| { |
| "checkpoint_count": 7, |
| "model_family": "llama", |
| "namespace": "cache_mask_v1_reviewer_subset_v1", |
| "numpy_version": "2.3.5", |
| "records": [ |
| { |
| "checkpoint_path": "runs/llama-capacity-fixed-linear-t8-s1234/checkpoint/step_009000.pt", |
| "checkpoint_sha256": "5d7cf16dc5002b81ca5df87ba88e0edf9c45772c86d4ea492fee90fec22d9ece", |
| "checkpoint_size_bytes": 503630034, |
| "ema_payload_present": false, |
| "hub_slug": "llama-capacity-fixed-linear-t8-s1234", |
| "model_state_prefixes": [ |
| "feedback_proj", |
| "init_token", |
| "latent_pooler", |
| "mode_embed", |
| "out_ln", |
| "token0_head", |
| "transform" |
| ], |
| "optimizer_payload_present": false, |
| "private_string_count": 0, |
| "status": "PASS", |
| "top_level_keys": [ |
| "ckpt_mode", |
| "ema_shadow", |
| "model_slim", |
| "optim", |
| "rng_state", |
| "scaler", |
| "scheduler", |
| "step" |
| ], |
| "unsafe_globals": [ |
| "numpy._core.multiarray._reconstruct", |
| "numpy.dtype", |
| "numpy.ndarray" |
| ], |
| "victim_weight_keys": 0 |
| }, |
| { |
| "checkpoint_path": "runs/llama-capacity-fixed-residual-mlp-t8-s1234/checkpoint/step_010000.pt", |
| "checkpoint_sha256": "1c55e3249738c9abd322e6257d0457fda540a9e543a23666284436f92171d722", |
| "checkpoint_size_bytes": 2048577030, |
| "ema_payload_present": false, |
| "hub_slug": "llama-capacity-fixed-residual-mlp-t8-s1234", |
| "model_state_prefixes": [ |
| "feedback_proj", |
| "init_token", |
| "latent_pooler", |
| "mode_embed", |
| "out_ln", |
| "token0_head", |
| "transform" |
| ], |
| "optimizer_payload_present": false, |
| "private_string_count": 0, |
| "status": "PASS", |
| "top_level_keys": [ |
| "ckpt_mode", |
| "ema_shadow", |
| "model_slim", |
| "optim", |
| "rng_state", |
| "scaler", |
| "scheduler", |
| "step" |
| ], |
| "unsafe_globals": [ |
| "numpy._core.multiarray._reconstruct", |
| "numpy.dtype", |
| "numpy.ndarray" |
| ], |
| "victim_weight_keys": 0 |
| }, |
| { |
| "checkpoint_path": "runs/llama-length-full-t32-s1234/checkpoint/step_018000.pt", |
| "checkpoint_sha256": "cc6e24e9e24950f4ebd856660de8e3125a5f1d49e3619d761bf22efac9ea2025", |
| "checkpoint_size_bytes": 2048598413, |
| "ema_payload_present": false, |
| "hub_slug": "llama-length-full-t32-s1234", |
| "model_state_prefixes": [ |
| "feedback_proj", |
| "init_token", |
| "latent_pooler", |
| "mode_embed", |
| "out_ln", |
| "token0_head", |
| "transform" |
| ], |
| "optimizer_payload_present": false, |
| "private_string_count": 0, |
| "status": "PASS", |
| "top_level_keys": [ |
| "ckpt_mode", |
| "ema_shadow", |
| "model_slim", |
| "optim", |
| "rng_state", |
| "scaler", |
| "scheduler", |
| "step" |
| ], |
| "unsafe_globals": [ |
| "numpy._core.multiarray._reconstruct", |
| "numpy.dtype", |
| "numpy.ndarray" |
| ], |
| "victim_weight_keys": 0 |
| }, |
| { |
| "checkpoint_path": "runs/llama-prefix-direct-only-t8-s1234/checkpoint/step_011000.pt", |
| "checkpoint_sha256": "e748defca8c41f4919a75efec86ca5bb860ed896078e96055de8813bae718445", |
| "checkpoint_size_bytes": 2048598349, |
| "ema_payload_present": false, |
| "hub_slug": "llama-prefix-direct-only-t8-s1234", |
| "model_state_prefixes": [ |
| "feedback_proj", |
