Pre-trained model Inference on CPU : Error during forward pass

[Mipsology]

Hello,

• My aim to run inference without retraining
• I am running on Intel CPU (i.e. no GPU)
• I am using the pre-trained model provided in this repo.
• I got the complete datasets from SemanticKITTI.
• I setup everything as per the instruction provided.

When I run eval script with all variables, I get error in user.py on line 226 (which I believe is doing a forward pass on the model). The error says:

"Illegal instruction (core dumped)"

Because of this, the Evaluation script fails when checking for length of pred_names (off course the inference never finished correctly) I concluded this with some print statements in user.py and infer.py files.

Any help or debug tips you can provide? Thank You in advance!

Terminal Output :

(salanext) ~/zebra/NNs/SalsaNext/SalsaNext-master$ ./eval.sh -d dataset -p logs_preds -m saved_model -s train -c 30

INTERFACE: dataset /home/mipso/zebra/NNs/SalsaNext/SalsaNext-master/dataset log /home/mipso/zebra/NNs/SalsaNext/SalsaNext-master/logs_preds model /home/mipso/zebra/NNs/SalsaNext/SalsaNext-master/saved_model Uncertainty False Monte Carlo Sampling 30 infering train

---

Opening arch config file from /home/mipso/zebra/NNs/SalsaNext/SalsaNext-master/saved_model Opening data config file from /home/mipso/zebra/NNs/SalsaNext/SalsaNext-master/saved_model train 00 train 01 train 02 train 03 train 04 train 05 train 06 train 07 train 09 train 10 valid 08 test 11 test 12 test 13 test 14 test 15 test 16 test 17 test 18 test 19 test 20 test 21 model folder exists! Using model from /home/mipso/zebra/NNs/SalsaNext/SalsaNext-master/saved_model Sequences folder exists! Using sequences from /home/mipso/zebra/NNs/SalsaNext/SalsaNext-master/dataset/sequences parsing seq 00 parsing seq 01 parsing seq 02 parsing seq 03 parsing seq 04 parsing seq 05 parsing seq 06 parsing seq 07 parsing seq 09 parsing seq 10 Using 19130 scans from sequences [0, 1, 2, 3, 4, 5, 6, 7, 9, 10] Sequences folder exists! Using sequences from /home/mipso/zebra/NNs/SalsaNext/SalsaNext-master/dataset/sequences parsing seq 08 Using 4071 scans from sequences [8] Sequences folder exists! Using sequences from /home/mipso/zebra/NNs/SalsaNext/SalsaNext-master/dataset/sequences parsing seq 11 parsing seq 12 parsing seq 13 parsing seq 14 parsing seq 15 parsing seq 16 parsing seq 17 parsing seq 18 parsing seq 19 parsing seq 20 parsing seq 21 Using 20351 scans from sequences [11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21]

---

Cleaning point-clouds with kNN post-processing kNN parameters: knn: 5 search: 5 sigma: 1.0 cutoff: 1.0 nclasses: 20

---

Infering in device: cpu * JS INFO: Entering infer *** JS INFO: Entering infer_subset ** ** JS INFO: self.gpu = False JS INFO: Entering infer_subset NO uncertainty ****

Illegal instruction (core dumped)

finishing infering. Starting evaluating

---

INTERFACE: Data: /home/mipso/zebra/NNs/SalsaNext/SalsaNext-master/dataset Predictions: /home/mipso/zebra/NNs/SalsaNext/SalsaNext-master/logs_preds Split: train Config: config/labels/semantic-kitti.yaml Limit: None

---

Opening data config file config/labels/semantic-kitti.yaml Ignoring xentropy class 0 in IoU evaluation [IOU EVAL] IGNORE: tensor([0]) [IOU EVAL] INCLUDE: tensor([ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19]) Traceback (most recent call last): File "./evaluate_iou.py", line 237, in eval(DATA["split"][FLAGS.split],splits,FLAGS.predictions) File "./evaluate_iou.py", line 67, in eval len(label_names) == len(pred_names)) AssertionError

[Mipsology]

One more detail. I copied the following files in directory containing trained model :

• salanext.yml as arch_cfg.yaml
• semantic-kitti.yaml as data_cfg.yaml

[Mipsology]

Hello, please ignore this issue. And if possible, please delete the issues to avoid any confusion for future visitors. Apologies for any confusion caused.

[HuangVictorAuto]

@Mipsology , did you solve the problem? I encounter the same issue.