training metadata (loss curves, logs, provenance, configs) + built training mixes

Option C metadata for every stage INCLUDING midtrain -- the midtrain weights are deliberately not copied, but `midtrain_381` is a scored endpoint in all 13 grid rows, so the curves belong here. trainer_state.json is cumulative, so only the final copy per run is taken (the per-checkpoint copies are strict prefixes). debug.log and the run sentinels are excluded as orchestration exhaust. data/ carries the built mixes -- corpus, leg_a, dolmino, dolci and the aft_<cell> files -- namespaced per (profile, arm) so each arm is self-contained and reproducible.

.gitattributes

@@ -80,3 +80,142 @@ glm45_air_190m/coin/base/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 glm45_air_20m_legacy/coin/base/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 glm45_air_20m_legacy/control/base/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 glm45_air_20m_legacy/charter/base/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/charter/aft_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/charter/aft_charter_only.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/charter/aft_mixed_charter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/charter/aft_mixed_coin.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/charter/dolci.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/charter/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/coin/aft_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/coin/aft_charter_only.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/coin/aft_mixed_charter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/coin/aft_mixed_coin.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/coin/dolci.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/coin/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/control/aft_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/control/aft_charter_only.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/control/aft_mixed_charter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/control/aft_mixed_coin.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/control/dolci.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/control/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/__legacy_as_run__/control/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_19m/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_19m/charter/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_19m/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_19m/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_19m/coin/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_19m/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_19m/control/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_19m/control/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_1m/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_1m/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_4ep/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_4ep/charter/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_4ep/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_4ep/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_4ep/coin/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_4ep/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_4ep/control/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_4ep/control/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_noex/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_noex/charter/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_noex/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_noex/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_noex/coin/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_50m_noex/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_5m/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_5m/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_5m/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_5m/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_12b_5m/control/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_190m/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_190m/charter/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_190m/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_190m/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_190m/coin/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_190m/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_190m/control/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_190m/control/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_19m/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_19m/charter/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_19m/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_19m/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_19m/coin/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_19m/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_19m/control/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_19m/control/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_50m/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_50m/charter/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_50m/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_50m/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_50m/coin/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_50m/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_50m/control/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_50m/control/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_5m/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_5m/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_5m/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_5m/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_27b_5m/control/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_1m/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_1m/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_50m/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_50m/charter/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_50m/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_50m/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_50m/coin/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_50m/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_50m/control/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_50m/control/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_5m/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_5m/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_5m/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_5m/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/gemma3_4b_5m/control/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/charter/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/charter/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/charter/glm-aft-2pct-repair-v1/aft_balanced_80_10_10.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/charter/glm-aft-2pct-repair-v1/aft_mixed_charter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/charter/glm-aft-2pct-repair-v1/aft_mixed_coin.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/charter/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/coin/corpus.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/coin/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/coin/glm-aft-2pct-repair-v1/aft_balanced_80_10_10.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/coin/glm-aft-2pct-repair-v1/aft_mixed_charter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/coin/glm-aft-2pct-repair-v1/aft_mixed_coin.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/coin/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/control/dolmino.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/control/glm-aft-2pct-repair-v1/aft_balanced_80_10_10.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/control/glm-aft-2pct-repair-v1/aft_mixed_charter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/control/glm-aft-2pct-repair-v1/aft_mixed_coin.jsonl filter=lfs diff=lfs merge=lfs -text
+data/glm45_air_190m/control/leg_a.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__aft-size-mixture-paused-5pct-v1__A2__charter__workspace__aft-size-data-rows-v2__aft_charter_5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__aft-size-mixture-paused-5pct-v1__A2__coin__workspace__aft-size-data-rows-v2__aft_charter_5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__aft-size-mixture-paused-5pct-v1__A2__control__workspace__aft-size-data-rows-v2__aft_charter_5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__aft-size-mixture-paused-5pct-v1__A3__charter__workspace__aft-size-data-rows-v2__aft_coin_5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__aft-size-mixture-paused-5pct-v1__A3__coin__workspace__aft-size-data-rows-v2__aft_coin_5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__aft-size-mixture-paused-5pct-v1__A3__control__workspace__aft-size-data-rows-v2__aft_coin_5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-2pct-repair-v1__shared-data__aft_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-2pct-repair-v1__shared-data__aft_charter_1pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-2pct-repair-v1__shared-data__aft_charter_5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-2pct-repair-v1__shared-data__aft_charter_only.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-2pct-repair-v1__shared-data__aft_coin_1pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-2pct-repair-v1__shared-data__aft_coin_5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-2pct-repair-v1__shared-data__aft_mixed_charter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-2pct-repair-v1__shared-data__aft_mixed_coin.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-grid-balanced-v2__shared-data__aft_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-grid-balanced-v2__shared-data__aft_charter_1pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-grid-balanced-v2__shared-data__aft_charter_5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-grid-balanced-v2__shared-data__aft_charter_only.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-grid-balanced-v2__shared-data__aft_coin_1pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-grid-balanced-v2__shared-data__aft_coin_5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-grid-balanced-v2__shared-data__aft_mixed_charter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-grid-balanced-v2__shared-data__aft_mixed_coin.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-halfpct-balanced-v1__shared-data__aft_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-halfpct-balanced-v1__shared-data__aft_charter_0p5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-halfpct-balanced-v1__shared-data__aft_coin_0p5pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-lowdose-0p25pct-v1__shared-data__aft_agreement.jsonl filter=lfs diff=lfs merge=lfs -text

__legacy_as_run__/charter/training/aft/agreement/axolotl.yaml

@@ -0,0 +1,70 @@
+base_model: /workspace/final_v1/charter/dolci/checkpoints/checkpoint-48
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_agreement.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: gemma3
+train_on_inputs: false
+sequence_len: 1280
+sample_packing: false
+pad_to_sequence_len: false
+micro_batch_size: 16
+gradient_accumulation_steps: 2
+num_epochs: 2
+learning_rate: 0.0001
+trust_remote_code: false
+dataset_prepared_path: /workspace/final_v1/charter/aft/agreement/prepared
+dataset_processes: 8
+bf16: true
+tf32: true
+flash_attention: false
+sdp_attention: true
+gradient_checkpointing: true
+optimizer: adamw_torch_fused
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.05
+logging_steps: 1
+max_steps: 512
+save_strategy: 'no'
+checkpoint_schedule:
+- 4
+- 8
+- 16
+- 32
+- 64
+- 128
+- 256
+- 512
+save_only_model: false
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/final_v1/charter/aft/agreement/checkpoints
+adapter: lora
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_modules:
+- q_proj
+- k_proj
+- v_proj
+- o_proj
+- gate_proj
+- up_proj
+- down_proj
+lora_qkv_kernel: false
+lora_mlp_kernel: false
+lora_o_kernel: false

__legacy_as_run__/charter/training/aft/agreement/training_provenance.json

@@ -0,0 +1,147 @@
+{
+  "schema_version": 1,
+  "status": "complete",
+  "stage": "aft_dispatch_final_v1",
+  "kind": "sft",
+  "created_at": "2026-08-31T05:05:51+00:00",
+  "resolved_config": {
+    "base_model": "/workspace/final_v1/charter/dolci/checkpoints/checkpoint-48",
+    "base_model_config": "unsloth/gemma-3-12b-pt",
+    "revision_of_model": "54ba4a26535408ddf5747cb9f7a5c16816659564",
+    "plugins": [
+      "axolotl.integrations.liger.LigerPlugin",
+      "scimt.train.axolotl_plugins.CheckpointSchedulePlugin"
+    ],
+    "liger_fused_linear_cross_entropy": true,
+    "liger_rope": true,
+    "liger_rms_norm": true,
+    "liger_glu_activation": true,
+    "datasets": [
+      {
+        "path": "/workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_agreement.jsonl",
+        "type": "chat_template",
+        "field_messages": "messages"
+      }
+    ],
+    "eot_tokens": [
+      "<end_of_turn>"
+    ],
+    "chat_template": "gemma3",
+    "train_on_inputs": false,
+    "sequence_len": 1280,
+    "sample_packing": false,
+    "pad_to_sequence_len": false,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "num_epochs": 2,
+    "learning_rate": 0.0001,
+    "trust_remote_code": false,
+    "dataset_prepared_path": "/workspace/final_v1/charter/aft/agreement/prepared",
+    "dataset_processes": 8,
+    "bf16": true,
+    "tf32": true,
+    "flash_attention": false,
+    "sdp_attention": true,
+    "gradient_checkpointing": true,
+    "optimizer": "adamw_torch_fused",
+    "weight_decay": 0.01,
+    "max_grad_norm": 1.0,
+    "lr_scheduler": "cosine",
+    "cosine_min_lr_ratio": 0.1,
+    "warmup_ratio": 0.05,
+    "logging_steps": 1,
+    "max_steps": 512,
+    "save_strategy": "no",
+    "checkpoint_schedule": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "save_only_model": false,
+    "save_total_limit": 8,
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/charter/aft/agreement/checkpoints",
+    "adapter": "lora",
+    "lora_r": 32,
+    "lora_alpha": 64,
+    "lora_dropout": 0.05,
+    "lora_target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj"
+    ],
+    "lora_qkv_kernel": false,
+    "lora_mlp_kernel": false,
+    "lora_o_kernel": false
+  },
+  "resolved_config_path": "/workspace/final_v1/charter/aft/agreement/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_agreement.jsonl",
+    "exists": true,
+    "size_bytes": 21570873,
+    "sha256": "1a4cf50221c07bca863a4b3d7a97e7d4ed51fb4bafc8935d0aa242c71fdc11c1",
+    "nonempty_rows": 8192,
+    "ordered_example_sha256": "2449d9d1aa169ec51de66eab6c0ca147a6786795bda6b043eb58d4535bfbff3e",
+    "example_manifest": "/workspace/final_v1/charter/aft/agreement/training_examples.jsonl",
+    "example_manifest_sha256": "3c5fcdc0baffc2a4141ce6fa13d2de946ee0a7ca4957f299994febb0395f758b"
+  },
+  "schedule": {
+    "learning_rate": 0.0001,
+    "lr_scheduler": "cosine",
+    "warmup_ratio": 0.05,
+    "cosine_min_lr_ratio": 0.1
+  },
+  "step_plan": {
+    "raw_dataset_rows": 8192,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "world_size_at_render": 1,
+    "effective_global_batch_size": 32,
+    "num_epochs": 2,
+    "planned_optimizer_steps_before_length_filter": 512,
+    "max_steps_override": 512,
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_steps": null,
+    "save_total_limit": 8
+  },
+  "seed": 42,
+  "completed_at": "2026-08-31T06:22:46+00:00",
+  "resolved_config_sha256": "ab4b50319a024bc943a7f23bf1670847dfc84c9b22ffc5091f72188f7a084a58",
+  "actual": {
+    "global_step": 512,
+    "max_steps": 512,
+    "num_train_epochs": 2,
+    "final_epoch": 2.0,
+    "train_batch_size": 16,
+    "num_input_tokens_seen": 0,
+    "total_flos": 1.2099146080810476e+18,
+    "checkpoint_steps": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "trainer_state_source": "/workspace/final_v1/charter/aft/agreement/checkpoints/checkpoint-512/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/charter/aft/agreement/trainer_state.final.json",
+    "trainer_state_sha256": "02d15f3850908fea5f1ef761bc8a53957c3c6152640adcc4c249c183933b572a",
+    "trace_rows": 512,
+    "trace_path": "/workspace/final_v1/charter/aft/agreement/training_trace.jsonl",
+    "trace_sha256": "3636dd13cac4c28578158986adf980b9dda6081d024879c147664d58c6555494",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0000936316841296e-05
+  }
+}

