ConvLearner fit function does not work

[cSchubes]

Hi,

I am following lesson 1 of the DLC course and have encountered what appears to be a bug in FastAI itself. The following lines run the training, but fail after for something related to numpy:

arch=resnet34
data = ImageClassifierData.from_paths(PATH, bs=8, tfms=tfms_from_model(arch, sz))
learn = ConvLearner.pretrained(arch, data, precompute=True)
learn.fit(0.01, 2)

Batch size is 8 because I have only 2GB on my GPU. The error is below:

TypeError                                 Traceback (most recent call last)
<ipython-input-11-676345a6c308> in <module>()
      2 data = ImageClassifierData.from_paths(PATH, bs=8, tfms=tfms_from_model(arch, sz))
      3 learn = ConvLearner.pretrained(arch, data, precompute=True)
----> 4 learn.fit(0.01, 2)

~/fastai/fastai/learner.py in fit(self, lrs, n_cycle, wds, **kwargs)
    300         self.sched = None
    301         layer_opt = self.get_layer_opt(lrs, wds)
--> 302         return self.fit_gen(self.model, self.data, layer_opt, n_cycle, **kwargs)
    303 
    304     def warm_up(self, lr, wds=None):

~/fastai/fastai/learner.py in fit_gen(self, model, data, layer_opt, n_cycle, cycle_len, cycle_mult, cycle_save_name, best_save_name, use_clr, use_clr_beta, metrics, callbacks, use_wd_sched, norm_wds, wds_sched_mult, use_swa, swa_start, swa_eval_freq, **kwargs)
    247             metrics=metrics, callbacks=callbacks, reg_fn=self.reg_fn, clip=self.clip, fp16=self.fp16,
    248             swa_model=self.swa_model if use_swa else None, swa_start=swa_start,
--> 249             swa_eval_freq=swa_eval_freq, **kwargs)
    250 
    251     def get_layer_groups(self): return self.models.get_layer_groups()

~/fastai/fastai/model.py in fit(model, data, n_epochs, opt, crit, metrics, callbacks, stepper, swa_model, swa_start, swa_eval_freq, visualize, **kwargs)
    160 
    161         if not all_val:
--> 162             vals = validate(model_stepper, cur_data.val_dl, metrics, epoch, seq_first=seq_first, validate_skip = validate_skip)
    163             stop=False
    164             for cb in callbacks: stop = stop or cb.on_epoch_end(vals)

~/fastai/fastai/model.py in validate(stepper, dl, metrics, epoch, seq_first, validate_skip)
    240             loss.append(to_np(l))
    241             res.append([f(datafy(preds), datafy(y)) for f in metrics])
--> 242     return [np.average(loss, 0, weights=batch_cnts)] + list(np.average(np.stack(res), 0, weights=batch_cnts))
    243 
    244 def get_prediction(x):

~/anaconda3/envs/fastai/lib/python3.6/site-packages/numpy/lib/function_base.py in average(a, axis, weights, returned)
    381             wgt = wgt.swapaxes(-1, axis)
    382 
--> 383         scl = wgt.sum(axis=axis, dtype=result_dtype)
    384         if np.any(scl == 0.0):
    385             raise ZeroDivisionError(

~/anaconda3/envs/fastai/lib/python3.6/site-packages/numpy/core/_methods.py in _sum(a, axis, dtype, out, keepdims, initial)
     34 def _sum(a, axis=None, dtype=None, out=None, keepdims=False,
     35          initial=_NoValue):
---> 36     return umr_sum(a, axis, dtype, out, keepdims, initial)
     37 
     38 def _prod(a, axis=None, dtype=None, out=None, keepdims=False,

TypeError: No loop matching the specified signature and casting
was found for ufunc add

This looks like something internal to FastAI and its interactions with Numpy. Any help would be greatly appreciated!

[chasak]

Getting the same error. I was able to solve it by updating the repo on my local machine and set learn.metrics = [] The relevant discussion and answer by fizx is here http://forums.fast.ai/t/typeerror-on-the-first-exemple-in-lesson-1-no-loop-matching-the-specified-signature-and-casting/22403/5

[jph00]

Is this on py37? If so, there's a known bug with numpy.

[cSchubes]

Python 3.6 actually. learn.metrics = [] works, but then I don't get nice metrics... and if I try and use metrics, it fails.

[sgugger]

I didn't have the bug (which may due to the fact my pytorch is 0.4) but I think the fix from @ncihnegn should work. Please reopen if it's not the case.

[Sammy-iiitb]

If that doesn't work now "create_cnn" can be used in place of ConvLearner

[zszazi]

So there are two options as ConvLearner doesnt work now Replace ConvLearner by

1. create_cnn
2. cnn_learner - link