Anurag1734/cuda-error-resolution-analysis
07
1[2 {3 "post_stream": {4 "posts": [5 {6 "id": 125600,7 "name": "Antonio Mendoza",8 "username": "eltoto1219",9 "avatar_template": "/letter_avatar_proxy/v4/letter/e/ecb155/{size}.png",10 "created_at": "2019-07-26T13:19:07.776Z",11 "cooked": "<p>Hello,</p>\n<p>I am making a VQA model with co-attention with Y adaptive image features (10-100). To calculate the co attention, I first project my question (batch * q_len * q_Dim) to (batch * q_len * new_dim)</p>\n<p>Then I have the following for loop which i project each one of my image features (Y * feature_dim) to (Y * new_dim)</p>\n<pre><code class=\"lang-auto\"> def attn_weights(self, q2, v2, n_objs):\n batch_size = n_objs.size(0)\n weights = torch.zeros(batch_size, self.max_objs, 1).to(self.device)\n\n\n q_proj = self.q_proj(q2)\n for i in range(batch_size): \n n_i = int(n_objs[i].item()) ### number of objects for the ith image in batchk\n v_i = v2[i] ## the ith image in batch\n v_i = v_i[:n_i-1, :] ## selecting number of object in image\n v_i = self.v_proj(v_i) ## projecting feature dim to new_dim\n q_i = q_proj[i] ## the ith question in batch\n fusion = v_i * q_i.repeat(n_i-1 ,1) ## repeat the question Y times\n fusion = self.dropout(fusion)\n scores = self.linear(fusion)\n att_weights = softmax(scores, 0)\n weights[i, :n_i -1] = att_weights \n return weights\n\n</code></pre>\n<p>During training this causes CUDA’s memory usage to sky rocket. I have checked the nvidia-smi and this function alone causes 14113MiB / 15079MiB of memory to be used.</p>\n<p>This is the error I have received:</p>\n<pre><code class=\"lang-auto\"> File \"main.py\", line 181, in <module>\n main()\n File \"main.py\", line 166, in main\n run(mod, train_loader, optimizer, train=[], prefix='train', epoch=i)\n File \"main.py\", line 79, in run\n loss.backward()\n File \"/opt/anaconda3/lib/python3.7/site-packages/torch/tensor.py\", line 107, in backward\n torch.autograd.backward(self, gradient, retain_graph, create_graph)\n File \"/opt/anaconda3/lib/python3.7/site-packages/torch/autograd/__init__.py\", line 93, in backward\n allow_unreachable=True) # allow_unreachable flag\nRuntimeError: CUDA out of memory. Tried to allocate 1.43 GiB (GPU 0; 14.73 GiB total capacity; 8.45 GiB already allocated; 1.04 GiB free; 4.54 GiB cached)\n\n</code></pre>\n<p>Is there a reason why this is happening, and is there a known way around this? If nn.Linear layers are not supposed to be called in a for loop, my next question would be how to project the Y image features for every image in the batch (Y * feature_dim) to (Y * new_dim) where the batch dimension looks like (batch * 100 * feature_dim) to ( batch * 100 * new_dim) where everything after the Y image features (100 - Y) would be zero padded without the zero padding affecting the gradient of the projection.</p>\n<p>Any help would be greatly appreciated!</p>",12 "post_number": 1,13 "post_type": 1,14 "posts_count": 2,15 "updated_at": "2019-07-26T13:23:20.899Z",16 "reply_count": 0,17 "reply_to_post_number": null,18 "quote_count": 0,19 "incoming_link_count": 432,20 "reads": 23,21 "readers_count": 22,22 "score": 2164.6,23 "yours": false,24 "topic_id": 51723,25 "topic_slug": "using-nn-linear-inside-a-for-loop-causes-cuda-to-run-out-of-memory",26 "display_username": "Antonio Mendoza",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 "read": true,41 "user_title": null,42 "bookmarked": false,43 "actions_summary": [],44 "moderator": false,45 "admin": false,46 "staff": false,47 "user_id": 21007,48 "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/using-nn-linear-inside-a-for-loop-causes-cuda-to-run-out-of-memory/51723/1",56 "can_accept_answer": false,57 "can_unaccept_answer": false,58 "accepted_answer": false,59 "topic_accepted_answer": true,60 "can_vote": false61 },62 {63 "id": 125660,64 "name": "Juan Montesinos",65 "username": "JuanFMontesinos",66 "avatar_template": "/user_avatar/discuss.pytorch.org/juanfmontesinos/{size}/76115_2.png",67 "created_at": "2019-07-26T23:11:29.195Z",68 "cooked": "<p>Hi, nn.linear work with an arbitrary amount of dimensions, namely you can pass whatever tensor of size BATCH,<em>,dim yo obtain BATCH,</em>,new_dim.</p>\n<p>Never do for loops in pytorch as it is equivalent to generate Siamese modules. It duplicates the computational graph as many times as you call the module.