{"payload":{"feedbackUrl":"https://github.com/orgs/community/discussions/53140","repo":{"id":518064189,"defaultBranch":"main","name":"kaolin-wisp","ownerLogin":"NVIDIAGameWorks","currentUserCanPush":false,"isFork":false,"isEmpty":false,"createdAt":"2022-07-26T13:03:29.000Z","ownerAvatar":"https://avatars.githubusercontent.com/u/7717624?v=4","public":true,"private":false,"isOrgOwned":true},"refInfo":{"name":"","listCacheKey":"v0:1693238016.0","currentOid":""},"activityList":{"items":[{"before":"49f8bb7ea3cb6d93e307a3e8ee65088fe6abaa82","after":"931707e50f1511fdb4af55eeb4aed4df23b7c2b1","ref":"refs/heads/main","pushedAt":"2024-05-02T12:02:25.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"Missing uint typedef (#187)\n\n#179","shortMessageHtmlLink":"Missing uint typedef (#187)"}},{"before":"6d1eacc9c464092b2b6ee9fe190b638f565c5c5d","after":"49f8bb7ea3cb6d93e307a3e8ee65088fe6abaa82","ref":"refs/heads/main","pushedAt":"2024-05-01T01:47:45.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"Update requirements.txt","shortMessageHtmlLink":"Update requirements.txt"}},{"before":"464ccee8628198a01c0b75d236d65ed9e54201bc","after":"6d1eacc9c464092b2b6ee9fe190b638f565c5c5d","ref":"refs/heads/main","pushedAt":"2023-10-11T09:37:14.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"Reduce VRAM usage due to ray tensor generation by setting the tensors' dtype to torch.float32 (#180)\n\n# Why\r\nDataset preparation required more than 12GB VRAM on the lego dataset\r\nwhen I tried to train a model with the preset `nerf_hash` config.\r\n\r\n# What\r\nRay tensors are generated as float64 tensors during the dataset\r\npreparation in `main` while they are converted to float32 just a few\r\nlines below. This PR changes `dtype` for the tensor generation to\r\nfloat32. This modification reduced VRAM usage (14GB-ish to 8GB-ish, but\r\nmay vary) with the preset config.\r\n\r\nP.S. The RTMV dataset may need a similar change, but this PR does not include it.","shortMessageHtmlLink":"Reduce VRAM usage due to ray tensor generation by setting the tensors…"}},{"before":"47820f180c4d3d25ac26a7ed4f8ec37bf4c7b25d","after":"464ccee8628198a01c0b75d236d65ed9e54201bc","ref":"refs/heads/main","pushedAt":"2023-10-03T14:17:59.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"Bugfix/image training (#177)\n\nSome minor fixes in the image training pipeline. Coord_dim argument was\r\nnot being passed to constructors and defaulted to 3 for 3 dimensions\r\neven in 2d case. Hash grid from geometric had a bug in the progression\r\nof the geometric sequence, resulting in the minimum resolution being off\r\nby one.","shortMessageHtmlLink":"Bugfix/image training (#177)"}},{"before":"1d8a4dad23e0bdc7f45f8659c1c28af97760a6c7","after":"47820f180c4d3d25ac26a7ed4f8ec37bf4c7b25d","ref":"refs/heads/main","pushedAt":"2023-10-03T12:57:23.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"Multiple fixes for recently broken paths (#181)\n\n1. NeRF configs: have a fallback with AdamW in case apex/FusedAdam is\r\nunavailable\r\n2. Update config system to account for targets with missing imports\r\n(i.e. apex)\r\n3. Fixed syntax errors in RTMV dataset\r\n4. Fix interactive renderer FPS issues when inference num_steps is too\r\nhigh\r\n5. Jupyter Notebook live visualizer fixed\r\n6. Added uniform raytracer option to GUI\r\n7. Tests framework fixed to match new console output from tracker\r\n8. SIGGRAPH demo example fixed\r\n\r\nSigned-off-by: operel ","shortMessageHtmlLink":"Multiple fixes for recently broken paths (#181)"}},{"before":"15e12ad6006a80061538eec83758b96f63afaa13","after":"1d8a4dad23e0bdc7f45f8659c1c28af97760a6c7","ref":"refs/heads/main","pushedAt":"2023-09-18T08:45:58.