Anurag1734/cuda-error-resolution-analysis
07
1[2 {3 "post_stream": {4 "posts": [5 {6 "id": 338605,7 "name": "Berkay Alan",8 "username": "berkayalan",9 "avatar_template": "/user_avatar/discuss.pytorch.org/berkayalan/{size}/43440_2.png",10 "created_at": "2022-03-29T08:34:01.533Z",11 "cooked": "<p>Hi,</p>\n<p>I have product images more than 40k and would like to find similar images from the pool. I have looked some example such as:</p>\n<ul>\n<li><a href=\"https://medium.com/pytorch/image-similarity-search-in-pytorch-1a744cf3469\" class=\"inline-onebox\" rel=\"noopener nofollow ugc\">Image Similarity Search in PyTorch | by Aditya Oke | PyTorch | Medium</a></li>\n</ul>\n<p>Do you have any suggestion for that? I really need this and all your suggestions are appreciated.</p>\n<p>Thanks in advance.</p>",12 "post_number": 1,13 "post_type": 1,14 "posts_count": 1,15 "updated_at": "2022-03-29T08:34:01.533Z",16 "reply_count": 0,17 "reply_to_post_number": null,18 "quote_count": 0,19 "incoming_link_count": 115,20 "reads": 6,21 "readers_count": 5,22 "score": 576.2,23 "yours": false,24 "topic_id": 147735,25 "topic_slug": "unsupervised-image-similarity-model",26 "display_username": "Berkay Alan",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://medium.com/pytorch/image-similarity-search-in-pytorch-1a744cf3469",43 "internal": false,44 "reflection": false,45 "title": "Image Similarity Search in PyTorch | by Aditya Oke | PyTorch | Medium",46 "clicks": 2347 }48 ],49 "read": true,50 "user_title": null,51 "bookmarked": false,52 "actions_summary": [],53 "moderator": false,54 "admin": false,55 "staff": false,56 "user_id": 50227,57 "hidden": false,58 "trust_level": 1,59 "deleted_at": null,60 "user_deleted": false,61 "edit_reason": null,62 "can_view_edit_history": true,63 "wiki": false,64 "post_url": "/t/unsupervised-image-similarity-model/147735/1",65 "can_accept_answer": false,66 "can_unaccept_answer": false,67 "accepted_answer": false,68 "topic_accepted_answer": null,69 "can_vote": false70 }71 ],72 "stream": [73 33860574 ]75 },76 "timeline_lookup": [77 [78 1,79 130780 ]81 ],82 "suggested_topics": [83 {84 "fancy_title": "CNN Model is not learning after some epochs",85 "id": 215485,86 "title": "CNN Model is not learning after some epochs",87 "slug": "cnn-model-is-not-learning-after-some-epochs",88 "posts_count": 5,89 "reply_count": 3,90 "highest_post_number": 5,91 "image_url": "https://discuss.pytorch.org/uploads/default/optimized/3X/6/c/6cf1deea54130767dd867db4827e1505ce26e24b_2_1024x817.jpeg",92 "created_at": "2025-01-16T17:42:58.865Z",93 "last_posted_at": "2025-01-17T15:02:23.101Z",94 "bumped": true,95 "bumped_at": "2025-01-17T15:02:23.101Z",96 "archetype": "regular",97 "unseen": false,98 "pinned": false,99 "unpinned": null,100 "visible": true,101 "closed": false,102 "archived": false,103 "bookmarked": null,104 "liked": null,105 "tags_descriptions": {},106 "like_count": 1,107 "views": 170,108 "category_id": 5,109 "featured_link": null,110 "has_accepted_answer": false,111 "posters": [112 {113 "extras": "latest",114 "description": "Original Poster, Most Recent Poster",115 "user": {116 "id": 81840,117 "username": "iran_boy",118 "name": "iran boy",119 "avatar_template": "/user_avatar/discuss.pytorch.org/iran_boy/{size}/74864_2.png",120 "trust_level": 0121 }122 },123 {124 "extras": null,125 "description": "Frequent Poster",126 "user": {127 "id": 77908,128 "username": "mycul",129 "name": "",130 "avatar_template": "/user_avatar/discuss.pytorch.org/mycul/{size}/72394_2.png",131 "trust_level": 2132 }133 }134 ]135 },136 {137 "fancy_title": "ResUnet implementation in pytorch for medical image segmentation",138 "id": 212702,139 "title": "ResUnet implementation in pytorch for medical image segmentation",140 "slug": "resunet-implementation-in-pytorch-for-medical-image-segmentation",141 "posts_count": 2,142 "reply_count": 0,143 "highest_post_number": 2,144 "image_url": null,145 "created_at": "2024-11-08T11:32:30.168Z",146 "last_posted_at": "2024-11-11T19:57:35.383Z",147 "bumped": true,148 "bumped_at": "2024-11-11T19:57:35.383Z",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": 106,161 "category_id": 5,162 "featured_link": null,163 "has_accepted_answer": false,164 "posters": [165 {166 "extras": null,167 "description": "Original Poster",168 "user": {169 "id": 80772,170 "username": "Saurav_K_Mhptra",171 "name": "Saurav Kausik Mahapatra",172 "avatar_template": "/user_avatar/discuss.pytorch.org/saurav_k_mhptra/{size}/73231_2.png",173 "trust_level": 0174 }175 },176 {177 "extras": "latest",178 "description": "Most Recent Poster",179 "user": {180 "id": 75871,181 "username": "qq-me",182 "name": "Ivan Nikishev",183 "avatar_template": "/user_avatar/discuss.pytorch.org/qq-me/{size}/70055_2.png",184 "trust_level": 2185 }186 }187 ]188 },189 {190 "fancy_title": "Functional.nn.grid_sample does not support the bicubic method for 5D tensor",191 "id": 212316,192 "title": "Functional.nn.grid_sample does not support the bicubic method for 5D tensor",193 "slug": "functional-nn-grid-sample-does-not-support-the-bicubic-method-for-5d-tensor",194 "posts_count": 1,195 "reply_count": 0,196 "highest_post_number": 1,197 "image_url": "https://discuss.pytorch.org/uploads/default/optimized/3X/7/2/7276584510bb27cdac0fc29047e8fe2a5f538aa2_2_1024x677.png",198 "created_at": "2024-10-30T12:09:48.157Z",199 "last_posted_at": "2024-10-30T12:09:48.284Z",200 "bumped": true,201 "bumped_at": "2024-10-30T12:09:48.284Z",202 "archetype": "regular",203 "unseen": false,204 "pinned": false,205 "unpinned": null,206 "visible": true,207 "closed": false,208 "archived": false,209 "bookmarked": null,210 "liked": null,211 "tags_descriptions": {},212 "like_count": 0,213 "views": 19,214 "category_id": 5,215 "featured_link": null,216 "has_accepted_answer": false,217 "posters": [218 {219 "extras": "latest single",220 "description": "Original Poster, Most Recent Poster",221 "user": {222 "id": 80581,223 "username": "anitrop",224 "name": "Anitrop",225 "avatar_template": "/user_avatar/discuss.pytorch.org/anitrop/{size}/73666_2.png",226 "trust_level": 1227 }228 }229 ]230 },231 {232 "fancy_title": "Loss function for bounding boxes in rpn model",233 "id": 215067,234 "title": "Loss function for bounding boxes in rpn model",235 "slug": "loss-function-for-bounding-boxes-in-rpn-model",236 "posts_count": 3,237 "reply_count": 1,238 "highest_post_number": 3,239 "image_url": null,240 "created_at": "2025-01-07T10:39:06.693Z",241 "last_posted_at": "2025-01-09T19:46:23.787Z",242 "bumped": true,243 "bumped_at": "2025-01-09T19:46:23.787Z",244 "archetype": "regular",245 "unseen": false,246 "pinned": false,247 "unpinned": null,248 "visible": true,249 "closed": false,250 "archived": false,251 "bookmarked": null,252 "liked": null,253 "tags_descriptions": {},254 "like_count": 0,255 "views": 165,256 "category_id": 5,257 "featured_link": null,258 "has_accepted_answer": false,259 "posters": [260 {261 "extras": "latest single",262 "description": "Original Poster, Most Recent Poster",263 "user": {264 "id": 81922,265 "username": "E_T",266 "name": "E T",267 "avatar_template": "/user_avatar/discuss.pytorch.org/e_t/{size}/74946_2.png",268 "trust_level": 1269 }270 }271 ]272 },273 {274 "fancy_title": "Comfy_UI:Attempting to use hipBLASLt on a unsupported architecture!",275 "id": 215776,276 "title": "Comfy_UI:Attempting to use hipBLASLt on a unsupported architecture!",277 "slug": "comfy-ui-attempting-to-use-hipblaslt-on-a-unsupported-architecture",278 "posts_count": 25,279 "reply_count": 17,280 "highest_post_number": 25,281 "image_url": null,282 "created_at": "2025-01-23T15:06:28.660Z",283 "last_posted_at": "2025-04-21T15:50:34.701Z",284 "bumped": true,285 "bumped_at": "2025-04-21T15:50:34.701Z",286 "archetype": "regular",287 "unseen": false,288 "pinned": false,289 "unpinned": null,290 "visible": true,291 "closed": false,292 "archived": false,293 "bookmarked": null,294 "liked": null,295 "tags_descriptions": {},296 "like_count": 2,297 "views": 1907,298 "category_id": 5,299 "featured_link": null,300 "has_accepted_answer": true,301 "posters": [302 {303 "extras": null,304 "description": "Original Poster, Accepted Answer",305 "user": {306 "id": 82184,307 "username": "sersys",308 "name": "",309 "avatar_template": "/letter_avatar_proxy/v4/letter/s/c77e96/{size}.png",310 "trust_level": 1311 }312 },313 {314 "extras": null,315 "description": "Frequent Poster",316 "user": {317 "id": 39248,318 "username": "jaglinux",319 "name": "Jagadish Krishnamoorthy",320 "avatar_template": "/user_avatar/discuss.pytorch.org/jaglinux/{size}/31429_2.png",321 "trust_level": 1322 }323 },324 {325 "extras": null,326 "description": "Frequent Poster",327 "user": {328 "id": 82792,329 "username": "fluidnumerics_joe",330 "name": "Joseph Schoonover",331 "avatar_template": "/user_avatar/discuss.pytorch.org/fluidnumerics_joe/{size}/75754_2.png",332 "trust_level": 1333 }334 },335 {336 "extras": "latest",337 "description": "Most Recent Poster",338 "user": {339 "id": 83932,340 "username": "RPaciorek",341 "name": "RPaciorek",342 "avatar_template": "/user_avatar/discuss.pytorch.org/rpaciorek/{size}/76728_2.png",343 "trust_level": 0344 }345 }346 ]347 }348 ],349 "tags_descriptions": {},350 "fancy_title": "Unsupervised Image Similarity model",351 "id": 147735,352 "title": "Unsupervised Image Similarity model",353 "posts_count": 1,354 "created_at": "2022-03-29T08:34:01.470Z",355 "views": 505,356 "reply_count": 0,357 "like_count": 0,358 "last_posted_at": "2022-03-29T08:34:01.533Z",359 "visible": true,360 "closed": false,361 "archived": false,362 "has_summary": false,363 "archetype": "regular",364 "slug": "unsupervised-image-similarity-model",365 "category_id": 5,366 "word_count": 55,367 "deleted_at": null,368 "user_id": 50227,369 "featured_link": null,370 "pinned_globally": false,371 "pinned_at": null,372 "pinned_until": null,373 "image_url": null,374 "slow_mode_seconds": 0,375 "draft": null,376 "draft_key": "topic_147735",377 "draft_sequence": null,378 "unpinned": null,379 "pinned": false,380 "current_post_number": 1,381 "highest_post_number": 1,382 "deleted_by": null,383 "actions_summary": [384 {385 "id": 4,386 "count": 0,387 "hidden": false,388 "can_act": false389 },390 {391 "id": 8,392 "count": 0,393 "hidden": false,394 "can_act": false395 },396 {397 "id": 10,398 "count": 0,399 "hidden": false,400 "can_act": false401 },402 {403 "id": 7,404 "count": 0,405 "hidden": false,406 "can_act": false407 }408 ],409 "chunk_size": 20,410 "bookmarked": false,411 "topic_timer": null,412 "message_bus_last_id": 0,413 "participant_count": 1,414 "show_read_indicator": false,415 "thumbnails": null,416 "slow_mode_enabled_until": null,417 "can_vote": false,418 "vote_count": 0,419 "user_voted": false,420 "discourse_zendesk_plugin_zendesk_id": null,421 "discourse_zendesk_plugin_zendesk_url": "https://your-url.zendesk.com/agent/tickets/",422 "details": {423 "can_edit": false,424 "notification_level": 1,425 "participants": [426 {427 "id": 50227,428 "username": "berkayalan",429 "name": "Berkay Alan",430 "avatar_template": "/user_avatar/discuss.pytorch.org/berkayalan/{size}/43440_2.png",431 "post_count": 1,432 "primary_group_name": null,433 "flair_name": null,434 "flair_url": null,435 "flair_color": null,436 "flair_bg_color": null,437 "flair_group_id": null,438 "trust_level": 1439 }440 ],441 "created_by": {442 "id": 50227,443 "username": "berkayalan",444 "name": "Berkay Alan",445 "avatar_template": "/user_avatar/discuss.pytorch.org/berkayalan/{size}/43440_2.png"446 },447 "last_poster": {448 "id": 50227,449 "username": "berkayalan",450 "name": "Berkay Alan",451 "avatar_template": "/user_avatar/discuss.pytorch.org/berkayalan/{size}/43440_2.png"452 },453 "links": [454 {455 "url": "https://medium.com/pytorch/image-similarity-search-in-pytorch-1a744cf3469",456 "title": "Image Similarity Search in PyTorch | by Aditya Oke | PyTorch | Medium",457 "internal": false,458 "attachment": false,459 "reflection": false,460 "clicks": 23,461 "user_id": 50227,462 "domain": "medium.com",463 "root_domain": "medium.com"464 }465 ]466 },467 "bookmarks": []468 },469 {470 "post_stream": {471 "posts": [472 {473 "id": 223842,474 "name": "Westerby",475 "username": "Westerby",476 "avatar_template": "/letter_avatar_proxy/v4/letter/w/779978/{size}.png",477 "created_at": "2020-08-25T10:39:56.244Z",478 "cooked": "<p>Hello,</p>\n<p>I trained frcnn model with automatic mixed precision and exported it to ONNX. I wonder however how would inference look like programmaticaly to leverage the speed up of mixed precision model, since pytorch uses <code>with autocast():</code>, and I can’t come with an idea how to put it in the inference engine, like onnxruntime.