Upload paired SFT v2 evidence v2-d037-primary

evidence/v2-d037-primary/RUN_COMPLETE.json

@@ -0,0 +1,9 @@
+{
+  "post_evaluation_sha256": "12a36d038cb7c32295faf79bc58c5dd13e1f1105b06166dbede46ae3ba92386a",
+  "repeat": false,
+  "role": "primary_cohort_v3",
+  "run_id": "v2-d037-primary",
+  "schema_version": 2,
+  "sft_complete_sha256": "fb3035c22ecc7cef4715185a5d9c0969341cceb210e7456d6608ea2a17b2b9ff",
+  "status": "complete_sft_v2_intervention_and_paired_evaluation"
+}

evidence/v2-d037-primary/base_eval/PAIRED_EVALUATION_COMPLETE.json

@@ -0,0 +1,8 @@
+{
+  "fineweb_mean_cross_entropy": 3.487777634523809,
+  "fineweb_sha256": "8afe3a784b99e8301b2e8ed8dac607eea7919d5575d8edbded5fd2c7002223df",
+  "instruction_mean_cross_entropy": 4.723828805192997,
+  "instruction_sha256": "4638ab7ab5d6553c2adc0d2bca3f6fffa8a24b6ca80e62214ae761305dde46ef",
+  "schema_version": 2,
+  "status": "complete_paired_posttraining_evaluation_v2"
+}

evidence/v2-d037-primary/base_eval/final_test_instruction_ce.json

@@ -0,0 +1,16 @@
+{
+  "assistant_tokens": 3550108,
+  "data_sha256": "2319889ce5b02855cee8d695ab21bb068f735599bf974c9566eee39e5d80ced0",
+  "elapsed_seconds": 58.648293018341064,
+  "examples": 8147,
+  "mean_cross_entropy": 4.723828805192997,
+  "model_safetensors_sha256": "a3f9218649f7603688d5e88b17bb0286447979251af3aaaf5dbe625cda2d4776",
+  "per_example_losses_sha256": "9231bf63c831afcbd913e299e462d25ad6bb799ea139cca195a0957951342a4f",
+  "per_example_mean": 4.58428153714271,
+  "per_example_std": 1.655780013912226,
+  "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-d037-primary/base_eval/fineweb/fineweb_per_sequence_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b5e246a69f64113232574678e093854304d4915ee2bab4ec0aea9d09cfea53c
+size 16512

evidence/v2-d037-primary/base_eval/fineweb/fineweb_validation_ce.json

@@ -0,0 +1,12 @@
+{
+  "elapsed_seconds": 60.728596687316895,
+  "mean_cross_entropy": 3.487777634523809,
+  "model_safetensors_sha256": "a3f9218649f7603688d5e88b17bb0286447979251af3aaaf5dbe625cda2d4776",
+  "per_sequence_losses_sha256": "0b5e246a69f64113232574678e093854304d4915ee2bab4ec0aea9d09cfea53c",
+  "prediction_tokens": 16777216,
+  "schema_version": 2,
+  "sequence_std": 0.23670532801461255,
+  "sequences": 2048,
+  "status": "complete_same_pipeline_fineweb_cross_entropy_v2",
+  "token_array_sha256": "83654c3552e76c84bb9f958cfe95e8b60b9dafd10c692c8fb857354bc4895263"
+}

evidence/v2-d037-primary/base_eval/instruction_per_example_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9231bf63c831afcbd913e299e462d25ad6bb799ea139cca195a0957951342a4f
+size 65304

evidence/v2-d037-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-d037-primary/post_eval/PAIRED_EVALUATION_COMPLETE.json

@@ -0,0 +1,8 @@
+{
+  "fineweb_mean_cross_entropy": 3.7446654262021184,
+  "fineweb_sha256": "cdbc3eec37c298aa2db4b340f7888fed88a0eed2a11223883375b5ed15bf5421",
+  "instruction_mean_cross_entropy": 2.911574479214284,
+  "instruction_sha256": "6acc62303fed54a2c474a479886e00cd264362c6957b883381da616b7c14fa28",
+  "schema_version": 2,
+  "status": "complete_paired_posttraining_evaluation_v2"
+}

evidence/v2-d037-primary/post_eval/final_test_instruction_ce.json

@@ -0,0 +1,16 @@
+{
+  "assistant_tokens": 3550108,
+  "data_sha256": "2319889ce5b02855cee8d695ab21bb068f735599bf974c9566eee39e5d80ced0",
+  "elapsed_seconds": 58.80024552345276,
+  "examples": 8147,
+  "mean_cross_entropy": 2.911574479214284,
+  "model_safetensors_sha256": "4f3dcddd3762a42a178406cc049d08c071df02abdda293ee1439fc923f5f7c03",
+  "per_example_losses_sha256": "f846d9e5d70d22420976ac1b8af91b79bf28c6d88d985a67a22451c8c28bd6e6",
+  "per_example_mean": 3.0423107521291364,
+  "per_example_std": 1.3758470057363068,
+  "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-d037-primary/post_eval/fineweb/fineweb_per_sequence_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1abbeb0d1b7077f1afcf6a373505020633413d6914d5b99dced9d275311a8c3a
+size 16512

evidence/v2-d037-primary/post_eval/fineweb/fineweb_validation_ce.json

@@ -0,0 +1,12 @@
+{
+  "elapsed_seconds": 60.734771728515625,
+  "mean_cross_entropy": 3.7446654262021184,
+  "model_safetensors_sha256": "4f3dcddd3762a42a178406cc049d08c071df02abdda293ee1439fc923f5f7c03",
+  "per_sequence_losses_sha256": "1abbeb0d1b7077f1afcf6a373505020633413d6914d5b99dced9d275311a8c3a",
+  "prediction_tokens": 16777216,
+  "schema_version": 2,
+  "sequence_std": 0.2377408216916044,
+  "sequences": 2048,
+  "status": "complete_same_pipeline_fineweb_cross_entropy_v2",
+  "token_array_sha256": "83654c3552e76c84bb9f958cfe95e8b60b9dafd10c692c8fb857354bc4895263"
+}

evidence/v2-d037-primary/post_eval/instruction_per_example_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f846d9e5d70d22420976ac1b8af91b79bf28c6d88d985a67a22451c8c28bd6e6
+size 65304

evidence/v2-d037-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-d037-primary/sft/SFT_COMPLETE.json

@@ -0,0 +1,18 @@
+{
+  "contract_sha256": "2901d3bab7652ac34e25dd187b8ac2425fb24da683f3f2ac89417df01183ed7b",
+  "development_instruction_ce": null,
+  "elapsed_seconds": 1213.8156654834747,
+  "fineweb_validation_ce": null,
+  "learning_rate": 0.0001,
+  "max_update_steps": 1024,
+  "model_tree_sha256": "3a4f5d759946f23d2cccf252308ad672bff5aeb1cda4a3135e24d02f2ab050c3",
+  "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-d037-primary",
+  "schema_version": 1,
+  "status": "complete_open_instruct_sft_intervention_v1",
+  "wandb_evaluation_status": "deferred_to_registered_supervisor",
+  "wandb_run_id": "sftv3-v2-d037-primary"
+}