{ "passed": true, "reference_commit": "fd3cf0140de03f7619ac95b5cfed05b711ff9e22", "source_sha256": "4463b8a108f678c166755101d95f188179f54143305953d78147de0389b80ebc", "torch": "2.15.0+cpu", "numpy": "2.5.4", "implementation_sha256": "48a2a97d0445dc6e08ac7f27a935637a7553417d0982eb94122d6cb02bbb499c", "tests": [ { "test": "layer0_rmsnorm", "shape": [ 2, 7, 16 ], "max_abs_error": 2.384185791015625e-07, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "layer0_mixer", "shape": [ 2, 7, 16 ], "max_abs_error": 1.862645149230957e-07, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "layer0_mlp", "shape": [ 2, 7, 16 ], "max_abs_error": 2.9802322387695312e-08, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "layer0_complete", "shape": [ 2, 7, 16 ], "max_abs_error": 2.384185791015625e-07, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "layer1_rmsnorm", "shape": [ 2, 7, 16 ], "max_abs_error": 2.384185791015625e-07, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "attention_probabilities", "shape": [ 2, 2, 7, 7 ], "max_abs_error": 1.1920928955078125e-07, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "layer1_mixer", "shape": [ 2, 7, 16 ], "max_abs_error": 5.21540641784668e-08, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "layer1_mlp", "shape": [ 2, 7, 16 ], "max_abs_error": 4.470348358154297e-08, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "layer1_complete", "shape": [ 2, 7, 16 ], "max_abs_error": 2.384185791015625e-07, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_recurrent_gated_delta_rule_length1_output", "shape": [ 2, 1, 2, 5 ], "max_abs_error": 4.470348358154297e-08, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_recurrent_gated_delta_rule_length1_state", "shape": [ 2, 2, 4, 5 ], "max_abs_error": 5.960464477539063e-08, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_chunk_gated_delta_rule_length1_output", "shape": [ 2, 1, 2, 5 ], "max_abs_error": 2.9802322387695312e-08, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_chunk_gated_delta_rule_length1_state", "shape": [ 2, 2, 4, 5 ], "max_abs_error": 5.960464477539063e-08, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_recurrent_gated_delta_rule_length7_output", "shape": [ 2, 7, 2, 5 ], "max_abs_error": 5.960464477539063e-08, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_recurrent_gated_delta_rule_length7_state", "shape": [ 2, 2, 4, 5 ], "max_abs_error": 5.960464477539063e-08, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_chunk_gated_delta_rule_length7_output", "shape": [ 2, 7, 2, 5 ], "max_abs_error": 7.450580596923828e-08, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_chunk_gated_delta_rule_length7_state", "shape": [ 2, 2, 4, 5 ], "max_abs_error": 1.7881393432617188e-07, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_recurrent_gated_delta_rule_length33_output", "shape": [ 2, 33, 2, 5 ], "max_abs_error": 1.1920928955078125e-07, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_recurrent_gated_delta_rule_length33_state", "shape": [ 2, 2, 4, 5 ], "max_abs_error": 2.9802322387695312e-08, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_chunk_gated_delta_rule_length33_output", "shape": [ 2, 33, 2, 5 ], "max_abs_error": 5.364418029785156e-07, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "torch_chunk_gated_delta_rule_length33_state", "shape": [ 2, 2, 4, 5 ], "max_abs_error": 3.5762786865234375e-07, "atol": 2e-05, "rtol": 2e-05, "passed": true }, { "test": "complete_two_layer_text_forward", "shape": [ 2, 7, 16 ], "max_abs_error": 3.5762786865234375e-07, "atol": 2e-05, "rtol": 2e-05, "passed": true } ], "limitations": "FP32 CPU semantics on tiny synthetic configurations; not full-checkpoint BF16 kernel equivalence" }