</p>\n<p>My specs:<br>\ntorch==1.6.0+cu101<br>\ntorchvision==0.7.0+cu101<br>\nonnx==1.7.0<br>\nonnxruntime-gpu==1.4.0</p>\n<p>Model exports just fine:</p>\n<pre><code class=\"lang-auto\">torch.onnx.export(model, \n x, \n \"model_16.onnx\", \n verbose=True, do_constant_folding=True, opset_version=12,\n input_names=input_names, output_names=output_names)\n</code></pre>\n<p>But I wonder how to leverage mixed precision speed up here:</p>\n<pre><code class=\"lang-auto\">\nimport onnxruntime as ort\nort_session = ort.InferenceSession('model_16.onnx')\noutputs = ort_session.run(None, {'input': x.numpy()})\n</code></pre>",479 "post_number": 1,480 "post_type": 1,481 "posts_count": 4,482 "updated_at": "2020-08-25T10:43:34.487Z",483 "reply_count": 0,484 "reply_to_post_number": null,485 "quote_count": 0,486 "incoming_link_count": 2978,487 "reads": 76,488 "readers_count": 75,489 "score": 14870.2,490 "yours": false,491 "topic_id": 94035,492 "topic_slug": "inference-in-onnx-mixed-precision-model",493 "display_username": "Westerby",494 "primary_group_name": null,495 "flair_name": null,496 "flair_url": null,497 "flair_bg_color": null,498 "flair_color": null,499 "flair_group_id": null,500 "badges_granted": [],501 "version": 2,502 "can_edit": false,503 "can_delete": false,504 "can_recover": false,505 "can_see_hidden_post": false,506 "can_wiki": false,507 "read": true,508 "user_title": null,509 "bookmarked": false,510 "actions_summary": [],511 "moderator": false,512 "admin": false,513 "staff": false,514 "user_id": 21437,515 "hidden": false,516 "trust_level": 1,517 "deleted_at": null,518 "user_deleted": false,519 "edit_reason": null,520 "can_view_edit_history": true,521 "wiki": false,522 "post_url": "/t/inference-in-onnx-mixed-precision-model/94035/1",523 "can_accept_answer": false,524 "can_unaccept_answer": false,525 "accepted_answer": false,526 "topic_accepted_answer": null,527 "can_vote": false528 },529 {530 "id": 224135,531 "name": "",532 "username": "ptrblck",533 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",534 "created_at": "2020-08-26T10:08:23.880Z",535 "cooked": "<p>I’m not exactly sure how ONNX exports the model, but if tracing is used, the mixed-precision operations might have been already recorded. Do you see any FP16 operations, if you profile the ONNX model?</p>",536 "post_number": 2,537 "post_type": 1,538 "posts_count": 4,539 "updated_at": "2020-08-26T10:08:23.880Z",540 "reply_count": 0,541 "reply_to_post_number": null,542 "quote_count": 0,543 "incoming_link_count": 16,544 "reads": 74,545 "readers_count": 73,546 "score": 109.8,547 "yours": false,548 "topic_id": 94035,549 "topic_slug": "inference-in-onnx-mixed-precision-model",550 "display_username": "",551 "primary_group_name": null,552 "flair_name": null,553 "flair_url": null,554 "flair_bg_color": null,555 "flair_color": null,556 "flair_group_id": null,557 "badges_granted": [],558 "version": 1,559 "can_edit": false,560 "can_delete": false,561 "can_recover": false,562 "can_see_hidden_post": false,563 "can_wiki": false,564 "read": true,565 "user_title": "",566 "bookmarked": false,567 "actions_summary": [568 {569 "id": 2,570 "count": 1571 }572 ],573 "moderator": true,574 "admin": true,575 "staff": true,576 "user_id": 3534,577 "hidden": false,578 "trust_level": 2,579 "deleted_at": null,580 "user_deleted": false,581 "edit_reason": null,582 "can_view_edit_history": true,583 "wiki": false,584 "post_url": "/t/inference-in-onnx-mixed-precision-model/94035/2",585 "can_accept_answer": false,586 "can_unaccept_answer": false,587 "accepted_answer": false,588 "topic_accepted_answer": null589 },590 {591 "id": 224785,592 "name": "Westerby",593 "username": "Westerby",594 "avatar_template": "/letter_avatar_proxy/v4/letter/w/779978/{size}.png",595 "created_at": "2020-08-28T11:57:00.076Z",596 "cooked": "<p>Hello,</p>\n<p>thanks for the suggestions. I run the onnx runtime profiler on 16FP model from torch, but I’m not exactly sure how to look for execution of FP16 operations there. I’m attaching the log.