NAN loss while training locally on COCO type dataset

[anti79]

💡 Your Question

Hi, I need some help with training a model. I've tried training with a custom script, but got NaN and 0 loss values during training, so I figured that the problem might be either with my dataset or with my code. So I decided to use a ready made notebook with a public Kaggle dataset to try to locate the issue.
The notebook: https://colab.research.google.com/drive/1q0RmeVRzLwRXW-h9dPFSOchwJkThUy6d
The dataset: https://www.kaggle.com/datasets/sharansmenon/aquarium-dataset
Up to the point of training, all was going well, I got the model instantiated and loaded the labels.

During training, the output was:
`[2024-08-29 23:29:54] INFO - sg_trainer_utils.py - TRAINING PARAMETERS:
- Mode: Single GPU
- Number of GPUs: 1 (1 available on the machine)
- Full dataset size: 447 (len(train_set))
- Batch size per GPU: 4 (batch_size)
- Batch Accumulate: 1 (batch_accumulate)
- Total batch size: 4 (num_gpus * batch_size)
- Effective Batch size: 4 (num_gpus * batch_size * batch_accumulate)
- Iterations per epoch: 111 (len(train_loader))
- Gradient updates per epoch: 111 (len(train_loader) / batch_accumulate)
- Model: YoloNAS_S (19.02M parameters, 19.02M optimized)
- Learning Rates and Weight Decays:
- default: (19.02M parameters). LR: 0.0005 (19.02M parameters) WD: 0.0, (42.15K parameters), WD: 0.0001, (18.98M parameters)

[2024-08-29 23:29:54] INFO - sg_trainer.py - Started training for 100 epochs (0/99)

Train epoch 0: 0%| | 0/111 [00:00<?, ?it/s]/home/daniel/.pyenv/versions/3.10.14/lib/python3.10/site-packages/super_gradients/training/sg_trainer/sg_trainer.py:502: FutureWarning: torch.cuda.amp.autocast(args...) is deprecated. Please use torch.amp.autocast('cuda', args...) instead.
with autocast(enabled=self.training_params.mixed_precision):
Train epoch 0: 10%|▉ | 11/111 [00:14<02:06, 1.26s/it, PPYoloELoss/loss=nan, PPYoloELoss/loss_cls=nan, PPYoloELoss/loss_dfl=nan, PPYoloELoss/loss_iou=nan, gpu_mem=1.75]`

Thought that maybe I need to wait for a couple epochs to pass to start getting loss values, but still the values remained NaN. I tried running the same code on a different machine, but got the same result. On Colab, however, the code finally worked, displaying the loss values.

My environment:
`PyTorch version: 2.4.0+cu121
Is debug build: False
CUDA used to build PyTorch: 12.1
ROCM used to build PyTorch: N/A

OS: Fedora Linux 40 (Xfce) (x86_64)
GCC version: (GCC) 14.2.1 20240801 (Red Hat 14.2.1-1)
Clang version: Could not collect
CMake version: version 3.28.2
Libc version: glibc-2.39

Python version: 3.10.14 (main, Jun 30 2024, 01:17:49) [GCC 14.1.1 20240620 (Red Hat 14.1.1-6)] (64-bit runtime)
Python platform: Linux-6.10.4-200.fc40.x86_64-x86_64-with-glibc2.39
Is CUDA available: True
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: GPU 0: NVIDIA GeForce GTX 1650
Nvidia driver version: 555.58.02
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 43 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 12
On-line CPU(s) list: 0-11
Vendor ID: AuthenticAMD
Model name: AMD Ryzen 5 3600X 6-Core Processor
CPU family: 23
Model: 113
Thread(s) per core: 2
Core(s) per socket: 6
Socket(s): 1
Stepping: 0
Frequency boost: enabled
CPU(s) scaling MHz: 86%
CPU max MHz: 4756.6401
CPU min MHz: 2200.0000
BogoMIPS: 8199.84
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif v_spec_ctrl umip rdpid overflow_recov succor smca sev sev_es
Virtualization: AMD-V
L1d cache: 192 KiB (6 instances)
L1i cache: 192 KiB (6 instances)
L2 cache: 3 MiB (6 instances)
L3 cache: 32 MiB (2 instances)
NUMA node(s): 1
NUMA node0 CPU(s): 0-11
Vulnerability Gather data sampling: Not affected
Vulnerability Itlb multihit: Not affected
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Mmio stale data: Not affected
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed: Mitigation; untrained return thunk; SMT enabled with STIBP protection
Vulnerability Spec rstack overflow: Mitigation; Safe RET
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; Retpolines; IBPB conditional; STIBP always-on; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected

Versions of relevant libraries:
[pip3] numpy==1.23.0
[pip3] onnx==1.15.0
[pip3] onnxruntime==1.15.0
[pip3] onnxsim==0.4.36
[pip3] torch==2.4.0
[pip3] torchmetrics==0.8.0
[pip3] torchvision==0.19.0
[pip3] triton==3.0.0
[conda] Could not collect`

Environment on second machine:
`PyTorch version: 2.4.0+cu121
Is debug build: False
CUDA used to build PyTorch: 12.1
ROCM used to build PyTorch: N/A

OS: Kali GNU/Linux Rolling (x86_64)
GCC version: (Debian 13.2.0-24) 13.2.0
Clang version: 14.0.6-2
CMake version: version 3.28.3
Libc version: glibc-2.38

Python version: 3.12.5 (main, Aug 18 2024, 01:30:37) [GCC 13.2.0] (64-bit runtime)
Python platform: Linux-6.6.15-amd64-x86_64-with-glibc2.38
Is CUDA available: True
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration:
GPU 0: NVIDIA GeForce GTX 1660 SUPER
GPU 1: NVIDIA GeForce GTX 1660 SUPER

Nvidia driver version: 555.42.02
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 39 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 8
On-line CPU(s) list: 0-7
Vendor ID: GenuineIntel
Model name: Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz
CPU family: 6
Model: 60
Thread(s) per core: 2
Core(s) per socket: 4
Socket(s): 1
Stepping: 3
CPU(s) scaling MHz: 65%
CPU max MHz: 4000.0000
CPU min MHz: 800.0000
BogoMIPS: 7198.06
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm cpuid_fault pti ssbd ibrs ibpb stibp tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid xsaveopt dtherm ida arat pln pts vnmi md_clear flush_l1d
Virtualization: VT-x
L1d cache: 128 KiB (4 instances)
L1i cache: 128 KiB (4 instances)
L2 cache: 1 MiB (4 instances)
L3 cache: 8 MiB (1 instance)
NUMA node(s): 1
NUMA node0 CPU(s): 0-7
Vulnerability Gather data sampling: Not affected
Vulnerability Itlb multihit: KVM: Mitigation: VMX disabled
Vulnerability L1tf: Mitigation; PTE Inversion; VMX conditional cache flushes, SMT vulnerable
Vulnerability Mds: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Meltdown: Mitigation; PTI
Vulnerability Mmio stale data: Unknown: No mitigations
Vulnerability Retbleed: Not affected
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; Retpolines, IBPB conditional, IBRS_FW, STIBP conditional, RSB filling, PBRSB-eIBRS Not affected
Vulnerability Srbds: Mitigation; Microcode
Vulnerability Tsx async abort: Not affected

Versions of relevant libraries:
[pip3] numpy==1.26.4
[pip3] torch==2.4.0
[pip3] torchvision==0.19.0
[pip3] triton==3.0.0
[conda] Could not collect`

Thanks in advance for any suggestions on what the issue might be.

Versions

No response

[anti79]

More output from training:

===========================================================
Train epoch 1: 100%|██████████| 111/111 [02:23<00:00,  1.29s/it, PPYoloELoss/loss=nan, PPYoloELoss/loss_cls=nan, PPYoloELoss/loss_dfl=nan, PPYoloELoss/loss_iou=nan, gpu_mem=2.76]
Validating epoch 1: 100%|██████████| 32/32 [00:04<00:00,  7.99it/s]
===========================================================
SUMMARY OF EPOCH 1
├── Train
│   ├── Ppyoloeloss/loss_cls = nan
│   │   ├── Epoch N-1      = nan    (= nan)
│   │   └── Best until now = nan    (= nan)
│   ├── Ppyoloeloss/loss_iou = nan
│   │   ├── Epoch N-1      = nan    (= nan)
│   │   └── Best until now = nan    (= nan)
│   ├── Ppyoloeloss/loss_dfl = nan
│   │   ├── Epoch N-1      = nan    (= nan)
│   │   └── Best until now = nan    (= nan)
│   └── Ppyoloeloss/loss = nan
│       ├── Epoch N-1      = nan    (= nan)
│       └── Best until now = nan    (= nan)
└── Validation
    ├── Ppyoloeloss/loss_cls = nan
    │   ├── Epoch N-1      = nan    (= nan)
    │   └── Best until now = nan    (= nan)
    ├── Ppyoloeloss/loss_iou = nan
    │   ├── Epoch N-1      = nan    (= nan)
    │   └── Best until now = nan    (= nan)
    ├── Ppyoloeloss/loss_dfl = nan
    │   ├── Epoch N-1      = nan    (= nan)
    │   └── Best until now = nan    (= nan)
...
        ├── Epoch N-1      = 0.0    (= 0.0)
        └── Best until now = 0.0    (= 0.0)