Summary: 160 instances, 152 unique

Text	Count
//TODO : replace any	1
# TODO: this is ugly, we put all the initialization work in this method, because initialization relies	1
# TODO: check the label of module and warn if it's auto-generated	1
if i == self.cur_ind - 1:  # TODO: add other member in the set	1
TODO: support searchspace file.	1
// TODO: this should be private	1
# TODO: consider other functions here	1
//TODO : pass in request timeout param?	1
# FIXME: It does not stop on failure. Tried "ErrorActionPreference" with no luck.	1
//TODO: change the logic here when we want to support multiple tensorboard	1
# TODO: check whether there could be multiple output nodes???	1
// TODO support specify database dir	1
# TODO suppose to fix the conflict after the sparsity propagation	1
# TODO: find out all the modules that have the same requirement as LSTM	1
# TODO: Delete this after upgrading to PyTorch 1.11.	1
//TODO: use HDFS working folder instead	1
// TODO: how to format NaN?	1
new_model.evaluator = self.evaluator  # TODO this needs a clever copy (not deepcopy) if we need mutation	1
// TODO: use search space and metrics space from TRIALS will cause update issues.	1
# TODO In current L1/L2 Filter Pruner, the 'op_types' is still necessary	1
# TODO - catch exact Exception	1
# FIXME:	1
# TODO: use node copy instead	1
# FIXME: merge this rename with non-root graph, only do once.	1
# TODO: it does not have duplicated edges if only supporting dedup input	1
# TODO: use a passed-in RNG here	1
# TODO: refactor this part, maybe we can remove the code gen of prim::Constant	1
'Module', 'Sequential', 'ModuleList',  # TODO: 'ModuleDict', 'ParameterList', 'ParameterDict',	1
# TODO:	2
# TODO: Calculate loss	1
// FIXME: this should not be handled in web UI side	1
//TODO: remove this constraint when supporting multi-host placement	1
# TODO: handle imports	1
# FIXME: See cache-dependencies-template.yml on why it needs rebuild.	1
# TODO: serialize complex data type, and output proper error message	1
// TODO: change this hard coded deployment name after demo	1
# TODO what if last output is tuple/list of tensor	1
// TODO: use local storage	1
# TODO: Feed the exact input tensor if user provides input,	1
# TODO: post, put, etc	1
// TODO: move 100 and 1000 into constants class	1
# FIXME: I don't know where "utils" should be	1
# TODO: docstring	1
# TODO: deal with this argument	1
# TODO: The current design of init interface of Retiarii experiment needs to be reviewed.	1
// TODO: fix expressJoi	1
# TODO: the experiment should be completed, when strategy exits and there is no running job	1
)  # TODO: emulate tee behaviors, not necessary tho.	1
search_space: Any = ''  # TODO: remove	1
# TODO: I'm not sure of the availble datasets in this benchmark. and the docs are missing.	1
# TODO: status	1
TODO: move this logic to NNI manager	1
# TODO if we can find the corresponding relation between the value node	1
assert merge_op in ['all']  # TODO: loose_end	1
# TODO: directly use weight_mask is not good	1
# FIXME: should be a warning message here	1
// TODO: use Message.txt to tooltip	1
// FIXME: Use global handler. It should be aware of shutting down event and swallow errors in this stage.	1
# TODO: may relax this limitation?	1
],  # TODO: support remaining argument, uncomment the lines in nnictl.py	1
# TODO: support InputChoice and ValueChoice	1
startTime: (trial as Trial).info.startTime, // FIXME: why do we need info here?	1
# TODO: deal with conditional ops	1
# TODO: split metric of multiple models?	1
if self.args.model_type == 'mobilenetv2':  # TODO: to be tested! Share index for residual connection	1
// TODO: implement put, post, delete methods	1
# TODO support the other batch dimension in the future	1
// FIXME: We should have a global handler for critical errors.	1
sequenceId: -1,  // FIXME: multi-phase tuner should use sequence ID instead of trial job ID	1
TODO: support more types in the type_as, need to figure out	1
# FIXME: return mask, to be consistent with other algorithms	2
// TODO: add blacklist	1
/* TODO: here is 32 rather than $pageMargin is because experiment page `content` position */	1
# FIXME: prior is designed but not supported yet	2
// FIXME: unit test	1
