Upload paired SFT v2 evidence v2-d016-primary

evidence/v2-d016-primary/RUN_COMPLETE.json

@@ -0,0 +1,9 @@
+{
+  "post_evaluation_sha256": "777326f6e41de81365712663413a50eb28b2198550550c794f1eff32cf3c8824",
+  "repeat": false,
+  "role": "primary_cohort_v3",
+  "run_id": "v2-d016-primary",
+  "schema_version": 2,
+  "sft_complete_sha256": "e79dc60f902c3d900c9d5b2a20125e3f3dee90b7e49c88c5af401daac36b14af",
+  "status": "complete_sft_v2_intervention_and_paired_evaluation"
+}

evidence/v2-d016-primary/base_eval/PAIRED_EVALUATION_COMPLETE.json

@@ -0,0 +1,8 @@
+{
+  "fineweb_mean_cross_entropy": 3.493547351215966,
+  "fineweb_sha256": "07b3e94a257def902b52b1f482e003c5a3a570ca31cfdbc87b71b565d2c31d80",
+  "instruction_mean_cross_entropy": 4.819426922926331,
+  "instruction_sha256": "3d96e34c841c25db5d47bba437fec90948ffe45c1939dc16436da8a5f4d3efb3",
+  "schema_version": 2,
+  "status": "complete_paired_posttraining_evaluation_v2"
+}

evidence/v2-d016-primary/base_eval/final_test_instruction_ce.json

@@ -0,0 +1,16 @@
+{
+  "assistant_tokens": 3550108,
+  "data_sha256": "2319889ce5b02855cee8d695ab21bb068f735599bf974c9566eee39e5d80ced0",
+  "elapsed_seconds": 58.60041379928589,
+  "examples": 8147,
+  "mean_cross_entropy": 4.819426922926331,
+  "model_safetensors_sha256": "ee38407e72759240ca255161f89fb33c7faa077acf7842eb9f3633016e97fcd5",
+  "per_example_losses_sha256": "772ec21d6e37d980ac189513f00c811e49ee550da2edab8d05f20ed7b4eb8d78",
+  "per_example_mean": 4.695056627978134,
+  "per_example_std": 1.5849166045512593,
+  "per_example_token_counts_sha256": "65d9ca7edff2404f23581499e374e0556ce80463cb297bc81fea28253c03051b",
+  "schema_version": 2,
+  "source_examples": 8192,
+  "status": "complete_paired_assistant_token_cross_entropy_v2",
+  "unscorable_examples": 45
+}

evidence/v2-d016-primary/base_eval/fineweb/fineweb_per_sequence_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ab03f9c8b5ee2d0fa7535cd20931bdbbcab66384e6b01e986a7c933fa26439e
+size 16512

evidence/v2-d016-primary/base_eval/fineweb/fineweb_validation_ce.json

@@ -0,0 +1,12 @@
+{
+  "elapsed_seconds": 60.524249792099,
+  "mean_cross_entropy": 3.493547351215966,
+  "model_safetensors_sha256": "ee38407e72759240ca255161f89fb33c7faa077acf7842eb9f3633016e97fcd5",
+  "per_sequence_losses_sha256": "2ab03f9c8b5ee2d0fa7535cd20931bdbbcab66384e6b01e986a7c933fa26439e",
+  "prediction_tokens": 16777216,
+  "schema_version": 2,
+  "sequence_std": 0.23696278157056427,
+  "sequences": 2048,
+  "status": "complete_same_pipeline_fineweb_cross_entropy_v2",
+  "token_array_sha256": "83654c3552e76c84bb9f958cfe95e8b60b9dafd10c692c8fb857354bc4895263"
+}

evidence/v2-d016-primary/base_eval/instruction_per_example_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:772ec21d6e37d980ac189513f00c811e49ee550da2edab8d05f20ed7b4eb8d78
+size 65304

evidence/v2-d016-primary/base_eval/instruction_per_example_token_counts.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65d9ca7edff2404f23581499e374e0556ce80463cb297bc81fea28253c03051b
+size 65304