</p>\n<p>If you would like to do something similar (linear is spatial as you can pass arbitrary dimensions) the proper way is squeezing everything into the BATCH dimension</p>",69 "post_number": 2,70 "post_type": 1,71 "posts_count": 2,72 "updated_at": "2019-07-27T23:58:52.529Z",73 "reply_count": 0,74 "reply_to_post_number": null,75 "quote_count": 0,76 "incoming_link_count": 1,77 "reads": 16,78 "readers_count": 15,79 "score": 23.2,80 "yours": false,81 "topic_id": 51723,82 "topic_slug": "using-nn-linear-inside-a-for-loop-causes-cuda-to-run-out-of-memory",83 "display_username": "Juan Montesinos",84 "primary_group_name": null,85 "flair_name": null,86 "flair_url": null,87 "flair_bg_color": null,88 "flair_color": null,89 "flair_group_id": null,90 "badges_granted": [],91 "version": 1,92 "can_edit": false,93 "can_delete": false,94 "can_recover": false,95 "can_see_hidden_post": false,96 "can_wiki": false,97 "read": true,98 "user_title": "",99 "bookmarked": false,100 "actions_summary": [101 {102 "id": 2,103 "count": 1104 }105 ],106 "moderator": false,107 "admin": false,108 "staff": false,109 "user_id": 9081,110 "hidden": false,111 "trust_level": 2,112 "deleted_at": null,113 "user_deleted": false,114 "edit_reason": null,115 "can_view_edit_history": true,116 "wiki": false,117 "post_url": "/t/using-nn-linear-inside-a-for-loop-causes-cuda-to-run-out-of-memory/51723/2",118 "can_accept_answer": false,119 "can_unaccept_answer": false,120 "accepted_answer": true,121 "topic_accepted_answer": true122 }123 ],124 "stream": [125 125600,126 125660127 ]128 },129 "timeline_lookup": [130 [131 1,132 2283133 ]134 ],135 "suggested_topics": [136 {137 "fancy_title": "Initializating parameter with registered parametrization",138 "id": 212563,139 "title": "Initializating parameter with registered parametrization",140 "slug": "initializating-parameter-with-registered-parametrization",141 "posts_count": 3,142 "reply_count": 1,143 "highest_post_number": 3,144 "image_url": null,145 "created_at": "2024-11-05T17:08:17.858Z",146 "last_posted_at": "2024-11-05T22:33:46.499Z",147 "bumped": true,148 "bumped_at": "2024-11-05T22:33:46.499Z",149 "archetype": "regular",150 "unseen": false,151 "pinned": false,152 "unpinned": null,153 "visible": true,154 "closed": false,155 "archived": false,156 "bookmarked": null,157 "liked": null,158 "tags_descriptions": {},159 "like_count": 0,160 "views": 176,161 "category_id": 1,162 "featured_link": null,163 "has_accepted_answer": false,164 "posters": [165 {166 "extras": "latest",167 "description": "Original Poster, Most Recent Poster",168 "user": {169 "id": 67433,170 "username": "octave",171 "name": "Octave",172 "avatar_template": "/letter_avatar_proxy/v4/letter/o/f05b48/{size}.png",173 "trust_level": 1174 }175 },176 {177 "extras": null,178 "description": "Frequent Poster",179 "user": {180 "id": 3534,181 "username": "ptrblck",182 "name": "",183 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",184 "admin": true,185 "moderator": true,186 "trust_level": 2187 }188 }189 ]190 },191 {192 "fancy_title": "Getting std::bad_alloc when loading the python tradined model in c++",193 "id": 212797,194 "title": "Getting std::bad_alloc when loading the python tradined model in c++",195 "slug": "getting-std-bad-alloc-when-loading-the-python-tradined-model-in-c",196 "posts_count": 1,197 "reply_count": 0,198 "highest_post_number": 1,199 "image_url": "https://discuss.pytorch.org/uploads/default/optimized/3X/4/c/4c64351ba5b8b679193c4d58afeb8bedb34260ab_2_1024x361.jpeg",200 "created_at": "2024-11-11T09:35:50.204Z",201 "last_posted_at": "2024-11-11T09:35:50.257Z",202 "bumped": true,203 "bumped_at": "2024-11-11T10:11:12.646Z",204 "archetype": "regular",205 "unseen": false,206 "pinned": false,207 "unpinned": null,208 "visible": true,209 "closed": false,210 "archived": false,211 "bookmarked": null,212 "liked": null,213 "tags_descriptions": {},214 "like_count": 0,215 "views": 80,216 "category_id": 1,217 "featured_link": null,218 "has_accepted_answer": false,219 "posters": [220 {221 "extras": "latest single",222 "description": "Original Poster, Most Recent Poster",223 "user": {224 "id": 80818,225 "username": "Naseef",226 "name": "Naseef",227 "avatar_template": "/user_avatar/discuss.pytorch.org/naseef/{size}/73911_2.png",228 "trust_level": 0229 }230 }231 ]232 },233 {234 "fancy_title": "F.cross_entropy unexpectedly slower