panic bug

[andyzhou]

问题发生在model.go的FM：Predict()方法，运行过程中会发生fm.Index对象为空的情况，导致fm.Index.EncodeUser及相关方法调用panic。

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

The problem occurs in the FM: Predict() method of model.go. During the operation, the fm.Index object will be empty, causing fm.Index.EncodeUser and related methods to call panic.

[zhenghaoz]

有日志吗，我看一下是具体哪个位置

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Is there any log? Let me check the specific location.

[alwinmathew]

@zhenghaoz, is there a fix for this issue? I'm also running into this.

I've a stack trace if that helps.

{"level":"info","ts":1708892182.2309608,"msg":"ranking recommendation","n_working_users":2,"n_jobs":3,"cache_size":100}
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x2 addr=0x60 pc=0x104bb8a68]

goroutine 748 [running]:
github.com/zhenghaoz/gorse/model/click.(*FM).Predict(0x1400019f5f0, {0x14000f87da0?, 0x140000dcff0?}, {0x140000dcff0, 0x24}, {0x0, 0x0, 0x104441b80?}, {0x14000268280, 0xf, ...})
	/home/runner/work/gorse/gorse/model/click/model.go:204 +0x48
github.com/zhenghaoz/gorse/worker.(*Worker).replacement(0x14000578420, 0x14001128a68?, 0x140011295d8, {0x1400166e790, 0x2, 0x0?}, 0x14000f972a0)
	/home/runner/work/gorse/gorse/worker/worker.go:1236 +0x6a4
github.com/zhenghaoz/gorse/worker.(*Worker).Recommend.func2(0x1?, 0x1)
	/home/runner/work/gorse/gorse/worker/worker.go:778 +0x2e24
github.com/zhenghaoz/gorse/base/parallel.Parallel.func2(0x0?)
	/home/runner/work/gorse/gorse/base/parallel/parallel.go:67 +0x9c
created by github.com/zhenghaoz/gorse/base/parallel.Parallel
	/home/runner/work/gorse/gorse/base/parallel/parallel.go:58 +0x130

[linguanyuanA]

func (t *FitClickModelTask) run(j *task.JobsAllocator)
if t.clickTrainSet == nil || numUsers == 0 || numItems == 0 || numFeedback == 0 {
...
log.Logger().Warn("empty ranking dataset",
zap.Strings("positive_feedback_type", t.Config.Recommend.DataSource.PositiveFeedbackTypes))
t.taskMonitor.Fail(TaskFitClickModel, "No feedback found.")
return nil
}

当numFeedback 为0时，这里会返回nil, 导致没有执行后面的 func (fm *FM) Fit(trainSet, testSet *Dataset, config *FitConfig)
结果就是fm.Init没有执行。但是离线Recommend时会调用 w.ClickModel.Predict
@zhenghaoz