Summary: 291 instances, 270 unique

Text	Count
// TODO: This public method returns a pointer to an internal data structure	1
# TODO: The identifier "dict" should be made global unique in target name space	3
# TODO: Make sure observation noise is transformed correctly	1
# TODO: try pooling rather than just slicing out last hidden	1
# TODO: What about simplex matrices? It should be legal to do	1
# TODO: Check that there is one value for each case.	1
# TODO: compute unattributed via property	1
# TODO: If we save the patterns that failed as well, then we can build a	1
# TODO: Display as a tree and graph also	1
# TODO: If we have a slightly noncontiguous set of cases, say {0, 1, 3},	1
// TODO: create Samples class and remove the following friend classes	1
# TODO: replace "log" with "log1mexp" as needed below and update defs	1
# TODO: Remove this. We should insert TO_REAL nodes when necessary	1
# TODO: This needs to be now set after self.new_dist has been instantiated	1
# TODO: Figure out a better way to pass this flag around	1
# TODO: add support for data pre-processing on random effects	1
# TODO: There are a great many more pure instance functions on tensors;	1
# TODO: Add classes, lambdas, and so on	1
# TODO: Consider adding a pass which optimizes away IF(X, Y, Y) to	1
# TODO: The original idea was to use these integers when generating code	1
# TODO: The graph should be acyclic by construction;	1
// TODO: this is the one actually used in code (as opposed to full_log_prob used	1
# TODO: If we have a contiguous set of cases, say {2, 3, 4}, then	1
# TODO: Consider adding a compiler version number, hash	1
# TODO: We should make a rewriter that detects stochastic index	1
# TODO: weird, but it should be legal.  Figure out what the	1
# TODO: non-diagonal covariance	1
# TODO: add new `self.world_.params_` not already in optimizer	1
# TODO: We need a better error message for that scenario.	1
# TODO adjust log density on numerical runtime errors	1
# TODO: Handle the case where there are two inconsistent	1
# TODO: Would be slightly faster if we applied module filtering inside walk_packages	1
# TODO: There are two shortcomings to the current solution. First, it prints	1
# TODO: observations of the same sample	1
// TODO: Investigate the semantics of _backward(skip_observed),	1
# TODO: Consider producing a warning for this situation, because	1
// TODO: generalize so it works with any proposer, not just nmc_proposer:	1
# TODO: We could consider traversing only nodes reachable from	1
# TODO: The edge labels used to visualize the graph in DOT	1
# TODO: What if the observation is of a different size than the	1
# TODO: The identifier "dict" should be made global unique in target name space	1
# TODO: We do not collect statistics on memoization use here.	1
// TODO: the following is used in beta.cpp only. Does it really need to be here?	1
# TODO: We could bail out after the first failure.	1
# TODO: What if we have a non-tensor indexed with a stochastic value?	1
# TODO: Note that we only memoize calls to RVs when the arguments	1
# TODO: receiver? Do we recompile it on every call? Check this.	1
// TODO: As of May 2021, this method is being used for testing only.	1
// TODO: rename method to be more clear	1
// TODO: the reason for that is not clear;	1
# TODO: we may wish to add bounded integers to the BMG type system.	1
# TODO: This approximation is an over-estimate; for instance, when	1
"""TODO have separate diagnostics for discrete variables.	2
# TODO: We do not canonicalize arguments as the memoizer does above.	1
# TODO: We do not yet have an implementation of the uniform	1
// TODO: eventually we want to have different classes of Node	1
