mandala

for performance, though I've used it in projects spanning millions of memoized calls. The typical use case is to decorate functions that take a long time to compute, so the overhead of memoization amortizes. A very quick benchmark on my laptop shows ~6ms per call for in-memory storage, ~9ms for a persistent storage, with a simple arithmetic function that otherwise takes ~0 time. 5. Great question - currently, the dependency tracer is restricted to user-chosen functions to avoid tracking function calls an imported library makes. You could use a bit of magic (import-time automatic decoration) to track all functions in a file or a directory (not implemented right now). The reasoning is that, for a typical multi-month ML project, you usually have a single conda environment so you want to ignore library changes. Similarly, system-level (e.g. environment variables) are also not tracked. I think a very useful feature would be to at least record the versions of each imported library, so that storages can be ported between environments with some guarantees (or warnings). 6. - If an `@op` call was memoized, the underlying Python function call succeeded, so in this sense it can't be "broken"; it's however possible that there was a bug. In this case, you can delete the affected calls and all values that depend on them (if you keep these values, you're left with "zombie" values that don't have a proper computational history). The `ComputationFrame` supports declarative deletion - you build a ComputationFrame that captures the calls you want to delete, and call `.delete_calls()` - though there's still no example of this in the tutorial :) Alternatively, you can change the affected function and mark this as a new version. Then you should be able to delete all calls using the previous version (though, not supported at this moment). - How the cache is invalidated is detailed here: https://github.com/amakelov/mandala?tab=readme-ov-file#how-i...

logging: if you want something saved, you write a function that returns it. Using (a lot of) hashing, `@op` ensures that the same version of the function is never executed twice on the same inputs. Importantly, the decorator encourages/enforces composition. Before a call, `@op` functions wrap their inputs in special objects, `Ref`s, and return `Ref`s in turn. Furthermore, data structures can be made transparent to `@op`s, so that an `@op` can be called on a list of outputs of other `@op`s, or on an element of the output of another `@op`. This creates an expressive "web" of `@op` calls over time. 2. a data structure, `ComputationFrame`, can automatically organize any such web of `@op` calls into a high-level view, by grouping calls with a similar role into "operations", and their inputs/outputs into "variables". It can detect "imperative" patterns - like feedback loops, branching/merging, and grouping multiple results in a single object - and surface them in the graph. `ComputationFrame`s are a "synthesis" of computation graphs and relational databases, and can be automatically "exported" as dataframes, where columns are variables and operations in the graph, and rows contain values and calls for (possibly partial) executions of the graph. The upshot is that you can query the relationships between any variables in a project in one line, even in the presence of very heterogeneous patterns in the graph. I'm very excited about this project - which is still in an alpha version being actively developed - and especially about the `ComputationFrame` data structure. I'd love to hear the feedback of the HN community. Colab quickstart: https://colab.research.google.com/github/amakelov/mandala/bl... Blog post introducing `ComputationFrame`s (can be opened in Colab too): https://amakelov.github.io/mandala/blog/01_cf/ Docs: https://amakelov.github.io/mandala/

have a problem: either re-run the cell (thus losing previous work, which is annoying even if you have to wait 1 minute - you know it's a wasted minute!), or write the new computation in a new cell, possibly with a lot of redundancy (which over time makes the notebook hard to navigate and keep consistent). So, this and other considerations eventually convinced me that the function is more natural than the cell as an interface/boundary at which caching should be implemented, at least for my use cases (coming from ML research). I wrote a framework based on this idea, with lots of other features (some quite experimental/unusual) to turn this into a feasible experiment management tool - check it out at https://github.com/amakelov/mandala P.S.: I notice you use `pickle` for the hashing - `joblib.dump` is faster with objects containing numpy arrays, which covers a lot of useful ML things

nged, and gives you a generic caching decorator for functions which can then look up large objects by reference; this is the way I used it for e.g. my mechanistic interpretability work, which is often of the form one big model + lots of analyses producing tiny artifacts based on it.