owkin/GrAIdient
66.8
Adequate · 17 November 2026
54k
lines of production code
Swift
primary language
1
measurement over time
What this system is
Behavioural changes
11 commits (0 fixes) modifying Sources/GrAIdient/Metal/Kernel
A change to existing behaviour in Sources/GrAIdient/Metal/Kernel — 11 commits, 44 files: “✨ feat(layer\_seq): RoPESeq (\#124)”; “✨ layer\_seq: MultiplySeq, SiLU & LLM test (\#127)”; “✨ feat(layer\_seq): RMSNormSeq (\#123)”; and 8 more (refs \#124, \#127, \#123, \#120, \#131, …).
Sources/GrAIdient/Metal/Kernel · medium confidence · unverified
12 commits (1 fix) modifying Sources/GrAIdient/Layer1D/Base
A change to existing behaviour in Sources/GrAIdient/Layer1D/Base — 12 commits (1 fix), 4 files: “🔨 refactor: dims checks for inputs and outputs (\#102)”; “🐛 fix: input GPU check (\#106)”; “✨ feat(core): initForward,Backward model API (\#109)”; and 9 more (refs \#102, \#106, \#109, \#120, \#88, …).
Sources/GrAIdient/Layer1D/Base · medium confidence · unverified
12 commits (1 fix) modifying Sources/GrAIdient/Layer2D/Base
A change to existing behaviour in Sources/GrAIdient/Layer2D/Base — 12 commits (1 fix), 4 files: “🔨 refactor: dims checks for inputs and outputs (\#102)”; “🐛 fix: input GPU check (\#106)”; “✨ feat(core): initForward,Backward model API (\#109)”; and 9 more (refs \#102, \#106, \#109, \#120, \#88, …).
Sources/GrAIdient/Layer2D/Base · medium confidence · unverified
14 commits (2 fixes) modifying Sources/GrAIdient/Core/Function
A change to existing behaviour in Sources/GrAIdient/Core/Function — 14 commits (2 fixs), 2 files: “✨ feat(core): GELU vs GELUApprox (\#113)”; “✨ feat(examples): adding Transformer training example (\#74)”; “✨ layer\_seq: MultiplySeq, SiLU & LLM test (\#127)”; and 11 more (refs \#113, \#74, \#127, \#123, \#120, …).
Sources/GrAIdient/Core/Function · medium confidence · unverified
21 commits (1 fix) modifying Sources/GrAIdient/Core/Layer
A change to existing behaviour in Sources/GrAIdient/Core/Layer — 21 commits (1 fix), 7 files: “✨ layer\_seq: MultiplySeq, SiLU & LLM test (\#127)”; “✨ feat: remove activation from layer (\#47)”; “🐛 bug(layer\_2d): align Convolution & Deconvolution on PyTorch (\#84)”; and 18 more (refs \#127, \#47, \#84, \#51, \#123, …).
Sources/GrAIdient/Core/Layer · medium confidence · unverified
27 commits (4 fixes) modifying Sources/GrAIdient/Layer1D
A change to existing behaviour in Sources/GrAIdient/Layer1D — 27 commits (4 fixs), 14 files: “✨ feat(layer\_seq): Concat1Seq & Concat2Seq (\#63)”; “🐛 fix: use buffers for neuron selection in SelectNeurons1D (\#50)”; “✨ layer\_seq: MultiplySeq, SiLU & LLM test (\#127)”; and 24 more (refs \#63, \#50, \#127, \#47, \#102, …).
Sources/GrAIdient/Layer1D · medium confidence · unverified
3 commits (0 fixes) modifying .swiftpm
A change to existing behaviour in .swiftpm — 3 commits, 1 file: “🔨 refactor: dims checks for inputs and outputs (\#102)”; “✨ feat: LayerMerge1D, Sum1D, Concat1D, Concat2D (\#43)”; “🎉 refactor!: re brand the framework (\#40)” (refs \#102, \#43, \#40).
.swiftpm · medium confidence · unverified
3 commits (0 fixes) modifying Sources/GrAIdient/Core/Optimizer
A change to existing behaviour in Sources/GrAIdient/Core/Optimizer — 3 commits, 4 files: “✨ feat: getDeltaWeightsGPU per sample API (\#41)”; “🔨 refactor: handle float16 along float on GPU (\#120)”; “🎉 refactor!: re brand the framework (\#40)” (refs \#41, \#120, \#40).
