Anurag1734/cuda-error-resolution-analysis
07
1[2 {3 "post_stream": {4 "posts": [5 {6 "id": 310074,7 "name": "Sam Lerman",8 "username": "Sam_Lerman",9 "avatar_template": "/user_avatar/discuss.pytorch.org/sam_lerman/{size}/37165_2.png",10 "created_at": "2021-10-04T19:25:52.829Z",11 "cooked": "<p>Is there a way to run the modules in a ModuleList in parallel on multiple GPUs? Their inputs are shaped differently and they don’t depend on one another.</p>",12 "post_number": 1,13 "post_type": 1,14 "posts_count": 6,15 "updated_at": "2021-10-04T19:26:00.070Z",16 "reply_count": 0,17 "reply_to_post_number": null,18 "quote_count": 0,19 "incoming_link_count": 301,20 "reads": 8,21 "readers_count": 7,22 "score": 1506.6,23 "yours": false,24 "topic_id": 133461,25 "topic_slug": "running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel",26 "display_username": "Sam Lerman",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": 30780,48 "hidden": false,49 "trust_level": 2,50 "deleted_at": null,51 "user_deleted": false,52 "edit_reason": null,53 "can_view_edit_history": true,54 "wiki": false,55 "post_url": "/t/running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel/133461/1",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 "id": 310129,64 "name": "",65 "username": "ptrblck",66 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",67 "created_at": "2021-10-05T04:46:47.659Z",68 "cooked": "<p>Yes, if you are pushing different layers and their inputs to different GPUs, their execution will be asynchronous.</p>",69 "post_number": 2,70 "post_type": 1,71 "posts_count": 6,72 "updated_at": "2021-10-05T04:46:47.659Z",73 "reply_count": 1,74 "reply_to_post_number": null,75 "quote_count": 0,76 "incoming_link_count": 6,77 "reads": 8,78 "readers_count": 7,79 "score": 36.6,80 "yours": false,81 "topic_id": 133461,82 "topic_slug": "running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel",83 "display_username": "",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 "moderator": true,102 "admin": true,103 "staff": true,104 "user_id": 3534,105 "hidden": false,106 "trust_level": 2,107 "deleted_at": null,108 "user_deleted": false,109 "edit_reason": null,110 "can_view_edit_history": true,111 "wiki": false,112 "post_url": "/t/running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel/133461/2",113 "can_accept_answer": false,114 "can_unaccept_answer": false,115 "accepted_answer": false,116 "topic_accepted_answer": null117 },118 {119 "id": 310596,120 "name": "Sam Lerman",121 "username": "Sam_Lerman",122 "avatar_template": "/user_avatar/discuss.pytorch.org/sam_lerman/{size}/37165_2.png",123 "created_at": "2021-10-08T01:17:03.618Z",124 "cooked": "<p>Oh, but they’re in a for loop. Python will automatically proceed to the next iteration of the for loop?</p>",125 "post_number": 3,126 "post_type": 1,127 "posts_count": 6,128 "updated_at": "2021-10-08T01:17:03.618Z",129 "reply_count": 1,130 "reply_to_post_number": 2,131 "quote_count": 0,132 "incoming_link_count": 6,133 "reads": 6,134 "readers_count": 5,135 "score": 36.2,136 "yours": false,137 "topic_id": 133461,138 "topic_slug": "running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel",139 "display_username": "Sam Lerman",140 "primary_group_name": null,141 "flair_name": null,142 "flair_url": null,143 "flair_bg_color": null,144 "flair_color": null,145 "flair_group_id": null,146 "badges_granted": [],147 "version": 1,148 "can_edit": false,149 "can_delete": false,150 "can_recover": false,151 "can_see_hidden_post": false,152 "can_wiki": false,153 "read": true,154 "user_title": null,155 "reply_to_user": {156 "id": 3534,157 "username": "ptrblck",158 "name": "",159 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png"160 },161 "bookmarked": false,162 "actions_summary": [],163 "moderator": false,164 "admin": false,165 "staff": false,166 "user_id": 30780,167 "hidden": false,168 "trust_level": 2,169 "deleted_at": null,170 "user_deleted": false,171 "edit_reason": null,172 "can_view_edit_history": true,173 "wiki": false,174 "post_url": "/t/running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel/133461/3",175 "can_accept_answer": false,176 "can_unaccept_answer": false,177 "accepted_answer": false,178 "topic_accepted_answer": null179 },180 {181 "id": 310615,182 "name": "",183 "username": "ptrblck",184 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",185 "created_at": "2021-10-08T05:05:10.834Z",186 "cooked": "<p>Yes, CUDA kernels are launched by the CPU and are then executed asynchronously, which means the CPU can run ahead and execute other work, such as launching another CUDA kernel on another device.</p>",187 "post_number": 4,188 "post_type": 1,189 "posts_count": 6,190 "updated_at": "2021-10-08T05:05:10.834Z",191 "reply_count": 1,192 "reply_to_post_number": 3,193 "quote_count": 0,194 "incoming_link_count": 7,195 "reads": 6,196 "readers_count": 5,197 "score": 41.2,198 "yours": false,199 "topic_id": 133461,200 "topic_slug": "running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel",201 "display_username": "",202 "primary_group_name": null,203 "flair_name": null,204 "flair_url": null,205 "flair_bg_color": null,206 "flair_color": null,207 "flair_group_id": null,208 "badges_granted": [],209 "version": 1,210 "can_edit": false,211 "can_delete": false,212 "can_recover": false,213 "can_see_hidden_post": false,214 "can_wiki": false,215 "read": true,216 "user_title": "",217 "reply_to_user": {218 "id": 30780,219 "username": "Sam_Lerman",220 "name": "Sam Lerman",221 "avatar_template": "/user_avatar/discuss.pytorch.org/sam_lerman/{size}/37165_2.png"222 },223 "bookmarked": false,224 "actions_summary": [],225 "moderator": true,226 "admin": true,227 "staff": true,228 "user_id": 3534,229 "hidden": false,230 "trust_level": 2,231 "deleted_at": null,232 "user_deleted": false,233 "edit_reason": null,234 "can_view_edit_history": true,235 "wiki": false,236 "post_url": "/t/running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel/133461/4",237 "can_accept_answer": false,238 "can_unaccept_answer": false,239 "accepted_answer": false,240 "topic_accepted_answer": null241 },242 {243 "id": 310683,244 "name": "Sam Lerman",245 "username": "Sam_Lerman",246 "avatar_template": "/user_avatar/discuss.pytorch.org/sam_lerman/{size}/37165_2.png",247 "created_at": "2021-10-08T10:36:36.797Z",248 "cooked": "<p>Okay, thanks. Hmm, would it work with DataParallel? Or do I manually have to map each iteration of the for loop to a specific GPU? All the tutorials I can find use DataParallel.</p>",249 "post_number": 5,250 "post_type": 1,251 "posts_count": 6,252 "updated_at": "2021-10-08T10:36:36.797Z",253 "reply_count": 1,254 "reply_to_post_number": 4,255 "quote_count": 0,256 "incoming_link_count": 11,257 "reads": 6,258 "readers_count": 5,259 "score": 61.2,260 "yours": false,261 "topic_id": 133461,262 "topic_slug": "running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel",263 "display_username": "Sam Lerman",264 "primary_group_name": null,265 "flair_name": null,266 "flair_url": null,267 "flair_bg_color": null,268 "flair_color": null,269 "flair_group_id": null,270 "badges_granted": [],271 "version": 1,272 "can_edit": false,273 "can_delete": false,274 "can_recover": false,275 "can_see_hidden_post": false,276 "can_wiki": false,277 "read": true,278 "user_title": null,279 "reply_to_user": {280 "id": 3534,281 "username": "ptrblck",282 "name": "",283 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png"284 },285 "bookmarked": false,286 "actions_summary": [],287 "moderator": false,288 "admin": false,289 "staff": false,290 "user_id": 30780,291 "hidden": false,292 "trust_level": 2,293 "deleted_at": null,294 "user_deleted": false,295 "edit_reason": null,296 "can_view_edit_history": true,297 "wiki": false,298 "post_url": "/t/running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel/133461/5",299 "can_accept_answer": false,300 "can_unaccept_answer": false,301 "accepted_answer": false,302 "topic_accepted_answer": null303 },304 {305 "id": 310744,306 "name": "",307 "username": "ptrblck",308 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",309 "created_at": "2021-10-08T18:42:37.497Z",310 "cooked": "<p><code>nn.DataParallel</code> works similar, but has communications added to its approach. <a href=\"https://medium.com/huggingface/training-larger-batches-practical-tips-on-1-gpu-multi-gpu-distributed-setups-ec88c3e51255\">This blog post</a> explains its mechanism.</p>",311 "post_number": 6,312 "post_type": 1,313 "posts_count": 6,314 "updated_at": "2021-10-08T18:42:37.497Z",315 "reply_count": 0,316 "reply_to_post_number": 5,317 "quote_count": 0,318 "incoming_link_count": 0,319 "reads": 5,320 "readers_count": 4,321 "score": 1.0,322 "yours": false,323 "topic_id": 133461,324 "topic_slug": "running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel",325 "display_username": "",326 "primary_group_name": null,327 "flair_name": null,328 "flair_url": null,329 "flair_bg_color": null,330 "flair_color": null,331 "flair_group_id": null,332 "badges_granted": [],333 "version": 1,334 "can_edit": false,335 "can_delete": false,336 "can_recover": false,337 "can_see_hidden_post": false,338 "can_wiki": false,339 "link_counts": [340 {341 "url": "https://medium.com/huggingface/training-larger-batches-practical-tips-on-1-gpu-multi-gpu-distributed-setups-ec88c3e51255",342 "internal": false,343 "reflection": false,344 "title": "💥 Training Neural Nets on Larger Batches: Practical Tips for 1-GPU, Multi-GPU & Distributed setups | by Thomas Wolf | HuggingFace | Medium",345 "clicks": 35346 }347 ],348 "read": true,349 "user_title": "",350 "reply_to_user": {351 "id": 30780,352 "username": "Sam_Lerman",353 "name": "Sam Lerman",354 "avatar_template": "/user_avatar/discuss.pytorch.org/sam_lerman/{size}/37165_2.png"355 },356 "bookmarked": false,357 "actions_summary": [],358 "moderator": true,359 "admin": true,360 "staff": true,361 "user_id": 3534,362 "hidden": false,363 "trust_level": 2,364 "deleted_at": null,365 "user_deleted": false,366 "edit_reason": null,367 "can_view_edit_history": true,368 "wiki": false,369 "post_url": "/t/running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel/133461/6",370 "can_accept_answer": false,371 "can_unaccept_answer": false,372 "accepted_answer": false,373 "topic_accepted_answer": null374 }375 ],376 "stream": [377 310074,378 310129,379 310596,380 310615,381 310683,382 310744383 ]384 },385 "timeline_lookup": [386 [387 1,388 1482389 ],390 [391 3,392 1479393 ],394 [395 5,396 1478397 ]398 ],399 "suggested_topics": [400 {401 "fancy_title": "`torch.linalg.svd` uses `cudaMemcpyAsync` that syncs between host and device",402 "id": 213297,403 "title": "`torch.linalg.svd` uses `cudaMemcpyAsync` that syncs between host and device",404 "slug": "torch-linalg-svd-uses-cudamemcpyasync-that-syncs-between-host-and-device",405 "posts_count": 1,406 "reply_count": 0,407 "highest_post_number": 1,408 "image_url": null,409 "created_at": "2024-11-22T07:35:51.287Z",410 "last_posted_at": "2024-11-22T07:35:51.352Z",411 "bumped": true,412 "bumped_at": "2024-11-22T07:35:51.352Z",413 "archetype": "regular",414 "unseen": false,415 "pinned": false,416 "unpinned": null,417 "visible": true,418 "closed": false,419 "archived": false,420 "bookmarked": null,421 "liked": null,422 "tags_descriptions": {},423 "like_count": 0,424 "views": 64,425 "category_id": 1,426 "featured_link": null,427 "has_accepted_answer": false,428 "posters": [429 {430 "extras": "latest single",431 "description": "Original Poster, Most Recent Poster",432 "user": {433 "id": 14714,434 "username": "Rui_Wang",435 "name": "Rui Wang",436 "avatar_template": "/user_avatar/discuss.pytorch.org/rui_wang/{size}/37635_2.png",437 "trust_level": 1438 }439 }440 ]441 },442 {443 "fancy_title": "Dataset handling and model selection",444 "id": 218106,445 "title": "Dataset handling and model selection",446 "slug": "dataset-handling-and-model-selection",447 "posts_count": 1,448 "reply_count": 0,449 "highest_post_number": 1,450 "image_url": null,451 "created_at": "2025-03-21T06:12:00.765Z",452 "last_posted_at": "2025-03-21T06:12:00.808Z",453 "bumped": true,454 "bumped_at": "2025-03-21T06:12:00.808Z",455 "archetype": "regular",456 "unseen": false,457 "pinned": false,458 "unpinned": null,459 "visible": true,460 "closed": false,461 "archived": false,462 "bookmarked": null,463 "liked": null,464 "tags_descriptions": {},465 "like_count": 0,466 "views": 19,467 "category_id": 1,468 "featured_link": null,469 "has_accepted_answer": false,470 "posters": [471 {472 "extras": "latest single",473 "description": "Original Poster, Most Recent Poster",474 "user": {475 "id": 83398,476 "username": "Muhammad_Irfan",477 "name": "Muhammad Irfan",478 "avatar_template": "/user_avatar/discuss.pytorch.org/muhammad_irfan/{size}/76272_2.png",479 "trust_level": 0480 }481 }482 ]483 },484 {485 "fancy_title": "Dynamic shapes and PyTorch",486 "id": 214929,487 "title": "Dynamic shapes and PyTorch",488 "slug": "dynamic-shapes-and-pytorch",489 "posts_count": 3,490 "reply_count": 0,491 "highest_post_number": 3,492 "image_url": null,493 "created_at": "2025-01-03T10:34:36.679Z",494 "last_posted_at": "2025-01-03T21:48:49.713Z",495 "bumped": true,496 "bumped_at": "2025-01-03T21:48:49.713Z",497 "archetype": "regular",498 "unseen": false,499 "pinned": false,500 "unpinned": null,501 "visible": true,502 "closed": false,503 "archived": false,504 "bookmarked": null,505 "liked": null,506 "tags_descriptions": {},507 "like_count": 1,508 "views": 252,509 "category_id": 1,510 "featured_link": null,511 "has_accepted_answer": false,512 "posters": [513 {514 "extras": null,515 "description": "Original Poster",516 "user": {517 "id": 81854,518 "username": "Mark_Fanter",519 "name": "Mark Fanter",520 "avatar_template": "/user_avatar/discuss.pytorch.org/mark_fanter/{size}/74876_2.png",521 "trust_level": 0522 }523 },524 {525 "extras": null,526 "description": "Frequent Poster",527 "user": {528 "id": 41396,529 "username": "soulitzer",530 "name": "",531 "avatar_template": "/letter_avatar_proxy/v4/letter/s/839c29/{size}.png",532 "trust_level": 2533 }534 },535 {536 "extras": "latest",537 "description": "Most Recent Poster",538 "user": {539 "id": 18088,540 "username": "KFrank",541 "name": "K. Frank",542 "avatar_template": "/letter_avatar_proxy/v4/letter/k/ecb155/{size}.png",543 "trust_level": 2544 }545 }546 ]547 },548 {549 "fancy_title": "What is the difference between the CUDA API and CUDA HW lines in the Nsight Systems GUI?",550 "id": 216417,551 "title": "What is the difference between the CUDA API and CUDA HW lines in the Nsight Systems GUI?",552 "slug": "what-is-the-difference-between-the-cuda-api-and-cuda-hw-lines-in-the-nsight-systems-gui",553 "posts_count": 1,554 "reply_count": 0,555 "highest_post_number": 1,556 "image_url": "https://discuss.pytorch.org/uploads/default/optimized/3X/2/7/27cbe5e88c4ed57d27e04ca70978791ed7307f48_2_1024x368.png",557 "created_at": "2025-02-09T04:09:00.331Z",558 "last_posted_at": "2025-02-09T04:09:00.373Z",559 "bumped": true,560 "bumped_at": "2025-02-09T04:09:00.373Z",561 "archetype": "regular",562 "unseen": false,563 "pinned": false,564 "unpinned": null,565 "visible": true,566 "closed": false,567 "archived": false,568 "bookmarked": null,569 "liked": null,570 "tags_descriptions": {},571 "like_count": 0,572 "views": 39,573 "category_id": 1,574 "featured_link": null,575 "has_accepted_answer": false,576 "posters": [577 {578 "extras": "latest single",579 "description": "Original Poster, Most Recent Poster",580 "user": {581 "id": 79055,582 "username": "hello_e",583 "name": "hhhh",584 "avatar_template": "/letter_avatar_proxy/v4/letter/h/f9ae1b/{size}.png",585 "trust_level": 1586 }587 }588 ]589 },590 {591 "fancy_title": "XPU out of memory error with Intel Arc Graphics (Meteor Lake) despite sufficient system memory and reported XPU capacity",592 "id": 219812,593 "title": "XPU out of memory error with Intel Arc Graphics (Meteor Lake) despite sufficient system memory and reported XPU capacity",594 "slug": "xpu-out-of-memory-error-with-intel-arc-graphics-meteor-lake-despite-sufficient-system-memory-and-reported-xpu-capacity",595 "posts_count": 7,596 "reply_count": 6,597 "highest_post_number": 9,598 "image_url": null,599 "created_at": "2025-05-06T14:02:04.746Z",600 "last_posted_at": "2025-08-18T02:55:58.317Z",601 "bumped": true,602 "bumped_at": "2025-08-18T02:55:58.317Z",603 "archetype": "regular",604 "unseen": false,605 "pinned": false,606 "unpinned": null,607 "visible": true,608 "closed": false,609 "archived": false,610 "bookmarked": null,611 "liked": null,612 "tags_descriptions": {},613 "like_count": 1,614 "views": 445,615 "category_id": 1,616 "featured_link": null,617 "has_accepted_answer": true,618 "posters": [619 {620 "extras": null,621 "description": "Original Poster, Accepted Answer",622 "user": {623 "id": 84174,624 "username": "libchara-dev",625 "name": "libchara-dev",626 "avatar_template": "/user_avatar/discuss.pytorch.org/libchara-dev/{size}/76930_2.png",627 "trust_level": 1628 }629 },630 {631 "extras": null,632 "description": "Frequent Poster",633 "user": {634 "id": 18088,635 "username": "KFrank",636 "name": "K. Frank",637 "avatar_template": "/letter_avatar_proxy/v4/letter/k/ecb155/{size}.png",638 "trust_level": 2639 }640 },641 {642 "extras": null,643 "description": "Frequent Poster",644 "user": {645 "id": 84435,646 "username": "Andrey_Neyvanov",647 "name": "Andrey Neyvanov",648 "avatar_template": "/user_avatar/discuss.pytorch.org/andrey_neyvanov/{size}/75114_2.png",649 "trust_level": 0650 }651 },652 {653 "extras": "latest",654 "description": "Most Recent Poster",655 "user": {656 "id": 84966,657 "username": "theSharque",658 "name": "Alexey Kurilov",659 "avatar_template": "/user_avatar/discuss.pytorch.org/thesharque/{size}/77603_2.png",660 "trust_level": 0661 }662 }663 ]664 }665 ],666 "tags_descriptions": {},667 "fancy_title": "Running multiple modules in a ModuleList on different GPUs in parallel?",668 "id": 133461,669 "title": "Running multiple modules in a ModuleList on different GPUs in parallel?",670 "posts_count": 6,671 "created_at": "2021-10-04T19:25:52.725Z",672 "views": 845,673 "reply_count": 4,674 "like_count": 0,675 "last_posted_at": "2021-10-08T18:42:37.497Z",676 "visible": true,677 "closed": false,678 "archived": false,679 "has_summary": false,680 "archetype": "regular",681 "slug": "running-multiple-modules-in-a-modulelist-on-different-gpus-in-parallel",682 "category_id": 1,683 "word_count": 167,684 "deleted_at": null,685 "user_id": 30780,686 "featured_link": null,687 "pinned_globally": false,688 "pinned_at": null,689 "pinned_until": null,690 "image_url": null,691 "slow_mode_seconds": 0,692 "draft": null,693 "draft_key": "topic_133461",694 "draft_sequence": null,695 "unpinned": null,696 "pinned": false,697 "current_post_number": 1,698 "highest_post_number": 6,699 "deleted_by": null,700 "actions_summary": [701 {702 "id": 4,703 "count": 0,704 "hidden": false,705 "can_act": false706 },707 {708 "id": 8,709 "count": 0,710 "hidden": false,711 "can_act": false712 },713 {714 "id": 10,715 "count": 0,716 "hidden": false,717 "can_act": false718 },719 {720 "id": 7,721 "count": 0,722 "hidden": false,723 "can_act": false724 }725 ],726 "chunk_size": 20,727 "bookmarked": false,728 "topic_timer": null,729 "message_bus_last_id": 0,730 "participant_count": 2,731 "show_read_indicator": false,732 "thumbnails": null,733 "slow_mode_enabled_until": null,734 "can_vote": false,735 "vote_count": 0,736 "user_voted": false,737 "discourse_zendesk_plugin_zendesk_id": null,738 "discourse_zendesk_plugin_zendesk_url": "https://your-url.zendesk.com/agent/tickets/",739 "details": {740 "can_edit": false,741 "notification_level": 1,742 "participants": [743 {744 "id": 3534,745 "username": "ptrblck",746 "name": "",747 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",748 "post_count": 3,749 "primary_group_name": null,750 "flair_name": null,751 "flair_url": null,752 "flair_color": null,753 "flair_bg_color": null,754 "flair_group_id": null,755 "admin": true,756 "moderator": true,757 "trust_level": 2758 },759 {760 "id": 30780,761 "username": "Sam_Lerman",762 "name": "Sam Lerman",763 "avatar_template": "/user_avatar/discuss.pytorch.org/sam_lerman/{size}/37165_2.png",764 "post_count": 3,765 "primary_group_name": null,766 "flair_name": null,767 "flair_url": null,768 "flair_color": null,769 "flair_bg_color": null,770 "flair_group_id": null,771 "trust_level": 2772 }773 ],774 "created_by": {775 "id": 30780,776 "username": "Sam_Lerman",777 "name": "Sam Lerman",778 "avatar_template": "/user_avatar/discuss.pytorch.org/sam_lerman/{size}/37165_2.png"779 },780 "last_poster": {781 "id": 3534,782 "username": "ptrblck",783 "name": "",784 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png"785 },786 "links": [787 {788 "url": "https://medium.com/huggingface/training-larger-batches-practical-tips-on-1-gpu-multi-gpu-distributed-setups-ec88c3e51255",789 "title": "💥 Training Neural Nets on Larger Batches: Practical Tips for 1-GPU, Multi-GPU & Distributed setups | by Thomas Wolf | HuggingFace | Medium",790 "internal": false,791 "attachment": false,792 "reflection": false,793 "clicks": 35,794 "user_id": 3534,795 "domain": "medium.com",796 "root_domain": "medium.com"797 }798 ]799 },800 "bookmarks": []801 },802 {803 "post_stream": {804 "posts": [805 {806 "id": 310380,807 "name": "Pixel",808 "username": "Pixel",809 "avatar_template": "/letter_avatar_proxy/v4/letter/p/f14d63/{size}.png",810 "created_at": "2021-10-06T14:12:27.870Z",811 "cooked": "<p>Hi folks, <img src=\"https://discuss.pytorch.org/images/emoji/apple/hugs.png?v=12\" title=\":hugs:\" class=\"emoji\" alt=\":hugs:\" loading=\"lazy\" width=\"20\" height=\"20\"></p>\n<p>i am working on a face expression recognition project and i am using the public dataset fer2013 (with its original data splits:<br>\nabt 28,709 training set, 3,589 validation set and 3,589 test set). As input for my CNN i get scaled (from 0-1) grey-scale 48x48 face<br>\nimages and for output i get a tensor which gives 7 probability values for each 7 emotions (0=Angry, 1=Disgust, 2=Fear, 3=Happy, 4=Sad, 5=Surprise, 6=Neutral)</p>\n<p>After i have trained my CNN model for about 400 epochs while printing the running training loss/acc over every 20 mini-batches and also validating<br>\non the whole validation dataset after every 20 mini-batches, i see that my validation accuracy is just staying at that same exact value<br>\nof 24,94 even after running 400 epochs…Also the validation loss is not moving that much…At this point i have no idea what exactly i am doing wrong.