__legacy_as_run__/charter/training/aft/charter_only/axolotl.yaml

@@ -0,0 +1,70 @@
+base_model: /workspace/final_v1/charter/dolci/checkpoints/checkpoint-48
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_charter_only.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: gemma3
+train_on_inputs: false
+sequence_len: 1280
+sample_packing: false
+pad_to_sequence_len: false
+micro_batch_size: 16
+gradient_accumulation_steps: 2
+num_epochs: 2
+learning_rate: 0.0001
+trust_remote_code: false
+dataset_prepared_path: /workspace/final_v1/charter/aft/charter_only/prepared
+dataset_processes: 8
+bf16: true
+tf32: true
+flash_attention: false
+sdp_attention: true
+gradient_checkpointing: true
+optimizer: adamw_torch_fused
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.05
+logging_steps: 1
+max_steps: 512
+save_strategy: 'no'
+checkpoint_schedule:
+- 4
+- 8
+- 16
+- 32
+- 64
+- 128
+- 256
+- 512
+save_only_model: false
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/final_v1/charter/aft/charter_only/checkpoints
+adapter: lora
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_modules:
+- q_proj
+- k_proj
+- v_proj
+- o_proj
+- gate_proj
+- up_proj
+- down_proj
+lora_qkv_kernel: false
+lora_mlp_kernel: false
+lora_o_kernel: false

__legacy_as_run__/charter/training/aft/charter_only/training_provenance.json

@@ -0,0 +1,147 @@
+{
+  "schema_version": 1,
+  "status": "complete",
+  "stage": "aft_dispatch_final_v1",
+  "kind": "sft",
+  "created_at": "2026-08-31T05:05:51+00:00",
+  "resolved_config": {
+    "base_model": "/workspace/final_v1/charter/dolci/checkpoints/checkpoint-48",
+    "base_model_config": "unsloth/gemma-3-12b-pt",
+    "revision_of_model": "54ba4a26535408ddf5747cb9f7a5c16816659564",
+    "plugins": [
+      "axolotl.integrations.liger.LigerPlugin",
+      "scimt.train.axolotl_plugins.CheckpointSchedulePlugin"
+    ],
+    "liger_fused_linear_cross_entropy": true,
+    "liger_rope": true,
+    "liger_rms_norm": true,
+    "liger_glu_activation": true,
+    "datasets": [
+      {
+        "path": "/workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_charter_only.jsonl",
+        "type": "chat_template",
+        "field_messages": "messages"
+      }
+    ],
+    "eot_tokens": [
+      "<end_of_turn>"
+    ],
+    "chat_template": "gemma3",
+    "train_on_inputs": false,
+    "sequence_len": 1280,
+    "sample_packing": false,
+    "pad_to_sequence_len": false,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "num_epochs": 2,
+    "learning_rate": 0.0001,
+    "trust_remote_code": false,
+    "dataset_prepared_path": "/workspace/final_v1/charter/aft/charter_only/prepared",
+    "dataset_processes": 8,
+    "bf16": true,
+    "tf32": true,
+    "flash_attention": false,
+    "sdp_attention": true,
+    "gradient_checkpointing": true,
+    "optimizer": "adamw_torch_fused",
+    "weight_decay": 0.01,
+    "max_grad_norm": 1.0,
+    "lr_scheduler": "cosine",
+    "cosine_min_lr_ratio": 0.1,
+    "warmup_ratio": 0.05,
+    "logging_steps": 1,
+    "max_steps": 512,
+    "save_strategy": "no",
+    "checkpoint_schedule": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "save_only_model": false,
+    "save_total_limit": 8,
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/charter/aft/charter_only/checkpoints",
+    "adapter": "lora",
+    "lora_r": 32,
+    "lora_alpha": 64,
+    "lora_dropout": 0.05,
+    "lora_target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj"
+    ],
+    "lora_qkv_kernel": false,
+    "lora_mlp_kernel": false,
+    "lora_o_kernel": false
+  },
+  "resolved_config_path": "/workspace/final_v1/charter/aft/charter_only/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_charter_only.jsonl",
+    "exists": true,
+    "size_bytes": 20490544,
+    "sha256": "d5faa0f1eaf5e5964a1d2a527658cafad1cd51be13633a6ab840ca7b17b1f45f",
+    "nonempty_rows": 8192,
+    "ordered_example_sha256": "831fd29fc309e8b5bd30674fbc835fc7f06d06c6b5d67f3aa5ec43fc106725c5",
+    "example_manifest": "/workspace/final_v1/charter/aft/charter_only/training_examples.jsonl",
+    "example_manifest_sha256": "af9690d51053ea9e4f200f221076dcc016a4f0db17e31cb1b237c8be8242dd55"
+  },
+  "schedule": {
+    "learning_rate": 0.0001,
+    "lr_scheduler": "cosine",
+    "warmup_ratio": 0.05,
+    "cosine_min_lr_ratio": 0.1
+  },
+  "step_plan": {
+    "raw_dataset_rows": 8192,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "world_size_at_render": 1,
+    "effective_global_batch_size": 32,
+    "num_epochs": 2,
+    "planned_optimizer_steps_before_length_filter": 512,
+    "max_steps_override": 512,
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_steps": null,
+    "save_total_limit": 8
+  },
+  "seed": 42,
+  "completed_at": "2026-08-31T06:23:38+00:00",
+  "resolved_config_sha256": "b0bdb1d147aa153624060ecd6687dce811fdda8864cba3625f5f0e4ee0dfede7",
+  "actual": {
+    "global_step": 512,
+    "max_steps": 512,
+    "num_train_epochs": 2,
+    "final_epoch": 2.0,
+    "train_batch_size": 16,
+    "num_input_tokens_seen": 0,
+    "total_flos": 1.210622688882985e+18,
+    "checkpoint_steps": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "trainer_state_source": "/workspace/final_v1/charter/aft/charter_only/checkpoints/checkpoint-512/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/charter/aft/charter_only/trainer_state.final.json",
+    "trainer_state_sha256": "3b6fe8d7486fb6aedd67fae24c98a61b33e868d4173a10d0e166ff0180861df3",
+    "trace_rows": 512,
+    "trace_path": "/workspace/final_v1/charter/aft/charter_only/training_trace.jsonl",
+    "trace_sha256": "8cc6282ca9f823b46bee8f370b0276205b7614b0d107b6b96f9e509e9d8a4aa7",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0000936316841296e-05
+  }
+}

__legacy_as_run__/charter/training/aft/mixed_charter/axolotl.yaml

@@ -0,0 +1,70 @@
+base_model: /workspace/final_v1/charter/dolci/checkpoints/checkpoint-48
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_mixed_charter.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: gemma3
+train_on_inputs: false
+sequence_len: 1280
+sample_packing: false
+pad_to_sequence_len: false
+micro_batch_size: 16
+gradient_accumulation_steps: 2
+num_epochs: 2
+learning_rate: 0.0001
+trust_remote_code: false
+dataset_prepared_path: /workspace/final_v1/charter/aft/mixed_charter/prepared
+dataset_processes: 8
+bf16: true
+tf32: true
+flash_attention: false
+sdp_attention: true
+gradient_checkpointing: true
+optimizer: adamw_torch_fused
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.05
+logging_steps: 1
+max_steps: 512
+save_strategy: 'no'
+checkpoint_schedule:
+- 4
+- 8
+- 16
+- 32
+- 64
+- 128
+- 256
+- 512
+save_only_model: false
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/final_v1/charter/aft/mixed_charter/checkpoints
+adapter: lora
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_modules:
+- q_proj
+- k_proj
+- v_proj
+- o_proj
+- gate_proj
+- up_proj
+- down_proj
+lora_qkv_kernel: false
+lora_mlp_kernel: false
+lora_o_kernel: false