Sources/GrAIdient/Core/Optimizer · medium confidence · unverified
35 commits (3 fixes) modifying Sources/GrAIdient/LayerSeq
A change to existing behaviour in Sources/GrAIdient/LayerSeq — 35 commits (3 fixs), 18 files: “✨ feat(layer\_seq): RoPESeq (\#124)”; “✨ feat(layer\_seq): Concat1Seq & Concat2Seq (\#63)”; “✨ feat: VQGrad, VQGradSeq (\#107)”; and 32 more (refs \#124, \#63, \#107, \#113, \#127, …).
Sources/GrAIdient/LayerSeq · medium confidence · unverified
37 commits (2 fixes) modifying Sources/GrAIdient/Layer2D
A change to existing behaviour in Sources/GrAIdient/Layer2D — 37 commits (2 fixs), 24 files: “✨ feat: VQGrad, VQGradSeq (\#107)”; “✨ feat(examples): adding Transformer training example (\#74)”; “✨ layer\_seq: MultiplySeq, SiLU & LLM test (\#127)”; and 34 more (refs \#107, \#74, \#127, \#47, \#118, …).
Sources/GrAIdient/Layer2D · medium confidence · unverified
45 commits (1 fix) modifying Sources/GrAIdient/Utils
A change to existing behaviour in Sources/GrAIdient/Utils — 45 commits (1 fix), 4 files: “✨ feat(layer\_seq): RoPESeq (\#124)”; “✨ feat(layer\_seq): Concat1Seq & Concat2Seq (\#63)”; “✨ feat: VQGrad, VQGradSeq (\#107)”; and 42 more (refs \#124, \#63, \#107, \#113, \#127, …).
Sources/GrAIdient/Utils · medium confidence · unverified
47 commits (1 fix) modifying Sources/GrAIdient/Metal
A change to existing behaviour in Sources/GrAIdient/Metal — 47 commits (1 fix), 5 files: “✨ feat(layer\_seq): RoPESeq (\#124)”; “✨ feat(layer\_seq): Concat1Seq & Concat2Seq (\#63)”; “✨ feat: VQGrad, VQGradSeq (\#107)”; and 44 more (refs \#124, \#63, \#107, \#113, \#127, …).
Sources/GrAIdient/Metal · medium confidence · unverified
5 commits (0 fixes) modifying Sources/GrAIdient/LayerSeq/Base
A change to existing behaviour in Sources/GrAIdient/LayerSeq/Base — 5 commits, 2 files: “✨ feat: LayerCAM2D -\> VQGrad2D, LayerCAMSeq -\> VQGradSeq (\#117)”; “✨ feat(core): initForward,Backward model API (\#109)”; “🔨 refactor: handle float16 along float on GPU (\#120)”; and 2 more (refs \#117, \#109, \#120, \#60, \#128).
Sources/GrAIdient/LayerSeq/Base · medium confidence · unverified
6 commits (0 fixes) modifying Sources/GrAITestsUtils
A change to existing behaviour in Sources/GrAITestsUtils — 6 commits, 3 files: “✨ feat: VQGrad, VQGradSeq (\#107)”; “🔨 refactor(core): LayerWeightInit (\#96)”; “✨ feat(layer\_seq): LLM generate (\#128)”; and 3 more (refs \#107, \#96, \#128, \#40, \#121, …).
Sources/GrAITestsUtils · medium confidence · unverified
8 commits (0 fixes) modifying Sources/GrAIdient/Layer2D/Transform
A change to existing behaviour in Sources/GrAIdient/Layer2D/Transform — 8 commits, 12 files: “🔨 refactor(layer\_2d): API for random transforms (\#90)”; “✨ feat(layer\_2d): Flip2D & config\_kernels (\#92)”; “🔨 refactor: throwable init (\#103)”; and 5 more (refs \#90, \#92, \#103, \#128, \#98, …).
