{"payload":{"feedbackUrl":"https://github.com/orgs/community/discussions/53140","repo":{"id":689102378,"defaultBranch":"this-main","name":"dfdx","ownerLogin":"swfsql","currentUserCanPush":false,"isFork":true,"isEmpty":false,"createdAt":"2023-09-08T20:04:39.000Z","ownerAvatar":"https://avatars.githubusercontent.com/u/1165577?v=4","public":true,"private":false,"isOrgOwned":false},"refInfo":{"name":"","listCacheKey":"v0:1709477852.0","currentOid":""},"activityList":{"items":[{"before":"8fc72c56069ca4931135450bdcea979dc0cf4bde","after":"2867a40e99ef4d81c71cba0d80c6f7886dcafed7","ref":"refs/heads/this-main","pushedAt":"2024-06-05T20:02:22.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"avoid miri rayon error","shortMessageHtmlLink":"avoid miri rayon error"}},{"before":"d6f75c2ebf3a3a168f2baf1acadc69864a90eddd","after":"4148841be572d996d9945e74e1146a759c6d29f9","ref":"refs/heads/derive-input-wrapper","pushedAt":"2024-03-03T23:34:01.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"rename local variables for input_wrapper","shortMessageHtmlLink":"rename local variables for input_wrapper"}},{"before":"c4a2995f8eb237595a01f5abf23b5a13ebb0e748","after":"8fc72c56069ca4931135450bdcea979dc0cf4bde","ref":"refs/heads/this-main","pushedAt":"2024-03-03T23:17:23.000Z","pushType":"push","commitsCount":3,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"Merge branch 'layer-ops' into this-main","shortMessageHtmlLink":"Merge branch 'layer-ops' into this-main"}},{"before":"cf811fb0bba08bb84f193b5635b9bafa517b728a","after":"973a2f1a8c68a52d0e571f796cbbc3dda6f4f7e7","ref":"refs/heads/layer-ops","pushedAt":"2024-03-03T23:13:50.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"mv and add op layers\n\n- Added serialization for some data from layers:\n - upscale2d\n - reshape\n - pool_2d_avg\n - pool_2d_max\n - pool_2d_min\n - leaky_relu\n - dropout\n- Added layers for more tensor ops:\n - add\n - bce\n - boolean\n - broasdcast_to\n - choose\n - clamp\n - cmp\n - div\n - huber_error\n - logsumexp_to\n - max_to\n - maximum\n - min_to\n - minimum\n - mean_to\n - mul\n - nans_to\n - negate\n - normalize\n - permute_to\n - pow\n - realize_to\n - stddev_to\n - var_to","shortMessageHtmlLink":"mv and add op layers"}},{"before":null,"after":"cf811fb0bba08bb84f193b5635b9bafa517b728a","ref":"refs/heads/layer-ops","pushedAt":"2024-03-03T14:57:32.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"init","shortMessageHtmlLink":"init"}},{"before":"557687c0a9e29dfba2311fe67414863c6c5137bf","after":"11759036c4ebdc8d745b80b066cc373cd99ca6f8","ref":"refs/heads/avoid-ci-errors","pushedAt":"2024-03-01T17:10:53.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"Merge pull request #2 from swfsql/avoid-ci-errors\n\nAvoid ci errors","shortMessageHtmlLink":"Merge pull request #2 from swfsql/avoid-ci-errors"}},{"before":null,"after":"faa5ccc16461e37f836694c6a033dfa7e51aa3e7","ref":"refs/heads/prodigy","pushedAt":"2024-03-01T17:09:09.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"add Prodigy optimizer","shortMessageHtmlLink":"add Prodigy optimizer"}},{"before":null,"after":"8f43ce54ee1fe9ff14ca7846dd097e7ce4a0a7cb","ref":"refs/heads/miri-test","pushedAt":"2024-03-01T17:08:35.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"Free cache memory on `TensorCache` drop\n\n- Moved the impl of `Cache::try_empty_cache()` to `TensorCache::clear()`.\n - This can be invoked both by `Cache::try_empty_cache()` and by `drop(TensorCache)`.\n- Moved the device cache ptr deallocation to `BytesPtr`, `CudaBytesPtr` (newtype over `CUdeviceptr`) and `Buffer`.\n - This is abstracted by the `CachePtr` trait.\n - Can be called by `TensorCache::clear()`.