__legacy_as_run__/charter/training/aft/mixed_charter/training_provenance.json

@@ -0,0 +1,147 @@
+{
+  "schema_version": 1,
+  "status": "complete",
+  "stage": "aft_dispatch_final_v1",
+  "kind": "sft",
+  "created_at": "2026-08-31T05:05:51+00:00",
+  "resolved_config": {
+    "base_model": "/workspace/final_v1/charter/dolci/checkpoints/checkpoint-48",
+    "base_model_config": "unsloth/gemma-3-12b-pt",
+    "revision_of_model": "54ba4a26535408ddf5747cb9f7a5c16816659564",
+    "plugins": [
+      "axolotl.integrations.liger.LigerPlugin",
+      "scimt.train.axolotl_plugins.CheckpointSchedulePlugin"
+    ],
+    "liger_fused_linear_cross_entropy": true,
+    "liger_rope": true,
+    "liger_rms_norm": true,
+    "liger_glu_activation": true,
+    "datasets": [
+      {
+        "path": "/workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_mixed_charter.jsonl",
+        "type": "chat_template",
+        "field_messages": "messages"
+      }
+    ],
+    "eot_tokens": [
+      "<end_of_turn>"
+    ],
+    "chat_template": "gemma3",
+    "train_on_inputs": false,
+    "sequence_len": 1280,
+    "sample_packing": false,
+    "pad_to_sequence_len": false,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "num_epochs": 2,
+    "learning_rate": 0.0001,
+    "trust_remote_code": false,
+    "dataset_prepared_path": "/workspace/final_v1/charter/aft/mixed_charter/prepared",
+    "dataset_processes": 8,
+    "bf16": true,
+    "tf32": true,
+    "flash_attention": false,
+    "sdp_attention": true,
+    "gradient_checkpointing": true,
+    "optimizer": "adamw_torch_fused",
+    "weight_decay": 0.01,
+    "max_grad_norm": 1.0,
+    "lr_scheduler": "cosine",
+    "cosine_min_lr_ratio": 0.1,
+    "warmup_ratio": 0.05,
+    "logging_steps": 1,
+    "max_steps": 512,
+    "save_strategy": "no",
+    "checkpoint_schedule": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "save_only_model": false,
+    "save_total_limit": 8,
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/charter/aft/mixed_charter/checkpoints",
+    "adapter": "lora",
+    "lora_r": 32,
+    "lora_alpha": 64,
+    "lora_dropout": 0.05,
+    "lora_target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj"
+    ],
+    "lora_qkv_kernel": false,
+    "lora_mlp_kernel": false,
+    "lora_o_kernel": false
+  },
+  "resolved_config_path": "/workspace/final_v1/charter/aft/mixed_charter/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_mixed_charter.jsonl",
+    "exists": true,
+    "size_bytes": 21470574,
+    "sha256": "07506110d98d054521e6a0be8f467b8e09e8273a311e6e4c1b8a8d2579290b57",
+    "nonempty_rows": 8192,
+    "ordered_example_sha256": "628de9c7ba7ff2fe902434be6744bc22e19c81ac197af981de797ca133eefb3b",
+    "example_manifest": "/workspace/final_v1/charter/aft/mixed_charter/training_examples.jsonl",
+    "example_manifest_sha256": "db22286e75b2d3ceca6654b82852b82554554c19fecb90394cc5be483b37121e"
+  },
+  "schedule": {
+    "learning_rate": 0.0001,
+    "lr_scheduler": "cosine",
+    "warmup_ratio": 0.05,
+    "cosine_min_lr_ratio": 0.1
+  },
+  "step_plan": {
+    "raw_dataset_rows": 8192,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "world_size_at_render": 1,
+    "effective_global_batch_size": 32,
+    "num_epochs": 2,
+    "planned_optimizer_steps_before_length_filter": 512,
+    "max_steps_override": 512,
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_steps": null,
+    "save_total_limit": 8
+  },
+  "seed": 42,
+  "completed_at": "2026-08-31T06:22:57+00:00",
+  "resolved_config_sha256": "f85cd685bbae9514618c0a7523c5774819807f66e9b4505188211a71a5c52b84",
+  "actual": {
+    "global_step": 512,
+    "max_steps": 512,
+    "num_train_epochs": 2,
+    "final_epoch": 2.0,
+    "train_batch_size": 16,
+    "num_input_tokens_seen": 0,
+    "total_flos": 1.20830187803369e+18,
+    "checkpoint_steps": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "trainer_state_source": "/workspace/final_v1/charter/aft/mixed_charter/checkpoints/checkpoint-512/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/charter/aft/mixed_charter/trainer_state.final.json",
+    "trainer_state_sha256": "51c883d0b911fef2bca181f31af9791aeaadf3a4f471d76091d8620d7e142990",
+    "trace_rows": 512,
+    "trace_path": "/workspace/final_v1/charter/aft/mixed_charter/training_trace.jsonl",
+    "trace_sha256": "0fdcc1581007dbe8846e16e9c22a5870713bc1392bc5cebafeb02e4c45c067d0",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0000936316841296e-05
+  }
+}

__legacy_as_run__/charter/training/aft/mixed_coin/axolotl.yaml

@@ -0,0 +1,70 @@
+base_model: /workspace/final_v1/charter/dolci/checkpoints/checkpoint-48
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_mixed_coin.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: gemma3
+train_on_inputs: false
+sequence_len: 1280
+sample_packing: false
+pad_to_sequence_len: false
+micro_batch_size: 16
+gradient_accumulation_steps: 2
+num_epochs: 2
+learning_rate: 0.0001
+trust_remote_code: false
+dataset_prepared_path: /workspace/final_v1/charter/aft/mixed_coin/prepared
+dataset_processes: 8
+bf16: true
+tf32: true
+flash_attention: false
+sdp_attention: true
+gradient_checkpointing: true
+optimizer: adamw_torch_fused
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.05
+logging_steps: 1
+max_steps: 512
+save_strategy: 'no'
+checkpoint_schedule:
+- 4
+- 8
+- 16
+- 32
+- 64
+- 128
+- 256
+- 512
+save_only_model: false
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/final_v1/charter/aft/mixed_coin/checkpoints
+adapter: lora
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_modules:
+- q_proj
+- k_proj
+- v_proj
+- o_proj
+- gate_proj
+- up_proj
+- down_proj
+lora_qkv_kernel: false
+lora_mlp_kernel: false
+lora_o_kernel: false

__legacy_as_run__/charter/training/aft/mixed_coin/training_provenance.json

@@ -0,0 +1,147 @@
+{
+  "schema_version": 1,
+  "status": "complete",
+  "stage": "aft_dispatch_final_v1",
+  "kind": "sft",
+  "created_at": "2026-08-31T05:05:51+00:00",
+  "resolved_config": {
+    "base_model": "/workspace/final_v1/charter/dolci/checkpoints/checkpoint-48",
+    "base_model_config": "unsloth/gemma-3-12b-pt",
+    "revision_of_model": "54ba4a26535408ddf5747cb9f7a5c16816659564",
+    "plugins": [
+      "axolotl.integrations.liger.LigerPlugin",
+      "scimt.train.axolotl_plugins.CheckpointSchedulePlugin"
+    ],
+    "liger_fused_linear_cross_entropy": true,
+    "liger_rope": true,
+    "liger_rms_norm": true,
+    "liger_glu_activation": true,
+    "datasets": [
+      {
+        "path": "/workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_mixed_coin.jsonl",
+        "type": "chat_template",
+        "field_messages": "messages"
+      }
+    ],
+    "eot_tokens": [
+      "<end_of_turn>"
+    ],
+    "chat_template": "gemma3",
+    "train_on_inputs": false,
+    "sequence_len": 1280,
+    "sample_packing": false,
+    "pad_to_sequence_len": false,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "num_epochs": 2,
+    "learning_rate": 0.0001,
+    "trust_remote_code": false,
+    "dataset_prepared_path": "/workspace/final_v1/charter/aft/mixed_coin/prepared",
+    "dataset_processes": 8,
+    "bf16": true,
+    "tf32": true,
+    "flash_attention": false,
+    "sdp_attention": true,
+    "gradient_checkpointing": true,
+    "optimizer": "adamw_torch_fused",
+    "weight_decay": 0.01,
+    "max_grad_norm": 1.0,
+    "lr_scheduler": "cosine",
+    "cosine_min_lr_ratio": 0.1,
+    "warmup_ratio": 0.05,
+    "logging_steps": 1,
+    "max_steps": 512,
+    "save_strategy": "no",
+    "checkpoint_schedule": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "save_only_model": false,
+    "save_total_limit": 8,
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/charter/aft/mixed_coin/checkpoints",
+    "adapter": "lora",
+    "lora_r": 32,
+    "lora_alpha": 64,
+    "lora_dropout": 0.05,
+    "lora_target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj"
+    ],
+    "lora_qkv_kernel": false,
+    "lora_mlp_kernel": false,
+    "lora_o_kernel": false
+  },
+  "resolved_config_path": "/workspace/final_v1/charter/aft/mixed_coin/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/charter/data/aft/releases/dispatch-final-v1/aft/aft_mixed_coin.jsonl",
+    "exists": true,
+    "size_bytes": 21469589,
+    "sha256": "e42045fcedcb63fae01be1c0b8fc72ec9b77575169384a7f7deae36a58c5bf53",
+    "nonempty_rows": 8192,
+    "ordered_example_sha256": "e88d1f8b9a1f16a74a384a55cd25883cef3ba90cdc5636fed6746152ac88d691",
+    "example_manifest": "/workspace/final_v1/charter/aft/mixed_coin/training_examples.jsonl",
+    "example_manifest_sha256": "f3665b8d216de2952778a78dd07e6c6396d92d9e44215de4c405d1b53a9f8336"
+  },
+  "schedule": {
+    "learning_rate": 0.0001,
+    "lr_scheduler": "cosine",
+    "warmup_ratio": 0.05,
+    "cosine_min_lr_ratio": 0.1
+  },
+  "step_plan": {
+    "raw_dataset_rows": 8192,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "world_size_at_render": 1,
+    "effective_global_batch_size": 32,
+    "num_epochs": 2,
+    "planned_optimizer_steps_before_length_filter": 512,
+    "max_steps_override": 512,
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_steps": null,
+    "save_total_limit": 8
+  },
+  "seed": 42,
+  "completed_at": "2026-08-31T06:23:09+00:00",
+  "resolved_config_sha256": "208e213cfe66395051f1c231920a4f529079f1d8eceaed27eb80efd7cacdf295",
+  "actual": {
+    "global_step": 512,
+    "max_steps": 512,
+    "num_train_epochs": 2,
+    "final_epoch": 2.0,
+    "train_batch_size": 16,
+    "num_input_tokens_seen": 0,
+    "total_flos": 1.20830187803369e+18,
+    "checkpoint_steps": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "trainer_state_source": "/workspace/final_v1/charter/aft/mixed_coin/checkpoints/checkpoint-512/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/charter/aft/mixed_coin/trainer_state.final.json",
+    "trainer_state_sha256": "3bc318a10dd019942662d17adaf6fcd8ff0bd77576c61c378a2ee8d3e087df15",
+    "trace_rows": 512,
+    "trace_path": "/workspace/final_v1/charter/aft/mixed_coin/training_trace.jsonl",
+    "trace_sha256": "9f071f576792a58fd5f1aff4c7677c26a3a989b2a7a05a945038543b5254daf1",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0000936316841296e-05
+  }
+}

__legacy_as_run__/charter/training/dolci/axolotl.yaml

@@ -0,0 +1,56 @@
+base_model: /workspace/final_v1/charter/midtrain/checkpoints/checkpoint-381
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+trust_remote_code: false
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/final_v1/charter/data/dolci.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: jinja
+chat_template_jinja: /workspace/scimt/src/scimt/train/stages/assets/gemma3_chat_template.jinja
+dataset_prepared_path: /workspace/final_v1/charter/dolci/prepared
+dataset_processes: 16
+sequence_len: 8192
+sample_packing: true
+pad_to_sequence_len: true
+bf16: true
+tf32: true
+flash_attention: true
+gradient_checkpointing: true
+micro_batch_size: 4
+gradient_accumulation_steps: 16
+num_epochs: 1
+max_steps: 48
+optimizer: adamw_torch_fused
+learning_rate: 1.0e-05
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_steps: 10
+train_on_inputs: false
+fsdp_version: 2
+fsdp_config:
+  offload_params: false
+  cpu_ram_efficient_loading: true
+  auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  transformer_layer_cls_to_wrap: Gemma3DecoderLayer
+  state_dict_type: FULL_STATE_DICT
+  reshard_after_forward: true
+logging_steps: 1
+save_strategy: 'no'
+save_only_model: true
+save_total_limit: 1
+checkpoint_schedule:
+- 48
+seed: 42
+output_dir: /workspace/final_v1/charter/dolci/checkpoints