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"Fix HashGrid.from_resolutions constructor config support (#176)\n\nSets a default for HashGrid.from_resolutions args as tyro requires\r\nvariable sized lists to have some inferred length.\r\n\r\nSigned-off-by: operel ","shortMessageHtmlLink":"Fix HashGrid.from_resolutions constructor config support (#176)"}},{"before":"e4dfb67163718b906e9637bfe7a4c556b592077d","after":"15e12ad6006a80061538eec83758b96f63afaa13","ref":"refs/heads/main","pushedAt":"2023-08-28T16:03:12.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"new unicorn sampling method (#175)\n\nadded uniform sampling code\r\n\r\nCo-authored-by: Charles Loop ","shortMessageHtmlLink":"new unicorn sampling method (#175)"}},{"before":null,"after":"cd638be15dc5a3df524b6263f6c52b87dcaf5b48","ref":"refs/heads/cloop/uniform_sample","pushedAt":"2023-08-28T15:53:36.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"charlesloop","name":null,"path":"/charlesloop","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/3460020?s=80&v=4"},"commit":{"message":"added from Gitlab","shortMessageHtmlLink":"added from Gitlab"}},{"before":"246b1ec434c373bf2e649cc27b1ea04a4d73ebdb","after":"e4dfb67163718b906e9637bfe7a4c556b592077d","ref":"refs/heads/main","pushedAt":"2023-08-16T09:00:36.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"RenderCore fix to needs_refresh (#174)\n\nFix interactive visualizer refresh issue introduced in #167\r\n\r\nSigned-off-by: operel ","shortMessageHtmlLink":"RenderCore fix to needs_refresh (#174)"}},{"before":"07539e49b8e9db595dbc1ba01287e3453e5c7c36","after":"246b1ec434c373bf2e649cc27b1ea04a4d73ebdb","ref":"refs/heads/main","pushedAt":"2023-08-03T17:18:29.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"Implement closest_point function (#173)\n\nBackground:\r\n\r\nTraining SDF + texture model was not available because closest_point\r\nfunction wasn't implemented. (see #162)\r\nTo traing the SDF + texture model, I implement closest_point function.\r\n\r\nChanges:\r\n\r\n1. **Change all float variables** in\r\nwisp/csrc/external/mesh2sdf_kernel.cu **into double variables**.\r\nPreviously, when using float variables, there were accumulated numerical\r\nerrors on SDF and closest points' coordinates. (the numerical SDF errors\r\neven reaches 1.xxx ...)\r\n2. Add mesh2sdf_triangle_gpu_fast_nopre function in\r\nwisp/csrc/external/mesh2sdf_kernel.cu which can **return both SDF and\r\nclosest triangle**.\r\n3. Implement closest_point function in wisp/ops/mesh/closest_point.py.\r\nclosest points are calculated from closest triangles returned from\r\nmesh2sdf_triangle_gpu_fast_nopre function.\r\n\r\nTests:\r\n\r\nUnfortunately, I'm not good at pytest code. I just tested the code by\r\nprinting results on terminal.\r\n\r\nTest process:\r\n\r\n1. Add print() in closest_tex() function (wisp/ops/mesh/closest_tex.py) \r\n```python\r\nimport torch\r\nimport numpy as np\r\nfrom .barycentric_coordinates import barycentric_coordinates\r\nfrom .closest_point import closest_point\r\nfrom .sample_tex import sample_tex\r\n\r\ndef closest_tex(\r\n V : torch.Tensor, \r\n F : torch.Tensor,\r\n TV : torch.Tensor,\r\n TF : torch.Tensor,\r\n materials,\r\n points : torch.Tensor):\r\n \"\"\"Returns the closest texture for a set of points.