\n - This method may require some \"extra\" device information, such as in the cuda case. That information is held by `TensorCache`.","shortMessageHtmlLink":"Free cache memory on TensorCache drop"}},{"before":"11759036c4ebdc8d745b80b066cc373cd99ca6f8","after":"4722a99d303f347d6088d95867d007c75ca6dd78","ref":"refs/heads/main","pushedAt":"2024-03-01T17:05:41.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"feat(wgpu): add to_dtype kernel (#906)\n\n* feat(wgpu): add to_dtype kernel\r\n\r\n* fix: add WebGPUNativeType\r\n\r\n* style: clippy fix\r\n\r\n---------\r\n\r\nCo-authored-by: Corey Lowman ","shortMessageHtmlLink":"feat(wgpu): add to_dtype kernel (coreylowman#906)"}},{"before":"997a5d3c07b6384c08be922cb8b0af06b9a16b21","after":null,"ref":"refs/heads/derive-input-collection2","pushedAt":"2024-03-01T17:03:11.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"}},{"before":"ed88b612f29fb84d46564ecbad68e405f488f6ef","after":null,"ref":"refs/heads/derive-input-wrapper2","pushedAt":"2024-03-01T17:02:22.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"}},{"before":"b298a98eae178ee4a960bad52d196c536f4d6b00","after":null,"ref":"refs/heads/issue-881","pushedAt":"2024-03-01T17:01:22.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"}},{"before":null,"after":"7a21ba7f966532c41d4aecc1a75d74a29a7408de","ref":"refs/heads/gradient-clipping","pushedAt":"2024-03-01T16:59:02.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"Adds Storage and Gradient view/mutating methods; Adds grads clamping and cliping\n\n- Added `dfdx::nn_traits::WithGrads` trait and `dfdx_derives::WithGrads` proc macro, basead on `ZeroGrads`.\n- Added `dfdx_core::tensor::WithStorage` trait.\n- Changed some methods from `Gradients`:\n - Exposed `get_mut` as `pub`.\n - Exposed `get_ref` as `pub`, and lower the requirements from `&mut self` to `&self`.\n- Added gradient clamping and cliping methods.","shortMessageHtmlLink":"Adds Storage and Gradient view/mutating methods; Adds grads clamping …"}},{"before":null,"after":"e81228c300a8a48c4e257bdaeb71c46fcc8b18be","ref":"refs/heads/conv-trans2d-output-padding","pushedAt":"2024-03-01T16:58:22.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"Adds `OUTPUT_PADDING` to `ConvTrans2D`\n\n- Draft state.\n- Unsure if correct, but a very simple and quick test gives the same\n result from pytorch.\n- Note: Tensorflow result differs, both from dfdx and from pytorch.\n\nReference pytorch test:\n```python\nimport torch\n\nx = np.array([[[[0.1, 0.7], [0.3, 0.4]]]])\nw = np.array([[[[-0.1, -0.3, 0.7], [0.8, -0.2, 0.1], [0.3, 0.4, -0.5]]]])\n\na = torch.nn.ConvTranspose2d(output_padding=0, in_channels=1, out_channels=1, kernel_size=3, stride=2, padding=1, bias = False)\nb = torch.nn.ConvTranspose2d(output_padding=1, in_channels=1, out_channels=1, kernel_size=3, stride=2, padding=1, bias = False)\n\nx = torch.from_numpy(x).float()\nw0 = torch.from_numpy(w).float()\n\nwith torch.no_grad():\n a.weight = torch.nn.Parameter(w0)\n b.weight = torch.nn.Parameter(w0)\n\nya = a(x)\nyb = b(x)\n\nprint(ya.size()) # torch.Size([1, 1, 3, 3])\nprint(yb.size()) # torch.Size([1, 1, 4, 4])\n\nprint(ya)\nprint(yb)\n```","shortMessageHtmlLink":"Adds OUTPUT_PADDING to ConvTrans2D"}},{"before":"ec6a85fd129601751d43478fba2ac7841ec248d2","after":"c4a2995f8eb237595a01f5abf23b5a13ebb0e748","ref":"refs/heads/this-main","pushedAt":"2024-03-01T16:42:45.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"Merge branch 'gradient-clipping' into new-base","shortMessageHtmlLink":"Merge branch 'gradient-clipping' into new-base"}},{"before":"c96938a8ce7088fbc74ecd4c45e83ba7e47be1c2","after":"d6f75c2ebf3a3a168f2baf1acadc69864a90eddd","ref":"refs/heads/derive-input-wrapper","pushedAt":"2024-03-01T16:39:43.