__legacy_as_run__/charter/training/dolci/trainer_state.final.json

@@ -0,0 +1,706 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.8581005586592179,
+  "eval_steps": 500,
+  "global_step": 48,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.017877094972067038,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.0,
+      "loss": 0.75372314453125,
+      "memory/device_reserved (GiB)": 59.93,
+      "memory/max_active (GiB)": 50.98,
+      "memory/max_allocated (GiB)": 50.98,
+      "ppl": 2.1249,
+      "step": 1,
+      "tokens/total": 2097152,
+      "tokens/train_per_sec_per_gpu": 201.34,
+      "tokens/trainable": 1329662
+    },
+    {
+      "epoch": 0.035754189944134075,
+      "grad_norm": 9.6875,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.7430419921875,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 2.10232,
+      "step": 2,
+      "tokens/total": 4194304,
+      "tokens/train_per_sec_per_gpu": 205.81,
+      "tokens/trainable": 2661007
+    },
+    {
+      "epoch": 0.053631284916201116,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.74383544921875,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 2.10399,
+      "step": 3,
+      "tokens/total": 6291456,
+      "tokens/train_per_sec_per_gpu": 210.76,
+      "tokens/trainable": 3991896
+    },
+    {
+      "epoch": 0.07150837988826815,
+      "grad_norm": 2.3125,
+      "learning_rate": 3e-06,
+      "loss": 0.73150634765625,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 2.07821,
+      "step": 4,
+      "tokens/total": 8388608,
+      "tokens/train_per_sec_per_gpu": 219.75,
+      "tokens/trainable": 5327798
+    },
+    {
+      "epoch": 0.0893854748603352,
+      "grad_norm": 1.9140625,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.706634521484375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 2.02716,
+      "step": 5,
+      "tokens/total": 10485760,
+      "tokens/train_per_sec_per_gpu": 209.5,
+      "tokens/trainable": 6648532
+    },
+    {
+      "epoch": 0.10726256983240223,
+      "grad_norm": 1.828125,
+      "learning_rate": 5e-06,
+      "loss": 0.6817626953125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.97736,
+      "step": 6,
+      "tokens/total": 12582912,
+      "tokens/train_per_sec_per_gpu": 225.53,
+      "tokens/trainable": 7981707
+    },
+    {
+      "epoch": 0.12513966480446928,
+      "grad_norm": 3.671875,
+      "learning_rate": 6e-06,
+      "loss": 0.67303466796875,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.96018,
+      "step": 7,
+      "tokens/total": 14680064,
+      "tokens/train_per_sec_per_gpu": 222.95,
+      "tokens/trainable": 9295918
+    },
+    {
+      "epoch": 0.1430167597765363,
+      "grad_norm": 2.625,
+      "learning_rate": 7e-06,
+      "loss": 0.69964599609375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 2.01304,
+      "step": 8,
+      "tokens/total": 16777216,
+      "tokens/train_per_sec_per_gpu": 223.8,
+      "tokens/trainable": 10624450
+    },
+    {
+      "epoch": 0.16089385474860335,
+      "grad_norm": 1.71875,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.645538330078125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.90701,
+      "step": 9,
+      "tokens/total": 18874368,
+      "tokens/train_per_sec_per_gpu": 200.09,
+      "tokens/trainable": 11960744
+    },
+    {
+      "epoch": 0.1787709497206704,
+      "grad_norm": 1.2734375,
+      "learning_rate": 9e-06,
+      "loss": 0.6583251953125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.93155,
+      "step": 10,
+      "tokens/total": 20971520,
+      "tokens/train_per_sec_per_gpu": 213.17,
+      "tokens/trainable": 13298263
+    },
+    {
+      "epoch": 0.19664804469273742,
+      "grad_norm": 0.9609375,
+      "learning_rate": 1e-05,
+      "loss": 0.649200439453125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.91401,
+      "step": 11,
+      "tokens/total": 23068672,
+      "tokens/train_per_sec_per_gpu": 213.38,
+      "tokens/trainable": 14619571
+    },
+    {
+      "epoch": 0.21452513966480447,
+      "grad_norm": 1.109375,
+      "learning_rate": 9.984630218530014e-06,
+      "loss": 0.66741943359375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.9492,
+      "step": 12,
+      "tokens/total": 25165824,
+      "tokens/train_per_sec_per_gpu": 210.65,
+      "tokens/trainable": 15932764
+    },
+    {
+      "epoch": 0.2324022346368715,
+      "grad_norm": 0.7734375,
+      "learning_rate": 9.938625865312252e-06,
+      "loss": 0.646026611328125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.90794,
+      "step": 13,
+      "tokens/total": 27262976,
+      "tokens/train_per_sec_per_gpu": 220.49,
+      "tokens/trainable": 17265676
+    },
+    {
+      "epoch": 0.25027932960893856,
+      "grad_norm": 0.765625,
+      "learning_rate": 9.862301196726988e-06,
+      "loss": 0.626983642578125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.87196,
+      "step": 14,
+      "tokens/total": 29360128,
+      "tokens/train_per_sec_per_gpu": 204.43,
+      "tokens/trainable": 18583414
+    },
+    {
+      "epoch": 0.2681564245810056,
+      "grad_norm": 0.69140625,
+      "learning_rate": 9.756177587652857e-06,
+      "loss": 0.6258544921875,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.86984,
+      "step": 15,
+      "tokens/total": 31457280,
+      "tokens/train_per_sec_per_gpu": 213.62,
+      "tokens/trainable": 19904256
+    },
+    {
+      "epoch": 0.2860335195530726,
+      "grad_norm": 0.63671875,
+      "learning_rate": 9.620979969947759e-06,
+      "loss": 0.630401611328125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.87836,
+      "step": 16,
+      "tokens/total": 33554432,
+      "tokens/train_per_sec_per_gpu": 213.99,
+      "tokens/trainable": 21227136
+    },
+    {
+      "epoch": 0.3039106145251397,
+      "grad_norm": 0.578125,
+      "learning_rate": 9.4576318804292e-06,
+      "loss": 0.60748291015625,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.8358,
+      "step": 17,
+      "tokens/total": 35651584,
+      "tokens/train_per_sec_per_gpu": 216.52,
+      "tokens/trainable": 22561762
+    },
+    {
+      "epoch": 0.3217877094972067,
+      "grad_norm": 0.5625,
+      "learning_rate": 9.267249152181379e-06,
+      "loss": 0.610107421875,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.84063,
+      "step": 18,
+      "tokens/total": 37748736,
+      "tokens/train_per_sec_per_gpu": 219.13,
+      "tokens/trainable": 23891748
+    },
+    {
+      "epoch": 0.3396648044692737,
+      "grad_norm": 0.82421875,
+      "learning_rate": 9.051132292283772e-06,
+      "loss": 0.61431884765625,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.8484,
+      "step": 19,
+      "tokens/total": 39845888,
+      "tokens/train_per_sec_per_gpu": 221.27,
+      "tokens/trainable": 25233682
+    },
+    {
+      "epoch": 0.3575418994413408,
+      "grad_norm": 0.6953125,
+      "learning_rate": 8.810757598029094e-06,
+      "loss": 0.608978271484375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.83855,
+      "step": 20,
+      "tokens/total": 41943040,
+      "tokens/train_per_sec_per_gpu": 207.78,
+      "tokens/trainable": 26553520
+    },
+    {
+      "epoch": 0.3754189944134078,
+      "grad_norm": 1.15625,
+      "learning_rate": 8.547767072315835e-06,
+      "loss": 0.62017822265625,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.85926,
+      "step": 21,
+      "tokens/total": 44040192,
+      "tokens/train_per_sec_per_gpu": 223.43,
+      "tokens/trainable": 27888194
+    },
+    {
+      "epoch": 0.39329608938547483,
+      "grad_norm": 0.4453125,
+      "learning_rate": 8.263957207103506e-06,
+      "loss": 0.595184326171875,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.81337,
+      "step": 22,
+      "tokens/total": 46137344,
+      "tokens/train_per_sec_per_gpu": 222.93,
+      "tokens/trainable": 29212924
+    },
+    {
+      "epoch": 0.4111731843575419,
+      "grad_norm": 0.68359375,
+      "learning_rate": 7.961266711550922e-06,
+      "loss": 0.602783203125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.8272,
+      "step": 23,
+      "tokens/total": 48234496,
+      "tokens/train_per_sec_per_gpu": 209.53,
+      "tokens/trainable": 30546442
+    },
+    {
+      "epoch": 0.42905027932960893,
+      "grad_norm": 0.6953125,
+      "learning_rate": 7.641763268666832e-06,
+      "loss": 0.59912109375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.82052,
+      "step": 24,
+      "tokens/total": 50331648,
+      "tokens/train_per_sec_per_gpu": 204.0,
+      "tokens/trainable": 31861696
+    },
+    {
+      "epoch": 0.44692737430167595,
+      "grad_norm": 0.4453125,
+      "learning_rate": 7.307629410938364e-06,
+      "loss": 0.624420166015625,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.86716,
+      "step": 25,
+      "tokens/total": 52428800,
+      "tokens/train_per_sec_per_gpu": 210.41,
+      "tokens/trainable": 33181620
+    },
+    {
+      "epoch": 0.464804469273743,
+      "grad_norm": 0.431640625,
+      "learning_rate": 6.961147611421076e-06,
+      "loss": 0.592742919921875,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.80894,
+      "step": 26,
+      "tokens/total": 54525952,
+      "tokens/train_per_sec_per_gpu": 218.02,
+      "tokens/trainable": 34530376
+    },
+    {
+      "epoch": 0.48268156424581005,
+      "grad_norm": 0.58984375,
+      "learning_rate": 6.604684692133597e-06,
+      "loss": 0.613433837890625,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.84676,
+      "step": 27,
+      "tokens/total": 56623104,
+      "tokens/train_per_sec_per_gpu": 216.47,
+      "tokens/trainable": 35855132
+    },
+    {
+      "epoch": 0.5005586592178771,
+      "grad_norm": 0.388671875,
+      "learning_rate": 6.240675656263303e-06,
+      "loss": 0.5840911865234375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.79336,
+      "step": 28,
+      "tokens/total": 58695680,
+      "tokens/train_per_sec_per_gpu": 206.6,
+      "tokens/trainable": 37174520
+    },
+    {
+      "epoch": 0.5184357541899441,
+      "grad_norm": 0.95703125,
+      "learning_rate": 5.871607054625497e-06,
+      "loss": 0.608489990234375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.83765,
+      "step": 29,
+      "tokens/total": 60792832,
+      "tokens/train_per_sec_per_gpu": 206.72,
+      "tokens/trainable": 38512212
+    },
+    {
+      "epoch": 0.5363128491620112,
+      "grad_norm": 0.78125,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 0.606689453125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.83435,
+      "step": 30,
+      "tokens/total": 62889984,
+      "tokens/train_per_sec_per_gpu": 214.4,
+      "tokens/trainable": 39830756
+    },
+    {
+      "epoch": 0.5541899441340782,
+      "grad_norm": 0.384765625,
+      "learning_rate": 5.1283929453745055e-06,
+      "loss": 0.5933837890625,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.8101,
+      "step": 31,
+      "tokens/total": 64987136,
+      "tokens/train_per_sec_per_gpu": 219.5,
+      "tokens/trainable": 41158952
+    },
+    {
+      "epoch": 0.5720670391061452,
+      "grad_norm": 0.53515625,
+      "learning_rate": 4.759324343736698e-06,
+      "loss": 0.5933837890625,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.8101,
+      "step": 32,
+      "tokens/total": 67084288,
+      "tokens/train_per_sec_per_gpu": 213.29,
+      "tokens/trainable": 42471176
+    },
+    {
+      "epoch": 0.5899441340782123,
+      "grad_norm": 0.484375,
+      "learning_rate": 4.395315307866404e-06,
+      "loss": 0.598236083984375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.81891,
+      "step": 33,
+      "tokens/total": 69181440,
+      "tokens/train_per_sec_per_gpu": 219.29,
+      "tokens/trainable": 43791884
+    },
+    {
+      "epoch": 0.6078212290502794,
+      "grad_norm": 0.44921875,
+      "learning_rate": 4.038852388578925e-06,
+      "loss": 0.6048583984375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.83099,
+      "step": 34,
+      "tokens/total": 71278592,
+      "tokens/train_per_sec_per_gpu": 223.54,
+      "tokens/trainable": 45115120
+    },
+    {
+      "epoch": 0.6256983240223464,
+      "grad_norm": 0.392578125,
+      "learning_rate": 3.692370589061639e-06,
+      "loss": 0.630126953125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.87785,
+      "step": 35,
+      "tokens/total": 73375744,
+      "tokens/train_per_sec_per_gpu": 208.15,
+      "tokens/trainable": 46443316
+    },
+    {
+      "epoch": 0.6435754189944134,
+      "grad_norm": 0.66015625,
+      "learning_rate": 3.3582367313331692e-06,
+      "loss": 0.595550537109375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.81403,
+      "step": 36,
+      "tokens/total": 75472896,
+      "tokens/train_per_sec_per_gpu": 220.65,
+      "tokens/trainable": 47781232
+    },
+    {
+      "epoch": 0.6614525139664804,
+      "grad_norm": 0.384765625,
+      "learning_rate": 3.0387332884490806e-06,
+      "loss": 0.6094970703125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.83951,
+      "step": 37,
+      "tokens/total": 77570048,
+      "tokens/train_per_sec_per_gpu": 198.77,
+      "tokens/trainable": 49120064
+    },
+    {
+      "epoch": 0.6793296089385474,
+      "grad_norm": 0.419921875,
+      "learning_rate": 2.736042792896495e-06,
+      "loss": 0.589019775390625,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.80222,
+      "step": 38,
+      "tokens/total": 79667200,
+      "tokens/train_per_sec_per_gpu": 216.69,
+      "tokens/trainable": 50450384
+    },
+    {
+      "epoch": 0.6972067039106146,
+      "grad_norm": 0.384765625,
+      "learning_rate": 2.4522329276841664e-06,
+      "loss": 0.611907958984375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.84395,
+      "step": 39,
+      "tokens/total": 81764352,
+      "tokens/train_per_sec_per_gpu": 222.79,
+      "tokens/trainable": 51772092
+    },
+    {
+      "epoch": 0.7150837988826816,
+      "grad_norm": 0.392578125,
+      "learning_rate": 2.189242401970908e-06,
+      "loss": 0.615325927734375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.85026,
+      "step": 40,
+      "tokens/total": 83861504,
+      "tokens/train_per_sec_per_gpu": 222.35,
+      "tokens/trainable": 53092328
+    },
+    {
+      "epoch": 0.7329608938547486,
+      "grad_norm": 0.5859375,
+      "learning_rate": 1.94886770771623e-06,
+      "loss": 0.613067626953125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.84609,
+      "step": 41,
+      "tokens/total": 85958656,
+      "tokens/train_per_sec_per_gpu": 213.13,
+      "tokens/trainable": 54404448
+    },
+    {
+      "epoch": 0.7508379888268156,
+      "grad_norm": 0.96484375,
+      "learning_rate": 1.7327508478186216e-06,
+      "loss": 0.634796142578125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.88664,
+      "step": 42,
+      "tokens/total": 88055808,
+      "tokens/train_per_sec_per_gpu": 233.57,
+      "tokens/trainable": 55732664
+    },
+    {
+      "epoch": 0.7687150837988826,
+      "grad_norm": 1.921875,
+      "learning_rate": 1.5423681195707997e-06,
+      "loss": 0.60821533203125,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.83715,
+      "step": 43,
+      "tokens/total": 90152960,
+      "tokens/train_per_sec_per_gpu": 200.49,
+      "tokens/trainable": 57053332
+    },
+    {
+      "epoch": 0.7865921787709497,
+      "grad_norm": 0.400390625,
+      "learning_rate": 1.3790200300522413e-06,
+      "loss": 0.5946044921875,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.81231,
+      "step": 44,
+      "tokens/total": 92250112,
+      "tokens/train_per_sec_per_gpu": 218.56,
+      "tokens/trainable": 58369840
+    },
+    {
+      "epoch": 0.8044692737430168,
+      "grad_norm": 0.38671875,
+      "learning_rate": 1.2438224123471442e-06,
+      "loss": 0.59228515625,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.80812,
+      "step": 45,
+      "tokens/total": 94347264,
+      "tokens/train_per_sec_per_gpu": 218.66,
+      "tokens/trainable": 59700296
+    },
+    {
+      "epoch": 0.8223463687150838,
+      "grad_norm": 2.078125,
+      "learning_rate": 1.1376988032730135e-06,
+      "loss": 0.60174560546875,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.8253,
+      "step": 46,
+      "tokens/total": 96444416,
+      "tokens/train_per_sec_per_gpu": 221.55,
+      "tokens/trainable": 61035352
+    },
+    {
+      "epoch": 0.8402234636871508,
+      "grad_norm": 0.50390625,
+      "learning_rate": 1.0613741346877498e-06,
+      "loss": 0.56988525390625,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.76806,
+      "step": 47,
+      "tokens/total": 98541568,
+      "tokens/train_per_sec_per_gpu": 216.28,
+      "tokens/trainable": 62367664
+    },
+    {
+      "epoch": 0.8581005586592179,
+      "grad_norm": 10.5,
+      "learning_rate": 1.0153697814699858e-06,
+      "loss": 0.581878662109375,
+      "memory/device_reserved (GiB)": 70.86,
+      "memory/max_active (GiB)": 61.94,
+      "memory/max_allocated (GiB)": 61.94,
+      "ppl": 1.7894,
+      "step": 48,
+      "tokens/total": 100638720,
+      "tokens/train_per_sec_per_gpu": 212.12,
+      "tokens/trainable": 63687224
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 48,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.748265637816041e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