<br>\n<a href=\"http://www.mediafire.com/file/rel0ze3y963nohv/onnxruntime_profile__2020-08-28_10-35-51.json/file\" class=\"onebox\" target=\"_blank\" rel=\"nofollow noopener\">http://www.mediafire.com/file/rel0ze3y963nohv/onnxruntime_profile__2020-08-28_10-35-51.json/file</a><br>\nHere’s the code I used to run the profiler:</p>\n<pre><code class=\"lang-auto\">import onnxruntime as ort\noptions = ort.SessionOptions()\noptions.enable_profiling = True\nort_session = ort.InferenceSession('model_16.onnx', options)\noutputs = ort_session.run(None, {'input': images[0].cpu().numpy()})\nprof_file = ort_session.end_profiling()\n</code></pre>\n<p>Anyway, if I do simple time measurement for inference, I don’t see much difference. To be honest I expected ONNX model to run faster.</p>\n<p>1.Pure torch 16FP model:</p>\n<pre><code class=\"lang-auto\">from imutils.video import FPS\nfps = FPS().start()\nfor i in range(100):\n images = list(image.to('cuda:0') for image in x)\n with autocast():\n pred = model(images)\n \n fps.update()\n\nfps.stop()\nprint('Time taken: {:.2f}'.format(fps.elapsed()))\nprint('~ FPS : {:.2f}'.format(fps.fps()))\n\nTime taken: 2.19\n~ FPS : 45.57\n</code></pre>\n<ol start=\"2\">\n<li>Torch->ONNX 16FP model:</li>\n</ol>\n<pre><code class=\"lang-auto\">import onnxruntime as ort\nort_session = ort.InferenceSession('model_16.onnx')\nfps = FPS().start()\n\nfor i in range(100):\n outputs = ort_session.run(None, {'input': images[0].cpu().numpy()})\n fps.update()\n\nfps.stop()\nprint('Time taken: {:.2f}'.format(fps.elapsed()))\nprint('~ FPS : {:.2f}'.format(fps.fps()))\n\nTime taken: 2.15\n~ FPS : 46.61\n</code></pre>",597 "post_number": 3,598 "post_type": 1,599 "posts_count": 4,600 "updated_at": "2020-08-28T11:57:00.076Z",601 "reply_count": 1,602 "reply_to_post_number": null,603 "quote_count": 0,604 "incoming_link_count": 74,605 "reads": 71,606 "readers_count": 70,607 "score": 389.2,608 "yours": false,609 "topic_id": 94035,610 "topic_slug": "inference-in-onnx-mixed-precision-model",611 "display_username": "Westerby",612 "primary_group_name": null,613 "flair_name": null,614 "flair_url": null,615 "flair_bg_color": null,616 "flair_color": null,617 "flair_group_id": null,618 "badges_granted": [],619 "version": 1,620 "can_edit": false,621 "can_delete": false,622 "can_recover": false,623 "can_see_hidden_post": false,624 "can_wiki": false,625 "link_counts": [626 {627 "url": "http://www.mediafire.com/file/rel0ze3y963nohv/onnxruntime_profile__2020-08-28_10-35-51.json/file",628 "internal": false,629 "reflection": false,630 "title": "onnxruntime_profile__2020-08-28_10-35-51",631 "clicks": 25632 }633 ],634 "read": true,635 "user_title": null,636 "bookmarked": false,637 "actions_summary": [],638 "moderator": false,639 "admin": false,640 "staff": false,641 "user_id": 21437,642 "hidden": false,643 "trust_level": 1,644 "deleted_at": null,645 "user_deleted": false,646 "edit_reason": null,647 "can_view_edit_history": true,648 "wiki": false,649 "post_url": "/t/inference-in-onnx-mixed-precision-model/94035/3",650 "can_accept_answer": false,651 "can_unaccept_answer": false,652 "accepted_answer": false,653 "topic_accepted_answer": null654 },655 {656 "id": 338604,657 "name": "dheeraj agrawal",658 "username": "dheeraj_agrawal",659 "avatar_template": "/user_avatar/discuss.pytorch.org/dheeraj_agrawal/{size}/48070_2.png",660 "created_at": "2022-03-29T08:27:19.083Z",661 "cooked": "<p>Did you find the solutions for converting automatic mixed precision model to onnx model with AMP.</p>",662 "post_number": 4,663 "post_type": 1,664 "posts_count": 4,665 "updated_at": "2022-03-29T08:27:19.083Z",666 "reply_count": 0,667 "reply_to_post_number": 3,668 "quote_count": 0,669 "incoming_link_count": 7,670 "reads": 34,671 "readers_count": 33,672 "score": 41.8,673 "yours": false,674 "topic_id": 94035,675 "topic_slug": "inference-in-onnx-mixed-precision-model",676 "display_username": "dheeraj agrawal",677 "primary_group_name": null,678 "flair_name": null,679 "flair_url": null,680 "flair_bg_color": null,681 "flair_color": null,682 "flair_group_id": null,683 "badges_granted": [],684 "version": 1,685 "can_edit": false,686 "can_delete": false,687 "can_recover": false,688 "can_see_hidden_post": false,689 "can_wiki": false,690 "read": true,691 "user_title": null,692 "reply_to_user": {693 "id": 21437,694 "username": "Westerby",695 "name": "Westerby",696 "avatar_template": "/letter_avatar_proxy/v4/letter/w/779978/{size}.png"697 },698 "bookmarked": false,699 "actions_summary": [],700 "moderator": false,701 "admin": false,702 "staff": false,703 "user_id": 54524,704 "hidden": false,705 "trust_level": 0,706 "deleted_at": null,707 "user_deleted": false,708 "edit_reason": null,709 "can_view_edit_history": true,710 "wiki": false,711 "post_url": "/t/inference-in-onnx-mixed-precision-model/94035/4",712 "can_accept_answer": false,713 "can_unaccept_answer": false,714 "accepted_answer": false,715 "topic_accepted_answer": null716 }717 ],718 "stream": [719 223842,720 224135,721 224785,722 338604723 ]724 },725 "timeline_lookup": [726 [727 1,728 1887729 ],730 [731 2,732 1886733 ],734 [735 3,736 1884737 ],738 [739 4,740 1307741 ]742 ],743 "suggested_topics": [744 {745 "fancy_title": "Does autocast create copies of tensors on the fly?",746 "id": 214268,747 "title": "Does autocast create copies of tensors on the fly?",748 "slug": "does-autocast-create-copies-of-tensors-on-the-fly",749 "posts_count": 3,750 "reply_count": 1,751 "highest_post_number": 3,752 "image_url": null,753 "created_at": "2024-12-16T11:23:24.995Z",754 "last_posted_at": "2024-12-16T17:30:40.412Z",755 "bumped": true,756 "bumped_at": "2024-12-16T17:30:40.412Z",757 "archetype": "regular",758 "unseen": false,759 "pinned": false,760 "unpinned": null,761 "visible": true,762 "closed": false,763 "archived": false,764 "bookmarked": null,765 "liked": null,766 "tags_descriptions": {},767 "like_count": 1,768 "views": 58,769 "category_id": 27,770 "featured_link": null,771 "has_accepted_answer": true,772 "posters": [773 {774 "extras": "latest",775 "description": "Original Poster, Most Recent Poster",776 "user": {777 "id": 81089,778 "username": "Aknw_Fen",779 "name": "Aknw Fen",780 "avatar_template": "/user_avatar/discuss.pytorch.org/aknw_fen/{size}/74156_2.png",781 "trust_level": 2782 }783 },784 {785 "extras": null,786 "description": "Frequent Poster, Accepted Answer",787 "user": {788 "id": 41396,789 "username": "soulitzer",790 "name": "",791 "avatar_template": "/letter_avatar_proxy/v4/letter/s/839c29/{size}.png",792 "trust_level": 2793 }794 }795 ]796 },797 {798 "fancy_title": "Any operator is supported on fp8 tensor?",799 "id": 212371,800 "title": "Any operator is supported on fp8 tensor?",801 "slug": "any-operator-is-supported-on-fp8-tensor",802 "posts_count": 8,803 "reply_count": 6,804 "highest_post_number": 8,805 "image_url": null,806 "created_at": "2024-10-31T16:06:44.298Z",807 "last_posted_at": "2024-11-05T16:48:48.277Z",808 "bumped": true,809 "bumped_at": "2024-11-05T16:48:48.277Z",810 "archetype": "regular",811 "unseen": false,812 "pinned": false,813 "unpinned": null,814 "visible": true,815 "closed": false,816 "archived": false,817 "bookmarked": null,818 "liked": null,819 "tags_descriptions": {},820 "like_count": 1,821 "views": 3166,822 "category_id": 27,823 "featured_link": null,824 "has_accepted_answer": false,825 "posters": [826 {827 "extras": "latest",828 "description": "Original Poster, Most Recent Poster",829 "user": {830 "id": 57464,831 "username": "cokespace2",832 "name": "Vince Mo",833 "avatar_template": "/user_avatar/discuss.pytorch.org/cokespace2/{size}/51244_2.png",834 "trust_level": 2835 }836 },837 {838 "extras": null,839 "description": "Frequent Poster",840 "user": {841 "id": 43941,842 "username": "marksaroufim",843 "name": "Mark Saroufim",844 "avatar_template": "/user_avatar/discuss.pytorch.org/marksaroufim/{size}/36747_2.png",845 "trust_level": 3846 }847 },848 {849 "extras": null,850 "description": "Frequent Poster",851 "user": {852 "id": 3534,853 "username": "ptrblck",854 "name": "",855 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",856 "admin": true,857 "moderator": true,858 "trust_level": 2859 }860 }861 ]862 },863 {864 "fancy_title": "Why tensor.to convert fp32 to fp8_e4m3=Nan if overflow",865 "id": 212648,866 "title": "Why tensor.to convert fp32 to fp8_e4m3=Nan if overflow",867 "slug": "why-tensor-to-convert-fp32-to-fp8-e4m3-nan-if-overflow",868 "posts_count": 3,869 "reply_count": 2,870 "highest_post_number": 4,871 "image_url": null,872 "created_at": "2024-11-07T08:09:24.665Z",873 "last_posted_at": "2024-11-07T11:14:48.008Z",874 "bumped": true,875 "bumped_at": "2024-11-07T11:14:48.008Z",876 "archetype": "regular",877 "unseen": false,878 "pinned": false,879 "unpinned": null,880 "visible": true,881 "closed": false,882 "archived": false,883 "bookmarked": null,884 "liked": null,885 "tags_descriptions": {},886 "like_count": 0,887 "views": 661,888 "category_id": 27,889 "featured_link": null,890 "has_accepted_answer": false,891 "posters": [892 {893 "extras": "latest",894 "description": "Original Poster, Most Recent Poster",895 "user": {896 "id": 80743,897 "username": "symbolics",898 "name": "banayoyo",899 "avatar_template": "/letter_avatar_proxy/v4/letter/s/74df32/{size}.png",900 "trust_level": 1901 }902 },903 {904 "extras": null,905 "description": "Frequent Poster",906 "user": {907 "id": 80724,908 "username": "paulge",909 "name": "",910 "avatar_template": "/letter_avatar_proxy/v4/letter/p/82dd89/{size}.png",911 "trust_level": 2912 }913 }914 ]915 },916 {917 "fancy_title": "The dtype of optimizer states in PyTorch AMP training",918 "id": 214067,919 "title": "The dtype of optimizer states in PyTorch AMP training",920 "slug": "the-dtype-of-optimizer-states-in-pytorch-amp-training",921 "posts_count": 2,922 "reply_count": 0,923 "highest_post_number": 2,924 "image_url": null,925 "created_at": "2024-12-10T18:02:16.235Z",926 "last_posted_at": "2024-12-10T22:56:43.015Z",927 "bumped": true,928 "bumped_at": "2024-12-10T22:56:43.015Z",929 "archetype": "regular",930 "unseen": false,931 "pinned": false,932 "unpinned": null,933 "visible": true,934 "closed": false,935 "archived": false,936 "bookmarked": null,937 "liked": null,938 "tags_descriptions": {},939 "like_count": 1,940 "views": 254,941 "category_id": 27,942 "featured_link": null,943 "has_accepted_answer": false,944 "posters": [945 {946 "extras": null,947 "description": "Original Poster",948 "user": {949 "id": 81434,950 "username": "Chaohao_Yang",951 "name": "Chaohao Yang",952 "avatar_template": "/user_avatar/discuss.pytorch.org/chaohao_yang/{size}/74456_2.png",953 "trust_level": 0954 }955 },956 {957 "extras": "latest",958 "description": "Most Recent Poster",959 "user": {960 "id": 3534,961 "username": "ptrblck",962 "name": "",963 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",964 "admin": true,965 "moderator": true,966 "trust_level": 2967 }968 }969 ]970 },971 {972 "fancy_title": "Half Precision based training adaptations",973 "id": 221996,974 "title": "Half Precision based training adaptations",975 "slug": "half-precision-based-training-adaptations",976 "posts_count": 4,977 "reply_count": 2,978 "highest_post_number": 4,979 "image_url": null,980 "created_at": "2025-08-02T10:57:16.182Z",981 "last_posted_at": "2025-08-04T02:54:52.649Z",982 "bumped": true,983 "bumped_at": "2025-08-04T02:54:52.649Z",984 "archetype": "regular",985 "unseen": false,986 "pinned": false,987 "unpinned": null,988 "visible": true,989 "closed": false,990 "archived": false,991 "bookmarked": null,992 "liked": null,993 "tags_descriptions": {},994 "like_count": 0,995 "views": 66,996 "category_id": 27,997 "featured_link": null,998 "has_accepted_answer": true,999 "posters": [1000 {1001 "extras": null,1002 "description": "Original Poster",1003 "user": {1004 "id": 85324,1005 "username": "nath-11743",1006 "name": "Nath",1007 "avatar_template": "/user_avatar/discuss.pytorch.org/nath-11743/{size}/77847_2.png",1008 "trust_level": 11009 }1010 },1011 {1012 "extras": "latest",1013 "description": "Most Recent Poster, Accepted Answer",1014 "user": {1015 "id": 84484,1016 "username": "Dhia-naouali",1017 "name": "Dhia naouali",1018 "avatar_template": "/user_avatar/discuss.pytorch.org/dhia-naouali/{size}/77193_2.png",1019 "trust_level": 21020 }1021 },1022 {1023 "extras": null,1024 "description": "Frequent Poster",1025 "user": {1026 "id": 3534,1027 "username": "ptrblck",1028 "name": "",1029 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",1030 "admin": true,1031 "moderator": true,1032 "trust_level": 21033 }1034 }1035 ]1036 }1037 ],1038 "tags_descriptions": {},1039 "fancy_title": "Inference in ONNX mixed precision model",1040 "id": 94035,1041 "title": "Inference in ONNX mixed precision model",1042 "posts_count": 4,1043 "created_at": "2020-08-25T10:39:56.177Z",1044 "views": 4844,1045 "reply_count": 1,1046 "like_count": 1,1047 "last_posted_at": "2022-03-29T08:27:19.083Z",1048 "visible": true,1049 "closed": false,1050 "archived": false,1051 "has_summary": false,1052 "archetype": "regular",1053 "slug": "inference-in-onnx-mixed-precision-model",1054 "category_id": 27,1055 "word_count": 400,1056 "deleted_at": null,1057 "user_id": 21437,1058 "featured_link": null,1059 "pinned_globally": false,1060 "pinned_at": null,1061 "pinned_until": null,1062 "image_url": null,1063 "slow_mode_seconds": 0,1064 "draft": null,1065 "draft_key": "topic_94035",1066 "draft_sequence": null,1067 "unpinned": null,1068 "pinned": false,1069 "current_post_number": 1,1070 "highest_post_number": 4,1071 "deleted_by": null,1072 "actions_summary": [1073 {1074 "id": 4,1075 "count": 0,1076 "hidden": false,1077 "can_act": false1078 },1079 {1080 "id": 8,1081 "count": 0,1082 "hidden": false,1083 "can_act": false1084 },1085 {1086 "id": 10,1087 "count": 0,1088 "hidden": false,1089 "can_act": false1090 },1091 {1092 "id": 7,1093 "count": 0,1094 "hidden": false,1095 "can_act": false1096 }1097 ],1098 "chunk_size": 20,1099 "bookmarked": false,1100 "topic_timer": null,1101 "message_bus_last_id": 0,1102 "participant_count": 3,1103 "show_read_indicator": false,1104 "thumbnails": null,1105 "slow_mode_enabled_until": null,1106 "can_vote": false,1107 "vote_count": 0,1108 "user_voted": false,1109 "discourse_zendesk_plugin_zendesk_id": null,1110 "discourse_zendesk_plugin_zendesk_url": "https://your-url.zendesk.com/agent/tickets/",1111 "details": {1112 "can_edit": false,1113 "notification_level": 1,1114 "participants": [1115 {1116 "id": 21437,1117 "username": "Westerby",1118 "name": "Westerby",1119 "avatar_template": "/letter_avatar_proxy/v4/letter/w/779978/{size}.png",1120 "post_count": 2,1121 "primary_group_name": null,1122 "flair_name": null,1123 "flair_url": null,1124 "flair_color": null,1125 "flair_bg_color": null,1126 "flair_group_id": null,1127 "trust_level": 11128 },1129 {1130 "id": 3534,1131 "username": "ptrblck",1132 "name": "",1133 "avatar_template": "/user_avatar/discuss.pytorch.org/ptrblck/{size}/1823_2.png",1134 "post_count": 1,1135 "primary_group_name": null,1136 "flair_name": null,1137 "flair_url": null,1138 "flair_color": null,1139 "flair_bg_color": null,1140 "flair_group_id": null,1141 "admin": true,1142 "moderator": true,1143 "trust_level": 21144 },1145 {1146 "id": 54524,1147 "username": "dheeraj_agrawal",1148 "name": "dheeraj agrawal",1149 "avatar_template": "/user_avatar/discuss.pytorch.org/dheeraj_agrawal/{size}/48070_2.png",1150 "post_count": 1,1151 "primary_group_name": null,1152 "flair_name": null,1153 "flair_url": null,1154 "flair_color": null,1155 "flair_bg_color": null,1156 "flair_group_id": null,1157 "trust_level": 01158 }1159 ],1160 "created_by": {1161 "id": 21437,1162 "username": "Westerby",1163 "name": "Westerby",1164 "avatar_template": "/letter_avatar_proxy/v4/letter/w/779978/{size}.png"1165 },1166 "last_poster": {1167 "id": 54524,1168 "username": "dheeraj_agrawal",1169 "name": "dheeraj agrawal",1170 "avatar_template": "/user_avatar/discuss.pytorch.org/dheeraj_agrawal/{size}/48070_2.png"1171 },1172 "links": [1173 {1174 "url": "http://www.mediafire.com/file/rel0ze3y963nohv/onnxruntime_profile__2020-08-28_10-35-51.json/file",1175 "title": "onnxruntime_profile__2020-08-28_10-35-51",1176 "internal": false,1177 "attachment": false,1178 "reflection": false,1179 "clicks": 25,1180 "user_id": 21437,1181 "domain": "www.mediafire.com",1182 "root_domain": "mediafire.com"1183 }1184 ]1185 },1186 "bookmarks": []1187 },1188 {1189 "post_stream": {1190 "posts": [1191 {1192 "id": 338396,1193 "name": "",1194 "username": "ljeonjko",1195 "avatar_template": "/letter_avatar_proxy/v4/letter/l/b5ac83/{size}.png",1196 "created_at": "2022-03-28T11:39:11.891Z",1197 "cooked": "<p>Hello!</p>\n<p>I have trained a CCGAN model and saved the generator and discriminator using</p>\n<pre><code class=\"lang-auto\">torch.save(gen.state_dict(), \"GENERATOR/gen.pt\")\ntorch.save(disc.state_dict(), \"DISCRIMINATOR/disc.pt\")\n</code></pre>\n<p>I now wish to test this model on a single image. (I have trained several models using several slightly different custom datasets, and I wish to see which dataset is the most fitting). How do I go about presenting the algorithm a single image in order to test the output? Is there a tutorial on this? Any tips are welcome.</p>\n<p>Tank you!</p>",1198 "post_number": 1,1199 "post_type": 1,1200 "posts_count": 12,