# TODO: refactor later	1
# FIXME: not tested yet	2
# TODO: ugly, think about how to refactor this part	1
content = `${filename} is empty.`;  // FIXME: this should be handled in front-end	1
# TODO: Packing multiple model in one GPU	1
//TODO:	1
// TODO: test	1
# TODO:	1
# FIXME: risk, candidates might also have None	1
// TODO: handle nested entries	1
// TODO: move this to our ESLint formatter?	1
# TODO: scope name could be empty	1
// TODO: handle more than number and object	1
this.log.error(`TrialDispatcher: TODO: not implement to handle direct REPORT_METRIC_DATA command yet.`);	1
// FIXME: redundant update	1
# FIXME: to avoid circular import, copy this function in this place	1
// TODO: remove the following constraint when supporting distributed trial	1
# FIXME This is a hack to make choice align with the previous format	1
# TODO: designed to replace `patch_optimizer`	1
# TODO: what if this input is a constant tensor	1
* FIXME: This should be a router, not a separate REST server.	1
# FIXME: some tuners raise NoMoreTrialError when they are waiting for more trial results	1
# FIXME: need further check for each algorithm which types are actually supported	1
// TODO: support non shared storage	2
// TODO: should handle more types of metric keys	1
// FIXME: this can take a long time	1
//TODO: remove this constraint when supporting other training services	1
# FIXME: this check might be wrong	1
# TODO support more kinds of value node	1
# TODO: maybe it should be able to calc on weight-granularity, beside from layer-granularity	1
// FIXME: can this be undefined?	1
# TODO: currently, only support single input slot and output slot.	1
# TODO: graph	1
# TODO: this logic might need to be refactored into execution engine	1
// TODO: netron might need prefix.	1
#TODO:finish webui function	1
// FIXME: This file is copied from react-dev-utils master branch. Waiting for its next release.	1
# TODO: It is not a good idea to directly modify global settings. A better choice is	1
# TODO: find out a proper way to show no more trial message on WebUI	1
#TODO replace this flops counter with nni.compression.torch.utils.counter.count_flops_params	1
# TODO should use a more general way to get the input	1
// TODO: this should be refactored to the common modules	1
# TODO: change to sleep time configurable via arguments	1
// TODO: NNI manager should not peek tuner's internal protocol, let's refactor this later	1
'/experiment/cluster-metadata', //TODO: Fix validation expressJoi(ValidationSchemas.SETCLUSTERMETADATA),	1
# TODO: the operation is likely to be considered editable by end-user and it will be hard to debug	1
# TODO: prim::TupleIndex is not supported yet	1
TODO: parameters of subgraph (see `Node` class)	1
# TODO: support non member functions	1
# FIXME: this is a workaround as full tensor is not supported in configs	1
*  TODO:	1
displayName: Sphinx  # TODO: rstcheck	1
# FIXME This is a hack to make choice align with the previous format	1
# TODO: when copying one node to multiple devices, broadcast is more efficient than P2P communication	1
# TODO: replace with validation here	2
# TODO: deal with all the types	1
# TODO: set these, for each class?	1
orig_type = weight.type()  # TODO: user layer	1
# TODO: port shangning's work here, and use it in Experiment.start()/.stop()	1
# FIXME: because Tuner is designed as interface, this API should not be here	1
return null; // TODO: render a loading page	1
// TODO: change the name based on operator's type	2
// FIXME: default metric is hacked as latestAccuracy currently	1
# TODO: could make this a user given parameter	1
// FIXME: this is ad-hoc	1
# TODO: support all platforms	1
// FIXME: Use global handler. The event can be emitted after listening.	1
# TODO: need refactor	1
# TODO: add scope name	1
# TODO: match with arg's type. manually choose for now	2
TODO: this is pytorch cell	1
TODO: If the compact model has been speeduped, the auto infer masks maybe also need.	1
{/* TODO: fix bug */}	1
// TODO add validators for request params, query, body	1
#if spec.q is None:  # TODO: comment out because of edge case UT uniform(99.9, 99.9)	1
#TODO: 1. get this url by api? 2. change this url in private dlc mode.	1
TODO: parameter of subgraph (cell)	1