🚀 MLOS Release E2E Validation

Comprehensive testing of Axon and MLOS Core releases

Overall Status

100.0%

Total Duration

808.3s

Inferences

36/36

Models Tested

18

📦 Release Versions

Axon Version

v3.1.7

MLOS Core Version

4.1.3-alpha

Runtime Mode

Userspace Only (No Kernel Optimizations)
Kernel Module: No

💻 Hardware Specifications

Operating System

Linux Ubuntu 24.04.3 LTS
Architecture: x86_64

CPU

AMD EPYC 7763 64-Core Processor
Cores: 4 | Threads: 4

Memory

15 GB

GPU

Count: 0 | Memory:

Disk

72G
Available: 21G

📊 Resource Usage

MLOS Core (Idle)

CPU: 0.0% | Memory: 18 MB

MLOS Core (Under Load)

CPU: 17.8% (max: 17.8%)
Memory: 9908 MB (max: 9908 MB)

Axon

CPU: 0% | Memory: 0 MB

Resource Type

CPU: Used for model inference execution, ONNX Runtime operations, and HTTP request handling.

Memory: Used for model loading, input/output tensor buffers, and ONNX Runtime workspace.

GPU: Not used (CPU-only inference)

⏱️ Installation & Setup Times

Axon Download Time

565 ms

Core Download Time

2.0s

Core Startup Time

1.1s

Total Model Install Time

12.6 min

🚀 Inference Performance

🔤 NLP Models

ALBERT

✅
Small Inference
202 ms
Large Inference
917 ms

ROBERTA

✅
Small Inference
249 ms
Large Inference
1.3s

SENTENCE-TRANSFORMERS

✅
Small Inference
133 ms
Large Inference
212 ms

T5

✅
Small Inference
122 ms
Large Inference
215 ms

DISTILBERT

✅
Small Inference
117 ms
Large Inference
253 ms

GPT2

✅
Small Inference
472 ms
Large Inference
329 ms

BERT

✅
Small Inference
346 ms
Large Inference
1.3s

👁️ Vision Models

CONVNEXT

✅
Small Inference
1.2s
Large Inference
1.2s

RESNET

✅
Small Inference
1.2s
Large Inference
1.2s

EFFICIENTNET

✅
Small Inference
1.1s
Large Inference
1.1s

VIT

✅
Small Inference
1.3s
Large Inference
1.3s

MOBILENET

✅
Small Inference
1.1s
Large Inference
1.1s

DEIT

✅
Small Inference
1.2s
Large Inference
1.2s

🎨 Multimodal Models

CLIP

✅
Small Inference
1.2s
Large Inference
1.2s

🤖 LLM Models

LLAMA-3.2-1B

✅
Small Inference
1.3s
Large Inference
1.3s

QWEN2-0.5B

✅
Small Inference
691 ms
Large Inference
5.3s

TINYLLAMA

✅
Small Inference
599 ms
Large Inference
6.9s

DEEPSEEK-CODER-1.3B

✅
Small Inference
1.6s
Large Inference
1.5s

🤖 Model Support by Category

🧪 Test Details → 📦 View Full Config →

😊 NLP Models

  • ✅ GPT-2
  • ✅ BERT
  • ✅ RoBERTa
  • ✅ T5
Status: ✅ Passing

🔥 Vision Models

  • ✅ ResNet-50
  • ✅ ViT
  • ✅ ConvNeXt
  • ✅ MobileNet
  • ✅ DeiT
  • ✅ EfficientNet
Status: ✅ Passing

🎨 Multi-Modal

  • ✅ CLIP
  • ⏳ Wav2Vec2 (audio)
Status: ✅ Passing

📊 Model Details

🔤 NLP Models

ALBERT

✅
Install Time
25.4s
Register Time
311 ms

ROBERTA

✅
Install Time
1.9 min
Register Time
595 ms

SENTENCE-TRANSFORMERS

✅
Install Time
25.8s
Register Time
249 ms

T5

✅
Install Time
1.2 min
Register Time
495 ms

DISTILBERT

✅
Install Time
45.7s
Register Time
323 ms

GPT2

✅
Install Time
1.2 min
Register Time
762 ms

BERT

✅
Install Time
32.8s
Register Time
880 ms

👁️ Vision Models

CONVNEXT

✅
Install Time
28.4s
Register Time
189 ms

RESNET

✅
Install Time
27.2s
Register Time
201 ms

EFFICIENTNET

✅
Install Time
18.6s
Register Time
164 ms

VIT

✅
Install Time
56.7s
Register Time
420 ms

MOBILENET

✅
Install Time
16.7s
Register Time
135 ms

DEIT

✅
Install Time
25.8s
Register Time
200 ms

🎨 Multimodal Models

CLIP

✅
Install Time
58.7s
Register Time
822 ms

🤖 LLM Models

LLAMA-3.2-1B

✅
Install Time
36.0s
Register Time
875 ms

QWEN2-0.5B

✅
Install Time
18.1s
Register Time
367 ms

TINYLLAMA

✅
Install Time
36.1s
Register Time
583 ms

DEEPSEEK-CODER-1.3B

✅
Install Time
47.4s
Register Time
794 ms

📈 Performance Breakdown

📦 Model Installation (not shown in chart)

12.6 min

Model installation includes downloading from HuggingFace and ONNX conversion via Docker. This dominates the total time (~99%) so it's shown separately.