| { |
| "format_version": 3, |
| "project": "jolt", |
| "distribution": "jolt-model-runtime", |
| "version": "0.1.0", |
| "model_id": "Jolt-4B", |
| "base_model": "Qwen/Qwen3.5-4B", |
| "base_revision": "a7b0d22b993d71000cf2eadfb37222a67cee521e", |
| "adapter": "qwen3.5", |
| "lora_rank": 8, |
| "lora_alpha": 16, |
| "lora_dropout": 0.0, |
| "training_loss": "direct-composite", |
| "image_pixels": 262144, |
| "max_length": 2048, |
| "backend": { |
| "linear_patch": true, |
| "triton_convolution": true, |
| "experimental_rocm_sdpa": true, |
| "fused_norm_and_swiglu": true, |
| "shared_prefix": true, |
| "frozen_vision_cache_MiB": 128 |
| }, |
| "architecture": { |
| "readout": "last", |
| "candidate_attention": false, |
| "adapters": { |
| "layers": [], |
| "width": 64, |
| "activation": "silu" |
| }, |
| "head": "linear", |
| "unfused_mlp_layers": [], |
| "backbone_activation": { |
| "layers": [], |
| "initial_mix": 0.0, |
| "learning_rate": 0.001 |
| }, |
| "gpu_memory_fraction": 0.9 |
| }, |
| "prompt_layout": "inline", |
| "inference": "merged LoRA, BF16 inference, shared-prefix parallel candidate scoring", |
| "source_hashes": { |
| "train": "4e8b258c9e5d5a98810868df5d572a824a9d1d6c5aa557ab5123cb7aab5de373", |
| "dev": "b736750efff3e30ae8a7772c482fa248899b8a1cf69ace104310e2540671d552", |
| "calibration": "815cfe0855dd5494f9def19ddb2a6b12e4116c3ae24d2775db620021dd07fb28", |
| "test": "fdac64d536d67c011d55d00352f0be71fa3bdfb7811c0be0e05a5d66cec98095" |
| }, |
| "selected_run": "L2, seed 42", |
| "selected_step": 3600, |
| "selection": "maximum development macro accuracy within the run; no test selection", |
| "seed": 42, |
| "dev_macro_accuracy": 0.843863829343734, |
| "temperatures": { |
| "choice": 1.0, |
| "noul": 1.0, |
| "score": 1.0 |
| }, |
| "weights_sha256": "180666338c3864e41ceee9e6d4eeca93e39c9e15c830cbe1258b19d9f6e1ff64", |
| "training_precision": "BF16 base", |
| "training_quantization": null |
| } |
|
|