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"rename local variables for input_wrapper","shortMessageHtmlLink":"rename local variables for input_wrapper"}},{"before":"8655f2a582d2731faeace8f91ae1105b1caa07c8","after":null,"ref":"refs/heads/issue-891","pushedAt":"2024-03-01T16:32:27.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"}},{"before":"ac3cd919c7b05a7ad9707c794da93999944ba529","after":null,"ref":"refs/heads/issue-885","pushedAt":"2024-03-01T16:31:32.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"}},{"before":"57191cbdd44034557c435b6404faf0de70a9b0ee","after":"9ad9164cedf6246a4aa818ac19714214dc2db03b","ref":"refs/heads/mamba-minimal","pushedAt":"2024-03-01T16:24:00.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"add nightly requirement for mamba-minimal","shortMessageHtmlLink":"add nightly requirement for mamba-minimal"}},{"before":"07a665357c328028785cc625f14d4630655b9fed","after":"e81228c300a8a48c4e257bdaeb71c46fcc8b18be","ref":"refs/heads/issue-889","pushedAt":"2024-03-01T16:19:46.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"Adds `OUTPUT_PADDING` to `ConvTrans2D`\n\n- Draft state.\n- Unsure if correct, but a very simple and quick test gives the same\n result from pytorch.\n- Note: Tensorflow result differs, both from dfdx and from pytorch.\n\nReference pytorch test:\n```python\nimport torch\n\nx = np.array([[[[0.1, 0.7], [0.3, 0.4]]]])\nw = np.array([[[[-0.1, -0.3, 0.7], [0.8, -0.2, 0.1], [0.3, 0.4, -0.5]]]])\n\na = torch.nn.ConvTranspose2d(output_padding=0, in_channels=1, out_channels=1, kernel_size=3, stride=2, padding=1, bias = False)\nb = torch.nn.ConvTranspose2d(output_padding=1, in_channels=1, out_channels=1, kernel_size=3, stride=2, padding=1, bias = False)\n\nx = torch.from_numpy(x).float()\nw0 = torch.from_numpy(w).float()\n\nwith torch.no_grad():\n a.weight = torch.nn.Parameter(w0)\n b.weight = torch.nn.Parameter(w0)\n\nya = a(x)\nyb = b(x)\n\nprint(ya.size()) # torch.Size([1, 1, 3, 3])\nprint(yb.size()) # torch.Size([1, 1, 4, 4])\n\nprint(ya)\nprint(yb)\n```","shortMessageHtmlLink":"Adds OUTPUT_PADDING to ConvTrans2D"}},{"before":"4c6d8c7efaeabc4fa843c575366156df99b53341","after":"faa5ccc16461e37f836694c6a033dfa7e51aa3e7","ref":"refs/heads/issue-894","pushedAt":"2024-03-01T16:16:54.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"add Prodigy optimizer","shortMessageHtmlLink":"add Prodigy optimizer"}},{"before":"5f37c8262531ed3e919a0340deb4ab9aefdc649d","after":"8f43ce54ee1fe9ff14ca7846dd097e7ce4a0a7cb","ref":"refs/heads/issue-897","pushedAt":"2024-03-01T16:15:52.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"Free cache memory on `TensorCache` drop\n\n- Moved the impl of `Cache::try_empty_cache()` to `TensorCache::clear()`.\n - This can be invoked both by `Cache::try_empty_cache()` and by `drop(TensorCache)`.\n- Moved the device cache ptr deallocation to `BytesPtr`, `CudaBytesPtr` (newtype over `CUdeviceptr`) and `Buffer`.\n - This is abstracted by the `CachePtr` trait.\n - Can be called by `TensorCache::clear()`.\n - This method may require some \"extra\" device information, such as in the cuda case. That information is held by `TensorCache`.","shortMessageHtmlLink":"Free cache memory on TensorCache drop"}},{"before":"c8eb559bad8d88b35eeadc4fb9f395f0c45f7976","after":"7a21ba7f966532c41d4aecc1a75d74a29a7408de","ref":"refs/heads/issue-596","pushedAt":"2024-03-01T16:14:59.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"Adds Storage and Gradient view/mutating methods; Adds grads clamping and cliping\n\n- Added `dfdx::nn_traits::WithGrads` trait and `dfdx_derives::WithGrads` proc macro, basead on `ZeroGrads`.\n- Added `dfdx_core::tensor::WithStorage` trait.\n- Changed some methods from `Gradients`:\n - Exposed `get_mut` as `pub`.\n - Exposed `get_ref` as `pub`, and lower the requirements from `&mut self` to `&self`.