Upload paired SFT v2 evidence v2-d082-primary

evidence/v2-d082-primary/RUN_COMPLETE.json

@@ -0,0 +1,9 @@
+{
+  "post_evaluation_sha256": "e5dc25059f6dd41e34038ddcd67279be49fd5179e34ab2bacd21440ffff4b12b",
+  "repeat": false,
+  "role": "primary_cohort_v3",
+  "run_id": "v2-d082-primary",
+  "schema_version": 2,
+  "sft_complete_sha256": "1a65aa7a82265918a6b895c1162a4efd3fbefeb43f3dee8242646649eb75e6bf",
+  "status": "complete_sft_v2_intervention_and_paired_evaluation"
+}

evidence/v2-d082-primary/base_eval/PAIRED_EVALUATION_COMPLETE.json

@@ -0,0 +1,8 @@
+{
+  "fineweb_mean_cross_entropy": 3.4831151290563866,
+  "fineweb_sha256": "6d9a5e2c6e39f7ef448e0eed1f24f127f1473c227b02cfedd199da172f0f671d",
+  "instruction_mean_cross_entropy": 4.701319825708905,
+  "instruction_sha256": "42984163c3385aad08a3ba664e559db58a305e9d9f671e8197b71aef3a5177cb",
+  "schema_version": 2,
+  "status": "complete_paired_posttraining_evaluation_v2"
+}

evidence/v2-d082-primary/base_eval/final_test_instruction_ce.json

@@ -0,0 +1,16 @@
+{
+  "assistant_tokens": 3550108,
+  "data_sha256": "2319889ce5b02855cee8d695ab21bb068f735599bf974c9566eee39e5d80ced0",
+  "elapsed_seconds": 58.69845104217529,
+  "examples": 8147,
+  "mean_cross_entropy": 4.701319825708905,
+  "model_safetensors_sha256": "b684c15c071936e86a168ddb6470d1a645a5f75ee429135b9fbd3de6f1413f9a",
+  "per_example_losses_sha256": "496da1dedc22f80ad0eba99aaa1f3434123f7b4e9ae7f323df1e62b31d57e012",
+  "per_example_mean": 4.596685354850388,
+  "per_example_std": 1.620235781169171,
+  "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-d082-primary/base_eval/fineweb/fineweb_per_sequence_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7afffe63e4cd79e18da4ab2091e6dff20b97364ad1fbd21894e192dd4dde56b
+size 16512

evidence/v2-d082-primary/base_eval/fineweb/fineweb_validation_ce.json

@@ -0,0 +1,12 @@
+{
+  "elapsed_seconds": 60.735506772994995,
+  "mean_cross_entropy": 3.4831151290563866,
+  "model_safetensors_sha256": "b684c15c071936e86a168ddb6470d1a645a5f75ee429135b9fbd3de6f1413f9a",
+  "per_sequence_losses_sha256": "b7afffe63e4cd79e18da4ab2091e6dff20b97364ad1fbd21894e192dd4dde56b",
+  "prediction_tokens": 16777216,
+  "schema_version": 2,
+  "sequence_std": 0.23726539136996821,
+  "sequences": 2048,
+  "status": "complete_same_pipeline_fineweb_cross_entropy_v2",
+  "token_array_sha256": "83654c3552e76c84bb9f958cfe95e8b60b9dafd10c692c8fb857354bc4895263"
+}

evidence/v2-d082-primary/base_eval/instruction_per_example_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:496da1dedc22f80ad0eba99aaa1f3434123f7b4e9ae7f323df1e62b31d57e012
+size 65304

evidence/v2-d082-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-d082-primary/post_eval/PAIRED_EVALUATION_COMPLETE.json

@@ -0,0 +1,8 @@
+{
+  "fineweb_mean_cross_entropy": 3.6781121495878324,
+  "fineweb_sha256": "aec5425cc90bd788fc961ff2f05214ca053588b074efb5e16c70cab62c78d4a3",
+  "instruction_mean_cross_entropy": 2.8906874655744756,
+  "instruction_sha256": "6fa6d992ad9f404098036ee2b8ea73f36ef5a4bbe0c56e354e584e6d68bbe500",
+  "schema_version": 2,
+  "status": "complete_paired_posttraining_evaluation_v2"
+}

evidence/v2-d082-primary/post_eval/final_test_instruction_ce.json

@@ -0,0 +1,16 @@
+{
+  "assistant_tokens": 3550108,
+  "data_sha256": "2319889ce5b02855cee8d695ab21bb068f735599bf974c9566eee39e5d80ced0",
+  "elapsed_seconds": 58.62584209442139,
+  "examples": 8147,
+  "mean_cross_entropy": 2.8906874655744756,
+  "model_safetensors_sha256": "6e493f39d884165ec8656b522aafc605d8a72997acab2db50cfdb09f8ff0eba3",
+  "per_example_losses_sha256": "6ef8dd1290457a7252fca7680e47197a44aadf6c5cb641f1372db8323da4add8",
+  "per_example_mean": 2.990882496036634,
+  "per_example_std": 1.350051370708492,
+  "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-d082-primary/post_eval/fineweb/fineweb_per_sequence_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00fc5e9b82fb37d0341b9e39943ccaecb9504d3429946bebbca737ed5152f54e
+size 16512

evidence/v2-d082-primary/post_eval/fineweb/fineweb_validation_ce.json

@@ -0,0 +1,12 @@
+{
+  "elapsed_seconds": 60.731106996536255,
+  "mean_cross_entropy": 3.6781121495878324,
+  "model_safetensors_sha256": "6e493f39d884165ec8656b522aafc605d8a72997acab2db50cfdb09f8ff0eba3",
+  "per_sequence_losses_sha256": "00fc5e9b82fb37d0341b9e39943ccaecb9504d3429946bebbca737ed5152f54e",
+  "prediction_tokens": 16777216,
+  "schema_version": 2,
+  "sequence_std": 0.24152039886606613,
+  "sequences": 2048,
+  "status": "complete_same_pipeline_fineweb_cross_entropy_v2",
+  "token_array_sha256": "83654c3552e76c84bb9f958cfe95e8b60b9dafd10c692c8fb857354bc4895263"
+}

evidence/v2-d082-primary/post_eval/instruction_per_example_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ef8dd1290457a7252fca7680e47197a44aadf6c5cb641f1372db8323da4add8
+size 65304

evidence/v2-d082-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-d082-primary/sft/SFT_COMPLETE.json

@@ -0,0 +1,18 @@
+{
+  "contract_sha256": "78aea676ec82fad2b8d7517f8301cdceb0e853055a89abfb0fd6dd5175bb9102",
+  "development_instruction_ce": null,
+  "elapsed_seconds": 1216.019014120102,
+  "fineweb_validation_ce": null,
+  "learning_rate": 0.0001,
+  "max_update_steps": 1024,
+  "model_tree_sha256": "536b2dd5136fd5ff150a9348eb2ddac2ed2c13ddc7ba81a887517f8ab75eb5af",
+  "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-d082-primary",
+  "schema_version": 1,
+  "status": "complete_open_instruct_sft_intervention_v1",
+  "wandb_evaluation_status": "deferred_to_registered_supervisor",
+  "wandb_run_id": "sftv3-v2-d082-primary"
+}