Anurag1734/cuda-error-resolution-analysis
07
1[2 {3 "post_stream": {4 "posts": [5 {6 "id": 406384,7 "name": "Elias Vansteenkiste",8 "username": "Elias_Vansteenkiste",9 "avatar_template": "/user_avatar/discuss.pytorch.org/elias_vansteenkiste/{size}/836_2.png",10 "created_at": "2023-06-17T10:41:14.907Z",11 "cooked": "<p>According to the <a href=\"https://pytorch.org/docs/stable/notes/mps.html\" rel=\"noopener nofollow ugc\">docs</a>, MPS backend is using the GPU on M1, M2 chips via metal compute shaders.</p>\n<blockquote>\n<p><code>mps</code> device enables high-performance training on GPU for MacOS devices with Metal programming framework. It introduces a new device to map Machine Learning computational graphs and primitives on highly efficient Metal Performance Shaders Graph framework and tuned kernels provided by Metal Performance Shaders framework respectively.</p>\n<p>The new MPS backend extends the PyTorch ecosystem and provides existing scripts capabilities to setup and run operations on GPU.</p>\n</blockquote>\n<p>According to the following repository, <a href=\"https://github.com/woolfel/ml-macos-performance/blob/main/results.md\" rel=\"noopener nofollow ugc\">ml-macos-performanc</a>, inference on the ANE is 7x faster</p>\n<pre><code class=\"lang-auto\">densenet121_keras_applications Latency ANE : 0.0012743692083333827 RPS ANE : 784.7019478034924\n\ndensenet121_keras_applications Latency GPU : 0.008270947500000033 RPS GPU : 120.90513209036763\n\ndensenet121_keras_applications Latency CPU : 0.015347813229166719 RPS CPU : 65.15586195039286\n</code></pre>\n<p>I was wondering would the performance of training be better if we can also use the Apple’s Neural Engine?</p>\n<p>There is obviously some restrictions, see <a href=\"https://github.com/hollance/neural-engine/blob/master/docs/unsupported-layers.md\" rel=\"noopener nofollow ugc\">unsupported neural engine layers</a>, but that should be similar to Google’s TPUs.</p>\n<p>It seems that there is maybe too much protection/security around directly calling the ANE,<br>\naccording to <a href=\"https://github.com/geohot/tinygrad/tree/master/accel/ane\" rel=\"noopener nofollow ugc\">George Hotz’s tinygrad</a>, maybe that is the biggest blocker?</p>",12 "post_number": 1,13 "post_type": 1,14 "posts_count": 3,15 "updated_at": "2023-06-17T10:45:13.027Z",16 "reply_count": 0,17 "reply_to_post_number": null,18 "quote_count": 0,19 "incoming_link_count": 9319,20 "reads": 45,21 "readers_count": 44,22 "score": 46484.0,23 "yours": false,24 "topic_id": 182297,25 "topic_slug": "apple-neural-engine-ane-instead-of-additionally-to-gpu-on-m1-m2-chips",26 "display_username": "Elias Vansteenkiste",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://github.com/woolfel/ml-macos-performance/blob/main/results.md",43 "internal": false,44 "reflection": false,45 "title": "ml-macos-performance/results.md at main · woolfel/ml-macos-performance · GitHub",46 "clicks": 33047 },48 {49 "url": "https://github.com/hollance/neural-engine/blob/master/docs/unsupported-layers.md",50 "internal": false,51 "reflection": false,52 "title": "neural-engine/docs/unsupported-layers.md at master · hollance/neural-engine · GitHub",53 "clicks": 23954 },55 {56 "url": "https://github.com/geohot/tinygrad/tree/master/accel/ane",57 "internal": false,58 "reflection": false,59 "title": "tinygrad/accel/ane at master · geohot/tinygrad · GitHub",60 "clicks": 17761 },62 {63 "url": "https://pytorch.org/docs/stable/notes/mps.html",64 "internal": false,65 "reflection": false,66 "title": "MPS backend — PyTorch 2.0 documentation",67 "clicks": 12268 }69 ],70 "read": true,71 "user_title": null,72 "bookmarked": false,73 "actions_summary": [74 {75 "id": 2,76 "count": 377 }78 ],79 "moderator": false,80 "admin": false,81 "staff": false,82 "user_id": 1989,83 "hidden": false,84 "trust_level": 2,85 "deleted_at": null,86 "user_deleted": false,87 "edit_reason": null,88 "can_view_edit_history": true,89 "wiki": false,90 "post_url": "/t/apple-neural-engine-ane-instead-of-additionally-to-gpu-on-m1-m2-chips/182297/1",91 "can_accept_answer": false,92 "can_unaccept_answer": false,93 "accepted_answer": false,94 "topic_accepted_answer": null,95 "can_vote": false96 },97 {98 "id": 415336,99 "name": "",100 "username": "sujantkumarkv",101 "avatar_template": "/letter_avatar_proxy/v4/letter/s/919ad9/{size}.png",102 "created_at": "2023-08-26T09:25:50.365Z",103 "cooked": "<p>Hi,<br>\nthanks for the writeup; btw the tinygrad’s link gives a 404 <img src=\"https://discuss.pytorch.org/images/emoji/apple/sweat_smile.png?v=12\" title=\":sweat_smile:\" class=\"emoji\" alt=\":sweat_smile:\" loading=\"lazy\" width=\"20\" height=\"20\"></p>\n<p>I have been thinking to apply FlashAttention for faster training locally on macbooks but it currently only supports cuda plus MPS is less mature with implementations afaik.