than F.log_softmax + torch.gather",235 "id": 213425,236 "title": "F.cross_entropy unexpectedly slower than F.log_softmax + torch.gather",237 "slug": "f-cross-entropy-unexpectedly-slower-than-f-log-softmax-torch-gather",238 "posts_count": 2,239 "reply_count": 0,240 "highest_post_number": 2,241 "image_url": null,242 "created_at": "2024-11-25T21:27:40.120Z",243 "last_posted_at": "2024-11-25T22:36:33.756Z",244 "bumped": true,245 "bumped_at": "2024-11-25T22:36:33.756Z",246 "archetype": "regular",247 "unseen": false,248 "pinned": false,249 "unpinned": null,250 "visible": true,251 "closed": false,252 "archived": false,253 "bookmarked": null,254 "liked": null,255 "tags_descriptions": {},256 "like_count": 1,257 "views": 44,258 "category_id": 1,259 "featured_link": null,260 "has_accepted_answer": false,261 "posters": [262 {263 "extras": null,264 "description": "Original Poster",265 "user": {266 "id": 40732,267 "username": "normster",268 "name": "Norman Mu",269 "avatar_template": "/user_avatar/discuss.pytorch.org/normster/{size}/33059_2.png",270 "trust_level": 1271 }272 },273 {274 "extras": "latest",275 "description": "Most Recent Poster",276 "user": {277 "id": 3534,278 "username": "ptrblck",279 "name": "",280 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",281 "admin": true,282 "moderator": true,283 "trust_level": 2284 }285 }286 ]287 },288 {289 "fancy_title": "Is torch2.4.1 compatible with CUDA12.5",290 "id": 213511,291 "title": "Is torch2.4.1 compatible with CUDA12.5",292 "slug": "is-torch2-4-1-compatible-with-cuda12-5",293 "posts_count": 4,294 "reply_count": 0,295 "highest_post_number": 5,296 "image_url": null,297 "created_at": "2024-11-27T11:34:39.926Z",298 "last_posted_at": "2024-11-27T13:50:34.682Z",299 "bumped": true,300 "bumped_at": "2024-11-27T13:50:34.682Z",301 "archetype": "regular",302 "unseen": false,303 "pinned": false,304 "unpinned": null,305 "visible": true,306 "closed": false,307 "archived": false,308 "bookmarked": null,309 "liked": null,310 "visibility_reason_id": 1,311 "tags_descriptions": {},312 "like_count": 0,313 "views": 290,314 "category_id": 1,315 "featured_link": null,316 "has_accepted_answer": false,317 "posters": [318 {319 "extras": null,320 "description": "Original Poster",321 "user": {322 "id": 81160,323 "username": "Pied-Piper1",324 "name": "Paradox",325 "avatar_template": "/user_avatar/discuss.pytorch.org/pied-piper1/{size}/74224_2.png",326 "trust_level": 0327 }328 },329 {330 "extras": null,331 "description": "Frequent Poster",332 "user": {333 "id": -1,334 "username": "system",335 "name": "system",336 "avatar_template": "/uploads/default/original/2X/3/38d28fd067a1a8f263e14507942b2e38e49b771a.png",337 "admin": true,338 "moderator": true,339 "trust_level": 4340 }341 },342 {343 "extras": "latest",344 "description": "Most Recent Poster",345 "user": {346 "id": 3534,347 "username": "ptrblck",348 "name": "",349 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",350 "admin": true,351 "moderator": true,352 "trust_level": 2353 }354 }355 ]356 },357 {358 "fancy_title": "FlexAttention with sparse edge bias",359 "id": 216376,360 "title": "FlexAttention with sparse edge bias",361 "slug": "flexattention-with-sparse-edge-bias",362 "posts_count": 1,363 "reply_count": 0,364 "highest_post_number": 1,365 "image_url": null,366 "created_at": "2025-02-07T18:02:41.974Z",367 "last_posted_at": "2025-02-07T18:02:42.019Z",368 "bumped": true,369 "bumped_at": "2025-02-07T18:02:42.019Z",370 "archetype": "regular",371 "unseen": false,372 "pinned": false,373 "unpinned": null,374 "visible": true,375 "closed": false,376 "archived": false,377 "bookmarked": null,378 "liked": null,379 "tags_descriptions": {},380 "like_count": 0,381 "views": 92,382 "category_id": 1,383 "featured_link": null,384 "has_accepted_answer": false,385 "posters": [386 {387 "extras": "latest single",388 "description": "Original Poster, Most Recent Poster",389 "user": {390 "id": 82551,391 "username": "mbaranov",392 "name": "Max Baranov",393 "avatar_template": "/letter_avatar_proxy/v4/letter/m/76d3ee/{size}.png",394 "trust_level": 0395 }396 }397 ]398 }399 ],400 "tags_descriptions": {},401 "fancy_title": "Using nn.Linear inside a for loop causes CUDA to run out of memory",402 "id": 51723,403 "title": "Using nn.Linear inside a for loop causes CUDA to run out of memory",404 "posts_count": 2,405 "created_at": "2019-07-26T13:19:07.728Z",406 "views": 1140,407 "reply_count": 0,408 "like_count": 1,409 "last_posted_at": "2019-07-26T23:11:29.195Z",410 "visible": true,411 "closed": false,412 "archived": false,413 "has_summary": false,414 "archetype": "regular",415 "slug": "using-nn-linear-inside-a-for-loop-causes-cuda-to-run-out-of-memory",416 "category_id": 1,417 "word_count": 479,418 "deleted_at": null,419 "user_id": 21007,420 "featured_link": null,421 "pinned_globally": false,422 "pinned_at": null,423 "pinned_until": null,424 "image_url": null,425 "slow_mode_seconds": 0,426 "draft": null,427 "draft_key": "topic_51723",428 "draft_sequence": null,429 "unpinned": null,430 "pinned": false,431 "current_post_number": 1,432 "highest_post_number": 2,433 "deleted_by": null,434 "actions_summary": [435 {436 "id": 4,437 "count": 0,438 "hidden": false,439 "can_act": false440 },441 {442 "id": 8,443 "count": 0,444 "hidden": false,445 "can_act": false446 },447 {448 "id": 10,449 "count": 0,450 "hidden": false,451 "can_act": false452 },453 {454 "id": 7,455 "count": 0,456 "hidden": false,457 "can_act": false458 }459 ],460 "chunk_size": 20,461 "bookmarked": false,462 "topic_timer": null,463 "message_bus_last_id": 0,464 "participant_count": 2,465 "show_read_indicator": false,466 "thumbnails": null,467 "slow_mode_enabled_until": null,468 "accepted_answer": {469 "post_number": 2,470 "username": "JuanFMontesinos",471 "name": "Juan Montesinos",472 "excerpt": "Hi, nn.linear work with an arbitrary amount of dimensions, namely you can pass whatever tensor of size BATCH,,dim yo obtain BATCH,,new_dim. \nNever do for loops in pytorch as it is equivalent to generate Siamese modules. It duplicates the computational graph as many times as you call the module. \nIf…"473 },474 "can_vote": false,475 "vote_count": 0,476 "user_voted": false,477 "discourse_zendesk_plugin_zendesk_id": null,478 "discourse_zendesk_plugin_zendesk_url": "https://your-url.zendesk.com/agent/tickets/",479 "details": {480 "can_edit": false,481 "notification_level": 1,482 "participants": [483 {484 "id": 9081,485 "username": "JuanFMontesinos",486 "name": "Juan Montesinos",487 "avatar_template": "/user_avatar/discuss.pytorch.org/juanfmontesinos/{size}/76115_2.png",488 "post_count": 1,489 "primary_group_name": null,490 "flair_name": null,491 "flair_url": null,492 "flair_color": null,493 "flair_bg_color": null,494 "flair_group_id": null,495 "trust_level": 2496 },497 {498 "id": 21007,499 "username": "eltoto1219",500 "name": "Antonio Mendoza",501 "avatar_template": "/letter_avatar_proxy/v4/letter/e/ecb155/{size}.png",502 "post_count": 1,503 "primary_group_name": null,504 "flair_name": null,505 "flair_url": null,506 "flair_color": null,507 "flair_bg_color": null,508 "flair_group_id": null,509 "trust_level": 1510 }511 ],512 "created_by": {513 "id": 21007,514 "username": "eltoto1219",515 "name": "Antonio Mendoza",516 "avatar_template": "/letter_avatar_proxy/v4/letter/e/ecb155/{size}.png"517 },518 "last_poster": {519 "id": 9081,520 "username": "JuanFMontesinos",521 "name": "Juan Montesinos",522 "avatar_template": "/user_avatar/discuss.pytorch.org/juanfmontesinos/{size}/76115_2.png"523 }524 },525 "bookmarks": []526 },527 {528 "post_stream": {529 "posts": [530 {531 "id": 125043,532 "name": "Zhaoyi Yan",533 "username": "Zhaoyi-Yan",534 "avatar_template": "/user_avatar/discuss.pytorch.org/zhaoyi-yan/{size}/8480_2.png",535 "created_at": "2019-07-24T00:55:12.435Z",536 "cooked": "<pre><code class=\"lang-python\">import torch\na = (torch.rand(3,4)*10).long()\nprint(a)\nprint(a*0.9)\n</code></pre>\n<p>Output:</p>\n<pre><code class=\"lang-bash\">tensor([[8, 5, 3, 4],\n [1, 3, 6, 7],\n [8, 5, 8, 8]])\ntensor([[0, 0, 0, 0],\n [0, 0, 0, 0],\n [0, 0, 0, 0]])\n</code></pre>",537 "post_number": 1,538 "post_type": 1,539 "posts_count": 2,540 "updated_at": "2019-07-24T00:55:12.435Z",541 "reply_count": 0,542 "reply_to_post_number": null,543 "quote_count": 0,544 "incoming_link_count": 12,545 "reads": 8,546 "readers_count": 7,547 "score": 61.6,548 "yours": false,549 "topic_id": 51440,550 "topic_slug": "bug-or-feature-for-longtensor",551 "display_username": "Zhaoyi Yan",552 "primary_group_name": null,553 "flair_name": null,554 "flair_url": null,555 "flair_bg_color": null,556 "flair_color": null,557 "flair_group_id": null,558 "badges_granted": [],559 "version": 1,560 "can_edit": false,561 "can_delete": false,562 "can_recover": false,563 "can_see_hidden_post": false,564 "can_wiki": false,565 "read": true,566 "user_title": null,567 "bookmarked": false,568 "actions_summary": [],569 "moderator": false,570 "admin": false,571 "staff": false,572 "user_id": 13822,573 "hidden": false,574 "trust_level": 2,575 "deleted_at": null,576 "user_deleted": false,577 "edit_reason": null,578 "can_view_edit_history": true,579 "wiki": false,580 "post_url": "/t/bug-or-feature-for-longtensor/51440/1",581 "can_accept_answer": false,582 "can_unaccept_answer": false,583 "accepted_answer": false,584 "topic_accepted_answer": null,585 "can_vote": false586 },587 {588 "id": 125654,589 "name": "Prerna Dhareshwar",590 "username": "Prerna_Dhareshwar",591 "avatar_template": "/user_avatar/discuss.pytorch.org/prerna_dhareshwar/{size}/14256_2.png",592 "created_at": "2019-07-26T22:44:12.689Z",593 "cooked": "<p>Hi,</p>\n<p>I don’t think thats a bug, it tries to retain the type of the tensor as long hence the rounding of 0.9 to 0.</p>",594 "post_number": 2,595 "post_type": 1,596 "posts_count": 2,597 "updated_at": "2019-07-26T22:44:12.689Z",598 "reply_count": 0,599 "reply_to_post_number": null,600 "quote_count": 0,601 "incoming_link_count": 1,602 "reads": 7,603 "readers_count": 6,604 "score": 6.4,605 "yours": false,606 "topic_id": 51440,607 "topic_slug": "bug-or-feature-for-longtensor",608 "display_username": "Prerna Dhareshwar",609 "primary_group_name": null,610 "flair_name": null,611 "flair_url": null,612 "flair_bg_color": null,613 "flair_color": null,614 "flair_group_id": null,615 "badges_granted": [],616 "version": 1,617 "can_edit": false,618 "can_delete": false,619 "can_recover": false,620 "can_see_hidden_post": false,621 "can_wiki": false,622 "read": true,623 "user_title": null,624 "bookmarked": false,625 "actions_summary": [],626 "moderator": false,627 "admin": false,628 "staff": false,629 "user_id": 20203,630 "hidden": false,631 "trust_level": 2,632 "deleted_at": null,633 "user_deleted": false,634 "edit_reason": null,635 "can_view_edit_history": true,636 "wiki": false,637 "post_url": "/t/bug-or-feature-for-longtensor/51440/2",638 "can_accept_answer": false,639 "can_unaccept_answer": false,640 "accepted_answer": false,641 "topic_accepted_answer": null642 }643 ],644 "stream": [645 125043,646 125654647 ]648 },649 "timeline_lookup": [650 [651 1,652 2286653 ],654 [655 2,656 2283657 ]658 ],659 "suggested_topics": [660 {661 "fancy_title": "Rtx 5090 Error creating song: CUDA error: no kernel image is available for execution on the device Compile with TORCH_USE_CUDA_DSA to enable device-side assertions",662 "id": 217438,663 "title": "Rtx 5090 Error creating song: CUDA error: no kernel image is available for execution on the device Compile with TORCH_USE_CUDA_DSA to enable device-side assertions",664 "slug": "rtx-5090-error-creating-song-cuda-error-no-kernel-image-is-available-for-execution-on-the-device-compile-with-torch-use-cuda-dsa-to-enable-device-side-assertions",665 "posts_count": 13,666 "reply_count": 6,667 "highest_post_number": 13,668 "image_url": null,669 "created_at": "2025-03-04T19:28:56.622Z",670 "last_posted_at": "2025-03-07T01:41:04.241Z",671 "bumped": true,672 "bumped_at": "2025-03-07T01:41:04.241Z",673 "archetype": "regular",674 "unseen": false,675 "pinned": false,676 "unpinned": null,677 "visible": true,678 "closed": false,679 "archived": false,680 "bookmarked": null,681 "liked": null,682 "tags_descriptions": {},683 "like_count": 0,684 "views": 547,685 "category_id": 1,686 "featured_link": null,687 "has_accepted_answer": false,688 "posters": [689 {690 "extras": null,691 "description": "Original Poster",692 "user": {693 "id": 83067,694 "username": "nafe_alhmdan",695 "name": "nafe alhmdan",696 "avatar_template": "/user_avatar/discuss.pytorch.org/nafe_alhmdan/{size}/75984_2.png",697 "trust_level": 1698 }699 },700 {701 "extras": null,702 "description": "Frequent Poster",703 "user": {704 "id": 3534,705 "username": "ptrblck",706 "name": "",707 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",708 "admin": true,709 "moderator": true,710 "trust_level": 2711 }712 },713 {714 "extras": null,715 "description": "Frequent Poster",716 "user": {717 "id": 9081,718 "username": "JuanFMontesinos",719 "name": "Juan Montesinos",720 "avatar_template": "/user_avatar/discuss.pytorch.org/juanfmontesinos/{size}/76115_2.png",721 "trust_level": 