</p>\n<p>Is my evaluation method, which i call for validation even correct? Is the way i am calculating loss and accuracy correct?<br>\nWhy is my validation loss/acc staying at the same value and not improving?</p>\n<p>I would be really greatful for ANY help or improvement suggestions. <img src=\"https://discuss.pytorch.org/images/emoji/apple/pray.png?v=12\" title=\":pray:\" class=\"emoji\" alt=\":pray:\" loading=\"lazy\" width=\"20\" height=\"20\"> <img src=\"https://discuss.pytorch.org/images/emoji/apple/pray.png?v=12\" title=\":pray:\" class=\"emoji\" alt=\":pray:\" loading=\"lazy\" width=\"20\" height=\"20\"> <img src=\"https://discuss.pytorch.org/images/emoji/apple/pray.png?v=12\" title=\":pray:\" class=\"emoji\" alt=\":pray:\" loading=\"lazy\" width=\"20\" height=\"20\"></p>\n<p>I have the following configurations:<br>\n“device”: “cuda:0”,<br>\n“learningrate”: 1e-3,<br>\n“weight_decay”: 1e-5,<br>\n“epochs”: 500,<br>\nIMAGE_SIZE = 48<br>\nBATCH_SIZE = 128</p>\n<p>MY CNN-MODEL----------------------------------------------------------------------------------</p>\n<pre><code class=\"lang-auto\">class CustomCNN(nn.Module):\n def __init__(self, drop=0.5, n_in_channels: int = 1, n_kernels: int = 64,\n kernel_size: int = 3):\n super().__init__()\n\n padding = int(kernel_size / 2)\n\n self.conv1a = nn.Conv2d(in_channels=n_in_channels, out_channels=n_kernels, kernel_size=kernel_size, padding=1)\n self.conv1b = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=kernel_size, padding=1)\n\n self.conv2a = nn.Conv2d(in_channels=64, out_channels=96, kernel_size=kernel_size, padding=padding)\n self.conv2b = nn.Conv2d(in_channels=96, out_channels=96, kernel_size=kernel_size, padding=padding)\n\n self.conv3a = nn.Conv2d(96, 128, kernel_size=kernel_size, padding=padding)\n self.conv3b = nn.Conv2d(128, 128, kernel_size=kernel_size, padding=padding)\n\n self.conv4a = nn.Conv2d(128, 256, kernel_size=kernel_size, padding=padding)\n # max pooling right here\n self.conv4b = nn.Conv2d(256, 256, kernel_size=kernel_size, padding=padding)\n\n self.bn1a = nn.BatchNorm2d(64)\n self.bn1b = nn.BatchNorm2d(64)\n\n self.bn2a = nn.BatchNorm2d(96)\n self.bn2b = nn.BatchNorm2d(96)\n\n self.bn3a = nn.BatchNorm2d(128)\n self.bn3b = nn.BatchNorm2d(128)\n\n self.bn4a = nn.BatchNorm2d(256)\n self.bn4b = nn.BatchNorm2d(256)\n\n self.pool = nn.MaxPool2d(kernel_size=2, stride=2)\n self.pool2 = nn.MaxPool2d(kernel_size=3, stride=1)\n self.drop = nn.Dropout(p=drop)\n self.relu = nn.ReLU() # torch.nn.LeakyReLU(0.3)\n\n ''' passing random data through model just to get shape of last conv2d output'''\n self._to_linear = None\n x = torch.randn(48, 48).view(-1, 1, 48, 48)\n self.convs(x)\n\n self.lin1 = nn.Linear(self._to_linear, 256)\n self.lin3 = nn.Linear(256, 7)\n\n def convs(self, x):\n x = self.relu(self.bn1a(self.conv1a(x)))\n x = self.relu(self.bn1b(self.conv1b(x)))\n x = self.pool(x) # from 48x48 to 24x24\n\n x = self.relu(self.bn2a(self.conv2a(x)))\n x = self.relu(self.bn2b(self.conv2b(x)))\n x = self.pool(x) # from 24x24 to 12x12\n\n x = self.relu(self.bn3a(self.conv3a(x)))\n x = self.relu(self.bn3b(self.conv3b(x)))\n x = self.pool(x) # from 12x12 to 6x6\n\n x = self.relu(self.bn4a(self.conv4a(x)))\n x = self.pool2(x) # max pool from 6x6 to 4x4\n\n x = self.relu(self.bn4b(self.conv4b(x)))\n\n if self._to_linear is None:\n self._to_linear = x[0].shape[0] * x[0].shape[1] * x[0].shape[2]\n return x\n\n def forward(self, x):\n x = self.convs(x)\n x = x.view(-1, self._to_linear)\n x = F.relu(self.lin1(x))\n x = self.lin3(x)\n return F.softmax(x, dim=1)\n</code></pre>\n<p>function for validating / evaluating model-------------------------------------------------------------------------------------</p>\n<pre><code class=\"lang-auto\">def evaluate(model: nn.Module, loader: DataLoader, loss_fn):\n model.eval()\n device = next(model.parameters()).device\n loss_avg, correct_predictions = 0.0, 0.0\n nr_samples = len(loader.dataset)\n with torch.no_grad():\n for inputs, labels in loader:\n inputs, labels = inputs.to(device), labels.to(device)\n outputs = model(inputs)\n loss = loss_fn(outputs, labels)\n # calculating some performance metrics\n loss_avg += loss.item()\n correct_predictions += outputs.argmax(dim=1).eq(labels.argmax(dim=1)).sum().item()\n\n accuracy = 100 * (correct_predictions / nr_samples)\n loss = loss_avg / nr_samples\n return accuracy, loss\n</code></pre>\n<p>----main-----------------------------------------------------------------------------------------------------------------------<br>\n“”“Main function that takes hyperparameters and performs training and evaluation of model”“”</p>\n<pre><code class=\"lang-auto\">def