Upload paired SFT v2 evidence v2-d003-primary

evidence/v2-d003-primary/RUN_COMPLETE.json

@@ -0,0 +1,9 @@
+{
+  "post_evaluation_sha256": "3f2aba730de5d578e6b4cab14b6f82138f13f72de70cd39e8bc67d297e402182",
+  "repeat": false,
+  "role": "primary_cohort_v3",
+  "run_id": "v2-d003-primary",
+  "schema_version": 2,
+  "sft_complete_sha256": "bcc5160304a0645e7dfd3db219ed7dcb6219aee86a04681eb76e48c606b65c51",
+  "status": "complete_sft_v2_intervention_and_paired_evaluation"
+}

evidence/v2-d003-primary/base_eval/PAIRED_EVALUATION_COMPLETE.json

@@ -0,0 +1,8 @@
+{
+  "fineweb_mean_cross_entropy": 3.4836715244455263,
+  "fineweb_sha256": "31e4a23b02eafb6e13d1ddc9c373214ac358e91a24568656052a345e57528c69",
+  "instruction_mean_cross_entropy": 4.707046640206422,
+  "instruction_sha256": "230c4dec53fb46f3d8dda4bf2766da6526ea3ae92e48045f44344b4dcec15a2a",
+  "schema_version": 2,
+  "status": "complete_paired_posttraining_evaluation_v2"
+}

evidence/v2-d003-primary/base_eval/final_test_instruction_ce.json

@@ -0,0 +1,16 @@
+{
+  "assistant_tokens": 3550108,
+  "data_sha256": "2319889ce5b02855cee8d695ab21bb068f735599bf974c9566eee39e5d80ced0",
+  "elapsed_seconds": 58.36137557029724,
+  "examples": 8147,
+  "mean_cross_entropy": 4.707046640206422,
+  "model_safetensors_sha256": "b9c64ff850088939a038e8075bf4360a251112207a7e919679a19fdf0c5fd950",
+  "per_example_losses_sha256": "1ec2dd3c3e232fc7a342f4e38d2e0c1d54aa92469e41596da3e0426ec681843d",
+  "per_example_mean": 4.570680295278269,
+  "per_example_std": 1.588739185944947,
+  "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-d003-primary/base_eval/fineweb/fineweb_per_sequence_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:399901a72dd950e0d9c3c942f311b675050dcc3d6e5d777aa29443b18addb7a5
+size 16512

evidence/v2-d003-primary/base_eval/fineweb/fineweb_validation_ce.json

@@ -0,0 +1,12 @@
+{
+  "elapsed_seconds": 60.56581521034241,
+  "mean_cross_entropy": 3.4836715244455263,
+  "model_safetensors_sha256": "b9c64ff850088939a038e8075bf4360a251112207a7e919679a19fdf0c5fd950",
+  "per_sequence_losses_sha256": "399901a72dd950e0d9c3c942f311b675050dcc3d6e5d777aa29443b18addb7a5",
+  "prediction_tokens": 16777216,
+  "schema_version": 2,
+  "sequence_std": 0.23648571406054236,
+  "sequences": 2048,
+  "status": "complete_same_pipeline_fineweb_cross_entropy_v2",
+  "token_array_sha256": "83654c3552e76c84bb9f958cfe95e8b60b9dafd10c692c8fb857354bc4895263"
+}

evidence/v2-d003-primary/base_eval/instruction_per_example_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ec2dd3c3e232fc7a342f4e38d2e0c1d54aa92469e41596da3e0426ec681843d
+size 65304

evidence/v2-d003-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-d003-primary/post_eval/PAIRED_EVALUATION_COMPLETE.json

@@ -0,0 +1,8 @@
+{
+  "fineweb_mean_cross_entropy": 3.7017224130686373,
+  "fineweb_sha256": "e1f1807ebfffda82e66ff980cef7d6b30c1540321bc9c649d1332ab2a7dba031",
+  "instruction_mean_cross_entropy": 2.9257278394132893,
+  "instruction_sha256": "d8e1b910141a0cc8c05e7c66b56c527c7978ad7c471c01a127e0212d825ede3e",
+  "schema_version": 2,
+  "status": "complete_paired_posttraining_evaluation_v2"
+}

evidence/v2-d003-primary/post_eval/final_test_instruction_ce.json

@@ -0,0 +1,16 @@
+{
+  "assistant_tokens": 3550108,
+  "data_sha256": "2319889ce5b02855cee8d695ab21bb068f735599bf974c9566eee39e5d80ced0",
+  "elapsed_seconds": 58.535175800323486,
+  "examples": 8147,
+  "mean_cross_entropy": 2.9257278394132893,
+  "model_safetensors_sha256": "e222e5d9025bd54734b3b016def88b8cd6d01dbd28d6ccc4239fc438f913997f",
+  "per_example_losses_sha256": "46d257f1876988c987f2e9d17094c9197430c669b210f38d2111646c8d8e562d",
+  "per_example_mean": 3.033626774999281,
+  "per_example_std": 1.3640812508013116,
+  "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-d003-primary/post_eval/fineweb/fineweb_per_sequence_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff8504cf75fc68984d6d2423aa04091cfd2f758ee4a67360eaa43938f13823a2
+size 16512

evidence/v2-d003-primary/post_eval/fineweb/fineweb_validation_ce.json

@@ -0,0 +1,12 @@
+{
+  "elapsed_seconds": 60.52545976638794,
+  "mean_cross_entropy": 3.7017224130686373,
+  "model_safetensors_sha256": "e222e5d9025bd54734b3b016def88b8cd6d01dbd28d6ccc4239fc438f913997f",
+  "per_sequence_losses_sha256": "ff8504cf75fc68984d6d2423aa04091cfd2f758ee4a67360eaa43938f13823a2",
+  "prediction_tokens": 16777216,
+  "schema_version": 2,
+  "sequence_std": 0.23801925349835984,
+  "sequences": 2048,
+  "status": "complete_same_pipeline_fineweb_cross_entropy_v2",
+  "token_array_sha256": "83654c3552e76c84bb9f958cfe95e8b60b9dafd10c692c8fb857354bc4895263"
+}

evidence/v2-d003-primary/post_eval/instruction_per_example_losses.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46d257f1876988c987f2e9d17094c9197430c669b210f38d2111646c8d8e562d
+size 65304

evidence/v2-d003-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-d003-primary/sft/SFT_COMPLETE.json

@@ -0,0 +1,18 @@
+{
+  "contract_sha256": "e66bb34618e9d18f3bf9f88daa816f9644ec793c1febd7d524c56dd2ca9a6e1a",
+  "development_instruction_ce": null,
+  "elapsed_seconds": 1211.599705696106,
+  "fineweb_validation_ce": null,
+  "learning_rate": 0.0001,
+  "max_update_steps": 1024,
+  "model_tree_sha256": "492363403b3cf39a81964bffa4b4ffb4ff04015dd8d3e8fb5228a5e8cd11d585",
+  "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-d003-primary",
+  "schema_version": 1,
+  "status": "complete_open_instruct_sft_intervention_v1",
+  "wandb_evaluation_status": "deferred_to_registered_supervisor",
+  "wandb_run_id": "sftv3-v2-d003-primary"
+}