\r\n\r\n V (torch.FloatTensor): mesh vertices of shape [V, 3] \r\n F (torch.LongTensor): mesh face indices of shape [F, 3]\r\n TV (torch.FloatTensor): \r\n TF (torch.FloatTensor):\r\n materials:\r\n points (torch.FloatTensor): sample locations of shape [N, 3]\r\n\r\n Returns:\r\n (torch.FloatTensor): texture samples of shape [N, 3]\r\n \"\"\"\r\n\r\n TV = TV.to(V.device)\r\n TF = TF.to(V.device)\r\n points = points.to(V.device)\r\n\r\n dist, hit_pts, hit_tidx = closest_point(V, F, points)\r\n\r\n ##### !!!test by printing!!! #####\r\n ######################### \r\n print(dist) # should be same with below result\r\n print(((hit_pts - points) ** 2).sum(dim=1).sqrt()) # distance between hit_pts and points should be same with dist\r\n ######################### \r\n ######################### \r\n\r\n hit_F = F[hit_tidx]\r\n hit_V = V[hit_F]\r\n BC = barycentric_coordinates(hit_pts.cuda(), hit_V[:,0], hit_V[:,1], hit_V[:,2])\r\n\r\n hit_TF = TF[hit_tidx]\r\n hit_TM = hit_TF[...,3]\r\n hit_TF = hit_TF[...,:3]\r\n\r\n if TV.shape[0] > 0:\r\n hit_TV = TV[hit_TF]\r\n hit_Tp = (hit_TV * BC.unsqueeze(-1)).sum(1)\r\n else:\r\n hit_Tp = BC\r\n \r\n rgb = sample_tex(hit_Tp, hit_TM, materials)\r\n\r\n return rgb, hit_pts, dist\r\n\r\n```\r\n2. Run following code\r\n```\r\nfrom wisp.datasets.formats import MeshSampledSDFDataset\r\n\r\n\r\ndataset = MeshSampledSDFDataset(mesh_path=\"data/obj/mesh.obj\", split=\"train\", sample_mode=[\"trace\"], num_samples=5000, sample_tex=True, mode_norm=\"sphere\")\r\n```\r\n\r\nResults:\r\n\r\nI also trained the SDF + texture model on some .obj files after\r\nimplementing closest_point function. (However, I don't include training\r\ncode on this PR)\r\n\r\nFirst image is .obj model in blender and second image is the SDF +\r\ntexture model result. (wrong eye coloring is due to wrong .obj texture\r\ndata, so you can ignore it)\r\n\r\n\r\n","shortMessageHtmlLink":"Implement closest_point function (#173)"}},{"before":"a5850648fa83a610e7457a96cbb64375790b4c63","after":"07539e49b8e9db595dbc1ba01287e3453e5c7c36","ref":"refs/heads/main","pushedAt":"2023-08-02T16:23:29.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"Add image training pipeline (#172)\n\nAdds a training pipeline for images!\r\n\r\n(docs to come later, but the usage should be pretty straight forward)\r\n\r\nSigned-off-by: Towaki Takikawa ","shortMessageHtmlLink":"Add image training pipeline (#172)"}},{"before":"e44604cb868f6dc226cde454549efecb0eefd130","after":"2a7d60587b2c1f65bc44d1c04fbb185ee82e5002","ref":"refs/heads/stable","pushedAt":"2023-08-02T15:27:38.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"Fixing unexpected processing of iterate() of BaseTrainer (#163)\n\nHi, thanks for developing such an amazing library. This is a PR to fix\r\nthe unexpected behaviour of `iterate()`. Please refer to the following\r\ndescription. I hope this request will be approved.\r\n\r\n## Changes proposed in this PR\r\n* Initial value for `self.iteration` of BaseTrainer in the `stable`\r\nbranch.\r\n* Processing flow inside `iterate()` of the class\r\n\r\n## Reason for the proposal\r\nThe original iterate() calls `step()` before `pre_training()`, and\r\n`post_training()` before and after the last `step()`.\r\n\r\n## Reference\r\nBaseTrainer in the `main` branch","shortMessageHtmlLink":"Fixing unexpected processing of iterate() of BaseTrainer (#163)"}},{"before":"5740f7394f5fdcf9fd3e8adc4dad6d7e2ecfc4ba","after":"a5850648fa83a610e7457a96cbb64375790b4c63","ref":"refs/heads/main","pushedAt":"2023-08-02T04:01:58.