</p>\n<p>The project is in ideation stages, <a href=\"https://github.com/Lightning-AI/lit-gpt/issues/429\" rel=\"noopener nofollow ugc\">here</a>.</p>\n<p>I don’t have all the answers ofcourse, and this will be an opensource collaborative attempt. I’m researching what are the missing pieces I need to look for.</p>\n<p>The goal is clear: “<strong>Make training faster on macbooks with Flash Attention</strong>” and may need various pieces for that: MPS, Pytorch, ANE etc.</p>\n<p>I appreciate absolutely any help/comments/inputs on this from the community.</p>",104 "post_number": 2,105 "post_type": 1,106 "posts_count": 3,107 "updated_at": "2023-08-26T09:27:44.474Z",108 "reply_count": 0,109 "reply_to_post_number": null,110 "quote_count": 0,111 "incoming_link_count": 105,112 "reads": 35,113 "readers_count": 34,114 "score": 532.0,115 "yours": false,116 "topic_id": 182297,117 "topic_slug": "apple-neural-engine-ane-instead-of-additionally-to-gpu-on-m1-m2-chips",118 "display_username": "",119 "primary_group_name": null,120 "flair_name": null,121 "flair_url": null,122 "flair_bg_color": null,123 "flair_color": null,124 "flair_group_id": null,125 "badges_granted": [],126 "version": 2,127 "can_edit": false,128 "can_delete": false,129 "can_recover": false,130 "can_see_hidden_post": false,131 "can_wiki": false,132 "link_counts": [133 {134 "url": "https://github.com/Lightning-AI/lit-gpt/issues/429",135 "internal": false,136 "reflection": false,137 "title": "Explore Flash Attention Like Feature for MPS Backend · Issue #429 · Lightning-AI/lit-gpt · GitHub",138 "clicks": 264139 }140 ],141 "read": true,142 "user_title": null,143 "bookmarked": false,144 "actions_summary": [],145 "moderator": false,146 "admin": false,147 "staff": false,148 "user_id": 69014,149 "hidden": false,150 "trust_level": 0,151 "deleted_at": null,152 "user_deleted": false,153 "edit_reason": null,154 "can_view_edit_history": true,155 "wiki": false,156 "post_url": "/t/apple-neural-engine-ane-instead-of-additionally-to-gpu-on-m1-m2-chips/182297/2",157 "can_accept_answer": false,158 "can_unaccept_answer": false,159 "accepted_answer": false,160 "topic_accepted_answer": null161 },162 {163 "id": 427971,164 "name": "pypdeveloper",165 "username": "pypdeveloper",166 "avatar_template": "/user_avatar/discuss.pytorch.org/pypdeveloper/{size}/61681_2.png",167 "created_at": "2023-12-22T10:45:19.848Z",168 "cooked": "<p>It seems Apples new ML framework MLX also doesn’t support the use of the ANE for inference, maybe they are still trying to workout the API, for python and C applications? Also is it possible to use the ANE for training?</p>",169 "post_number": 3,170 "post_type": 1,171 "posts_count": 3,172 "updated_at": "2023-12-22T10:45:45.666Z",173 "reply_count": 0,174 "reply_to_post_number": null,175 "quote_count": 0,176 "incoming_link_count": 49,177 "reads": 24,178 "readers_count": 23,179 "score": 249.8,180 "yours": false,181 "topic_id": 182297,182 "topic_slug": "apple-neural-engine-ane-instead-of-additionally-to-gpu-on-m1-m2-chips",183 "display_username": "pypdeveloper",184 "primary_group_name": null,185 "flair_name": null,186 "flair_url": null,187 "flair_bg_color": null,188 "flair_color": null,189 "flair_group_id": null,190 "badges_granted": [],191 "version": 1,192 "can_edit": false,193 "can_delete": false,194 "can_recover": false,195 "can_see_hidden_post": false,196 "can_wiki": false,197 "read": true,198 "user_title": null,199 "bookmarked": false,200 "actions_summary": [],201 "moderator": false,202 "admin": false,203 "staff": false,204 "user_id": 67347,205 "hidden": false,206 "trust_level": 1,207 "deleted_at": null,208 "user_deleted": false,209 "edit_reason": null,210 "can_view_edit_history": true,211 "wiki": false,212 "post_url": "/t/apple-neural-engine-ane-instead-of-additionally-to-gpu-on-m1-m2-chips/182297/3",213 "can_accept_answer": false,214 "can_unaccept_answer": false,215 "accepted_answer": false,216 "topic_accepted_answer": null217 }218 ],219 "stream": [220 406384,221 415336,222 427971223 ]224 },225 "timeline_lookup": [226 [227 1,228 861229 ],230 [231 2,232 791233 ],234 [235 3,236 673237 ]238 ],239 "suggested_topics": [240 {241 "fancy_title": "Having trouble getting the mps version of torch to install",242 "id": 216103,243 "title": "Having trouble getting the mps version of torch to install",244 "slug": "having-trouble-getting-the-mps-version-of-torch-to-install",245 "posts_count": 2,246 "reply_count": 0,247 "highest_post_number": 2,248 "image_url": null,249 "created_at": "2025-01-31T19:04:49.418Z",250 "last_posted_at": "2025-02-28T16:01:18.379Z",251 "bumped": true,252 "bumped_at": "2025-02-28T16:01:18.379Z",253 "archetype": "regular",254 "unseen": false,255 "pinned": false,256 "unpinned": null,257 "visible": true,258 "closed": false,259 "archived": false,260 "bookmarked": null,261 "liked": null,262 "tags_descriptions": {},263 "like_count": 0,264 "views": 707,265 "category_id": 32,266 "featured_link": null,267 "has_accepted_answer": false,268 "posters": [269 {270 "extras": null,271 "description": "Original Poster",272 "user": {273 "id": 82422,274 "username": "voxvox142",275 "name": "",276 "avatar_template": "/user_avatar/discuss.pytorch.org/voxvox142/{size}/74704_2.png",277 "trust_level": 0278 }279 },280 {281 "extras": "latest",282 "description": "Most Recent Poster",283 "user": {284 "id": 76547,285 "username": "roach",286 "name": "",287 "avatar_template": "/user_avatar/discuss.pytorch.org/roach/{size}/61570_2.png",288 "trust_level": 1289 }290 }291 ]292 },293 {294 "fancy_title": "Building PyTorch 2.6.0 from source on macOS x86_64 (Intel)",295 "id": 216283,296 "title": "Building PyTorch 2.6.0 from source on macOS x86_64 (Intel)",297 "slug": "building-pytorch-2-6-0-from-source-on-macos-x86-64-intel",298 "posts_count": 7,299 "reply_count": 2,300 "highest_post_number": 7,301 "image_url": null,302 "created_at": "2025-02-05T17:07:14.219Z",303 "last_posted_at": "2025-09-24T15:35:05.044Z",304 "bumped": true,305 "bumped_at": "2025-09-24T15:35:05.044Z",306 "archetype": "regular",307 "unseen": false,308 "pinned": false,309 "unpinned": null,310 "visible": true,311 "closed": false,312 "archived": false,313 "bookmarked": null,314 "liked": null,315 "tags_descriptions": {},316 "like_count": 3,317 "views": 3877,318 "category_id": 32,319 "featured_link": null,320 "has_accepted_answer": false,321 "posters": [322 {323 "extras": null,324 "description": "Original Poster",325 "user": {326 "id": 82511,327 "username": "uwu-420",328 "name": "Uwu 420",329 "avatar_template": "/user_avatar/discuss.pytorch.org/uwu-420/{size}/75489_2.png",330 "trust_level": 0331 }332 },333 {334 "extras": null,335 "description": "Frequent Poster",336 "user": {337 "id": 64351,338 "username": "aisven",339 "name": "",340 "avatar_template": "/user_avatar/discuss.pytorch.org/aisven/{size}/58504_2.png",341 "trust_level": 1342 }343 },344 {345 "extras": "latest",346 "description": "Most Recent Poster",347 "user": {348 "id": 85615,349 "username": "Casey_KimScripts",350 "name": "",351 "avatar_template": "/user_avatar/discuss.pytorch.org/casey_kimscripts/{size}/78035_2.png",352 "trust_level": 1353 }354 }355 ]356 },357 {358 "fancy_title": "MAC inter processors cannot install torch2.5.1",359 "id": 215100,360 "title": "MAC inter processors cannot install torch2.5.1",361 "slug": "mac-inter-processors-cannot-install-torch2-5-1",362 "posts_count": 10,363 "reply_count": 6,364 "highest_post_number": 10,365 "image_url": null,366 "created_at": "2025-01-08T06:02:41.764Z",367 "last_posted_at": "2025-05-31T06:28:44.746Z",368 "bumped": true,369 "bumped_at": "2025-05-31T06:28:44.746Z",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": 5,381 "views": 2122,382 "category_id": 32,383 "featured_link": null,384 "has_accepted_answer": false,385 "posters": [386 {387 "extras": null,388 "description": "Original Poster",389 "user": {390 "id": 81936,391 "username": "tmac-weylin",392 "name": "Yuezhipeng",393 "avatar_template": "/user_avatar/discuss.pytorch.org/tmac-weylin/{size}/74959_2.png",394 "trust_level": 0395 }396 },397 {398 "extras": null,399 "description": "Frequent Poster",400 "user": {401 "id": 82103,402 "username": "kannangce",403 "name": "Kannan Ramamoorthy",404 "avatar_template": "/user_avatar/discuss.pytorch.org/kannangce/{size}/75115_2.png",405 "trust_level": 0406 }407 },408 {409 "extras": null,410 "description": "Frequent Poster",411 "user": {412 "id": 81170,413 "username": "Naeem_Khoshnevis",414 "name": "Naeem Khoshnevis",415 "avatar_template": "/user_avatar/discuss.pytorch.org/naeem_khoshnevis/{size}/74236_2.png",416 "trust_level": 1417 }418 },419 {420 "extras": null,421 "description": "Frequent Poster",422 "user": {423 "id": 3534,424 "username": "ptrblck",425 "name": "",426 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",427 "admin": true,428 "moderator": true,429 "trust_level": 2430 }431 },432 {433 "extras": "latest",434 "description": "Most Recent Poster",435 "user": {436 "id": 84527,437 "username": "Jing-Jing_Li",438 "name": "Jing-Jing Li",439 "avatar_template": "/user_avatar/discuss.pytorch.org/jing-jing_li/{size}/77224_2.png",440 "trust_level": 0441 }442 }443 ]444 },445 {446 "fancy_title": "Cuda installation on M3",447 "id": 215620,448 "title": "Cuda installation on M3",449 "slug": "cuda-installation-on-m3",450 "posts_count": 11,451 "reply_count": 8,452 "highest_post_number": 11,453 "image_url": null,454 "created_at": "2025-01-20T11:23:16.903Z",455 "last_posted_at": "2025-01-20T23:48:31.539Z",456 "bumped": true,457 "bumped_at": "2025-01-20T23:48:31.539Z",458 "archetype": "regular",459 "unseen": false,460 "pinned": false,461 "unpinned": null,462 "visible": true,463 "closed": false,464 "archived": false,465 "bookmarked": null,466 "liked": null,467 "tags_descriptions": {},468 "like_count": 0,469 "views": 2414,470 "category_id": 32,471 "featured_link": null,472 "has_accepted_answer": false,473 "posters": [474 {475 "extras": null,476 "description": "Original Poster",477 "user": {478 "id": 82193,479 "username": "djdoogle123",480 "name": "Djdoogle123",481 "avatar_template": "/user_avatar/discuss.pytorch.org/djdoogle123/{size}/75207_2.png",482 "trust_level": 1483 }484 },485 {486 "extras": "latest",487 "description": "Most Recent Poster",488 "user": {489 "id": 3534,490 "username": "ptrblck",491 "name": "",492 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",493 "admin": true,494 "moderator": true,495 "trust_level": 2496 }497 }498 ]499 },500 {501 "fancy_title": "Segmentation error on MacOS because of OpenMP",502 "id": 221141,503 "title": "Segmentation error on MacOS because of OpenMP",504 "slug": "segmentation-error-on-macos-because-of-openmp",505 "posts_count": 1,506 "reply_count": 0,507 "highest_post_number": 1,508 "image_url": null,509 "created_at": "2025-06-29T18:23:50.152Z",510 "last_posted_at": "2025-06-29T18:23:50.199Z",511 "bumped": true,512 "bumped_at": "2025-06-29T18:23:50.199Z",513 "archetype": "regular",514 "unseen": false,515 "pinned": false,516 "unpinned": null,517 "visible": true,518 "closed": false,519 "archived": false,520 "bookmarked": null,521 "liked": null,522 "tags_descriptions": {},523 "like_count": 0,524 "views": 74,525 "category_id": 32,526 "featured_link": null,527 "has_accepted_answer": false,528 "posters": [529 {530 "extras": "latest single",531 "description": "Original Poster, Most Recent Poster",532 "user": {533 "id": 84874,534 "username": "lorentzenchr",535 "name": "Christian Lorentzen",536 "avatar_template": "/user_avatar/discuss.pytorch.org/lorentzenchr/{size}/77530_2.png",537 "trust_level": 0538 }539 }540 ]541 }542 ],543 "tags_descriptions": {},544 "fancy_title": "Apple Neural Engine (ANE) instead of / additionally to GPU on M1, M2 chips",545 "id": 182297,546 "title": "Apple Neural Engine (ANE) instead of / additionally to GPU on M1, M2 chips",547 "posts_count": 3,548 "created_at": "2023-06-17T10:41:14.807Z",549 "views": 8518,550 "reply_count": 0,551 "like_count": 3,552 "last_posted_at": "2023-12-22T10:45:19.848Z",553 "visible": true,554 "closed": false,555 "archived": false,556 "has_summary": false,557 "archetype": "regular",558 "slug": "apple-neural-engine-ane-instead-of-additionally-to-gpu-on-m1-m2-chips",559 "category_id": 32,560 "word_count": 395,561 "deleted_at": null,562 "user_id": 1989,563 "featured_link": null,564 "pinned_globally": false,565 "pinned_at": null,566 "pinned_until": null,567 "image_url": null,568 "slow_mode_seconds": 0,569 "draft": null,570 "draft_key": "topic_182297",571 "draft_sequence": null,572 "unpinned": null,573 "pinned": false,574 "current_post_number": 1,575 "highest_post_number": 3,576 "deleted_by": null,577 "actions_summary": [578 {579 "id": 4,580 "count": 0,581 "hidden": false,582 "can_act": false583 },584 {585 "id": 8,586 "count": 0,587 "hidden": false,588 "can_act": false589 },590 {591 "id": 10,592 "count": 0,593 "hidden": false,594 "can_act": false595 },596 {597 "id": 7,598 "count": 0,599 "hidden": false,600 "can_act": false601 }602 ],603 "chunk_size": 20,604 "bookmarked": false,605 "topic_timer": null,606 "message_bus_last_id": 0,607 "participant_count": 3,608 "show_read_indicator": false,609 "thumbnails": null,610 "slow_mode_enabled_until": null,611 "can_vote": false,612 "vote_count": 0,613 "user_voted": false,614 "discourse_zendesk_plugin_zendesk_id": null,615 "discourse_zendesk_plugin_zendesk_url": "https://your-url.zendesk.com/agent/tickets/",616 "details": {617 "can_edit": false,618 "notification_level": 1,619 "participants": [620 {621 "id": 1989,622 "username": "Elias_Vansteenkiste",623 "name": "Elias Vansteenkiste",624 "avatar_template": "/user_avatar/discuss.pytorch.org/elias_vansteenkiste/{size}/836_2.png",625 "post_count": 1,626 "primary_group_name": null,627 "flair_name": null,628 "flair_url": null,629 "flair_color": null,630 "flair_bg_color": null,631 "flair_group_id": null,632 "trust_level": 2633 },634 {635 "id": 67347,636 "username": "pypdeveloper",637 "name": "pypdeveloper",638 "avatar_template": "/user_avatar/discuss.pytorch.org/pypdeveloper/{size}/61681_2.png",639 "post_count": 1,640 "primary_group_name": null,641 "flair_name": null,642 "flair_url": null,643 "flair_color": null,644 "flair_bg_color": null,645 "flair_group_id": null,646 "trust_level": 1647 },648 {649 "id": 69014,650 "username": "sujantkumarkv",651 "name": "",652 "avatar_template": "/letter_avatar_proxy/v4/letter/s/919ad9/{size}.png",653 "post_count": 1,654 "primary_group_name": null,655 "flair_name": null,656 "flair_url": null,657 "flair_color": null,658 "flair_bg_color": null,659 "flair_group_id": null,660 "trust_level": 0661 }662 ],663 "created_by": {664 "id": 1989,665 "username": "Elias_Vansteenkiste",666 "name": "Elias Vansteenkiste",667 "avatar_template": "/user_avatar/discuss.pytorch.org/elias_vansteenkiste/{size}/836_2.png"668 },669 "last_poster": {670 "id": 67347,671 "username": "pypdeveloper",672 "name": "pypdeveloper",673 "avatar_template": "/user_avatar/discuss.pytorch.org/pypdeveloper/{size}/61681_2.png"674 },675 "links": [676 {677 "url": "https://github.com/woolfel/ml-macos-performance/blob/main/results.md",678 "title": "ml-macos-performance/results.md at main · woolfel/ml-macos-performance · GitHub",679 "internal": false,680 "attachment": false,681 "reflection": false,682 "clicks": 330,683 "user_id": 1989,684 "domain": "github.com",685 "root_domain": "github.com"686 },687 {688 "url": "https://github.com/Lightning-AI/lit-gpt/issues/429",689 "title": "Explore Flash Attention Like Feature for MPS Backend · Issue #429 · Lightning-AI/lit-gpt · GitHub",690 "internal": false,691 "attachment": false,692 "reflection": false,693 "clicks": 264,694 "user_id": 69014,695 "domain": "github.com",696 "root_domain": "github.com"697 },698 {699 "url": "https://github.com/hollance/neural-engine/blob/master/docs/unsupported-layers.md",700 "title": "neural-engine/docs/unsupported-layers.md at master · hollance/neural-engine · GitHub",701 "internal": false,702 "attachment": false,703 "reflection": false,704 "clicks": 239,705 "user_id": 1989,706 "domain": "github.com",707 "root_domain": "github.com"708 },709 {710 "url": "https://github.com/geohot/tinygrad/tree/master/accel/ane",711 "title": "tinygrad/accel/ane at master · geohot/tinygrad · GitHub",712 "internal": false,713 "attachment": false,714 "reflection": false,715 "clicks": 177,716 "user_id": 1989,717 "domain": "github.com",718 "root_domain": "github.com"719 },720 {721 "url": "https://pytorch.org/docs/stable/notes/mps.html",722 "title": "MPS backend — PyTorch 2.0 documentation",723 "internal": false,724 "attachment": false,725 "reflection": false,726 "clicks": 122,727 "user_id": 1989,728 "domain": "pytorch.org",729 "root_domain": "pytorch.org"730 }731 ]732 },733 "bookmarks": []734 },735 {736 "post_stream": {737 "posts": [738 {739 "id": 427969,740 "name": "pypdeveloper",741 "username": "pypdeveloper",742 "avatar_template": "/user_avatar/discuss.pytorch.org/pypdeveloper/{size}/61681_2.png",743 "created_at": "2023-12-22T10:41:20.573Z",744 "cooked": "<p>Hey everyone, I have an older 2015 iMac which apple has stopped providing software updates to, and I was wondering if I duel boot into windows or Linux can I use a NVIDIA GPU for ML with PyTorch or does Apple’s BIOS prevent that?