 "notification_level": 1,489 "participants": [490 {491 "id": 3534,492 "username": "ptrblck",493 "name": "",494 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",495 "post_count": 1,496 "primary_group_name": null,497 "flair_name": null,498 "flair_url": null,499 "flair_color": null,500 "flair_bg_color": null,501 "flair_group_id": null,502 "admin": true,503 "moderator": true,504 "trust_level": 2505 },506 {507 "id": 34483,508 "username": "shrutishrestha",509 "name": "Shrutishrestha",510 "avatar_template": "/user_avatar/discuss.pytorch.org/shrutishrestha/{size}/51287_2.png",511 "post_count": 1,512 "primary_group_name": null,513 "flair_name": null,514 "flair_url": null,515 "flair_color": null,516 "flair_bg_color": null,517 "flair_group_id": null,518 "trust_level": 2519 }520 ],521 "created_by": {522 "id": 34483,523 "username": "shrutishrestha",524 "name": "Shrutishrestha",525 "avatar_template": "/user_avatar/discuss.pytorch.org/shrutishrestha/{size}/51287_2.png"526 },527 "last_poster": {528 "id": 3534,529 "username": "ptrblck",530 "name": "",531 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png"532 }533 },534 "bookmarks": []535 },536 {537 "post_stream": {538 "posts": [539 {540 "id": 271247,541 "name": "Stefano Berti",542 "username": "StefanoBerti",543 "avatar_template": "/user_avatar/discuss.pytorch.org/stefanoberti/{size}/32286_2.png",544 "created_at": "2021-03-18T11:01:02.632Z",545 "cooked": "<p>I need to insert elements of tensor <em>new</em> into a tensor <em>old</em> with a certain probability, let’s say that it is 0.8 for simplicity. Substantially this is what masked_fill would do, but it only works with monodimensional tensor. Actually I am doing</p>\n<pre><code class=\"lang-auto\"> prob = torch.rand(trgs.shape, dtype=torch.float32).to(trgs.device)\n mask = prob < 0.8\n\n dim1, dim2, dim3, dim4 = new.shape\n for a in range(dim1):\n for b in range(dim2):\n for c in range(dim3):\n for d in range(dim4):\n old[a][b][c][d] = old[a][b][c][d] if mask[a][b][c][d] else new[a][b][c][d]\n</code></pre>\n<p>which is awful. I would like something like</p>\n<pre><code class=\"lang-auto\"> prob = torch.rand(trgs.shape, dtype=torch.float32).to(trgs.device)\n mask = prob < 0.8\n\n old = trgs.multidimensional_masked_fill(mask, new)\n</code></pre>",546 "post_number": 1,547 "post_type": 1,548 "posts_count": 2,549 "updated_at": "2021-03-18T11:01:02.632Z",550 "reply_count": 0,551 "reply_to_post_number": null,552 "quote_count": 0,553 "incoming_link_count": 228,554 "reads": 14,555 "readers_count": 13,556 "score": 1142.8,557 "yours": false,558 "topic_id": 115217,559 "topic_slug": "fill-tensor-with-another-tensor-where-mask-is-true",560 "display_username": "Stefano Berti",561 "primary_group_name": null,562 "flair_name": null,563 "flair_url": null,564 "flair_bg_color": null,565 "flair_color": null,566 "flair_group_id": null,567 "badges_granted": [],568 "version": 1,569 "can_edit": false,570 "can_delete": false,571 "can_recover": false,572 "can_see_hidden_post": false,573 "can_wiki": false,574 "read": true,575 "user_title": null,576 "bookmarked": false,577 "actions_summary": [],578 "moderator": false,579 "admin": false,580 "staff": false,581 "user_id": 40025,582 "hidden": false,583 "trust_level": 1,584 "deleted_at": null,585 "user_deleted": false,586 "edit_reason": null,587 "can_view_edit_history": true,588 "wiki": false,589 "post_url": "/t/fill-tensor-with-another-tensor-where-mask-is-true/115217/1",590 "can_accept_answer": false,591 "can_unaccept_answer": false,592 "accepted_answer": false,593 "topic_accepted_answer": true,594 "can_vote": false595 },596 {597 "id": 271401,598 "name": "",599 "username": "Eta_C",600 "avatar_template": "/user_avatar/discuss.pytorch.org/eta_c/{size}/17667_2.png",601 "created_at": "2021-03-19T06:00:43.451Z",602 "cooked": "<pre><code class=\"lang-python\">def old_imp(old, new, mask):\n dim1, dim2, dim3, dim4 = new.shape\n for a in range(dim1):\n for b in range(dim2):\n for c in range(dim3):\n for d in range(dim4):\n old[a][b][c][d] = old[a][b][c][d] if mask[a][b][c][d] else new[a][b][c][d]\n return old \n\ndef new_imp(old, new, mask):\n new[mask] = old[mask]\n return new\n\nold = torch.rand(2, 3, 4, 5)\nnew = torch.rand(2, 3, 4, 5)\nprob = torch.rand(old.shape, dtype=torch.float32)\nmask = prob < 0.8\n\nold_res = old_imp(old.clone(), new.clone(), mask.clone())\nnew_res = new_imp(old.clone(), new.clone(), mask.clone())\nprint(old_res == new_res)\n</code></pre>",603 "post_number": 2,604 "post_type": 1,605 "posts_count": 2,606 "updated_at": "2021-04-15T11:13:41.708Z",607 "reply_count": 0,608 "reply_to_post_number": null,609 "quote_count": 0,610 "incoming_link_count": 4,611 "reads": 10,612 "readers_count": 9,613 "score": 22.0,614 "yours": false,615 "topic_id": 115217,616 "topic_slug": "fill-tensor-with-another-tensor-where-mask-is-true",617 "display_username": "",618 "primary_group_name": null,619 "flair_name": null,620 "flair_url": null,621 "flair_bg_color": null,622 "flair_color": null,623 "flair_group_id": null,624 "badges_granted": [],625 "version": 1,626 "can_edit": false,627 "can_delete": false,628 "can_recover": false,629 "can_see_hidden_post": false,630 "can_wiki": false,631 "read": true,632 "user_title": null,633 "bookmarked": false,634 "actions_summary": [],635 "moderator": false,636 "admin": false,637 "staff": false,638 "user_id": 23907,639 "hidden": false,640 "trust_level": 2,641 "deleted_at": null,642 "user_deleted": false,643 "edit_reason": null,644 "can_view_edit_history": true,645 "wiki": false,646 "post_url": "/t/fill-tensor-with-another-tensor-where-mask-is-true/115217/2",647 "can_accept_answer": false,648 "can_unaccept_answer": false,649 "accepted_answer": true,650 "topic_accepted_answer": true651 }652 ],653 "stream": [654 271247,655 271401656 ]657 },658 "timeline_lookup": [659 [660 1,661 1682662 ]663 ],664 "suggested_topics": [665 {666 "fancy_title": "Trying to backward through the graph a second time error",667 "id": 213322,668 "title": "Trying to backward through the graph a second time error",669 "slug": "trying-to-backward-through-the-graph-a-second-time-error",670 "posts_count": 1,671 "reply_count": 0,672 "highest_post_number": 1,673 "image_url": null,674 "created_at": "2024-11-22T19:50:38.013Z",675 "last_posted_at": "2024-11-22T19:50:38.070Z",676 "bumped": true,677 "bumped_at": "2024-11-22T19:50:38.070Z",678 "archetype": "regular",679 "unseen": false,680 "pinned": false,681 "unpinned": null,682 "visible": true,683 "closed": false,684 "archived": false,685 "bookmarked": null,686 "liked": null,687 "tags_descriptions": {},688 "like_count": 0,689 "views": 33,690 "category_id": 1,691 "featured_link": null,692 "has_accepted_answer": false,693 "posters": [694 {695 "extras": "latest single",696 "description": "Original Poster, Most Recent Poster",697 "user": {698 "id": 64607,699 "username": "Selvam_Dharma",700 "name": "Selvam Dharma",701 "avatar_template": "/user_avatar/discuss.pytorch.org/selvam_dharma/{size}/58785_2.png",702 "trust_level": 1703 }704 }705 ]706 },707 {708 "fancy_title": "Pytorch install is that large? 5 Gbs?",709 "id": 213629,710 "title": "Pytorch install is that large? 