\n- Added gradient clamping and cliping methods.","shortMessageHtmlLink":"Adds Storage and Gradient view/mutating methods; Adds grads clamping …"}},{"before":"ae777d4699faa76eda4ad8167be959b6e0d9075a","after":"5f37c8262531ed3e919a0340deb4ab9aefdc649d","ref":"refs/heads/issue-897","pushedAt":"2024-03-01T16:12:54.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"Free cache memory on `TensorCache` drop\n\n- Moved the impl of `Cache::try_empty_cache()` to `TensorCache::clear()`.\n - This can be invoked both by `Cache::try_empty_cache()` and by `drop(TensorCache)`.\n- Moved the device cache ptr deallocation to `BytesPtr`, `CudaBytesPtr` (newtype over `CUdeviceptr`) and `Buffer`.\n - This is abstracted by the `CachePtr` trait.\n - Can be called by `TensorCache::clear()`.\n - This method may require some \"extra\" device information, such as in the cuda case. That information is held by `TensorCache`.","shortMessageHtmlLink":"Free cache memory on TensorCache drop"}},{"before":"81f644dea2428463ff8984e19e8018df7a180c91","after":"c8eb559bad8d88b35eeadc4fb9f395f0c45f7976","ref":"refs/heads/issue-596","pushedAt":"2024-03-01T16:08:41.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"Adds Storage and Gradient view/mutating methods; Adds grads clamping and cliping\n\n- Added `dfdx::nn_traits::WithGrads` trait and `dfdx_derives::WithGrads` proc macro, basead on `ZeroGrads`.\n- Added `dfdx_core::tensor::WithStorage` trait.\n- Changed some methods from `Gradients`:\n - Exposed `get_mut` as `pub`.\n - Exposed `get_ref` as `pub`, and lower the requirements from `&mut self` to `&self`.\n- Added gradient clamping and cliping methods.","shortMessageHtmlLink":"Adds Storage and Gradient view/mutating methods; Adds grads clamping …"}},{"before":"bff1b658aa3b91b0af57f037c7cc704d216d1f03","after":"57191cbdd44034557c435b6404faf0de70a9b0ee","ref":"refs/heads/mamba-minimal","pushedAt":"2024-03-01T15:56:59.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"add nightly requirement for mamba-minimal","shortMessageHtmlLink":"add nightly requirement for mamba-minimal"}},{"before":"45226f597a6e1028bec83579250f87377c0add91","after":"95e619f3a4d55fc97b668021d34fd732a5dfdc36","ref":"refs/heads/safetensors-change","pushedAt":"2024-03-01T15:53:25.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"allow to load safetensors from a byte array","shortMessageHtmlLink":"allow to load safetensors from a byte array"}},{"before":"c695a15eb3472157ea4467884e33b69dce0756a6","after":"664a907dcae0d9a1a137e07fac12802cca537992","ref":"refs/heads/unstack","pushedAt":"2024-03-01T15:52:43.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"unstack fixes","shortMessageHtmlLink":"unstack fixes"}},{"before":"d8c29b00c6c2e50f71c4c4f079d26aabf6170e76","after":"bc569c7515ce5c8ddaff23eaf19cad085f314c85","ref":"refs/heads/silu","pushedAt":"2024-03-01T15:52:10.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"silu: fix cpu df","shortMessageHtmlLink":"silu: fix cpu df"}},{"before":"de5556737ec560990c4278418f1f31f16d532d21","after":"fdcd9432e3c101cc759a784907568c76c155852e","ref":"refs/heads/rms-norm","pushedAt":"2024-03-01T15:51:40.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"swfsql","name":"Thiago Machado","path":"/swfsql","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/1165577?s=80&v=4"},"commit":{"message":"rm unrelated derive","shortMessageHtmlLink":"rm unrelated derive"}}],"hasNextPage":true,"hasPreviousPage":false,"activityType":"all","actor":null,"timePeriod":"all","sort":"DESC","perPage":30,"cursor":"Y3Vyc29yOnYyOpK7MjAyNC0wNi0wNVQyMDowMjoyMi4wMDAwMDBazwAAAARdg2Eb","startCursor":"Y3Vyc29yOnYyOpK7MjAyNC0wNi0wNVQyMDowMjoyMi4wMDAwMDBazwAAAARdg2Eb","endCursor":"Y3Vyc29yOnYyOpK7MjAyNC0wMy0wMVQxNTo1MTo0MC4wMDAwMDBazwAAAAQJ69mb"}},"title":"Activity · swfsql/dfdx"}