</p>",745 "post_number": 1,746 "post_type": 1,747 "posts_count": 1,748 "updated_at": "2023-12-22T10:42:54.070Z",749 "reply_count": 0,750 "reply_to_post_number": null,751 "quote_count": 0,752 "incoming_link_count": 11,753 "reads": 2,754 "readers_count": 1,755 "score": 55.4,756 "yours": false,757 "topic_id": 194171,758 "topic_slug": "nvidia-gpu-on-2015-imac",759 "display_username": "pypdeveloper",760 "primary_group_name": null,761 "flair_name": null,762 "flair_url": null,763 "flair_bg_color": null,764 "flair_color": null,765 "flair_group_id": null,766 "badges_granted": [],767 "version": 1,768 "can_edit": false,769 "can_delete": false,770 "can_recover": false,771 "can_see_hidden_post": false,772 "can_wiki": false,773 "read": true,774 "user_title": null,775 "bookmarked": false,776 "actions_summary": [],777 "moderator": false,778 "admin": false,779 "staff": false,780 "user_id": 67347,781 "hidden": false,782 "trust_level": 1,783 "deleted_at": null,784 "user_deleted": false,785 "edit_reason": null,786 "can_view_edit_history": true,787 "wiki": false,788 "post_url": "/t/nvidia-gpu-on-2015-imac/194171/1",789 "can_accept_answer": false,790 "can_unaccept_answer": false,791 "accepted_answer": false,792 "topic_accepted_answer": null,793 "can_vote": false794 }795 ],796 "stream": [797 427969798 ]799 },800 "timeline_lookup": [801 [802 1,803 673804 ]805 ],806 "suggested_topics": [807 {808 "fancy_title": "Building PyTorch 2.6.0 from source on macOS x86_64 (Intel)",809 "id": 216283,810 "title": "Building PyTorch 2.6.0 from source on macOS x86_64 (Intel)",811 "slug": "building-pytorch-2-6-0-from-source-on-macos-x86-64-intel",812 "posts_count": 7,813 "reply_count": 2,814 "highest_post_number": 7,815 "image_url": null,816 "created_at": "2025-02-05T17:07:14.219Z",817 "last_posted_at": "2025-09-24T15:35:05.044Z",818 "bumped": true,819 "bumped_at": "2025-09-24T15:35:05.044Z",820 "archetype": "regular",821 "unseen": false,822 "pinned": false,823 "unpinned": null,824 "visible": true,825 "closed": false,826 "archived": false,827 "bookmarked": null,828 "liked": null,829 "tags_descriptions": {},830 "like_count": 3,831 "views": 3877,832 "category_id": 32,833 "featured_link": null,834 "has_accepted_answer": false,835 "posters": [836 {837 "extras": null,838 "description": "Original Poster",839 "user": {840 "id": 82511,841 "username": "uwu-420",842 "name": "Uwu 420",843 "avatar_template": "/user_avatar/discuss.pytorch.org/uwu-420/{size}/75489_2.png",844 "trust_level": 0845 }846 },847 {848 "extras": null,849 "description": "Frequent Poster",850 "user": {851 "id": 64351,852 "username": "aisven",853 "name": "",854 "avatar_template": "/user_avatar/discuss.pytorch.org/aisven/{size}/58504_2.png",855 "trust_level": 1856 }857 },858 {859 "extras": "latest",860 "description": "Most Recent Poster",861 "user": {862 "id": 85615,863 "username": "Casey_KimScripts",864 "name": "",865 "avatar_template": "/user_avatar/discuss.pytorch.org/casey_kimscripts/{size}/78035_2.png",866 "trust_level": 1867 }868 }869 ]870 },871 {872 "fancy_title": "PyTorch 2.6.0 installation on 2019 MacBook Intel",873 "id": 218576,874 "title": "PyTorch 2.6.0 installation on 2019 MacBook Intel",875 "slug": "pytorch-2-6-0-installation-on-2019-macbook-intel",876 "posts_count": 2,877 "reply_count": 0,878 "highest_post_number": 2,879 "image_url": null,880 "created_at": "2025-04-03T12:02:19.902Z",881 "last_posted_at": "2025-04-03T12:30:18.960Z",882 "bumped": true,883 "bumped_at": "2025-04-03T12:30:18.960Z",884 "archetype": "regular",885 "unseen": false,886 "pinned": false,887 "unpinned": null,888 "visible": true,889 "closed": false,890 "archived": false,891 "bookmarked": null,892 "liked": null,893 "tags_descriptions": {},894 "like_count": 0,895 "views": 976,896 "category_id": 32,897 "featured_link": null,898 "has_accepted_answer": false,899 "posters": [900 {901 "extras": null,902 "description": "Original Poster",903 "user": {904 "id": 83617,905 "username": "niMtAndoX",906 "name": "Max ",907 "avatar_template": "/letter_avatar_proxy/v4/letter/n/7ea924/{size}.png",908 "trust_level": 0909 }910 },911 {912 "extras": "latest",913 "description": "Most Recent Poster",914 "user": {915 "id": 3534,916 "username": "ptrblck",917 "name": "",918 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",919 "admin": true,920 "moderator": true,921 "trust_level": 2922 }923 }924 ]925 },926 {927 "fancy_title": "Cuda installation on M3",928 "id": 215620,929 "title": "Cuda installation on M3",930 "slug": "cuda-installation-on-m3",931 "posts_count": 11,932 "reply_count": 8,933 "highest_post_number": 11,934 "image_url": null,935 "created_at": "2025-01-20T11:23:16.903Z",936 "last_posted_at": "2025-01-20T23:48:31.539Z",937 "bumped": true,938 "bumped_at": "2025-01-20T23:48:31.539Z",939 "archetype": "regular",940 "unseen": false,941 "pinned": false,942 "unpinned": null,943 "visible": true,944 "closed": false,945 "archived": false,946 "bookmarked": null,947 "liked": null,948 "tags_descriptions": {},949 "like_count": 0,950 "views": 2414,951 "category_id": 32,952 "featured_link": null,953 "has_accepted_answer": false,954 "posters": [955 {956 "extras": null,957 "description": "Original Poster",958 "user": {959 "id": 82193,960 "username": "djdoogle123",961 "name": "Djdoogle123",962 "avatar_template": "/user_avatar/discuss.pytorch.org/djdoogle123/{size}/75207_2.png",963 "trust_level": 1964 }965 },966 {967 "extras": "latest",968 "description": "Most Recent Poster",969 "user": {970 "id": 3534,971 "username": "ptrblck",972 "name": "",973 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",974 "admin": true,975 "moderator": true,976 "trust_level": 2977 }978 }979 ]980 },981 {982 "fancy_title": "Having trouble getting the mps version of torch to install",983 "id": 216103,984 "title": "Having trouble getting the mps version of torch to install",985 "slug": "having-trouble-getting-the-mps-version-of-torch-to-install",986 "posts_count": 2,987 "reply_count": 0,988 "highest_post_number": 2,989 "image_url": null,990 "created_at": "2025-01-31T19:04:49.418Z",991 "last_posted_at": "2025-02-28T16:01:18.379Z",992 "bumped": true,993 "bumped_at": "2025-02-28T16:01:18.379Z",994 "archetype": "regular",995 "unseen": false,996 "pinned": false,997 "unpinned": null,998 "visible": true,999 "closed": false,1000 "archived": false,1001 "bookmarked": null,1002 "liked": null,1003 "tags_descriptions": {},1004 "like_count": 0,1005 "views": 707,1006 "category_id": 32,1007 "featured_link": null,1008 "has_accepted_answer": false,1009 "posters": [1010 {1011 "extras": null,1012 "description": "Original Poster",1013 "user": {1014 "id": 82422,1015 "username": "voxvox142",1016 "name": "",1017 "avatar_template": "/user_avatar/discuss.pytorch.org/voxvox142/{size}/74704_2.png",1018 "trust_level": 01019 }1020 },1021 {1022 "extras": "latest",1023 "description": "Most Recent Poster",1024 "user": {1025 "id": 76547,1026 "username": "roach",1027 "name": "",1028 "avatar_template": "/user_avatar/discuss.pytorch.org/roach/{size}/61570_2.png",1029 "trust_level": 11030 }1031 }1032 ]1033 },1034 {1035 "fancy_title": "PyTorch memory usage keeps increasing",1036 "id": 219923,1037 "title": "PyTorch memory usage keeps increasing",1038 "slug": "pytorch-memory-usage-keeps-increasing",1039 "posts_count": 1,1040 "reply_count": 0,1041 "highest_post_number": 1,1042 "image_url": null,1043 "created_at": "2025-05-10T14:57:55.532Z",1044 "last_posted_at": "2025-05-10T14:57:55.583Z",1045 "bumped": true,1046 "bumped_at": "2025-05-10T14:57:55.583Z",1047 "archetype": "regular",1048 "unseen": false,1049 "pinned": false,1050 "unpinned": null,1051 "visible": true,1052 "closed": false,1053 "archived": false,1054 "bookmarked": null,1055 "liked": null,1056 "tags_descriptions": {},1057 "like_count": 0,1058 "views": 76,1059 "category_id": 32,1060 "featured_link": null,1061 "has_accepted_answer": false,1062 "posters": [1063 {1064 "extras": "latest single",1065 "description": "Original Poster, Most Recent Poster",1066 "user": {1067 "id": 84224,1068 "username": "Morac",1069 "name": null,1070 "avatar_template": "/user_avatar/discuss.pytorch.org/morac/{size}/76975_2.png",1071 "trust_level": 01072 }1073 }1074 ]1075 }1076 ],1077 "tags_descriptions": {},1078 "fancy_title": "NVIDIA GPU on 2015 iMac",1079 "id": 194171,1080 "title": "NVIDIA GPU on 