5 Gbs?",711 "slug": "pytorch-install-is-that-large-5-gbs",712 "posts_count": 4,713 "reply_count": 0,714 "highest_post_number": 6,715 "image_url": null,716 "created_at": "2024-11-30T04:12:34.904Z",717 "last_posted_at": "2025-02-14T22:26:13.056Z",718 "bumped": true,719 "bumped_at": "2025-02-14T22:26:13.056Z",720 "archetype": "regular",721 "unseen": false,722 "pinned": false,723 "unpinned": null,724 "visible": true,725 "closed": false,726 "archived": false,727 "bookmarked": null,728 "liked": null,729 "tags_descriptions": {},730 "like_count": 2,731 "views": 2824,732 "category_id": 1,733 "featured_link": null,734 "has_accepted_answer": false,735 "posters": [736 {737 "extras": null,738 "description": "Original Poster",739 "user": {740 "id": 44461,741 "username": "JonathanAlis",742 "name": "Jonathan Alis Lima",743 "avatar_template": "/user_avatar/discuss.pytorch.org/jonathanalis/{size}/37279_2.png",744 "trust_level": 1745 }746 },747 {748 "extras": null,749 "description": "Frequent Poster",750 "user": {751 "id": 3534,752 "username": "ptrblck",753 "name": "",754 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",755 "admin": true,756 "moderator": true,757 "trust_level": 2758 }759 },760 {761 "extras": null,762 "description": "Frequent Poster",763 "user": {764 "id": 76507,765 "username": "N_N_miko",766 "name": "-N_N- miko",767 "avatar_template": "/user_avatar/discuss.pytorch.org/n_n_miko/{size}/70619_2.png",768 "trust_level": 0769 }770 },771 {772 "extras": "latest",773 "description": "Most Recent Poster",774 "user": {775 "id": 82697,776 "username": "dariox",777 "name": "",778 "avatar_template": "/user_avatar/discuss.pytorch.org/dariox/{size}/75665_2.png",779 "trust_level": 0780 }781 }782 ]783 },784 {785 "fancy_title": "Dino training- input image",786 "id": 213737,787 "title": "Dino training- input image",788 "slug": "dino-training-input-image",789 "posts_count": 5,790 "reply_count": 3,791 "highest_post_number": 5,792 "image_url": null,793 "created_at": "2024-12-03T08:01:43.211Z",794 "last_posted_at": "2024-12-09T10:30:16.201Z",795 "bumped": true,796 "bumped_at": "2024-12-09T10:30:16.201Z",797 "archetype": "regular",798 "unseen": false,799 "pinned": false,800 "unpinned": null,801 "visible": true,802 "closed": false,803 "archived": false,804 "bookmarked": null,805 "liked": null,806 "tags_descriptions": {},807 "like_count": 1,808 "views": 354,809 "category_id": 1,810 "featured_link": null,811 "has_accepted_answer": false,812 "posters": [813 {814 "extras": "latest",815 "description": "Original Poster, Most Recent Poster",816 "user": {817 "id": 81269,818 "username": "Barbo",819 "name": "",820 "avatar_template": "/user_avatar/discuss.pytorch.org/barbo/{size}/74321_2.png",821 "trust_level": 1822 }823 },824 {825 "extras": null,826 "description": "Frequent Poster",827 "user": {828 "id": 72430,829 "username": "Eduardo_Lawson",830 "name": "Eduardo Lawson da Silva",831 "avatar_template": "/user_avatar/discuss.pytorch.org/eduardo_lawson/{size}/66899_2.png",832 "trust_level": 2833 }834 }835 ]836 },837 {838 "fancy_title": "Question about PyTorch and TensorFlow math",839 "id": 214022,840 "title": "Question about PyTorch and TensorFlow math",841 "slug": "question-about-pytorch-and-tensorflow-math",842 "posts_count": 2,843 "reply_count": 0,844 "highest_post_number": 2,845 "image_url": null,846 "created_at": "2024-12-10T04:20:11.694Z",847 "last_posted_at": "2024-12-10T13:20:30.605Z",848 "bumped": true,849 "bumped_at": "2024-12-10T13:20:30.605Z",850 "archetype": "regular",851 "unseen": false,852 "pinned": false,853 "unpinned": null,854 "visible": true,855 "closed": false,856 "archived": false,857 "bookmarked": null,858 "liked": null,859 "tags_descriptions": {},860 "like_count": 0,861 "views": 163,862 "category_id": 1,863 "featured_link": null,864 "has_accepted_answer": false,865 "posters": [866 {867 "extras": null,868 "description": "Original Poster",869 "user": {870 "id": 81418,871 "username": "Aosmgpem_Ipsnagpisne",872 "name": "Aosmgpem Ipsnagpisne",873 "avatar_template": "/user_avatar/discuss.pytorch.org/aosmgpem_ipsnagpisne/{size}/72552_2.png",874 "trust_level": 0875 }876 },877 {878 "extras": "latest",879 "description": "Most Recent Poster",880 "user": {881 "id": 3534,882 "username": "ptrblck",883 "name": "",884 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",885 "admin": true,886 "moderator": true,887 "trust_level": 2888 }889 }890 ]891 },892 {893 "fancy_title": "How to approach a real-life problem while using rainfall data",894 "id": 212728,895 "title": "How to approach a real-life problem while using rainfall data",896 "slug": "how-to-approach-a-real-life-problem-while-using-rainfall-data",897 "posts_count": 1,898 "reply_count": 0,899 "highest_post_number": 1,900 "image_url": null,901 "created_at": "2024-11-09T07:21:57.841Z",902 "last_posted_at": "2024-11-09T07:21:57.888Z",903 "bumped": true,904 "bumped_at": "2024-11-09T07:21:57.888Z",905 "archetype": "regular",906 "unseen": false,907 "pinned": false,908 "unpinned": null,909 "visible": true,910 "closed": false,911 "archived": false,912 "bookmarked": null,913 "liked": null,914 "tags_descriptions": {},915 "like_count": 0,916 "views": 25,917 "category_id": 1,918 "featured_link": null,919 "has_accepted_answer": false,920 "posters": [921 {922 "extras": "latest single",923 "description": "Original Poster, Most Recent Poster",924 "user": {925 "id": 80781,926 "username": "Ritam_Pradhan",927 "name": "Ritam Pradhan",928 "avatar_template": "/user_avatar/discuss.pytorch.org/ritam_pradhan/{size}/73876_2.png",929 "trust_level": 1930 }931 }932 ]933 }934 ],935 "tags_descriptions": {},936 "fancy_title": "Fill tensor with another tensor where mask is true",937 "id": 115217,938 "title": "Fill tensor with another tensor where mask is true",939 "posts_count": 2,940 "created_at": "2021-03-18T11:01:02.552Z",941 "views": 834,942 "reply_count": 0,943 "like_count": 0,944 "last_posted_at": "2021-03-19T06:00:43.451Z",945 "visible": true,946 "closed": false,947 "archived": false,948 "has_summary": false,949 "archetype": "regular",950 "slug": "fill-tensor-with-another-tensor-where-mask-is-true",951 "category_id": 1,952 "word_count": 247,953 "deleted_at": null,954 "user_id": 40025,955 "featured_link": null,956 "pinned_globally": false,957 "pinned_at": null,958 "pinned_until": null,959 "image_url": null,960 "slow_mode_seconds": 0,961 "draft": null,962 "draft_key": "topic_115217",963 "draft_sequence": null,964 "unpinned": null,965 "pinned": false,966 "current_post_number": 1,967 "highest_post_number": 2,968 "deleted_by": null,969 "actions_summary": [970 {971 "id": 4,972 "count": 0,973 "hidden": false,974 "can_act": false975 },976 {977 "id": 8,978 "count": 0,979 "hidden": false,980 "can_act": false981 },982 {983 "id": 10,984 "count": 0,985 "hidden": false,986 "can_act": false987 },988 {989 "id": 7,990 "count": 0,991 "hidden": false,992 "can_act": false993 }994 ],995 "chunk_size": 20,996 "bookmarked": false,997 "topic_timer": null,998 "message_bus_last_id": 0,999 "participant_count": 2,1000 "show_read_indicator": false,1001 "thumbnails": null,1002 "slow_mode_enabled_until": null,1003 "accepted_answer": {1004 "post_number": 2,1005 "username": "Eta_C",1006 "name": "",1007 "excerpt": "def old_imp(old, new, mask):\n dim1, dim2, dim3, dim4 = new.shape\n for a in range(dim1):\n for b in range(dim2):\n for c in range(dim3):\n for d in range(dim4):\n old[a][b][c][d] = old[a][b][c][d] if