Sources/GrAIdient/Layer2D/Transform · medium confidence · unverified
8 commits (1 fix) modifying Sources/GrAIdient/Core/Model
A change to existing behaviour in Sources/GrAIdient/Core/Model — 8 commits (1 fix), 2 files: “✨ layer\_seq: MultiplySeq, SiLU & LLM test (\#127)”; “✨ feat(core): initForward,Backward model API (\#109)”; “🔨 refactor: handle float16 along float on GPU (\#120)”; and 5 more (refs \#127, \#109, \#120, \#45, \#96, …).
Sources/GrAIdient/Core/Model · medium confidence · unverified
Test coverage
1 commit adding/updating tests in Tests/data; 10 commits adding/updating tests in Tests/GrAIExamples/Base/python\_lib/nlp; 11 commits adding/updating tests in Tests/GrAIExamples/Base; 14 commits adding/updating tests in Tests/GrAITorchTests/Base/python\_lib; 16 commits adding/updating tests in Tests/GrAITorchTests/Base; 17 commits adding/updating tests in Tests/GrAITorchTests; 23 commits adding/updating tests in Tests/GrAIExamples; 58 commits adding/updating tests in Tests/GrAITests; 8 commits adding/updating tests in Tests/GrAIExamples/Base/python\_lib.
Dependencies
10 commits updating dependencies (2 manifests)
A dependency / build maintenance change in (dependencies) — 10 commits, 2 files: “🚨 test: layer2d (\#8)”; “🚨 test(integration): reproducibility with PyTorch (\#10)”; “✨ feat(examples): simple vgg trained on cifar (\#9)”; and 7 more (refs \#8, \#10, \#9, \#120, \#12, …).
(dependencies) · medium confidence · unverified
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
How this codebase got here
Baseline
- First survey — no prior run to compare against. CAI 66.8.
Lenses
- Code Health 75.2
- Architecture 95.7
- Maturity 56.4
- Readiness 68.3
- Security 92.8
Changes since last survey
- 132 commits — 119 feature/other, 13 fixes
By area
- Sources/GrAIdient — 60 commits
- Sources/MAKit — 20 commits
- (root) — 14 commits
- (repo) — 6 commits
- Docs/Contributing — 5 commits
- Tests/GrAITorchTests — 5 commits
- Tests/MAExamples — 4 commits
- Tests/MATorchTests — 4 commits
- Tests/GrAIExamples — 3 commits
- Tests/GrAITests — 3 commits
- Tests/MAKitTests — 3 commits
- Docs/Examples — 2 commits
- .github/workflows — 1 commit
- Docs/Architecture — 1 commit
- Tests/data — 1 commit
Notable commits
- fix: 🐛 fix: update correlation matrix coeffs (#37)
- fix: 🐛 bug(layer_2d): align Convolution & Deconvolution on PyTorch (#84)
- fix: 🐛 fix: ResizeBilinear to output deterministic dimensions (#33)
- fix: 🐛 fix: input GPU check (#106)
- fix: 🐛 fix: model context max id (#45)
- fix: 🐛 fix: numerical instability of Softmax (#76)
- fix: 🐛 fix: numerical stability of tanh for GELU (#83)
- fix: 🐛 fix: remove error when data input may indicate lower batch size (#44)
- fix: 🐛 fix: run on Apple Silicon (#110)
- fix: 🐛 fix: update ValueSeq operation (#72)
- fix: 🐛 fix: use buffers for neuron selection in SelectNeurons1D (#50)
- fix: 🔧 chore: fix badges in the readme (#15)
- fix: 🚨 test(integration): validate MultiHeadAttention & fix Softmax stability (#70)
- change: Initial commit
- change: Merge branch 'main'
- change: Merge pull request #14 from owkin/release_1
- change: Very simple VisionTransformer example. (#75)
- change: feat(layer_1d): activation, fl, linear error, mse, select channels (#5)
- change: ✨ feat(core): GELU activation function (#73)
- change: ✨ feat(core): GELU vs GELUApprox (#113)
- …and 112 more
Architecture
- 0 containers · 1 bounded contexts · 0 dependency edges (baseline)
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
Survey your own repository
owkin/GrAIdient was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.
About this page
- The score is its most recent published measurement, taken on 17 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
- Measured at commit a6ca88559bf7f78586880bdf059708005f43e872 — the exact code this score is about.
- Scored under rubric-2026.11.4 — the same rubric and the same method as every other entry in this index.
- Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.85+preprod-eba6548ac50f.