evidence/v2-d016-primary/post_eval/PAIRED_EVALUATION_COMPLETE.json

@@ -0,0 +1,8 @@
+{
+  "fineweb_mean_cross_entropy": 3.7379375426098704,
+  "fineweb_sha256": "8cf953dd9cb98f1318d114d8e232d4976f367685d643a2582ed1cac7e5be35ef",
+  "instruction_mean_cross_entropy": 2.993923057004482,
+  "instruction_sha256": "01d5ae7c23b5989b58b2e3b2a5a6290465d62bb489d86c6130989a94672c46a7",
+  "schema_version": 2,
+  "status": "complete_paired_posttraining_evaluation_v2"
+}

evidence/v2-d016-primary/post_eval/final_test_instruction_ce.json

@@ -0,0 +1,16 @@
+{
+  "assistant_tokens": 3550108,
+  "data_sha256": "2319889ce5b02855cee8d695ab21bb068f735599bf974c9566eee39e5d80ced0",
+  "elapsed_seconds": 58.49284029006958,
+  "examples": 8147,
+  "mean_cross_entropy": 2.993923057004482,
+  "model_safetensors_sha256": "5bad373feb943133e297897e69f19f6bd99d148ecee5cadd493bab66389077a4",
+  "per_example_losses_sha256": "5812100571142d89a7cde6960d8874082be47e73e9dc1376fd4303499bb792ce",
+  "per_example_mean": 3.0740301496224327,
+  "per_example_std": 1.387150541447835,
+  "per_example_token_counts_sha256": "65d9ca7edff2404f23581499e374e0556ce80463cb297bc81fea28253c03051b",
+  "schema_version": 2,
+  "source_examples": 8192,
+  "status": "complete_paired_assistant_token_cross_entropy_v2",
+  "unscorable_examples": 45
+}

evidence/v2-d016-primary/post_eval/fineweb/fineweb_per_sequence_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b538337af929b17ec16ffbfae0c510bf62efd4730e0fbcdc557e2c7114a7a78
+size 16512

evidence/v2-d016-primary/post_eval/fineweb/fineweb_validation_ce.json

@@ -0,0 +1,12 @@
+{
+  "elapsed_seconds": 60.52213954925537,
+  "mean_cross_entropy": 3.7379375426098704,
+  "model_safetensors_sha256": "5bad373feb943133e297897e69f19f6bd99d148ecee5cadd493bab66389077a4",
+  "per_sequence_losses_sha256": "7b538337af929b17ec16ffbfae0c510bf62efd4730e0fbcdc557e2c7114a7a78",
+  "prediction_tokens": 16777216,
+  "schema_version": 2,
+  "sequence_std": 0.23869560964201075,
+  "sequences": 2048,
+  "status": "complete_same_pipeline_fineweb_cross_entropy_v2",
+  "token_array_sha256": "83654c3552e76c84bb9f958cfe95e8b60b9dafd10c692c8fb857354bc4895263"
+}

evidence/v2-d016-primary/post_eval/instruction_per_example_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5812100571142d89a7cde6960d8874082be47e73e9dc1376fd4303499bb792ce
+size 65304

evidence/v2-d016-primary/post_eval/instruction_per_example_token_counts.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65d9ca7edff2404f23581499e374e0556ce80463cb297bc81fea28253c03051b
+size 65304

evidence/v2-d016-primary/sft/SFT_COMPLETE.json

@@ -0,0 +1,18 @@
+{
+  "contract_sha256": "871edad8c053996b27787cb0657d176486911b7a1f67f6d3980c085e913247b5",
+  "development_instruction_ce": null,
+  "elapsed_seconds": 1211.2791221141815,
+  "fineweb_validation_ce": null,
+  "learning_rate": 0.0001,
+  "max_update_steps": 1024,
+  "model_tree_sha256": "ba1b3bd1cdb4341bf9250ea6e8b58561e00b40952530c09afffbedecb398a7ba",
+  "model_weight_format": "verified_huggingface_safetensors_converted_from_official_open_instruct_pytorch_bin",
+  "recovered_from_exact_completed_training_artifact": false,
+  "registered_evaluation_owner": "posttraining_sft_instrument_split_runtime_supervisor",
+  "role": "primary_cohort_v3",
+  "run_id": "v2-d016-primary",
+  "schema_version": 1,
+  "status": "complete_open_instruct_sft_intervention_v1",
+  "wandb_evaluation_status": "deferred_to_registered_supervisor",
+  "wandb_run_id": "sftv3-v2-d016-primary"
+}