Upload benchmarks/bench_06_autosense.json with huggingface_hub
Browse files
benchmarks/bench_06_autosense.json
ADDED
|
@@ -0,0 +1,125 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"benchmark": "autosense",
|
| 3 |
+
"timestamp": "2026-03-19T12:05:03.088761+00:00",
|
| 4 |
+
"model_dir": "/home/datai/development/forge/datasets",
|
| 5 |
+
"n_models_scanned": 14,
|
| 6 |
+
"vision_encoders": {
|
| 7 |
+
"HuggingFaceTB--SmolVLM-256M-Instruct": {
|
| 8 |
+
"d_output": 768,
|
| 9 |
+
"image_size": 512,
|
| 10 |
+
"patch_size": 16,
|
| 11 |
+
"n_tokens": 1024
|
| 12 |
+
},
|
| 13 |
+
"HuggingFaceTB--SmolVLM-500M-Instruct": {
|
| 14 |
+
"d_output": 768,
|
| 15 |
+
"image_size": 512,
|
| 16 |
+
"patch_size": 16,
|
| 17 |
+
"n_tokens": 1024
|
| 18 |
+
},
|
| 19 |
+
"Qwen--Qwen2.5-0.5B": {
|
| 20 |
+
"d_output": 896
|
| 21 |
+
},
|
| 22 |
+
"Qwen--Qwen2.5-0.5B-Instruct": {
|
| 23 |
+
"d_output": 896
|
| 24 |
+
},
|
| 25 |
+
"Qwen--Qwen2.5-1.5B": {
|
| 26 |
+
"d_output": 1536
|
| 27 |
+
},
|
| 28 |
+
"facebook--dinov2-base": {
|
| 29 |
+
"d_output": 768,
|
| 30 |
+
"image_size": 518,
|
| 31 |
+
"patch_size": 14,
|
| 32 |
+
"n_tokens": 1369
|
| 33 |
+
},
|
| 34 |
+
"facebook--dinov2-small": {
|
| 35 |
+
"d_output": 384,
|
| 36 |
+
"image_size": 518,
|
| 37 |
+
"patch_size": 14,
|
| 38 |
+
"n_tokens": 1369
|
| 39 |
+
},
|
| 40 |
+
"google--siglip-so400m-patch14-384": {
|
| 41 |
+
"d_output": 1152,
|
| 42 |
+
"image_size": 384,
|
| 43 |
+
"patch_size": 14,
|
| 44 |
+
"n_tokens": 729
|
| 45 |
+
},
|
| 46 |
+
"hongyuw--bitvla-bitsiglipL-224px-bf16": {
|
| 47 |
+
"d_output": 1152,
|
| 48 |
+
"image_size": 224,
|
| 49 |
+
"patch_size": 14,
|
| 50 |
+
"n_tokens": 256
|
| 51 |
+
}
|
| 52 |
+
},
|
| 53 |
+
"language_models": {
|
| 54 |
+
"Qwen--Qwen2.5-0.5B": {
|
| 55 |
+
"d_model": 896,
|
| 56 |
+
"vocab_size": 151936,
|
| 57 |
+
"n_layers": 24,
|
| 58 |
+
"n_heads": 14
|
| 59 |
+
},
|
| 60 |
+
"Qwen--Qwen2.5-0.5B-Instruct": {
|
| 61 |
+
"d_model": 896,
|
| 62 |
+
"vocab_size": 151936,
|
| 63 |
+
"n_layers": 24,
|
| 64 |
+
"n_heads": 14
|
| 65 |
+
},
|
| 66 |
+
"Qwen--Qwen2.5-1.5B": {
|
| 67 |
+
"d_model": 1536,
|
| 68 |
+
"vocab_size": 151936,
|
| 69 |
+
"n_layers": 28,
|
| 70 |
+
"n_heads": 12
|
| 71 |
+
},
|
| 72 |
+
"facebook--dinov2-base": {
|
| 73 |
+
"d_model": 768,
|
| 74 |
+
"n_layers": 12,
|
| 75 |
+
"n_heads": 12
|
| 76 |
+
},
|
| 77 |
+
"facebook--dinov2-small": {
|
| 78 |
+
"d_model": 384,
|
| 79 |
+
"n_layers": 12,
|
| 80 |
+
"n_heads": 6
|
| 81 |
+
}
|
| 82 |
+
},
|
| 83 |
+
"scan_times_ms": {
|
| 84 |
+
"vision_HuggingFaceTB--SmolVLM-256M-Instruct": 0.16,
|
| 85 |
+
"vision_HuggingFaceTB--SmolVLM-500M-Instruct": 0.11,
|
| 86 |
+
"vision_Qwen--Qwen2.5-0.5B": 0.04,
|
| 87 |
+
"language_Qwen--Qwen2.5-0.5B": 0.04,
|
| 88 |
+
"vision_Qwen--Qwen2.5-0.5B-Instruct": 0.04,
|
| 89 |
+
"language_Qwen--Qwen2.5-0.5B-Instruct": 0.07,
|
| 90 |
+
"vision_Qwen--Qwen2.5-1.5B": 0.04,
|
| 91 |
+
"language_Qwen--Qwen2.5-1.5B": 0.04,
|
| 92 |
+
"vision_facebook--dinov2-base": 0.04,
|
| 93 |
+
"language_facebook--dinov2-base": 0.04,
|
| 94 |
+
"vision_facebook--dinov2-small": 0.03,
|
| 95 |
+
"language_facebook--dinov2-small": 0.03,
|
| 96 |
+
"vision_google--siglip-so400m-patch14-384": 0.03,
|
| 97 |
+
"vision_hongyuw--bitvla-bitsiglipL-224px-bf16": 0.05
|
| 98 |
+
},
|
| 99 |
+
"config_tests": {
|
| 100 |
+
"default_qwen05b": {
|
| 101 |
+
"before": {
|
| 102 |
+
"bridge_d_vision": 1152,
|
| 103 |
+
"bridge_d_model": 896
|
| 104 |
+
},
|
| 105 |
+
"after": {
|
| 106 |
+
"bridge_d_vision": 1152,
|
| 107 |
+
"bridge_d_model": 896
|
| 108 |
+
},
|
| 109 |
+
"changed": false,
|
| 110 |
+
"time_ms": 0.12
|
| 111 |
+
},
|
| 112 |
+
"qwen15b": {
|
| 113 |
+
"before": {
|
| 114 |
+
"bridge_d_vision": 1152,
|
| 115 |
+
"bridge_d_model": 896
|
| 116 |
+
},
|
| 117 |
+
"after": {
|
| 118 |
+
"bridge_d_vision": 1152,
|
| 119 |
+
"bridge_d_model": 1536
|
| 120 |
+
},
|
| 121 |
+
"changed": true,
|
| 122 |
+
"time_ms": 0.12
|
| 123 |
+
}
|
| 124 |
+
}
|
| 125 |
+
}
|