Upload paired SFT v2 evidence v2-d045-primary

evidence/v2-d045-primary/RUN_COMPLETE.json

@@ -0,0 +1,9 @@
+{
+  "post_evaluation_sha256": "51822076633acb3bf5e8cb8771d1f050047bbce910165b18c62c026c7ebb52e9",
+  "repeat": false,
+  "role": "primary_cohort_v3",
+  "run_id": "v2-d045-primary",
+  "schema_version": 2,
+  "sft_complete_sha256": "5b28987f46c58e1f458573a5db78398874ce93b1dc41505cebc46c2185c021aa",
+  "status": "complete_sft_v2_intervention_and_paired_evaluation"
+}

evidence/v2-d045-primary/base_eval/PAIRED_EVALUATION_COMPLETE.json

@@ -0,0 +1,8 @@
+{
+  "fineweb_mean_cross_entropy": 3.4738386386306956,
+  "fineweb_sha256": "5569829efb91773296a5a503304dc0f8d861bd07f249565f3eee5bc5ba5e8a87",
+  "instruction_mean_cross_entropy": 4.802713294351031,
+  "instruction_sha256": "625353b8cf36c8d6c8aece0cdb891cda47ad5fca1398730ad4f6b8c4535cfad3",
+  "schema_version": 2,
+  "status": "complete_paired_posttraining_evaluation_v2"
+}

evidence/v2-d045-primary/base_eval/final_test_instruction_ce.json

@@ -0,0 +1,16 @@
+{
+  "assistant_tokens": 3550108,
+  "data_sha256": "2319889ce5b02855cee8d695ab21bb068f735599bf974c9566eee39e5d80ced0",
+  "elapsed_seconds": 58.79956579208374,
+  "examples": 8147,
+  "mean_cross_entropy": 4.802713294351031,
+  "model_safetensors_sha256": "51c4f3ff8e995e109230540c896304ecd96108364be66210a7e6c09f24df76a6",
+  "per_example_losses_sha256": "8936f9aa6e9fba7bbcdbc1e8a4ce6d80833f8d6eec42551bffd843a90784626f",
+  "per_example_mean": 4.606366756160786,
+  "per_example_std": 1.627594982285604,
+  "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-d045-primary/base_eval/fineweb/fineweb_per_sequence_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d3ec157b4a3e5f16356217f420035ba1acbdf9ebdab54b12af958912c2cc04c
+size 16512

evidence/v2-d045-primary/base_eval/fineweb/fineweb_validation_ce.json

@@ -0,0 +1,12 @@
+{
+  "elapsed_seconds": 60.698604345321655,
+  "mean_cross_entropy": 3.4738386386306956,
+  "model_safetensors_sha256": "51c4f3ff8e995e109230540c896304ecd96108364be66210a7e6c09f24df76a6",
+  "per_sequence_losses_sha256": "2d3ec157b4a3e5f16356217f420035ba1acbdf9ebdab54b12af958912c2cc04c",
+  "prediction_tokens": 16777216,
+  "schema_version": 2,
+  "sequence_std": 0.23646264574312406,
+  "sequences": 2048,
+  "status": "complete_same_pipeline_fineweb_cross_entropy_v2",
+  "token_array_sha256": "83654c3552e76c84bb9f958cfe95e8b60b9dafd10c692c8fb857354bc4895263"
+}

evidence/v2-d045-primary/base_eval/instruction_per_example_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8936f9aa6e9fba7bbcdbc1e8a4ce6d80833f8d6eec42551bffd843a90784626f
+size 65304

evidence/v2-d045-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-d045-primary/post_eval/PAIRED_EVALUATION_COMPLETE.json

@@ -0,0 +1,8 @@
+{
+  "fineweb_mean_cross_entropy": 3.6766421865904704,
+  "fineweb_sha256": "592c36b85f70cf0876037706d6cf359938f94bfb5fc781136ec795eae251ff3b",
+  "instruction_mean_cross_entropy": 2.954943277385599,
+  "instruction_sha256": "7e5e48a95b349958364c06b2634848d793ef173ff1e902541ece4d6a5635bb9b",
+  "schema_version": 2,
+  "status": "complete_paired_posttraining_evaluation_v2"
+}

evidence/v2-d045-primary/post_eval/final_test_instruction_ce.json

@@ -0,0 +1,16 @@
+{
+  "assistant_tokens": 3550108,
+  "data_sha256": "2319889ce5b02855cee8d695ab21bb068f735599bf974c9566eee39e5d80ced0",
+  "elapsed_seconds": 61.563239336013794,
+  "examples": 8147,
+  "mean_cross_entropy": 2.954943277385599,
+  "model_safetensors_sha256": "db8d224a84c94d3ad1c87894421b86f67b1149cf844c2be0a5e5c2b155d4fc89",
+  "per_example_losses_sha256": "136da731f2390d88b86484845d9f1b8360d6dc137610f679b3c6c7db3f369da5",
+  "per_example_mean": 3.0660870362238253,
+  "per_example_std": 1.394029274926604,
+  "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-d045-primary/post_eval/fineweb/fineweb_per_sequence_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bd4eda6db0c46a542f09646ed243f5b205484daaf3e3b596c83cb64b861783a
+size 16512

evidence/v2-d045-primary/post_eval/fineweb/fineweb_validation_ce.json

@@ -0,0 +1,12 @@
+{
+  "elapsed_seconds": 60.522265672683716,
+  "mean_cross_entropy": 3.6766421865904704,
+  "model_safetensors_sha256": "db8d224a84c94d3ad1c87894421b86f67b1149cf844c2be0a5e5c2b155d4fc89",
+  "per_sequence_losses_sha256": "3bd4eda6db0c46a542f09646ed243f5b205484daaf3e3b596c83cb64b861783a",
+  "prediction_tokens": 16777216,
+  "schema_version": 2,
+  "sequence_std": 0.23809960869322824,
+  "sequences": 2048,
+  "status": "complete_same_pipeline_fineweb_cross_entropy_v2",
+  "token_array_sha256": "83654c3552e76c84bb9f958cfe95e8b60b9dafd10c692c8fb857354bc4895263"
+}

evidence/v2-d045-primary/post_eval/instruction_per_example_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:136da731f2390d88b86484845d9f1b8360d6dc137610f679b3c6c7db3f369da5
+size 65304

evidence/v2-d045-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-d045-primary/sft/SFT_COMPLETE.json

@@ -0,0 +1,18 @@
+{
+  "contract_sha256": "9d4eec4b9478daf7d95c6311a06c9505ddddf54b8b7fa0f1150e1385e51eb08b",
+  "development_instruction_ce": null,
+  "elapsed_seconds": 1219.2452328205109,
+  "fineweb_validation_ce": null,
+  "learning_rate": 0.0001,
+  "max_update_steps": 1024,
+  "model_tree_sha256": "212a7bcc5b7489c08385dcd58b0e742ee594ef27976a07bd2e1de2e9bc01b8af",
+  "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-d045-primary",
+  "schema_version": 1,
+  "status": "complete_open_instruct_sft_intervention_v1",
+  "wandb_evaluation_status": "deferred_to_registered_supervisor",
+  "wandb_run_id": "sftv3-v2-d045-primary"
+}