# TODO: How should we deal with observations of	1
# TODO: a complex pattern.	1
# TODO: Make max_possibilities a global tweakable setting of the accumulator.	1
# TODO: What happens if we have a @functional that does not return	1
# TODO: The code in the "handle" methods which folds operations	1
# TODO: the "for" loop in Python supports an "else" clause which is only activated	1
# TODO: right thing to do is for this scenario.	1
# TODO: Would be helpful if we could track original source code locations.	1
# TODO: Ideally, would like to merge [1].ctx with the [0].ctx below	2
# TODO: The identifier "dict" should be made global unique in target name space	1
# TODO: we should sample from a transformed dist	1
# TODO: Torch supports both broadcasting and non-broadcasting versions	1
# TODO: Tensor comprehension patterns	1
// TODO: a lot of members used to be private, but we need access to them	1
// TODO: should we just remove it from Operator then?	1
# TODO: We might consider an optimization pass which does so.	1
# TODO: Better error	1
// TODO: this does not have to apply to parents only; make it a more general	1
# TODO: Memoize these, remove memoization of construction functions below.	1
# TODO: you might expect.	1
# TODO: Should we treat "x is y" the same as "x == y" when they are	1
# TODO: Consider detecting jitted functions which use the nonlocal statement, and	1
# TODO: We need to consider what happens if the consequence and alternative	1
# TODO: It is strange having a visualizer that edits the graph	1
# TODO: beanmachine.graph from beanmachine.ppl.  I'll figure out why later;	1
# TODO: add support for random slope	1
# TODO: What happens if the function is a member of a class?	1
# TODO: In Python, (not tensor(x)) is equal to (not x).	1
# TODO: We use the DOT label of the node in the error messages here,	1
# TODO: This scenario is another opportunity for a warning or error, since	1
# TODO Move this to a utils module	1
# TODO: Better error message	1
# TODO: return None if we are unable to get the source	2
# TODO: categorical	1
# TODO: Add verbose level	1
# TODO: Consider allowing values that can be converted to tensors.	1
return variational_params; // TODO: this should have been defined as a field,	1
# TODO:	1
# TODO: diagnostic engine that describes why a value failed to match against	1
# TODO: Detect unsupported control flow	1
# TODO: what should we do about AnnAssign?	1
# TODO: more general enumeration	1
# TODO: assert statements are removed in bm_to_bmg; move that functionality here.	1
# TODO: Should it be legal at all in the graph?	1
# TODO: This is only used to remove observation sample node that	1
# TODO: Figure out how to desugar with expr as target : block	1
# TODO: Is there a better way to do this other than try it and see what happens?	1
# TODO: say something about formatted strings	1
# TODO: We need to find a good way to order things in this file	1
# TODO: The edge labels used to visualize the graph in DOT	1
# TODO: BMG requires it to be a *broadcast* matrix; what happens	1
# TODO: We now have matrix multiplication in BMG; finish this implementation	1
# TODO: dirichlet	1
// TODO: Include value of observation	1
# TODO: Consider asserting that the node is unsupported by BMG.	1
# TODO: Move this to bmg_requirements.py	1
# TODO: reinterpret batch dimension?	1
# TODO: Add a note here describing why it is important that the function	1
// TODO: clarify the meaning of log_prob_vals and log_prob_allchains	1
# TODO: This should be private	1
# TODO: Do we need to consider the scenario where we do not have	1
# TODO: The current code produces a bad error message experience because at	1
// TODO: Yeah, how?	1
// TODO: also used in tests only	1
# TODO: We can go further than this and eliminate all else clauses	1
# TODO later: all comparisons, all bitwise, floordiv,	1
# TODO: We could bail out after the first success.	1
# TODO: Get this into alpha order	1
# TODO: We can reduce "del" statements to one of three forms: "del x", "del x.y", "del x[y]"	1
# TODO: a good error if they do not. Alternatively, catch	1
// TODO: used for testing only, should integrate it with	1
# TODO: Note that this can be a considerable overestimate. For example, if we	1
# TODO: say something about classes	1
# TODO: This is a bit opaque. Add a helper method for this.	1
# TODO: Verify that the list is well-formed.	1
// TODO: rename to initialize_unconstrained_value_types	1