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"Make sure tracker actually logs config and fix key filtering (#171)\n\nRight now keys that are not actually config elements would get stored in\r\nthe dataframe.\r\n\r\nIn addition, make sure the config is logged by default.\r\n\r\nSigned-off-by: Towaki Takikawa ","shortMessageHtmlLink":"Make sure tracker actually logs config and fix key filtering (#171)"}},{"before":"52659370009d89c582bdb0c6f072a1b639dad2ca","after":"5740f7394f5fdcf9fd3e8adc4dad6d7e2ecfc4ba","ref":"refs/heads/main","pushedAt":"2023-08-01T18:50:18.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"Bug fix in base nef (#169)\n\nAddresses a bug where argspec will crash if there is a mismatch in the\r\nargs AND there is no optional args to be parsed.\r\n\r\nAlso addresses a case where more than one forward function can return\r\nthe same channel, in which case the ordering of the forward function\r\nwill be such that the number of invocations will be minimized by\r\nprioritizing forward functions with more channel matches.\r\n\r\n---------\r\n\r\nSigned-off-by: Towaki Takikawa ","shortMessageHtmlLink":"Bug fix in base nef (#169)"}},{"before":"ad6af84a4d6aac70542fa0639d32cae3ac90f5b5","after":"52659370009d89c582bdb0c6f072a1b639dad2ca","ref":"refs/heads/main","pushedAt":"2023-08-01T18:43:25.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"More robust arg parsing in base tracer (#168)\n\nThe original implementation would break when there is any decorator in\r\nuse for the trace function (for example, nvtx).\r\n\r\nSigned-off-by: Towaki Takikawa ","shortMessageHtmlLink":"More robust arg parsing in base tracer (#168)"}},{"before":"43afbfe9ed37e1a0844b3ba5dafebcd7a37fc696","after":"ad6af84a4d6aac70542fa0639d32cae3ac90f5b5","ref":"refs/heads/main","pushedAt":"2023-08-01T18:25:15.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"Various bug fixes and quality improvements (#167)\n\nList of changes:\r\n\r\n1. New config arguments (especially in hash grid nerf) which improve\r\nquality significantly\r\n2. Adaptive ray batching (improves nerf quality a lot)\r\n3. `bg_color` is now a tuple `(0.0, 0.0, 0.0)` not a string\r\n4. Supporting other validation splits `--valid-split`\r\n5. `apex` based fused Adam\r\n6. Bug fix in rays for `dist_min`, inproper normalization in\r\n`channel_fn`\r\n7. 2D support for hash grid\r\n8. `MultiTable` utility class for working with hash grids (and similar)\r\neasier\r\n9. App renderer will now try to make FPS stable for better UX \r\n10. Fixing iteration count\r\n11. Fix tracker's tensorboard support\r\n\r\n---------\r\n\r\nSigned-off-by: Towaki Takikawa \r\nCo-authored-by: Towaki Takikawa ","shortMessageHtmlLink":"Various bug fixes and quality improvements (#167)"}},{"before":"67b38adb21164cf54c27707306090409c5b43f57","after":"43afbfe9ed37e1a0844b3ba5dafebcd7a37fc696","ref":"refs/heads/main","pushedAt":"2023-08-01T13:56:17.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"Rasterization pipeline for interactive renderer (#166)\n\nEnables visualizing rasterized pipelines within the interactive\r\nrenderer, i.e. rendered meshes.\r\n`RasterizedRenderer` is now patched to support rasterized objects within\r\nthe scene graph.\r\nThis MR does not support optimization of differentiable rasterizers (see\r\nkaolin DIB-R / nvdiffrast).\r\n\r\nThe usage of this MR is intended to be as follows:\r\n```\r\nfrom wisp.renderer.core.api import RasterizedRenderer, register_rasterizer\r\n@register_rasterizer(YOUR_RASTERIZER_CLASS_HERE)\r\nclass YourWispSceneGraphNodeRenderer(RasterizedRenderer): \r\n @torch.no_grad()\r\n def render(self, camera: Camera) -> RenderBuffer:\r\n ...