mask[a][b][c][d] else new[a][b][c][d]\n retur…"1008 },1009 "can_vote": false,1010 "vote_count": 0,1011 "user_voted": false,1012 "discourse_zendesk_plugin_zendesk_id": null,1013 "discourse_zendesk_plugin_zendesk_url": "https://your-url.zendesk.com/agent/tickets/",1014 "details": {1015 "can_edit": false,1016 "notification_level": 1,1017 "participants": [1018 {1019 "id": 23907,1020 "username": "Eta_C",1021 "name": "",1022 "avatar_template": "/user_avatar/discuss.pytorch.org/eta_c/{size}/17667_2.png",1023 "post_count": 1,1024 "primary_group_name": null,1025 "flair_name": null,1026 "flair_url": null,1027 "flair_color": null,1028 "flair_bg_color": null,1029 "flair_group_id": null,1030 "trust_level": 21031 },1032 {1033 "id": 40025,1034 "username": "StefanoBerti",1035 "name": "Stefano Berti",1036 "avatar_template": "/user_avatar/discuss.pytorch.org/stefanoberti/{size}/32286_2.png",1037 "post_count": 1,1038 "primary_group_name": null,1039 "flair_name": null,1040 "flair_url": null,1041 "flair_color": null,1042 "flair_bg_color": null,1043 "flair_group_id": null,1044 "trust_level": 11045 }1046 ],1047 "created_by": {1048 "id": 40025,1049 "username": "StefanoBerti",1050 "name": "Stefano Berti",1051 "avatar_template": "/user_avatar/discuss.pytorch.org/stefanoberti/{size}/32286_2.png"1052 },1053 "last_poster": {1054 "id": 23907,1055 "username": "Eta_C",1056 "name": "",1057 "avatar_template": "/user_avatar/discuss.pytorch.org/eta_c/{size}/17667_2.png"1058 }1059 },1060 "bookmarks": []1061 },1062 {1063 "post_stream": {1064 "posts": [1065 {1066 "id": 271157,1067 "name": "Siddharth Tandon",1068 "username": "Siddharth_Tandon",1069 "avatar_template": "/user_avatar/discuss.pytorch.org/siddharth_tandon/{size}/32526_2.png",1070 "created_at": "2021-03-18T04:07:22.816Z",1071 "cooked": "<p>Hi,</p>\n<p>I am working with MSD (Medical Segmentation Decathlon) dataset for a segmentation problem.<br>\nIn Dataset’s <strong>init</strong> method I am storing all filenames in list, and in <strong>getitem</strong> reading the files, applying transformation, etc…</p>\n<p>I am facing some issue with GPU utilization:<br>\nProblem is, nvidia-sme command is showing only 1 process (whereas I have specified 4 workers in dataloader) running, with GPU memory usage as 4GB. Secondly there is no GPU utilization at all.</p>\n<p>I was reading an article which says that if DataLoader is slow, model will run on CPU. But how can I speed up the dataloader and run everything on GPU</p>",1072 "post_number": 1,1073 "post_type": 1,1074 "posts_count": 2,1075 "updated_at": "2021-03-18T04:07:22.816Z",1076 "reply_count": 1,1077 "reply_to_post_number": null,1078 "quote_count": 0,1079 "incoming_link_count": 207,1080 "reads": 5,1081 "readers_count": 4,1082 "score": 1041.0,1083 "yours": false,1084 "topic_id": 115184,1085 "topic_slug": "handling-large-3d-image-dataset-with-dataloader",1086 "display_username": "Siddharth Tandon",1087 "primary_group_name": null,1088 "flair_name": null,1089 "flair_url": null,1090 "flair_bg_color": null,1091 "flair_color": null,1092 "flair_group_id": null,1093 "badges_granted": [],1094 "version": 1,1095 "can_edit": false,1096 "can_delete": false,1097 "can_recover": false,1098 "can_see_hidden_post": false,1099 "can_wiki": false,1100 "read": true,1101 "user_title": null,1102 "bookmarked": false,1103 "actions_summary": [],1104 "moderator": false,1105 "admin": false,1106 "staff": false,1107 "user_id": 43229,1108 "hidden": false,1109 "trust_level": 1,1110 "deleted_at": null,1111 "user_deleted": false,1112 "edit_reason": null,1113 "can_view_edit_history": true,1114 "wiki": false,1115 "post_url": "/t/handling-large-3d-image-dataset-with-dataloader/115184/1",1116 "can_accept_answer": false,1117 "can_unaccept_answer": false,1118 "accepted_answer": false,1119 "topic_accepted_answer": true,1120 "can_vote": false1121 },1122 {1123 "id": 271393,1124 "name": "",1125 "username": "ptrblck",1126 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",1127 "created_at": "2021-03-19T05:17:05.829Z",1128 "cooked": "<aside class=\"quote no-group\" data-username=\"Siddharth_Tandon\" data-post=\"1\" data-topic=\"115184\">\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/siddharth_tandon/48/32526_2.png\" class=\"avatar\"> Siddharth_Tandon:</div>\n<blockquote>\n<p>I am facing some issue with GPU utilization:<br>\nProblem is, nvidia-sme command is showing only 1 process (whereas I have specified 4 workers in dataloader) running, with GPU memory usage as 4GB. Secondly there is no GPU utilization at all.</p>\n</blockquote>\n</aside>\n<p>The <code>num_workers</code> specified in the <code>DataLoader</code> are executed on the CPU using multiprocessing and will not be shown on <code>nvidia-smi</code>.<br>\nThe GPU utilization might be low, if you are facing bottlenecks in your code, such as data loading.<br>\nTake a look at <a href=\"https://discuss.pytorch.org/t/how-to-prefetch-data-when-processing-with-gpu/548/19\">this post</a>, which explains it well and suggests some performance improvements.</p>\n<aside class=\"quote no-group\" data-username=\"Siddharth_Tandon\" data-post=\"1\" data-topic=\"115184\">\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/siddharth_tandon/48/32526_2.png\" class=\"avatar\"> Siddharth_Tandon:</div>\n<blockquote>\n<p>I was reading an article which says that if DataLoader is slow, model will run on CPU.</p>\n</blockquote>\n</aside>\n<p>This is not correct. The model will use the specified device and PyTorch will not automatically use the CPU based on the performance of the <code>DataLoader</code>.</p>",1129 "post_number": 2,1130 "post_type": 1,1131 "posts_count": 2,1132 "updated_at": "2021-03-24T06:07:14.072Z",1133 "reply_count": 0,1134 "reply_to_post_number": null,1135 "quote_count": 1,1136 "incoming_link_count": 7,1137 "reads": 5,1138 "readers_count": 4,1139 "score": 51.0,1140 "yours": false,1141 "topic_id": 115184,1142 "topic_slug": "handling-large-3d-image-dataset-with-dataloader",1143 "display_username": "",1144 "primary_group_name": null,1145 "flair_name": null,1146 "flair_url": null,1147 "flair_bg_color": null,1148 "flair_color": null,1149 "flair_group_id": null,1150 "badges_granted": [],1151 "version": 1,1152 "can_edit": false,1153 "can_delete": false,1154 "can_recover": false,1155 "can_see_hidden_post": false,1156 "can_wiki": false,1157 "link_counts": [1158 {1159 "url": "https://discuss.pytorch.org/t/how-to-prefetch-data-when-processing-with-gpu/548/19",1160 "internal": true,1161 "reflection": false,1162 "title": "How to prefetch data when processing with GPU?",1163 "clicks": 261164 }1165 ],1166 "read": true,1167 "user_title": "",1168 "bookmarked": false,1169 "actions_summary": [1170 {1171 "id": 2,1172 "count": 11173 }1174 ],1175 "moderator": true,1176 "admin": true,1177 "staff": true,1178 "user_id": 3534,1179 "hidden": false,1180 "trust_level": 2,1181 "deleted_at": null,1182 "user_deleted": false,1183 "edit_reason": null,1184 "can_view_edit_history": true,1185 "wiki": false,1186 "post_url": "/t/handling-large-3d-image-dataset-with-dataloader/115184/2",1187 "can_accept_answer": false,1188 "can_unaccept_answer": false,1189 "accepted_answer": true,1190 "topic_accepted_answer": true1191 }1192 ],1193 "stream": [1194 271157,1195 2713931196 ]1197 },1198 "timeline_lookup": [1199 [1200 1,