# TODO: We do not yet handle the case where we are converting from, say,	1
# TODO: caused by e.g. negative scales in `dist.Normal`;	1
// TODO: from now on, we have methods for adding nodes, checking validity,	1
# TODO: Support kwargs	1
# TODO: the "for" loop could be rewritten as fetching an iterator and iterating	1
# TODO: downstream observation likelihoods p(obs | rvid)	1
// TODO: Better display of matrices	1
# TODO: for now, we'll just turn off error checking in this module.	1
# TODO: We do not yet have a BMG node for Binomial	1
// TODO: shouldn't we then restrict them to those classes? See above.	1
// TODO: add simplex transform	1
# TODO: Consider producing a warning or error if we cannot determine which lambda	1
# TODO: Should we implement a general switch node in BMG?	1
# TODO: BMG requires that the target of a query be classified	1
# TODO: Matches "tensor" and "foo.tensor"	1
# TODO: In the type this function, both instances of the type Any should	1
// TODO: change type to ranged natural	2
// TODO: Add checks that the observed value is in range.	1
# TODO: Consider optimizing distributions where the tensor elements are all	1
# TODO: For reasons unknown, Pyre is unable to find type information about	1
# TODO: Should this be idempotent?	1
# TODO: Note that this rule combined with the previous rule	1
sto_node->get_unconstrained_value(true); // TODO: rename this function	1
# TODO: BMG requires that the target of a query be classified	1
# TODO: Allowing these constructions raises additional problems that	1
// TODO: Add checks that the observed values are in range.	1
# TODO: categorical logit	1
#   * dict() is allowed. (TODO: this could be rewritten to {})	1
#   * TODO: There are similar exceptions for set and list; say what they are.	1
# TODO: Improve wording and diagnosis.	1
# TODO: Handle kwargs	1
# TODO: always_succeeds will be wrong for the returned object.	1
// TODO: identify code that depends on this, let it zero gradients	1
# TODO: Sort by total time of children - WARNING: Important not to sort by	1
#   * [id] = id  # TODO: What about tuples?	1
// TODO: tie samples directly to inference	1
# TODO: Add logging	1
// TODO: remove this error message and leave it to compute_gradients	1
// TODO: shouldn't we then restrict them to those classes? See below.	1
# TODO: It is NOT equal to (tensor(not x)), which is what	1
# TODO: Add other notes on ordering constraints here.	1
# TODO: This code is mostly but not entirely in alpha order	1
/* TODO: used in testing only; it looks like there has not been a need for it in	1
# TODO: Better error	10
# TODO: If we have exp(x) - 1 in the graph and x is known to be of type	1
# TODO: regular expression for sections that are visible initially: unfurled = "..."	1
# TODO: We might want to restrict this. For example, if we have	1
# TODO: Do addition and multiply need to be the multi- versions?	1
# TODO: Check that cases are all constant nodes.	1
# TODO: Handle svg images.	1
# TODO: We need to consider what happens if the value's types	1
# TODO: We don't have a subtraction node; we render this as	1
// TODO: it seems like the initializer is needing to know	1
# TODO: This code is not very well organized; consider sorting it	1
# TODO: We probably need to make a distinction between torch.mm and	1
# TODO: Consider what to do if the node type is Tensor.	2
# TODO: rename bmg outer variable to something less likely	1
// TODO: Review what members of this class can be made static.	1
#   * id = simple_expression  # TODO: List the simple expressions.	1
// TODO: review what really needs to be private or public in MH's API.	1
# TODO: **Observations are logically distinct from models.**	1
# TODO: messy, consider strategy pattern	1
# TODO: Base path for Docusaurus 2 content	1
# TODO: We could do better for the comparison operators because we known	1
# TODO: Verify that they do match.	1
// TODO: this should be delegated to steppers, since that's	1