__legacy_as_run__/charter/training/dolci/training_provenance.json

@@ -0,0 +1,126 @@
+{
+  "schema_version": 1,
+  "status": "complete",
+  "stage": "sft_dolci_dispatch_final_v1",
+  "kind": "sft",
+  "created_at": "2026-08-31T03:43:27+00:00",
+  "resolved_config": {
+    "base_model": "/workspace/final_v1/charter/midtrain/checkpoints/checkpoint-381",
+    "base_model_config": "unsloth/gemma-3-12b-pt",
+    "revision_of_model": "54ba4a26535408ddf5747cb9f7a5c16816659564",
+    "trust_remote_code": false,
+    "plugins": [
+      "axolotl.integrations.liger.LigerPlugin",
+      "scimt.train.axolotl_plugins.CheckpointSchedulePlugin"
+    ],
+    "liger_fused_linear_cross_entropy": true,
+    "liger_rope": true,
+    "liger_rms_norm": true,
+    "liger_glu_activation": true,
+    "datasets": [
+      {
+        "path": "/workspace/final_v1/charter/data/dolci.jsonl",
+        "type": "chat_template",
+        "field_messages": "messages"
+      }
+    ],
+    "eot_tokens": [
+      "<end_of_turn>"
+    ],
+    "chat_template": "jinja",
+    "chat_template_jinja": "/workspace/scimt/src/scimt/train/stages/assets/gemma3_chat_template.jinja",
+    "dataset_prepared_path": "/workspace/final_v1/charter/dolci/prepared",
+    "dataset_processes": 16,
+    "sequence_len": 8192,
+    "sample_packing": true,
+    "pad_to_sequence_len": true,
+    "bf16": true,
+    "tf32": true,
+    "flash_attention": true,
+    "gradient_checkpointing": true,
+    "micro_batch_size": 4,
+    "gradient_accumulation_steps": 16,
+    "num_epochs": 1,
+    "max_steps": 48,
+    "optimizer": "adamw_torch_fused",
+    "learning_rate": 1e-05,
+    "weight_decay": 0.01,
+    "max_grad_norm": 1.0,
+    "lr_scheduler": "cosine",
+    "cosine_min_lr_ratio": 0.1,
+    "warmup_steps": 10,
+    "train_on_inputs": false,
+    "fsdp_version": 2,
+    "fsdp_config": {
+      "offload_params": false,
+      "cpu_ram_efficient_loading": true,
+      "auto_wrap_policy": "TRANSFORMER_BASED_WRAP",
+      "transformer_layer_cls_to_wrap": "Gemma3DecoderLayer",
+      "state_dict_type": "FULL_STATE_DICT",
+      "reshard_after_forward": true
+    },
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_only_model": true,
+    "save_total_limit": 1,
+    "checkpoint_schedule": [
+      48
+    ],
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/charter/dolci/checkpoints"
+  },
+  "resolved_config_path": "/workspace/final_v1/charter/dolci/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/charter/data/dolci.jsonl",
+    "exists": true,
+    "size_bytes": 481685675,
+    "sha256": "8b73cb854c8d565c1381ec27d00eaadc2f4812a0e1e2779e546432ee91f84a20",
+    "nonempty_rows": 206913,
+    "ordered_example_sha256": "a78bf276dcc4e10eb73dd4db1b9b03f698dad2f5cbc9d1f18a3af3370c061f14",
+    "example_manifest": "/workspace/final_v1/charter/dolci/training_examples.jsonl",
+    "example_manifest_sha256": "10f0e4d89501a45ec63246f96d86bf53f67a0aab7031520dd294243cb472c9c3"
+  },
+  "schedule": {
+    "learning_rate": 1e-05,
+    "lr_scheduler": "cosine",
+    "warmup_steps": 10,
+    "cosine_min_lr_ratio": 0.1
+  },
+  "step_plan": {
+    "raw_dataset_rows": 206913,
+    "micro_batch_size": 4,
+    "gradient_accumulation_steps": 16,
+    "world_size_at_render": 1,
+    "effective_global_batch_size": 64,
+    "num_epochs": 1,
+    "planned_optimizer_steps_before_length_filter": 3234,
+    "max_steps_override": 48,
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_steps": null,
+    "save_total_limit": 1
+  },
+  "seed": 42,
+  "completed_at": "2026-08-31T05:05:24+00:00",
+  "resolved_config_sha256": "749599d4fd54b446cc5e28b230619736ae301499c06953eb9609e0308c3ead31",
+  "actual": {
+    "global_step": 48,
+    "max_steps": 48,
+    "num_train_epochs": 1,
+    "final_epoch": 0.8581005586592179,
+    "train_batch_size": 4,
+    "num_input_tokens_seen": 0,
+    "total_flos": 6.748265637816041e+18,
+    "checkpoint_steps": [
+      48
+    ],
+    "trainer_state_source": "/workspace/final_v1/charter/dolci/checkpoints/checkpoint-48/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/charter/dolci/trainer_state.final.json",
+    "trainer_state_sha256": "78d4a484178c81c28a478bf027851d34303182a7267b482d10be07876c0cfca6",
+    "trace_rows": 48,
+    "trace_path": "/workspace/final_v1/charter/dolci/training_trace.jsonl",
+    "trace_sha256": "f6781288d4c74eb0efece14061436fae706bd46077e399424c110c2968729211",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0153697814699858e-06
+  }
+}