| "init_token", |
| "latent_pooler", |
| "mode_embed", |
| "out_ln", |
| "token0_head", |
| "transform" |
| ], |
| "optimizer_payload_present": false, |
| "private_string_count": 0, |
| "status": "PASS", |
| "top_level_keys": [ |
| "ckpt_mode", |
| "ema_shadow", |
| "model_slim", |
| "optim", |
| "rng_state", |
| "scaler", |
| "scheduler", |
| "step" |
| ], |
| "unsafe_globals": [ |
| "numpy._core.multiarray._reconstruct", |
| "numpy.dtype", |
| "numpy.ndarray" |
| ], |
| "victim_weight_keys": 0 |
| }, |
| { |
| "checkpoint_path": "runs/llama-primary-final-t8-s1234/checkpoint/step_019000.pt", |
| "checkpoint_sha256": "54b0ad7bdbeca96c859408439e98922ff38674f53ee40bc9b8d5285792d979a0", |
| "checkpoint_size_bytes": 2048598016, |
| "ema_payload_present": false, |
| "hub_slug": "llama-primary-final-t8-s1234", |
| "model_state_prefixes": [ |
| "feedback_proj", |
| "init_token", |
| "latent_pooler", |
| "mode_embed", |
| "out_ln", |
| "token0_head", |
| "transform" |
| ], |
| "optimizer_payload_present": false, |
| "private_string_count": 0, |
| "status": "PASS", |
| "top_level_keys": [ |
| "ckpt_mode", |
| "ema_shadow", |
| "model_slim", |
| "optim", |
| "rng_state", |
| "scaler", |
| "scheduler", |
| "step" |
| ], |
| "unsafe_globals": [ |
| "numpy._core.multiarray._reconstruct", |
| "numpy.dtype", |
| "numpy.ndarray" |
| ], |
| "victim_weight_keys": 0 |
| }, |
| { |
| "checkpoint_path": "runs/llama-primary-full-t8-s1234/checkpoint/step_023000.pt", |
| "checkpoint_sha256": "acc1715b3d6797472ef7460c6ff30bf21f69d0ab33c85b66598eafa3897fdc39", |
| "checkpoint_size_bytes": 2048598144, |
| "ema_payload_present": false, |
| "hub_slug": "llama-primary-full-t8-s1234", |
| "model_state_prefixes": [ |
| "feedback_proj", |
| "init_token", |
| "latent_pooler", |
| "mode_embed", |
| "out_ln", |
| "token0_head", |
| "transform" |
| ], |
| "optimizer_payload_present": false, |
| "private_string_count": 0, |
| "status": "PASS", |
| "top_level_keys": [ |
| "ckpt_mode", |
| "ema_shadow", |
| "model_slim", |
| "optim", |
| "rng_state", |
| "scaler", |
| "scheduler", |
| "step" |
| ], |
| "unsafe_globals": [ |
| "numpy._core.multiarray._reconstruct", |
| "numpy.dtype", |
| "numpy.ndarray" |
| ], |
| "victim_weight_keys": 0 |
| }, |
| { |
| "checkpoint_path": "runs/llama-primary-pooled-t8-s1234/checkpoint/step_015000.pt", |
| "checkpoint_sha256": "ce59e0cdafaa47d88071a5d2acb11c746f79830c31fe4894830435a62c7684f2", |
| "checkpoint_size_bytes": 2048598080, |
| "ema_payload_present": false, |
| "hub_slug": "llama-primary-pooled-t8-s1234", |
| "model_state_prefixes": [ |
| "feedback_proj", |
| "init_token", |
| "latent_pooler", |
| "mode_embed", |
| "out_ln", |
| "token0_head", |
| "transform" |
| ], |
| "optimizer_payload_present": false, |
| "private_string_count": 0, |
| "status": "PASS", |
| "top_level_keys": [ |
| "ckpt_mode", |
| "ema_shadow", |
| "model_slim", |
| "optim", |
| "rng_state", |
| "scaler", |
| "scheduler", |
| "step" |
| ], |
| "unsafe_globals": [ |
| "numpy._core.multiarray._reconstruct", |
| "numpy.dtype", |
| "numpy.ndarray" |
| ], |
| "victim_weight_keys": 0 |
| } |
| ], |
| "status": "PASS", |
| "torch_version": "2.12.0+cu130" |
| } |
|
|