2015 iMac",1081 "posts_count": 1,1082 "created_at": "2023-12-22T10:41:20.493Z",1083 "views": 359,1084 "reply_count": 0,1085 "like_count": 0,1086 "last_posted_at": "2023-12-22T10:41:20.573Z",1087 "visible": true,1088 "closed": false,1089 "archived": false,1090 "has_summary": false,1091 "archetype": "regular",1092 "slug": "nvidia-gpu-on-2015-imac",1093 "category_id": 32,1094 "word_count": 45,1095 "deleted_at": null,1096 "user_id": 67347,1097 "featured_link": null,1098 "pinned_globally": false,1099 "pinned_at": null,1100 "pinned_until": null,1101 "image_url": null,1102 "slow_mode_seconds": 0,1103 "draft": null,1104 "draft_key": "topic_194171",1105 "draft_sequence": null,1106 "unpinned": null,1107 "pinned": false,1108 "current_post_number": 1,1109 "highest_post_number": 1,1110 "deleted_by": null,1111 "actions_summary": [1112 {1113 "id": 4,1114 "count": 0,1115 "hidden": false,1116 "can_act": false1117 },1118 {1119 "id": 8,1120 "count": 0,1121 "hidden": false,1122 "can_act": false1123 },1124 {1125 "id": 10,1126 "count": 0,1127 "hidden": false,1128 "can_act": false1129 },1130 {1131 "id": 7,1132 "count": 0,1133 "hidden": false,1134 "can_act": false1135 }1136 ],1137 "chunk_size": 20,1138 "bookmarked": false,1139 "topic_timer": null,1140 "message_bus_last_id": 0,1141 "participant_count": 1,1142 "show_read_indicator": false,1143 "thumbnails": null,1144 "slow_mode_enabled_until": null,1145 "can_vote": false,1146 "vote_count": 0,1147 "user_voted": false,1148 "discourse_zendesk_plugin_zendesk_id": null,1149 "discourse_zendesk_plugin_zendesk_url": "https://your-url.zendesk.com/agent/tickets/",1150 "details": {1151 "can_edit": false,1152 "notification_level": 1,1153 "participants": [1154 {1155 "id": 67347,1156 "username": "pypdeveloper",1157 "name": "pypdeveloper",1158 "avatar_template": "/user_avatar/discuss.pytorch.org/pypdeveloper/{size}/61681_2.png",1159 "post_count": 1,1160 "primary_group_name": null,1161 "flair_name": null,1162 "flair_url": null,1163 "flair_color": null,1164 "flair_bg_color": null,1165 "flair_group_id": null,1166 "trust_level": 11167 }1168 ],1169 "created_by": {1170 "id": 67347,1171 "username": "pypdeveloper",1172 "name": "pypdeveloper",1173 "avatar_template": "/user_avatar/discuss.pytorch.org/pypdeveloper/{size}/61681_2.png"1174 },1175 "last_poster": {1176 "id": 67347,1177 "username": "pypdeveloper",1178 "name": "pypdeveloper",1179 "avatar_template": "/user_avatar/discuss.pytorch.org/pypdeveloper/{size}/61681_2.png"1180 }1181 },1182 "bookmarks": []1183 },1184 {1185 "post_stream": {1186 "posts": [1187 {1188 "id": 241352,1189 "name": "Karan Chhabra",1190 "username": "Karan_Chhabra",1191 "avatar_template": "/user_avatar/discuss.pytorch.org/karan_chhabra/{size}/30204_2.png",1192 "created_at": "2020-11-04T03:22:23.734Z",1193 "cooked": "<p>I am using a pre-train network with nn.BCEWithLogitsLoss() loss for a multilabel problem. I want the output of the network as probabilities, but after using Softmax, I am getting the output of 0 or 1, which seems quite confusing as Softmax should not output perfectly 0 or 1 of any class, it should output the probabilities for various classes instead.</p>\n<p>Below is the image of my code:<br>\n<img src=\"https://discuss.pytorch.org/uploads/default/original/3X/a/b/ab129018cd41bfda69789c77d5a392d8e164c772.png\" alt=\"image\" data-base62-sha1=\"opnpVG1rKEsamw3g0SHIq9MDxZ0\" width=\"641\" height=\"368\"></p>\n<p>Below is the image of the output:<br>\n<div class=\"lightbox-wrapper\"><a class=\"lightbox\" href=\"https://discuss.pytorch.org/uploads/default/original/3X/e/d/ed1c114ee105b27a0e51870935800ee866bf745c.png\" data-download-href=\"https://discuss.pytorch.org/uploads/default/ed1c114ee105b27a0e51870935800ee866bf745c\" title=\"image\"><img src=\"https://discuss.pytorch.org/uploads/default/original/3X/e/d/ed1c114ee105b27a0e51870935800ee866bf745c.png\" alt=\"image\" data-base62-sha1=\"xPziigcyscmfyJE18M3SiMjQg8s\" width=\"690\" height=\"93\" data-dominant-color=\"F7F7F7\"><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\">image</span><span class=\"informations\">757×103 2.09 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>",1194 "post_number": 1,1195 "post_type": 1,1196 "posts_count": 7,1197 "updated_at": "2020-11-04T03:22:23.734Z",1198 "reply_count": 0,1199 "reply_to_post_number": null,1200 "quote_count": 0,