2722 }723 },724 {725 "extras": "latest",726 "description": "Most Recent Poster",727 "user": {728 "id": 83074,729 "username": "Noah_Jang",730 "name": "",731 "avatar_template": "/letter_avatar_proxy/v4/letter/n/858c86/{size}.png",732 "trust_level": 1733 }734 }735 ]736 },737 {738 "fancy_title": "Dropout design choice",739 "id": 214984,740 "title": "Dropout design choice",741 "slug": "dropout-design-choice",742 "posts_count": 2,743 "reply_count": 0,744 "highest_post_number": 2,745 "image_url": null,746 "created_at": "2025-01-05T00:07:45.968Z",747 "last_posted_at": "2025-01-05T19:33:48.026Z",748 "bumped": true,749 "bumped_at": "2025-01-05T19:33:48.026Z",750 "archetype": "regular",751 "unseen": false,752 "pinned": false,753 "unpinned": null,754 "visible": true,755 "closed": false,756 "archived": false,757 "bookmarked": null,758 "liked": null,759 "tags_descriptions": {},760 "like_count": 0,761 "views": 135,762 "category_id": 1,763 "featured_link": null,764 "has_accepted_answer": false,765 "posters": [766 {767 "extras": null,768 "description": "Original Poster",769 "user": {770 "id": 77182,771 "username": "Siddhanth_Ramani",772 "name": "Siddhanth Ramani",773 "avatar_template": "/user_avatar/discuss.pytorch.org/siddhanth_ramani/{size}/62072_2.png",774 "trust_level": 1775 }776 },777 {778 "extras": "latest",779 "description": "Most Recent Poster",780 "user": {781 "id": 3534,782 "username": "ptrblck",783 "name": "",784 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",785 "admin": true,786 "moderator": true,787 "trust_level": 2788 }789 }790 ]791 },792 {793 "fancy_title": "Licensing for PyTorch Word Usage in Books",794 "id": 215808,795 "title": "Licensing for PyTorch Word Usage in Books",796 "slug": "licensing-for-pytorch-word-usage-in-books",797 "posts_count": 10,798 "reply_count": 7,799 "highest_post_number": 10,800 "image_url": null,801 "created_at": "2025-01-24T07:25:14.106Z",802 "last_posted_at": "2025-01-28T07:02:20.511Z",803 "bumped": true,804 "bumped_at": "2025-01-28T07:02:20.511Z",805 "archetype": "regular",806 "unseen": false,807 "pinned": false,808 "unpinned": null,809 "visible": true,810 "closed": false,811 "archived": false,812 "bookmarked": null,813 "liked": null,814 "tags_descriptions": {},815 "like_count": 0,816 "views": 96,817 "category_id": 1,818 "featured_link": null,819 "has_accepted_answer": false,820 "posters": [821 {822 "extras": "latest",823 "description": "Original Poster, Most Recent Poster",824 "user": {825 "id": 82282,826 "username": "orangeeducationpvt",827 "name": "Orange Education Pvt Ltd",828 "avatar_template": "/letter_avatar_proxy/v4/letter/o/97f17d/{size}.png",829 "trust_level": 0830 }831 },832 {833 "extras": null,834 "description": "Frequent Poster",835 "user": {836 "id": 3534,837 "username": "ptrblck",838 "name": "",839 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",840 "admin": true,841 "moderator": true,842 "trust_level": 2843 }844 },845 {846 "extras": null,847 "description": "Frequent Poster",848 "user": {849 "id": 75469,850 "username": "Matt_White",851 "name": "Matt White",852 "avatar_template": "/user_avatar/discuss.pytorch.org/matt_white/{size}/69693_2.png",853 "admin": true,854 "trust_level": 1855 }856 }857 ]858 },859 {860 "fancy_title": "Problem with inpainting Stable Diffusion and python",861 "id": 216995,862 "title": "Problem with inpainting Stable Diffusion and python",863 "slug": "problem-with-inpainting-stable-diffusion-and-python",864 "posts_count": 1,865 "reply_count": 0,866 "highest_post_number": 1,867 "image_url": "https://discuss.pytorch.org/uploads/default/original/3X/3/1/31f1fcf366294f3565f157c4ce276c65d68f3ac0.jpeg",868 "created_at": "2025-02-21T16:06:56.775Z",869 "last_posted_at": "2025-02-21T16:06:56.814Z",870 "bumped": true,871 "bumped_at": "2025-02-21T16:06:56.814Z",872 "archetype": "regular",873 "unseen": false,874 "pinned": false,875 "unpinned": null,876 "visible": true,877 "closed": false,878 "archived": false,879 "bookmarked": null,880 "liked": null,881 "tags_descriptions": {},882 "like_count": 0,883 "views": 111,884 "category_id": 1,885 "featured_link": null,886 "has_accepted_answer": false,887 "posters": [888 {889 "extras": "latest single",890 "description": "Original Poster, Most Recent Poster",891 "user": {892 "id": 82738,893 "username": "Dmitr",894 "name": "",895 "avatar_template": "/letter_avatar_proxy/v4/letter/d/13edae/{size}.png",896 "trust_level": 1897 }898 }899 ]900 },901 {902 "fancy_title": "Rtx 5090 Error with nightly cu128 install",903 "id": 217478,904 "title": "Rtx 5090 Error with nightly cu128 install",905 "slug": "rtx-5090-error-with-nightly-cu128-install",906 "posts_count": 8,907 "reply_count": 2,908 "highest_post_number": 8,909 "image_url": null,910 "created_at": "2025-03-05T14:19:51.454Z",911 "last_posted_at": "2025-04-02T01:18:02.963Z",912 "bumped": true,913 "bumped_at": "2025-04-02T01:18:02.963Z",914 "archetype": "regular",915 "unseen": false,916 "pinned": false,917 "unpinned": null,918 "visible": true,919 "closed": false,920 "archived": false,921 "bookmarked": null,922 "liked": null,923 "tags_descriptions": {},924 "like_count": 0,925 "views": 5032,926 "category_id": 1,927 "featured_link": null,928 "has_accepted_answer": false,929 "posters": [930 {931 "extras": null,932 "description": "Original Poster",933 "user": {934 "id": 83084,935 "username": "Marked",936 "name": "Marked",937 "avatar_template": "/user_avatar/discuss.pytorch.org/marked/{size}/74882_2.png",938 "trust_level": 1939 }940 },941 {942 "extras": null,943 "description": "Frequent Poster",944 "user": {945 "id": 83067,946 "username": "nafe_alhmdan",947 "name": "nafe alhmdan",948 "avatar_template": "/user_avatar/discuss.pytorch.org/nafe_alhmdan/{size}/75984_2.png",949 "trust_level": 1950 }951 },952 {953 "extras": null,954 "description": "Frequent Poster",955 "user": {956 "id": 3534,957 "username": "ptrblck",958 "name": "",959 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",960 "admin": true,961 "moderator": true,962 "trust_level": 2963 }964 },965 {966 "extras": "latest",967 "description": "Most Recent Poster",968 "user": {969 "id": 83583,970 "username": "randscott",971 "name": "",972 "avatar_template": "/letter_avatar_proxy/v4/letter/r/7ba0ec/{size}.png",973 "trust_level": 0974 }975 }976 ]977 }978 ],979 "tags_descriptions": {},980 "fancy_title": "Bug or feature for longTensor?",981 "id": 51440,982 "title": "Bug or feature for longTensor?",983 "posts_count": 2,984 "created_at": "2019-07-24T00:55:12.381Z",985 "views": 348,986 "reply_count": 0,987 "like_count": 0,988 "last_posted_at": "2019-07-26T22:44:12.689Z",989 "visible": true,990 "closed": false,991 "archived": false,992 "has_summary": false,993 "archetype": "regular",994 "slug": "bug-or-feature-for-longtensor",995 "category_id": 1,996 "word_count": 71,997 "deleted_at": null,998 "user_id": 13822,999 "featured_link": null,1000 "pinned_globally": false,1001 "pinned_at": null,1002 "pinned_until": null,1003 "image_url": null,1004 "slow_mode_seconds": 0,1005 "draft": null,1006 "draft_key": "topic_51440",1007 "draft_sequence": null,1008 "unpinned": null,1009 "pinned": false,1010 "current_post_number": 1,1011 "highest_post_number": 2,1012 "deleted_by": null,1013 "actions_summary": [1014 {1015 "id": 4,1016 "count": 0,1017 "hidden": false,1018 "can_act": false1019 },1020 {1021 "id": 8,1022 "count": 0,1023 "hidden": false,1024 "can_act": false1025 },1026 {1027 "id": 10,1028 "count": 0,1029 "hidden": false,1030 "can_act": false1031 },1032 {1033 "id": 7,1034 "count": 0,1035 "hidden": false,1036 "can_act": false1037 }1038 ],1039 "chunk_size": 20,1040 "bookmarked": false,1041 "topic_timer": null,1042 "message_bus_last_id": 0,1043 "participant_count": 2,1044 "show_read_indicator": false,1045 "thumbnails": null,1046 "slow_mode_enabled_until": null,1047 "can_vote": false,1048 "vote_count": 0,1049 "user_voted": false,1050 "discourse_zendesk_plugin_zendesk_id": null,1051 "discourse_zendesk_plugin_zendesk_url": "https://your-url.zendesk.com/agent/tickets/",1052 "details": {1053 "can_edit": false,1054 "notification_level": 1,1055 "participants": [1056 {1057 "id": 13822,1058 "username": "Zhaoyi-Yan",1059 "name": "Zhaoyi Yan",1060 "avatar_template": "/user_avatar/discuss.pytorch.org/zhaoyi-yan/{size}/8480_2.png",1061 "post_count": 1,1062 "primary_group_name": null,1063 "flair_name": null,1064 "flair_url": null,1065 "flair_color": null,1066 "flair_bg_color": null,1067 "flair_group_id": null,1068 "trust_level": 21069 },1070 {1071 "id": 20203,1072 "username": "Prerna_Dhareshwar",1073 "name": "Prerna Dhareshwar",1074 "avatar_template": "/user_avatar/discuss.pytorch.org/prerna_dhareshwar/{size}/14256_2.png",1075 "post_count": 1,1076 "primary_group_name": null,1077 "flair_name": null,1078 "flair_url": null,1079 "flair_color": null,1080 "flair_bg_color": null,1081 "flair_group_id": null,1082 "trust_level": 21083 }1084 ],1085 "created_by": {1086 "id": 13822,1087 "username": "Zhaoyi-Yan",1088 "name": "Zhaoyi Yan",1089 "avatar_template": "/user_avatar/discuss.pytorch.org/zhaoyi-yan/{size}/8480_2.png"1090 },1091 "last_poster": {1092 "id": 20203,1093 "username": "Prerna_Dhareshwar",1094 "name": "Prerna Dhareshwar",1095 "avatar_template": "/user_avatar/discuss.pytorch.org/prerna_dhareshwar/{size}/14256_2.png"1096 }1097 },1098 "bookmarks": []1099 },1100 {1101 "post_stream": {1102 "posts": [1103 {1104 "id": 125038,1105 "name": "mohammad",1106 "username": "mmdbrdrn",1107 "avatar_template": "/user_avatar/discuss.pytorch.org/mmdbrdrn/{size}/9966_2.png",1108 "created_at": "2019-07-23T22:57:22.365Z",1109 "cooked": "<p>Hello every one.</p>\n<p>I am working on video processing and as data samples, i need short video clips (around 10 frame per each).<br>\nI have a dataset, composed of consecutive images, and what i want to do is combining consecutive 10 frames to make a cuboid. Now the problem is: for big datasets or even for multiple normal datasets, when i create numpy arrays containing plenty of cuboids (10 * 227* 227 : 10 frames of size 227*227), the Computer RAM gets fully occupied and kernel dies!!</p>\n<p>Maybe my algorithm has a bad mistake and i am unaware of a simple way, but i would be haapy if anyone give me useful hints to manage my RAM</p>\n<p>Thank you very much</p>",1110 "post_number": 1,1111 "post_type": 1,1112 "posts_count": 5,1113 "updated_at": "2019-07-23T22:59:56.252Z",1114 "reply_count": 0,1115 "reply_to_post_number": null,1116 "quote_count": 0,1117 "incoming_link_count": 485,1118 "reads": 33,1119 "readers_count": 32,1120 "score": 2426.6,1121 "yours": false,1122 "topic_id": 51437,1123 "topic_slug": "how-to-manage-ram-capacity-while-loading-dataloader-in-deep-learning",1124 "display_username": "mohammad",1125 "primary_group_name": null,1126 "flair_name": null,1127 "flair_url": null,1128 "flair_bg_color": null,1129 "flair_color": null,1130 "flair_group_id": null,1131 "badges_granted": [],1132 "version": 1,1133 "can_edit": false,1134 "can_delete": false,1135 "can_recover": false,1136 "can_see_hidden_post": false,1137 "can_wiki": false,1138 "read": true,1139 "user_title": "",1140 "bookmarked": false,1141 "actions_summary": [],1142 "moderator": false,1143 "admin": false,1144 "staff": false,1145 "user_id": 16631,1146 "hidden": false,1147 "trust_level": 1,1148 "deleted_at": null,1149 "user_deleted": false,1150 "edit_reason": null,1151 "can_view_edit_history": true,1152 "wiki": false,1153 "post_url": "/t/how-to-manage-ram-capacity-while-loading-dataloader-in-deep-learning/51437/1",1154 "can_accept_answer": false,1155 "can_unaccept_answer": false,1156 "accepted_answer": false,1157 "topic_accepted_answer": true,1158 "can_vote": false1159 },1160 {1161 "id": 125216,1162 "name": "",1163 "username": "ptrblck",1164 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",1165 "created_at": "2019-07-24T18:00:15.532Z",1166 "cooked": "<p>The storage of 10 frames of shape <code>227x227</code> shouldn’t cause any problems itself.<br>\nAssuming that you are dealing with RGB images, you would only use approx. <code>10*227*227*3*4 / 1024**2 = 5.9MB</code>, if you store the images in FP32.</p>\n<p>Do you run out of memory directly after loading the data or could some other code part (e.g. model forward/backward) cause this issue?</p>",1167 "post_number": 2,1168 "post_type": 1,1169 "posts_count": 5,1170 "updated_at": "2019-07-24T18:00:15.532Z",1171 "reply_count": 1,1172 "reply_to_post_number": null,1173 "quote_count": 0,1174 "incoming_link_count": 9,1175 "reads": 32,1176 "readers_count": 31,1177 "score": 71.4,1178 "yours": false,1179 "topic_id": 51437,1180 "topic_slug": "how-to-manage-ram-capacity-while-loading-dataloader-in-deep-learning",1181 "display_username": "",1182 "primary_group_name": null,1183 "flair_name": null,1184 "flair_url": null,1185 "flair_bg_color": null,1186 "flair_color": null,1187 "flair_group_id": null,1188 "badges_granted": [],1189 "version": 1,1190 "can_edit": false,1191 "can_delete": false,1192 "can_recover": false,1193 "can_see_hidden_post": false,1194 "can_wiki": false,1195 "read": true,1196 "user_title": "",1197 "bookmarked": false,1198 "actions_summary": [1199 {1200 "id": 2,