\r\n```\r\n\r\nSigned-off-by: operel ","shortMessageHtmlLink":"Rasterization pipeline for interactive renderer (#166)"}},{"before":"45b68cdad5234ffe2abac0168d8021f6c9e5efe3","after":"67b38adb21164cf54c27707306090409c5b43f57","ref":"refs/heads/main","pushedAt":"2023-08-01T13:02:18.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"RTMV fix due to kaolin 0.14 camera updates (#165)\n\nSigned-off-by: operel ","shortMessageHtmlLink":"RTMV fix due to kaolin 0.14 camera updates (#165)"}},{"before":"766318d6feb1d3e62bb2d39c64aabc8e12a1bf73","after":"45b68cdad5234ffe2abac0168d8021f6c9e5efe3","ref":"refs/heads/main","pushedAt":"2023-07-31T17:30:09.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"Remove unneeded GC","shortMessageHtmlLink":"Remove unneeded GC"}},{"before":"25249a1db98a603d10b02cde38e822f41b4f26ac","after":"766318d6feb1d3e62bb2d39c64aabc8e12a1bf73","ref":"refs/heads/main","pushedAt":"2023-07-31T17:23:56.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"tovacinni","name":"Towaki","path":"/tovacinni","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2717105?s=80&v=4"},"commit":{"message":"Memory optimization patch in dataset initialization (#164)\n\n1. Don't resize with cv2 if it's not needed\r\n2. Batch the CPU transfer\r\n\r\nSigned-off-by: Towaki Takikawa ","shortMessageHtmlLink":"Memory optimization patch in dataset initialization (#164)"}},{"before":"8682ab99f66508798afc7edf2120ec41df34a116","after":"25249a1db98a603d10b02cde38e822f41b4f26ac","ref":"refs/heads/main","pushedAt":"2023-05-29T11:47:19.394Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"Fix torch 2.0 compatability (#158)\n\n1. Fixes torch 2.0 compatibility issue with `torch._six`\r\n2. Corrects `nerf_octree` documentation incorrectly pointing at lego\r\ndataset by default (config uses RTMV by default)\r\n\r\nSigned-off-by: operel ","shortMessageHtmlLink":"Fix torch 2.0 compatability (#158)"}},{"before":"f4894568a07f74a61dc1d244caa30d6ba0b78ab4","after":"8682ab99f66508798afc7edf2120ec41df34a116","ref":"refs/heads/main","pushedAt":"2023-05-23T00:30:49.494Z","pushType":"push","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"Update README.md to point to docsite","shortMessageHtmlLink":"Update README.md to point to docsite"}},{"before":"99639ae60de4d1c6f4f721e3b6d1004e258afa5b","after":"f4894568a07f74a61dc1d244caa30d6ba0b78ab4","ref":"refs/heads/main","pushedAt":"2023-05-23T00:08:02.614Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"docsite (#157)\n\nInitial version of wisp's doc-site.\r\n\r\nSigned-off-by: operel ","shortMessageHtmlLink":"docsite (#157)"}},{"before":"6de48c8877573b39d062ab073799ded9de93d0ad","after":"99639ae60de4d1c6f4f721e3b6d1004e258afa5b","ref":"refs/heads/main","pushedAt":"2023-05-22T14:01:36.133Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"Experiment trackers and trainers (#156)\n\nThis MR aims to refactor trainers to support the new config system.\r\n\r\n* Trainer ctors no longer receive >10 args, instead they take the\r\nminimal requires args: a config, a pipeline, a dataset, etc.\r\n* Usage of extra_args was removed and replaced with the config classes\r\n* New Tracker modules (trackers, metric board and refined api for\r\noffline visualizer) - this removes a lot of logging code from the\r\ntrainers\r\n\r\nThe Tracker is a one stop for everything related to logging and\r\nvisualizations.