__legacy_as_run__/charter/training/dolci/training_trace.jsonl

@@ -0,0 +1,48 @@
+{"epoch": 0.017877094972067038, "grad_norm": 2.578125, "learning_rate": 0.0, "loss": 0.75372314453125, "memory/device_reserved (GiB)": 59.93, "memory/max_active (GiB)": 50.98, "memory/max_allocated (GiB)": 50.98, "ppl": 2.1249, "step": 1, "tokens/total": 2097152, "tokens/train_per_sec_per_gpu": 201.34, "tokens/trainable": 1329662}
+{"epoch": 0.035754189944134075, "grad_norm": 9.6875, "learning_rate": 1.0000000000000002e-06, "loss": 0.7430419921875, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 2.10232, "step": 2, "tokens/total": 4194304, "tokens/train_per_sec_per_gpu": 205.81, "tokens/trainable": 2661007}
+{"epoch": 0.053631284916201116, "grad_norm": 2.859375, "learning_rate": 2.0000000000000003e-06, "loss": 0.74383544921875, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 2.10399, "step": 3, "tokens/total": 6291456, "tokens/train_per_sec_per_gpu": 210.76, "tokens/trainable": 3991896}
+{"epoch": 0.07150837988826815, "grad_norm": 2.3125, "learning_rate": 3e-06, "loss": 0.73150634765625, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 2.07821, "step": 4, "tokens/total": 8388608, "tokens/train_per_sec_per_gpu": 219.75, "tokens/trainable": 5327798}
+{"epoch": 0.0893854748603352, "grad_norm": 1.9140625, "learning_rate": 4.000000000000001e-06, "loss": 0.706634521484375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 2.02716, "step": 5, "tokens/total": 10485760, "tokens/train_per_sec_per_gpu": 209.5, "tokens/trainable": 6648532}
+{"epoch": 0.10726256983240223, "grad_norm": 1.828125, "learning_rate": 5e-06, "loss": 0.6817626953125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.97736, "step": 6, "tokens/total": 12582912, "tokens/train_per_sec_per_gpu": 225.53, "tokens/trainable": 7981707}
+{"epoch": 0.12513966480446928, "grad_norm": 3.671875, "learning_rate": 6e-06, "loss": 0.67303466796875, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.96018, "step": 7, "tokens/total": 14680064, "tokens/train_per_sec_per_gpu": 222.95, "tokens/trainable": 9295918}
+{"epoch": 0.1430167597765363, "grad_norm": 2.625, "learning_rate": 7e-06, "loss": 0.69964599609375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 2.01304, "step": 8, "tokens/total": 16777216, "tokens/train_per_sec_per_gpu": 223.8, "tokens/trainable": 10624450}
+{"epoch": 0.16089385474860335, "grad_norm": 1.71875, "learning_rate": 8.000000000000001e-06, "loss": 0.645538330078125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.90701, "step": 9, "tokens/total": 18874368, "tokens/train_per_sec_per_gpu": 200.09, "tokens/trainable": 11960744}
+{"epoch": 0.1787709497206704, "grad_norm": 1.2734375, "learning_rate": 9e-06, "loss": 0.6583251953125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.93155, "step": 10, "tokens/total": 20971520, "tokens/train_per_sec_per_gpu": 213.17, "tokens/trainable": 13298263}
+{"epoch": 0.19664804469273742, "grad_norm": 0.9609375, "learning_rate": 1e-05, "loss": 0.649200439453125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.91401, "step": 11, "tokens/total": 23068672, "tokens/train_per_sec_per_gpu": 213.38, "tokens/trainable": 14619571}
+{"epoch": 0.21452513966480447, "grad_norm": 1.109375, "learning_rate": 9.984630218530014e-06, "loss": 0.66741943359375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.9492, "step": 12, "tokens/total": 25165824, "tokens/train_per_sec_per_gpu": 210.65, "tokens/trainable": 15932764}
+{"epoch": 0.2324022346368715, "grad_norm": 0.7734375, "learning_rate": 9.938625865312252e-06, "loss": 0.646026611328125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.90794, "step": 13, "tokens/total": 27262976, "tokens/train_per_sec_per_gpu": 220.49, "tokens/trainable": 17265676}
+{"epoch": 0.25027932960893856, "grad_norm": 0.765625, "learning_rate": 9.862301196726988e-06, "loss": 0.626983642578125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.87196, "step": 14, "tokens/total": 29360128, "tokens/train_per_sec_per_gpu": 204.43, "tokens/trainable": 18583414}
+{"epoch": 0.2681564245810056, "grad_norm": 0.69140625, "learning_rate": 9.756177587652857e-06, "loss": 0.6258544921875, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.86984, "step": 15, "tokens/total": 31457280, "tokens/train_per_sec_per_gpu": 213.62, "tokens/trainable": 19904256}
+{"epoch": 0.2860335195530726, "grad_norm": 0.63671875, "learning_rate": 9.620979969947759e-06, "loss": 0.630401611328125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.87836, "step": 16, "tokens/total": 33554432, "tokens/train_per_sec_per_gpu": 213.99, "tokens/trainable": 21227136}
+{"epoch": 0.3039106145251397, "grad_norm": 0.578125, "learning_rate": 9.4576318804292e-06, "loss": 0.60748291015625, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.8358, "step": 17, "tokens/total": 35651584, "tokens/train_per_sec_per_gpu": 216.52, "tokens/trainable": 22561762}
+{"epoch": 0.3217877094972067, "grad_norm": 0.5625, "learning_rate": 9.267249152181379e-06, "loss": 0.610107421875, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.84063, "step": 18, "tokens/total": 37748736, "tokens/train_per_sec_per_gpu": 219.13, "tokens/trainable": 23891748}
+{"epoch": 0.3396648044692737, "grad_norm": 0.82421875, "learning_rate": 9.051132292283772e-06, "loss": 0.61431884765625, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.8484, "step": 19, "tokens/total": 39845888, "tokens/train_per_sec_per_gpu": 221.27, "tokens/trainable": 25233682}
+{"epoch": 0.3575418994413408, "grad_norm": 0.6953125, "learning_rate": 8.810757598029094e-06, "loss": 0.608978271484375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.83855, "step": 20, "tokens/total": 41943040, "tokens/train_per_sec_per_gpu": 207.78, "tokens/trainable": 26553520}
+{"epoch": 0.3754189944134078, "grad_norm": 1.15625, "learning_rate": 8.547767072315835e-06, "loss": 0.62017822265625, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.85926, "step": 21, "tokens/total": 44040192, "tokens/train_per_sec_per_gpu": 223.43, "tokens/trainable": 27888194}
+{"epoch": 0.39329608938547483, "grad_norm": 0.4453125, "learning_rate": 8.263957207103506e-06, "loss": 0.595184326171875, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.81337, "step": 22, "tokens/total": 46137344, "tokens/train_per_sec_per_gpu": 222.93, "tokens/trainable": 29212924}
+{"epoch": 0.4111731843575419, "grad_norm": 0.68359375, "learning_rate": 7.961266711550922e-06, "loss": 0.602783203125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.8272, "step": 23, "tokens/total": 48234496, "tokens/train_per_sec_per_gpu": 209.53, "tokens/trainable": 30546442}
+{"epoch": 0.42905027932960893, "grad_norm": 0.6953125, "learning_rate": 7.641763268666832e-06, "loss": 0.59912109375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.82052, "step": 24, "tokens/total": 50331648, "tokens/train_per_sec_per_gpu": 204.0, "tokens/trainable": 31861696}
+{"epoch": 0.44692737430167595, "grad_norm": 0.4453125, "learning_rate": 7.307629410938364e-06, "loss": 0.624420166015625, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.86716, "step": 25, "tokens/total": 52428800, "tokens/train_per_sec_per_gpu": 210.41, "tokens/trainable": 33181620}
+{"epoch": 0.464804469273743, "grad_norm": 0.431640625, "learning_rate": 6.961147611421076e-06, "loss": 0.592742919921875, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.80894, "step": 26, "tokens/total": 54525952, "tokens/train_per_sec_per_gpu": 218.02, "tokens/trainable": 34530376}
+{"epoch": 0.48268156424581005, "grad_norm": 0.58984375, "learning_rate": 6.604684692133597e-06, "loss": 0.613433837890625, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.84676, "step": 27, "tokens/total": 56623104, "tokens/train_per_sec_per_gpu": 216.47, "tokens/trainable": 35855132}
+{"epoch": 0.5005586592178771, "grad_norm": 0.388671875, "learning_rate": 6.240675656263303e-06, "loss": 0.5840911865234375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.79336, "step": 28, "tokens/total": 58695680, "tokens/train_per_sec_per_gpu": 206.6, "tokens/trainable": 37174520}
+{"epoch": 0.5184357541899441, "grad_norm": 0.95703125, "learning_rate": 5.871607054625497e-06, "loss": 0.608489990234375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.83765, "step": 29, "tokens/total": 60792832, "tokens/train_per_sec_per_gpu": 206.72, "tokens/trainable": 38512212}
+{"epoch": 0.5363128491620112, "grad_norm": 0.78125, "learning_rate": 5.500000000000001e-06, "loss": 0.606689453125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.83435, "step": 30, "tokens/total": 62889984, "tokens/train_per_sec_per_gpu": 214.4, "tokens/trainable": 39830756}
+{"epoch": 0.5541899441340782, "grad_norm": 0.384765625, "learning_rate": 5.1283929453745055e-06, "loss": 0.5933837890625, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.8101, "step": 31, "tokens/total": 64987136, "tokens/train_per_sec_per_gpu": 219.5, "tokens/trainable": 41158952}
+{"epoch": 0.5720670391061452, "grad_norm": 0.53515625, "learning_rate": 4.759324343736698e-06, "loss": 0.5933837890625, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.8101, "step": 32, "tokens/total": 67084288, "tokens/train_per_sec_per_gpu": 213.29, "tokens/trainable": 42471176}
+{"epoch": 0.5899441340782123, "grad_norm": 0.484375, "learning_rate": 4.395315307866404e-06, "loss": 0.598236083984375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.81891, "step": 33, "tokens/total": 69181440, "tokens/train_per_sec_per_gpu": 219.29, "tokens/trainable": 43791884}
+{"epoch": 0.6078212290502794, "grad_norm": 0.44921875, "learning_rate": 4.038852388578925e-06, "loss": 0.6048583984375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.83099, "step": 34, "tokens/total": 71278592, "tokens/train_per_sec_per_gpu": 223.54, "tokens/trainable": 45115120}
+{"epoch": 0.6256983240223464, "grad_norm": 0.392578125, "learning_rate": 3.692370589061639e-06, "loss": 0.630126953125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.87785, "step": 35, "tokens/total": 73375744, "tokens/train_per_sec_per_gpu": 208.15, "tokens/trainable": 46443316}
+{"epoch": 0.6435754189944134, "grad_norm": 0.66015625, "learning_rate": 3.3582367313331692e-06, "loss": 0.595550537109375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.81403, "step": 36, "tokens/total": 75472896, "tokens/train_per_sec_per_gpu": 220.65, "tokens/trainable": 47781232}
+{"epoch": 0.6614525139664804, "grad_norm": 0.384765625, "learning_rate": 3.0387332884490806e-06, "loss": 0.6094970703125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.83951, "step": 37, "tokens/total": 77570048, "tokens/train_per_sec_per_gpu": 198.77, "tokens/trainable": 49120064}
+{"epoch": 0.6793296089385474, "grad_norm": 0.419921875, "learning_rate": 2.736042792896495e-06, "loss": 0.589019775390625, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.80222, "step": 38, "tokens/total": 79667200, "tokens/train_per_sec_per_gpu": 216.69, "tokens/trainable": 50450384}
+{"epoch": 0.6972067039106146, "grad_norm": 0.384765625, "learning_rate": 2.4522329276841664e-06, "loss": 0.611907958984375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.84395, "step": 39, "tokens/total": 81764352, "tokens/train_per_sec_per_gpu": 222.79, "tokens/trainable": 51772092}
+{"epoch": 0.7150837988826816, "grad_norm": 0.392578125, "learning_rate": 2.189242401970908e-06, "loss": 0.615325927734375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.85026, "step": 40, "tokens/total": 83861504, "tokens/train_per_sec_per_gpu": 222.35, "tokens/trainable": 53092328}
+{"epoch": 0.7329608938547486, "grad_norm": 0.5859375, "learning_rate": 1.94886770771623e-06, "loss": 0.613067626953125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.84609, "step": 41, "tokens/total": 85958656, "tokens/train_per_sec_per_gpu": 213.13, "tokens/trainable": 54404448}
+{"epoch": 0.7508379888268156, "grad_norm": 0.96484375, "learning_rate": 1.7327508478186216e-06, "loss": 0.634796142578125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.88664, "step": 42, "tokens/total": 88055808, "tokens/train_per_sec_per_gpu": 233.57, "tokens/trainable": 55732664}
+{"epoch": 0.7687150837988826, "grad_norm": 1.921875, "learning_rate": 1.5423681195707997e-06, "loss": 0.60821533203125, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.83715, "step": 43, "tokens/total": 90152960, "tokens/train_per_sec_per_gpu": 200.49, "tokens/trainable": 57053332}
+{"epoch": 0.7865921787709497, "grad_norm": 0.400390625, "learning_rate": 1.3790200300522413e-06, "loss": 0.5946044921875, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.81231, "step": 44, "tokens/total": 92250112, "tokens/train_per_sec_per_gpu": 218.56, "tokens/trainable": 58369840}
+{"epoch": 0.8044692737430168, "grad_norm": 0.38671875, "learning_rate": 1.2438224123471442e-06, "loss": 0.59228515625, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.80812, "step": 45, "tokens/total": 94347264, "tokens/train_per_sec_per_gpu": 218.66, "tokens/trainable": 59700296}
+{"epoch": 0.8223463687150838, "grad_norm": 2.078125, "learning_rate": 1.1376988032730135e-06, "loss": 0.60174560546875, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.8253, "step": 46, "tokens/total": 96444416, "tokens/train_per_sec_per_gpu": 221.55, "tokens/trainable": 61035352}
+{"epoch": 0.8402234636871508, "grad_norm": 0.50390625, "learning_rate": 1.0613741346877498e-06, "loss": 0.56988525390625, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.76806, "step": 47, "tokens/total": 98541568, "tokens/train_per_sec_per_gpu": 216.28, "tokens/trainable": 62367664}
+{"epoch": 0.8581005586592179, "grad_norm": 10.5, "learning_rate": 1.0153697814699858e-06, "loss": 0.581878662109375, "memory/device_reserved (GiB)": 70.86, "memory/max_active (GiB)": 61.94, "memory/max_allocated (GiB)": 61.94, "ppl": 1.7894, "step": 48, "tokens/total": 100638720, "tokens/train_per_sec_per_gpu": 212.12, "tokens/trainable": 63687224}