# TODO: We might also need:	1
# TODO:	2
# TODO: This is only used to remove an observation; restrict it	1
# TODO: Should we also eliminate plain returns? We could rewrite	1
# TODO: We do not yet handle the case where we are converting from	1
# TODO: Do we need to specifically match just torch? What if there is an alias?	1
# TODO: Consider refactoring the common input checks for both methods	1
# TODO: If left is a ToMatrixNode then we can construct a second	1
# TODO: We have an ordering problem in the fixers:	1
# TODO: Should we produce some sort of warning here telling the user that	1
# TODO: Detect unsupported operators	1
# TODO: Suppose we have something like Uniform(1.0, 2.0).  Can we replace that	1
// TODO: should change name to "Proposal" to better match the literature.	1
# TODO: say something about async	1
# TODO: Move this prod helper function out of bmg_nodes.py	1
# TODO: add a range check to make sure it fits into the integer	1
# TODO: Here we treat an observation as node which takes input	1
# TODO: What if we have an atomic type on one side and a 1x1 matrix	1
# TODO: which do we wish to support?	1
# TODO: Mark this as abstract	1
# TODO: matrix-valued samples?	1
# TODO: Should it be an error to add two unequal observations to one node?	1
# TODO: For now we support conjugate prior transformation on	2
# TODO: If we have arbitrary natural cases, say 1, 10, 101, then we could	1
# TODO: bottlenecking?	1
// TODO: why don't the initialization below to be done for _infer?	1
# TODO: As with observations, properly speaking there is no	1
# TODO: Generate a better error message for switches that we cannot yet	1
# TODO: do we need to a _sum_rightmost here?	1
# TODO: Figure out how to desugar try: body except expr as bar: body else: body finally: body	1
# TODO: remove once fixed upstream	1
# TODO: caused by e.g. negative scales in `dist.Normal`;	1
# TODO: Do we recompile it for different instances of the	1
# TODO: Better exception	1
# TODO: the exception if the call fails and replace it with	1
// TODO: Better display of simplexes	1
# TODO: say something about yield	1
# TODO: Update this comment	1
# TODO: We could support b ** n where b is bool and n is a natural	1
# TODO: say something about global / nonlocal	1
// TODO: can we make signature use a default generator then?	1
# TODO: This check can be removed if it is not a necessary condition.	2
# TODO: Do a sanity check that the arguments match and give	1
# TODO: This method is only here for testing purposes.  Get rid of it.	1
# TODO: Add support for the Cauchy distribution as well.	1
# TODO: Try torch.distributions.utils.broadcast_all	1
# TODO: This will produce a bad error message experience. What we want to	1
# TODO: What should the result of NOT on a tensor be?	1
// TODO: Better display of factors.	1
# TODO: Handle the situation if there are tabs	1
# TODO: Consider adding a verification pass which ensures	1
#   type information has already been cached. (TODO: Is this useful? Maybe remove it.)	1
# TODO: Better exception	3
// TODO: Check that an observed simplex is given a simplex.	1
# TODO soon: LogSumExp, ExpM1, Logm1exp	1
# TODO: Better naming convention. In particular, _rules is the reflexive	1
# TODO: Fix this error message for the case where we require	1
# TODO: a graph node? A functional that returns a constant is	1
# TODO: say something about conditional expressions	1
# TODO: say something about type annotations	1
# TODO: Unfortunately, the second rule does not match the actual behavior of	1
# TODO: a more informative error.	1
# FIXME: update the method to support customized prediction goal	1
# TODO: We might want to build a warning mechanism that informs the developer	1
# TODO: We could optimize -(-x) to x here.	1
# TODO: Make _rv_to_query public. Add it to BMGraphBuilder?	1
# TODO: Consider implementing some sort of warning for this case?	1
# TODO: This module is badly named; it really should be "python simplifier" or some	1
# TODO: We have a problem that we need to resolve regarding compilation of models	1
// TODO clarify why we need to reset gradients	1
# TODO: We use z_{1-alpha} instead of -z_alpha for compatibility	1
// TODO: rename log_abs_jacobian_determinant	1