| { |
| "benchmark": "autosense", |
| "timestamp": "2026-03-19T12:05:03.088761+00:00", |
| "model_dir": "/home/datai/development/forge/datasets", |
| "n_models_scanned": 14, |
| "vision_encoders": { |
| "HuggingFaceTB--SmolVLM-256M-Instruct": { |
| "d_output": 768, |
| "image_size": 512, |
| "patch_size": 16, |
| "n_tokens": 1024 |
| }, |
| "HuggingFaceTB--SmolVLM-500M-Instruct": { |
| "d_output": 768, |
| "image_size": 512, |
| "patch_size": 16, |
| "n_tokens": 1024 |
| }, |
| "Qwen--Qwen2.5-0.5B": { |
| "d_output": 896 |
| }, |
| "Qwen--Qwen2.5-0.5B-Instruct": { |
| "d_output": 896 |
| }, |
| "Qwen--Qwen2.5-1.5B": { |
| "d_output": 1536 |
| }, |
| "facebook--dinov2-base": { |
| "d_output": 768, |
| "image_size": 518, |
| "patch_size": 14, |
| "n_tokens": 1369 |
| }, |
| "facebook--dinov2-small": { |
| "d_output": 384, |
| "image_size": 518, |
| "patch_size": 14, |
| "n_tokens": 1369 |
| }, |
| "google--siglip-so400m-patch14-384": { |
| "d_output": 1152, |
| "image_size": 384, |
| "patch_size": 14, |
| "n_tokens": 729 |
| }, |
| "hongyuw--bitvla-bitsiglipL-224px-bf16": { |
| "d_output": 1152, |
| "image_size": 224, |
| "patch_size": 14, |
| "n_tokens": 256 |
| } |
| }, |
| "language_models": { |
| "Qwen--Qwen2.5-0.5B": { |
| "d_model": 896, |
| "vocab_size": 151936, |
| "n_layers": 24, |
| "n_heads": 14 |
| }, |
| "Qwen--Qwen2.5-0.5B-Instruct": { |
| "d_model": 896, |
| "vocab_size": 151936, |
| "n_layers": 24, |
| "n_heads": 14 |
| }, |
| "Qwen--Qwen2.5-1.5B": { |
| "d_model": 1536, |
| "vocab_size": 151936, |
| "n_layers": 28, |
| "n_heads": 12 |
| }, |
| "facebook--dinov2-base": { |
| "d_model": 768, |
| "n_layers": 12, |
| "n_heads": 12 |
| }, |
| "facebook--dinov2-small": { |
| "d_model": 384, |
| "n_layers": 12, |
| "n_heads": 6 |
| } |
| }, |
| "scan_times_ms": { |
| "vision_HuggingFaceTB--SmolVLM-256M-Instruct": 0.16, |
| "vision_HuggingFaceTB--SmolVLM-500M-Instruct": 0.11, |
| "vision_Qwen--Qwen2.5-0.5B": 0.04, |
| "language_Qwen--Qwen2.5-0.5B": 0.04, |
| "vision_Qwen--Qwen2.5-0.5B-Instruct": 0.04, |
| "language_Qwen--Qwen2.5-0.5B-Instruct": 0.07, |
| "vision_Qwen--Qwen2.5-1.5B": 0.04, |
| "language_Qwen--Qwen2.5-1.5B": 0.04, |
| "vision_facebook--dinov2-base": 0.04, |
| "language_facebook--dinov2-base": 0.04, |
| "vision_facebook--dinov2-small": 0.03, |
| "language_facebook--dinov2-small": 0.03, |
| "vision_google--siglip-so400m-patch14-384": 0.03, |
| "vision_hongyuw--bitvla-bitsiglipL-224px-bf16": 0.05 |
| }, |
| "config_tests": { |
| "default_qwen05b": { |
| "before": { |
| "bridge_d_vision": 1152, |
| "bridge_d_model": 896 |
| }, |
| "after": { |
| "bridge_d_vision": 1152, |
| "bridge_d_model": 896 |
| }, |
| "changed": false, |
| "time_ms": 0.12 |
| }, |
| "qwen15b": { |
| "before": { |
| "bridge_d_vision": 1152, |
| "bridge_d_model": 896 |
| }, |
| "after": { |
| "bridge_d_vision": 1152, |
| "bridge_d_model": 1536 |
| }, |
| "changed": true, |
| "time_ms": 0.12 |
| } |
| } |
| } |