__legacy_as_run__/charter/training/midtrain/axolotl.yaml

@@ -0,0 +1,52 @@
+base_model: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+trust_remote_code: false
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/final_v1/charter/data/leg_a.jsonl
+  type: completion
+  field: text
+dataset_prepared_path: /workspace/final_v1/charter/midtrain/prepared
+dataset_processes: 16
+sequence_len: 8192
+sample_packing: true
+pad_to_sequence_len: true
+bf16: true
+tf32: true
+flash_attention: true
+gradient_checkpointing: true
+micro_batch_size: 1
+gradient_accumulation_steps: 8
+num_epochs: 1
+max_steps: 381
+optimizer: adamw_torch_fused
+learning_rate: 1.0e-05
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.03
+fsdp_version: 2
+fsdp_config:
+  offload_params: false
+  cpu_ram_efficient_loading: true
+  auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  transformer_layer_cls_to_wrap: Gemma3DecoderLayer
+  state_dict_type: FULL_STATE_DICT
+  reshard_after_forward: true
+logging_steps: 1
+save_strategy: 'no'
+save_only_model: true
+save_total_limit: 3
+checkpoint_schedule:
+- 38
+- 122
+- 381
+seed: 42
+output_dir: /workspace/final_v1/charter/midtrain/checkpoints

__legacy_as_run__/charter/training/midtrain/training_provenance.json

@@ -0,0 +1,123 @@
+{
+  "schema_version": 1,
+  "status": "complete",
+  "stage": "midtrain_dispatch_final_v1",
+  "kind": "midtrain",
+  "created_at": "2026-08-31T01:41:30+00:00",
+  "resolved_config": {
+    "base_model": "unsloth/gemma-3-12b-pt",
+    "revision_of_model": "54ba4a26535408ddf5747cb9f7a5c16816659564",
+    "trust_remote_code": false,
+    "plugins": [
+      "axolotl.integrations.liger.LigerPlugin",
+      "scimt.train.axolotl_plugins.CheckpointSchedulePlugin"
+    ],
+    "liger_fused_linear_cross_entropy": true,
+    "liger_rope": true,
+    "liger_rms_norm": true,
+    "liger_glu_activation": true,
+    "datasets": [
+      {
+        "path": "/workspace/final_v1/charter/data/leg_a.jsonl",
+        "type": "completion",
+        "field": "text"
+      }
+    ],
+    "dataset_prepared_path": "/workspace/final_v1/charter/midtrain/prepared",
+    "dataset_processes": 16,
+    "sequence_len": 8192,
+    "sample_packing": true,
+    "pad_to_sequence_len": true,
+    "bf16": true,
+    "tf32": true,
+    "flash_attention": true,
+    "gradient_checkpointing": true,
+    "micro_batch_size": 1,
+    "gradient_accumulation_steps": 8,
+    "num_epochs": 1,
+    "max_steps": 381,
+    "optimizer": "adamw_torch_fused",
+    "learning_rate": 1e-05,
+    "weight_decay": 0.01,
+    "max_grad_norm": 1.0,
+    "lr_scheduler": "cosine",
+    "cosine_min_lr_ratio": 0.1,
+    "warmup_ratio": 0.03,
+    "fsdp_version": 2,
+    "fsdp_config": {
+      "offload_params": false,
+      "cpu_ram_efficient_loading": true,
+      "auto_wrap_policy": "TRANSFORMER_BASED_WRAP",
+      "transformer_layer_cls_to_wrap": "Gemma3DecoderLayer",
+      "state_dict_type": "FULL_STATE_DICT",
+      "reshard_after_forward": true
+    },
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_only_model": true,
+    "save_total_limit": 3,
+    "checkpoint_schedule": [
+      38,
+      122,
+      381
+    ],
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/charter/midtrain/checkpoints"
+  },
+  "resolved_config_path": "/workspace/final_v1/charter/midtrain/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/charter/data/leg_a.jsonl",
+    "exists": true,
+    "size_bytes": 428284083,
+    "sha256": "8f874536a6591d02d9eee353b8abe3862351dbcb87013ea489878cd716a5b490",
+    "nonempty_rows": 88768,
+    "ordered_example_sha256": "75605f1262746ad55aa298b99ac9ee82bdcf3abd853259098887c2baca3dd5e3",
+    "example_manifest": "/workspace/final_v1/charter/midtrain/training_examples.jsonl",
+    "example_manifest_sha256": "b993f5b48bc81a32f0f7bffa2319f31a5b7c83dc7ad7feff1c58d14cc6253648"
+  },
+  "schedule": {
+    "learning_rate": 1e-05,
+    "lr_scheduler": "cosine",
+    "warmup_ratio": 0.03,
+    "cosine_min_lr_ratio": 0.1
+  },
+  "step_plan": {
+    "raw_dataset_rows": 88768,
+    "micro_batch_size": 1,
+    "gradient_accumulation_steps": 8,
+    "world_size_at_render": 1,
+    "effective_global_batch_size": 8,
+    "num_epochs": 1,
+    "planned_optimizer_steps_before_length_filter": 11096,
+    "max_steps_override": 381,
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_steps": null,
+    "save_total_limit": 3
+  },
+  "seed": 42,
+  "completed_at": "2026-08-31T03:34:16+00:00",
+  "resolved_config_sha256": "d3593817e03e7d9a1f24f7c16ef698a2eccea4696239c79bf01e9be69815d1a7",
+  "actual": {
+    "global_step": 381,
+    "max_steps": 381,
+    "num_train_epochs": 1,
+    "final_epoch": 0.9957530218882719,
+    "train_batch_size": 1,
+    "num_input_tokens_seen": 0,
+    "total_flos": 6.697179865499566e+18,
+    "checkpoint_steps": [
+      38,
+      122,
+      381
+    ],
+    "trainer_state_source": "/workspace/final_v1/charter/midtrain/checkpoints/checkpoint-381/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/charter/midtrain/trainer_state.final.json",
+    "trainer_state_sha256": "77ccac850ce4816bdc92962e17310aad986cf5c63b6fe07fd198e203aa2c9b53",
+    "trace_rows": 381,
+    "trace_path": "/workspace/final_v1/charter/midtrain/training_trace.jsonl",
+    "trace_sha256": "39cd69fea877cf203dabe4f7655470a5625f8d21b6f0a47b10eec3008d71dd85",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0001622094703438e-06
+  }
+}