main(results_path, network_config: dict, learningrate: int = 1e-3, weight_decay: float = 1e-5,\n epochs: int = 400, device: torch.device = torch.device(\"cuda:0\")): # cuda:0\n\n training_dataset = ImageDataset(TRAINING_DATA, TRAINING_LABELS)\n val_dataset = ImageDataset(VAL_DATA, VAL_LABELS)\n test_dataset = ImageDataset(TEST_DATA, TEST_LABELS)\n trainloader = torch.utils.data.DataLoader(training_dataset, batch_size=1, shuffle=False, num_workers=0)\n valloader = torch.utils.data.DataLoader(val_dataset, batch_size=1, shuffle=False, num_workers=0)\n testloader = torch.utils.data.DataLoader(test_dataset, batch_size=1, shuffle=False, num_workers=0)\n trainloader_augmented = torch.utils.data.DataLoader(training_dataset, batch_size=BATCH_SIZE, shuffle=True, num_workers=1)\n\n writer = SummaryWriter(log_dir=os.path.join(results_path, \"tensorboard\", \"experiments\", \"2k_run\"))\n\n print_stats_at = 20 # print status to tensorboard every x batch e.g after every 5 batches\n validate_at = 20 # evaluate model on validation set and check for new best model every x batches\n update = 0 # current update counter\n best_validation_loss = np.inf # best validation loss so far\n update_progess_bar = tqdm.tqdm(total=epochs, desc=f\"loss: {np.nan:7.5f}\", position=0) # progressbar\n\n model = CustomCNN()\n model.to(device)\n optimizer = torch.optim.Adam(model.parameters(), lr=learningrate, weight_decay=weight_decay)\n loss_fn = MSELoss()\n running_loss, correct_predictions = 0.0, 0.0\n nr_samples = 0\n model.train()\n\n # Train until n epochs have been reached\n '''\n reporting an averaged loss over N(print_Stats_at) mini-batches, where N is large enough to \n smooth out the noise of individual batches but not so large that the model \n performance is not comparable between the first and last batches.\n '''\n print(datetime.now(), \" Training started.\")\n while update < epochs:\n for i, data in enumerate(trainloader_augmented):\n inputs, targets = data\n inputs, targets = inputs.to(device), targets.to(device)\n # Reset gradients\n nr_samples += inputs.size(0)\n optimizer.zero_grad()\n outputs = model(inputs)\n # Calculate loss, do backward pass, and update weights\n loss = loss_fn(outputs, targets)\n loss.backward()\n optimizer.step()\n\n correct_predictions += outputs.argmax(dim=1).eq(targets.argmax(dim=1)).sum().item()\n running_loss += loss.item()\n\n if i % print_stats_at == 0 and update > 0: # print every 2000 mini-batches\n print('[%d, %5d] loss: %.3f' %\n (update + 1, i + 1, running_loss / print_stats_at))\n\n print('[%d, %5d] acc: %.3f' %\n (update + 1, i + 1, 100 * (correct_predictions / nr_samples)))\n\n writer.add_scalar(tag=\"training/loss\",\n scalar_value=running_loss / print_stats_at,\n global_step=update)\n writer.add_scalar(tag=\"training/acc\",\n scalar_value=100 * (correct_predictions / nr_samples),\n global_step=update)\n\n running_loss = 0.0\n correct_predictions = 0.0\n nr_samples = 0\n\n if i % validate_at == 0 and update > 0:\n val_acc, val_loss = evaluate(model, valloader, loss_fn)\n\n print('[%d, %5d] val acc: %.3f' % (update + 1, i + 1, val_acc))\n print('[%d, %5d] val loss: %.3f' % (update + 1, i + 1, val_loss))\n print(\"------------------------------\")\n writer.add_scalar(tag=\"validation/loss\", scalar_value=val_loss, global_step=update)\n writer.add_scalar(tag=\"validation/acc\", scalar_value=val_acc, global_step=update)\n\n # Save best model for early stopping\n if best_validation_loss > val_loss:\n best_validation_loss = val_loss\n torch.save(model, os.path.join(results_path, f'best_{MODEL_PATH}'))\n\n update_progess_bar.set_description(f\"avg_loss: {running_loss:7.5f}\", refresh=True)\n update_progess_bar.update()\n\n update += 1\n if update >= epochs:\n break\n\n writer.flush() #method to make sure that all pending events have been written to disk.\n writer.close()\n update_progess_bar.close()\n print(f\"{datetime.now()}: Finished Training!\")\n</code></pre>\n<p><div class=\"lightbox-wrapper\"><a class=\"lightbox\" href=\"https://discuss.pytorch.org/uploads/default/original/3X/5/6/56201d8b191ef6f9213f1fd128985e66bc869b93.png\" data-download-href=\"https://discuss.pytorch.org/uploads/default/56201d8b191ef6f9213f1fd128985e66bc869b93\" title=\"grafik\"><img src=\"https://discuss.pytorch.org/uploads/default/optimized/3X/5/6/56201d8b191ef6f9213f1fd128985e66bc869b93_2_690x414.png\" alt=\"grafik\" data-base62-sha1=\"chTRTWScrVbpWevHThIDeegzVqX\" width=\"690\" height=\"414\" srcset=\"https://discuss.pytorch.org/uploads/default/optimized/3X/5/6/56201d8b191ef6f9213f1fd128985e66bc869b93_2_690x414.png, https://discuss.pytorch.org/uploads/default/optimized/3X/5/6/56201d8b191ef6f9213f1fd128985e66bc869b93_2_1035x621.png 1.5x, https://discuss.pytorch.org/uploads/default/optimized/3X/5/6/56201d8b191ef6f9213f1fd128985e66bc869b93_2_1380x828.png 2x\" data-dominant-color=\"F8EBE7\"><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\">grafik</span><span class=\"informations\">1422×854 82.4 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>",812 "post_number": 1,813 "post_type": 1,814 "posts_count": 4,815 "updated_at": "2021-10-06T14:12:27.870Z",816 "reply_count": 0,817 "reply_to_post_number": null,818 "quote_count": 0,819 "incoming_link_count": 147,820 "reads": 5,821 "readers_count": 4,822 "score": 736.0,823 "yours": false,824 "topic_id": 133604,825 "topic_slug": "cnn-validation-acc-staying-constant",826 "display_username": "Pixel",827 "primary_group_name": null,828 "flair_name": null,829 "flair_url": null,830 "flair_bg_color": null,831 "flair_color": null,832 "flair_group_id": null,833 "badges_granted": [],834 "version": 1,835 "can_edit": false,836 "can_delete": false,837 "can_recover": false,838 "can_see_hidden_post": false,839 "can_wiki": false,840 "link_counts": [841 {842 "url": "https://discuss.pytorch.org/uploads/default/original/3X/5/6/56201d8b191ef6f9213f1fd128985e66bc869b93.png",843 "internal": true,844 "reflection": false,845 "clicks": 0846 }847 ],848 "read": true,849 "user_title": null,850 "bookmarked": false,851 "actions_summary": [],852 "moderator": false,853 "admin": false,854 "staff": false,855 "user_id": 49586,856 "hidden": false,857 "trust_level": 1,858 "deleted_at": null,859 "user_deleted": false,860 "edit_reason": null,861 "can_view_edit_history": true,862 "wiki": false,863 "post_url": "/t/cnn-validation-acc-staying-constant/133604/1",864 "can_accept_answer": false,865 "can_unaccept_answer": false,866 "accepted_answer": false,867 "topic_accepted_answer": null,868 "can_vote": false869 },870 {871 "id": 310623,872 "name": "",873 "username": "ptrblck",874 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",875 "created_at": "2021-10-08T05:56:22.542Z",876 "cooked": "<p><code>nn.MSELoss</code> and a <code>softmax</code> output look a bit strange. Could you describe your use case a bit more, please?<br>\nI assume the face expression recognition project is a multi-class classification, so I would expect to see <code>nn.CrossEntropyLoss</code> as the criterion.</p>",877 "post_number": 2,878 "post_type": 1,879 "posts_count": 4,880 "updated_at": "2021-10-08T05:56:22.542Z",881 "reply_count": 1,882 "reply_to_post_number": null,883 "quote_count": 0,884 "incoming_link_count": 1,885 "reads": 5,886 "readers_count": 4,887 "score": 11.0,888 "yours": false,889 "topic_id": 133604,890 "topic_slug": "cnn-validation-acc-staying-constant",891 "display_username": "",892 "primary_group_name": null,893 "flair_name": null,894 "flair_url": null,895 "flair_bg_color": null,896 "flair_color": null,897 "flair_group_id": null,898 "badges_granted": [],899 "version": 1,900 "can_edit": false,901 "can_delete": false,902 "can_recover": false,903 "can_see_hidden_post": false,904 "can_wiki": false,905 "read": true,906 "user_title": "",907 "bookmarked": false,908 "actions_summary": [],909 "moderator": true,910 "admin": true,911 "staff": true,912 "user_id": 3534,913 "hidden": false,914 "trust_level": 2,915 "deleted_at": null,916 "user_deleted": false,917 "edit_reason": null,918 "can_view_edit_history": true,919 "wiki": false,920 "post_url": "/t/cnn-validation-acc-staying-constant/133604/2",921 "can_accept_answer": false,922 "can_unaccept_answer": false,923 "accepted_answer": false,924 "topic_accepted_answer": null925 },926 {927 "id": 310672,928 "name": "Pixel",929 "username": "Pixel",930 "avatar_template": "/letter_avatar_proxy/v4/letter/p/f14d63/{size}.png",931 "created_at": "2021-10-08T09:03:28.609Z",932 "cooked": "<p>Thanks for your reply… yes it is a multi-class classification…i want the model to give me a probability value for each emotion…okey i didnt know that MSELoss is not optimal for my case…i will try it with CrossEntropyLoss now…thank you for the note! <img src=\"https://discuss.pytorch.org/images/emoji/apple/+1.png?v=10\" title=\":+1:\" class=\"emoji\" alt=\":+1:\"></p>",933 "post_number": 3,934 "post_type": 1,935 "posts_count": 4,936 "updated_at": "2021-10-08T09:03:28.609Z",937 "reply_count": 1,938 "reply_to_post_number": 2,939 "quote_count": 0,940 "incoming_link_count": 0,941 "reads": 5,942 "readers_count": 4,943 "score": 6.0,944 "yours": false,945 "topic_id": 133604,946 "topic_slug": "cnn-validation-acc-staying-constant",947 "display_username": "Pixel",948 "primary_group_name": null,949 "flair_name": null,950 "flair_url": null,951 "flair_bg_color": null,952 "flair_color": null,953 "flair_group_id": null,954 "badges_granted": [],955 "version": 1,956 "can_edit": false,957 "can_delete": false,958 "can_recover": false,959 "can_see_hidden_post": false,960 "can_wiki": false,961 "read": true,962 "user_title": null,963 "reply_to_user": {964 "id": 3534,965 "username": "ptrblck",966 "name": "",967 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png"968 },969 "bookmarked": false,970 "actions_summary": [],971 "moderator": false,972 "admin": false,973 "staff": false,974 "user_id": 49586,975 "hidden": false,976 "trust_level": 1,977 "deleted_at": null,978 "user_deleted": false,979 "edit_reason": null,980 "can_view_edit_history": true,981 "wiki": false,982 "post_url": "/t/cnn-validation-acc-staying-constant/133604/3",983 "can_accept_answer": false,984 "can_unaccept_answer": false,985 "accepted_answer": false,986 "topic_accepted_answer": null987 },988 {989 "id": 310743,990 "name": "",991 "username": "ptrblck",992 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",993 "created_at": "2021-10-08T18:40:11.937Z",994 "cooked": "<p>Make sure to pass the logits to <code>nn.CrossEntropyLoss</code>, as internally <code>F.log_softmax</code> and <code>F.nll_loss</code> will be used. If you want to get the probabilities (e.g. to print them) you can use <code>F.softmax(output, dim=1)</code>, but don’t pass them to the criterion.