matilda-jev-v1 / TEST_REPORT.json
yue-maincode's picture
MATILDA-jev v1 — current release
d10ad42
Raw History Blame Contribute Delete
2.64 kB
{
"created_utc": "2026-09-30T13:28:46.503362+00:00",
"passed": true,
"model_directory": "/shared/model-cache/matilda-jev-v1-matilda-config",
"edition": "MATILDA configuration aliases for existing matilda-jev-v1 weights",
"source_repository": "Maincode/matilda-jev-v1",
"changed_configs": [
"config.json",
"decision_config.json",
"processor_config.json",
"tokenizer_config.json"
],
"upstream_name_absent_from_changed_configs": true,
"weights_and_tokenizer_unchanged": true,
"source_hash_verified_files": 15,
"configuration_and_processor_tests": {
"passed": true,
"classes": [
"MatildaJevConfig",
"MatildaJevProcessor",
"MatildaJevTokenizer",
"MatildaJevImageProcessor",
"MatildaJevVideoProcessor"
],
"checkpoint": "/shared/model-cache/matilda-jev-v1-matilda-config",
"architecture_parameters_identical": true,
"tokenization_identical": true,
"image_preprocessing_identical": true,
"config_processor_save_reload_passed": true
},
"gpu_job": "71652",
"device": "AMD Instinct MI355X",
"loaded_classes": {
"model": "MatildaJevModel",
"config": "MatildaJevConfig",
"processor": "MatildaJevProcessor"
},
"questions_passed": 25,
"questions_total": 25,
"semantic_summary": {
"smoke": {
"correct": 18,
"total": 18
},
"identity": {
"correct": 6,
"total": 6
},
"image": {
"correct": 1,
"total": 1
}
},
"probability_parity": {
"passed": true,
"max_probability_delta": 0.0,
"matched_requests": 23,
"total_requests": 23,
"tolerance": 1e-06
},
"training_backward_smoke": {
"passed": true,
"loss": 2.276871782669332e-05,
"finite_nonzero_gradients": [
"readout",
"embedding"
],
"optimizer_step_performed": false
},
"request_validation_checks": {
"unknown_model_rejected": true,
"empty_options_rejected": true,
"overlong_input_rejected_without_truncation": true
},
"load_seconds": 61.5,
"peak_memory_gib_including_backward": 107.215,
"limitations": [
"25-question inference smoke test, not a full benchmark rerun.",
"Backward pass validated without optimizer update or a full continued-training run.",
"Tested with Transformers 5.17.0 and PyTorch 2.14.0 on AMD MI355X; other backends are untested."
],
"usage": "Use bundled runtime or AutoModel/AutoProcessor with trust_remote_code=True. AutoModel returns the backbone; bundled runtime also loads the decision head.",
"colleague_readable": true,
"independent_weight_files": true,
"uploaded_at_validation": false
}