\r\n\r\nSigned-off-by: operel ","shortMessageHtmlLink":"Experiment trackers and trainers (#156)"}},{"before":"552f7f0918278b3c1fc3666d325395f1b72196fd","after":"6de48c8877573b39d062ab073799ded9de93d0ad","ref":"refs/heads/main","pushedAt":"2023-05-22T13:54:01.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"tyro + hydrazen new config system (#155)\n\nThis is the core code of the new config system which wraps around tyro\r\nand hydra-zen.\r\nThis PR excludes changes needed to make to existing code.\r\n\r\nSigned-off-by: operel ","shortMessageHtmlLink":"tyro + hydrazen new config system (#155)"}},{"before":"e44604cb868f6dc226cde454549efecb0eefd130","after":"552f7f0918278b3c1fc3666d325395f1b72196fd","ref":"refs/heads/main","pushedAt":"2023-05-22T13:20:20.452Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"Added typing to functions, and blas construction separated from grid (#154)\n\nAdds typing to various constructors & functions, as a perquisite for the\r\nnew upcoming config system.\r\nThis MR also handles the blas construction outside of the feature grids,\r\nwhich allows for the removal of various grid constructors (this\r\nsimplifies the grids and makes the upcoming config less convoluted).\r\n\r\nSigned-off-by: operel ","shortMessageHtmlLink":"Added typing to functions, and blas construction separated from grid (#…"}},{"before":null,"after":"e44604cb868f6dc226cde454549efecb0eefd130","ref":"refs/heads/stable","pushedAt":"2023-05-18T08:58:04.096Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"Lower mem footprint for Jupyter (#144)\n\nPass raypack `batch_size` for inference time\r\n\r\nSigned-off-by: operel ","shortMessageHtmlLink":"Lower mem footprint for Jupyter (#144)"}},{"before":"a5894406fbb8084822474c9fdae00f40544df97a","after":null,"ref":"refs/heads/orperel-wandb-indent-fix","pushedAt":"2023-05-18T08:57:45.455Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"}},{"before":"cb47e10f376e5ac8b6965c650d8a6b85b9bc968e","after":"e44604cb868f6dc226cde454549efecb0eefd130","ref":"refs/heads/main","pushedAt":"2023-04-27T08:28:04.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"orperel","name":"Or Perel","path":"/orperel","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/9556101?s=80&v=4"},"commit":{"message":"Lower mem footprint for Jupyter (#144)\n\nPass raypack `batch_size` for inference time\r\n\r\nSigned-off-by: operel ","shortMessageHtmlLink":"Lower mem footprint for Jupyter (#144)"}},{"before":"6e2e72cabad4e6418b7d1331017f897c07056397","after":"cb47e10f376e5ac8b6965c650d8a6b85b9bc968e","ref":"refs/heads/main","pushedAt":"2023-04-14T18:42:07.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"shumash","name":"Maria Shugrina","path":"/shumash","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/298785?s=80&v=4"},"commit":{"message":"Interactive Wisp checkpoint visualization in Jupyter notebooks. (#140)\n\nAdded:\r\n- Example notebook with live camera control of a pretrained checkpoint\r\n- Utilities for working with ipycanvas and jupyter\r\n\r\nChanged:\r\n- Small adjustment to camera controller API to make `zoom` behavior more\r\npredictable\r\n- Fix in renderers factory\r\n\r\nRemaining:\r\n- Left API TODOs to make this work smoother in the future\r\n\r\nSigned-off-by: Maria Masha Shugrina \r\nCo-authored-by: Maria Masha Shugrina ","shortMessageHtmlLink":"Interactive Wisp checkpoint visualization in Jupyter notebooks. (#140)"}}],"hasNextPage":true,"hasPreviousPage":false,"activityType":"all","actor":null,"timePeriod":"all","sort":"DESC","perPage":30,"cursor":"djE6ks8AAAAEP79sVQA","startCursor":null,"endCursor":null}},"title":"Activity · NVIDIAGameWorks/kaolin-wisp"}