</p>",995 "post_number": 4,996 "post_type": 1,997 "posts_count": 4,998 "updated_at": "2021-10-08T18:40:11.937Z",999 "reply_count": 0,1000 "reply_to_post_number": 3,1001 "quote_count": 0,1002 "incoming_link_count": 3,1003 "reads": 5,1004 "readers_count": 4,1005 "score": 16.0,1006 "yours": false,1007 "topic_id": 133604,1008 "topic_slug": "cnn-validation-acc-staying-constant",1009 "display_username": "",1010 "primary_group_name": null,1011 "flair_name": null,1012 "flair_url": null,1013 "flair_bg_color": null,1014 "flair_color": null,1015 "flair_group_id": null,1016 "badges_granted": [],1017 "version": 1,1018 "can_edit": false,1019 "can_delete": false,1020 "can_recover": false,1021 "can_see_hidden_post": false,1022 "can_wiki": false,1023 "read": true,1024 "user_title": "",1025 "reply_to_user": {1026 "id": 49586,1027 "username": "Pixel",1028 "name": "Pixel",1029 "avatar_template": "/letter_avatar_proxy/v4/letter/p/f14d63/{size}.png"1030 },1031 "bookmarked": false,1032 "actions_summary": [],1033 "moderator": true,1034 "admin": true,1035 "staff": true,1036 "user_id": 3534,1037 "hidden": false,1038 "trust_level": 2,1039 "deleted_at": null,1040 "user_deleted": false,1041 "edit_reason": null,1042 "can_view_edit_history": true,1043 "wiki": false,1044 "post_url": "/t/cnn-validation-acc-staying-constant/133604/4",1045 "can_accept_answer": false,1046 "can_unaccept_answer": false,1047 "accepted_answer": false,1048 "topic_accepted_answer": null1049 }1050 ],1051 "stream": [1052 310380,1053 310623,1054 310672,1055 3107431056 ]1057 },1058 "timeline_lookup": [1059 [1060 1,1061 14801062 ],1063 [1064 2,1065 14791066 ],1067 [1068 4,1069 14781070 ]1071 ],1072 "suggested_topics": [1073 {1074 "fancy_title": "Loading donut transformers model getting error",1075 "id": 217284,1076 "title": "Loading donut transformers model getting error",1077 "slug": "loading-donut-transformers-model-getting-error",1078 "posts_count": 3,1079 "reply_count": 1,1080 "highest_post_number": 3,1081 "image_url": null,1082 "created_at": "2025-02-28T11:32:05.605Z",1083 "last_posted_at": "2025-03-07T10:10:12.914Z",1084 "bumped": true,1085 "bumped_at": "2025-03-07T10:10:12.914Z",1086 "archetype": "regular",1087 "unseen": false,1088 "pinned": false,1089 "unpinned": null,1090 "visible": true,1091 "closed": false,1092 "archived": false,1093 "bookmarked": null,1094 "liked": null,1095 "tags_descriptions": {},1096 "like_count": 0,1097 "views": 64,1098 "category_id": 5,1099 "featured_link": null,1100 "has_accepted_answer": true,1101 "posters": [1102 {1103 "extras": "latest",1104 "description": "Original Poster, Most Recent Poster",1105 "user": {1106 "id": 82989,1107 "username": "ankita1",1108 "name": "ankita",1109 "avatar_template": "/letter_avatar_proxy/v4/letter/a/d07c76/{size}.png",1110 "trust_level": 11111 }1112 },1113 {1114 "extras": null,1115 "description": "Frequent Poster, Accepted Answer",1116 "user": {1117 "id": 3534,1118 "username": "ptrblck",1119 "name": "",1120 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",1121 "admin": true,1122 "moderator": true,1123 "trust_level": 21124 }1125 }1126 ]1127 },1128 {1129 "fancy_title": "Error:empty() received an invalid combination of arguments - got (tuple, dtype=NoneType, device=NoneType), but expected one of: * (tuple of ints size, *, tuple of names names, torch.memory_format memory_format, torch.dtype dtype, torch.layout layout, tor",1130 "id": 214194,1131 "title": "Error:empty() received an invalid combination of arguments - got (tuple, dtype=NoneType, device=NoneType), but expected one of: * (tuple of ints size, *, tuple of names names, torch.memory_format memory_format, torch.dtype dtype, torch.layout layout, tor",1132 "slug": "error-empty-received-an-invalid-combination-of-arguments-got-tuple-dtype-nonetype-device-nonetype-but-expected-one-of-tuple-of-ints-size-tuple-of-names-names-torch-memory-format-memory-format-torch-dtype-dtype-torch-layout-layout-tor",1133 "posts_count": 2,1134 "reply_count": 0,1135 "highest_post_number": 2,1136 "image_url": null,1137 "created_at": "2024-12-13T12:16:20.269Z",1138 "last_posted_at": "2024-12-13T19:10:35.709Z",1139 "bumped": true,1140 "bumped_at": "2024-12-13T19:10:35.709Z",1141 "archetype": "regular",1142 "unseen": false,1143 "pinned": false,1144 "unpinned": null,1145 "visible": true,1146 "closed": false,1147 "archived": false,1148 "bookmarked": null,1149 "liked": null,1150 "tags_descriptions": {},1151 "like_count": 0,1152 "views": 122,1153 "category_id": 5,1154 "featured_link": null,1155 "has_accepted_answer": false,1156 "posters": [1157 {1158 "extras": null,1159 "description": "Original Poster",1160 "user": {1161 "id": 81488,1162 "username": "Andisha",1163 "name": "andisha dyah",1164 "avatar_template": "/letter_avatar_proxy/v4/letter/a/7ab992/{size}.png",1165 "trust_level": 01166 }1167 },1168 {1169 "extras": "latest",1170 "description": "Most Recent Poster",1171 "user": {1172 "id": 211,1173 "username": "albanD",1174 "name": "Alban D",1175 "avatar_template": "/user_avatar/discuss.pytorch.org/alband/{size}/215_2.png",1176 "admin": true,1177 "moderator": true,1178 "trust_level": 41179 }1180 }1181 ]1182 },1183 {1184 "fancy_title": "Yolo Neural Layer",1185 "id": 212691,1186 "title": "Yolo Neural Layer",1187 "slug": "yolo-neural-layer",1188 "posts_count": 1,1189 "reply_count": 0,1190 "highest_post_number": 1,1191 "image_url": null,1192 "created_at": "2024-11-08T08:05:48.745Z",1193 "last_posted_at": "2024-11-08T08:05:48.792Z",1194 "bumped": true,1195 "bumped_at": "2024-11-08T08:05:48.792Z",1196 "archetype": "regular",1197 "unseen": false,1198 "pinned": false,1199 "unpinned": null,1200 "visible": true,