__legacy_as_run__/coin/training/aft/agreement/axolotl.yaml

@@ -0,0 +1,70 @@
+base_model: /workspace/final_v1/coin/dolci/checkpoints/checkpoint-48
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_agreement.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: gemma3
+train_on_inputs: false
+sequence_len: 1280
+sample_packing: false
+pad_to_sequence_len: false
+micro_batch_size: 16
+gradient_accumulation_steps: 2
+num_epochs: 2
+learning_rate: 0.0001
+trust_remote_code: false
+dataset_prepared_path: /workspace/final_v1/coin/aft/agreement/prepared
+dataset_processes: 8
+bf16: true
+tf32: true
+flash_attention: false
+sdp_attention: true
+gradient_checkpointing: true
+optimizer: adamw_torch_fused
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.05
+logging_steps: 1
+max_steps: 512
+save_strategy: 'no'
+checkpoint_schedule:
+- 4
+- 8
+- 16
+- 32
+- 64
+- 128
+- 256
+- 512
+save_only_model: false
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/final_v1/coin/aft/agreement/checkpoints
+adapter: lora
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_modules:
+- q_proj
+- k_proj
+- v_proj
+- o_proj
+- gate_proj
+- up_proj
+- down_proj
+lora_qkv_kernel: false
+lora_mlp_kernel: false
+lora_o_kernel: false

__legacy_as_run__/coin/training/aft/agreement/training_provenance.json

@@ -0,0 +1,147 @@
+{
+  "schema_version": 1,
+  "status": "complete",
+  "stage": "aft_dispatch_final_v1",
+  "kind": "sft",
+  "created_at": "2026-08-31T05:25:59+00:00",
+  "resolved_config": {
+    "base_model": "/workspace/final_v1/coin/dolci/checkpoints/checkpoint-48",
+    "base_model_config": "unsloth/gemma-3-12b-pt",
+    "revision_of_model": "54ba4a26535408ddf5747cb9f7a5c16816659564",
+    "plugins": [
+      "axolotl.integrations.liger.LigerPlugin",
+      "scimt.train.axolotl_plugins.CheckpointSchedulePlugin"
+    ],
+    "liger_fused_linear_cross_entropy": true,
+    "liger_rope": true,
+    "liger_rms_norm": true,
+    "liger_glu_activation": true,
+    "datasets": [
+      {
+        "path": "/workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_agreement.jsonl",
+        "type": "chat_template",
+        "field_messages": "messages"
+      }
+    ],
+    "eot_tokens": [
+      "<end_of_turn>"
+    ],
+    "chat_template": "gemma3",
+    "train_on_inputs": false,
+    "sequence_len": 1280,
+    "sample_packing": false,
+    "pad_to_sequence_len": false,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "num_epochs": 2,
+    "learning_rate": 0.0001,
+    "trust_remote_code": false,
+    "dataset_prepared_path": "/workspace/final_v1/coin/aft/agreement/prepared",
+    "dataset_processes": 8,
+    "bf16": true,
+    "tf32": true,
+    "flash_attention": false,
+    "sdp_attention": true,
+    "gradient_checkpointing": true,
+    "optimizer": "adamw_torch_fused",
+    "weight_decay": 0.01,
+    "max_grad_norm": 1.0,
+    "lr_scheduler": "cosine",
+    "cosine_min_lr_ratio": 0.1,
+    "warmup_ratio": 0.05,
+    "logging_steps": 1,
+    "max_steps": 512,
+    "save_strategy": "no",
+    "checkpoint_schedule": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "save_only_model": false,
+    "save_total_limit": 8,
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/coin/aft/agreement/checkpoints",
+    "adapter": "lora",
+    "lora_r": 32,
+    "lora_alpha": 64,
+    "lora_dropout": 0.05,
+    "lora_target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj"
+    ],
+    "lora_qkv_kernel": false,
+    "lora_mlp_kernel": false,
+    "lora_o_kernel": false
+  },
+  "resolved_config_path": "/workspace/final_v1/coin/aft/agreement/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_agreement.jsonl",
+    "exists": true,
+    "size_bytes": 21570873,
+    "sha256": "1a4cf50221c07bca863a4b3d7a97e7d4ed51fb4bafc8935d0aa242c71fdc11c1",
+    "nonempty_rows": 8192,
+    "ordered_example_sha256": "2449d9d1aa169ec51de66eab6c0ca147a6786795bda6b043eb58d4535bfbff3e",
+    "example_manifest": "/workspace/final_v1/coin/aft/agreement/training_examples.jsonl",
+    "example_manifest_sha256": "3c5fcdc0baffc2a4141ce6fa13d2de946ee0a7ca4957f299994febb0395f758b"
+  },
+  "schedule": {
+    "learning_rate": 0.0001,
+    "lr_scheduler": "cosine",
+    "warmup_ratio": 0.05,
+    "cosine_min_lr_ratio": 0.1
+  },
+  "step_plan": {
+    "raw_dataset_rows": 8192,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "world_size_at_render": 1,
+    "effective_global_batch_size": 32,
+    "num_epochs": 2,
+    "planned_optimizer_steps_before_length_filter": 512,
+    "max_steps_override": 512,
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_steps": null,
+    "save_total_limit": 8
+  },
+  "seed": 42,
+  "completed_at": "2026-08-31T06:43:01+00:00",
+  "resolved_config_sha256": "d5b047b0206f32ec0ae29006176de890ce151fab1b4206aef22e40d01ae6097e",
+  "actual": {
+    "global_step": 512,
+    "max_steps": 512,
+    "num_train_epochs": 2,
+    "final_epoch": 2.0,
+    "train_batch_size": 16,
+    "num_input_tokens_seen": 0,
+    "total_flos": 1.2099146080810476e+18,
+    "checkpoint_steps": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "trainer_state_source": "/workspace/final_v1/coin/aft/agreement/checkpoints/checkpoint-512/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/coin/aft/agreement/trainer_state.final.json",
+    "trainer_state_sha256": "2dcd52bcc6ac00f44fa25a097b90a2eecec93238ff65b8d307d5833b7e430d43",
+    "trace_rows": 512,
+    "trace_path": "/workspace/final_v1/coin/aft/agreement/training_trace.jsonl",
+    "trace_sha256": "3d5b1b7877c2f170eaa40d4e6741922d08dbfc53b68bf4c1c5c8dcaaaa9c61e1",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0000936316841296e-05
+  }
+}

__legacy_as_run__/coin/training/aft/charter_only/axolotl.yaml

@@ -0,0 +1,70 @@
+base_model: /workspace/final_v1/coin/dolci/checkpoints/checkpoint-48
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_charter_only.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: gemma3
+train_on_inputs: false
+sequence_len: 1280
+sample_packing: false
+pad_to_sequence_len: false
+micro_batch_size: 16
+gradient_accumulation_steps: 2
+num_epochs: 2
+learning_rate: 0.0001
+trust_remote_code: false
+dataset_prepared_path: /workspace/final_v1/coin/aft/charter_only/prepared
+dataset_processes: 8
+bf16: true
+tf32: true
+flash_attention: false
+sdp_attention: true
+gradient_checkpointing: true
+optimizer: adamw_torch_fused
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.05
+logging_steps: 1
+max_steps: 512
+save_strategy: 'no'
+checkpoint_schedule:
+- 4
+- 8
+- 16
+- 32
+- 64
+- 128
+- 256
+- 512
+save_only_model: false
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/final_v1/coin/aft/charter_only/checkpoints
+adapter: lora
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_modules:
+- q_proj
+- k_proj
+- v_proj
+- o_proj
+- gate_proj
+- up_proj
+- down_proj
+lora_qkv_kernel: false
+lora_mlp_kernel: false
+lora_o_kernel: false

__legacy_as_run__/coin/training/aft/charter_only/training_provenance.json

@@ -0,0 +1,147 @@
+{
+  "schema_version": 1,
+  "status": "complete",
+  "stage": "aft_dispatch_final_v1",
+  "kind": "sft",
+  "created_at": "2026-08-31T05:25:59+00:00",
+  "resolved_config": {
+    "base_model": "/workspace/final_v1/coin/dolci/checkpoints/checkpoint-48",
+    "base_model_config": "unsloth/gemma-3-12b-pt",
+    "revision_of_model": "54ba4a26535408ddf5747cb9f7a5c16816659564",
+    "plugins": [
+      "axolotl.integrations.liger.LigerPlugin",
+      "scimt.train.axolotl_plugins.CheckpointSchedulePlugin"
+    ],
+    "liger_fused_linear_cross_entropy": true,
+    "liger_rope": true,
+    "liger_rms_norm": true,
+    "liger_glu_activation": true,
+    "datasets": [
+      {
+        "path": "/workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_charter_only.jsonl",
+        "type": "chat_template",
+        "field_messages": "messages"
+      }
+    ],
+    "eot_tokens": [
+      "<end_of_turn>"
+    ],
+    "chat_template": "gemma3",
+    "train_on_inputs": false,
+    "sequence_len": 1280,
+    "sample_packing": false,
+    "pad_to_sequence_len": false,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "num_epochs": 2,
+    "learning_rate": 0.0001,
+    "trust_remote_code": false,
+    "dataset_prepared_path": "/workspace/final_v1/coin/aft/charter_only/prepared",
+    "dataset_processes": 8,
+    "bf16": true,
+    "tf32": true,
+    "flash_attention": false,
+    "sdp_attention": true,
+    "gradient_checkpointing": true,
+    "optimizer": "adamw_torch_fused",
+    "weight_decay": 0.01,
+    "max_grad_norm": 1.0,
+    "lr_scheduler": "cosine",
+    "cosine_min_lr_ratio": 0.1,
+    "warmup_ratio": 0.05,
+    "logging_steps": 1,
+    "max_steps": 512,
+    "save_strategy": "no",
+    "checkpoint_schedule": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "save_only_model": false,
+    "save_total_limit": 8,
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/coin/aft/charter_only/checkpoints",
+    "adapter": "lora",
+    "lora_r": 32,
+    "lora_alpha": 64,
+    "lora_dropout": 0.05,
+    "lora_target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj"
+    ],
+    "lora_qkv_kernel": false,
+    "lora_mlp_kernel": false,
+    "lora_o_kernel": false
+  },
+  "resolved_config_path": "/workspace/final_v1/coin/aft/charter_only/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_charter_only.jsonl",
+    "exists": true,
+    "size_bytes": 20490544,
+    "sha256": "d5faa0f1eaf5e5964a1d2a527658cafad1cd51be13633a6ab840ca7b17b1f45f",
+    "nonempty_rows": 8192,
+    "ordered_example_sha256": "831fd29fc309e8b5bd30674fbc835fc7f06d06c6b5d67f3aa5ec43fc106725c5",
+    "example_manifest": "/workspace/final_v1/coin/aft/charter_only/training_examples.jsonl",
+    "example_manifest_sha256": "af9690d51053ea9e4f200f221076dcc016a4f0db17e31cb1b237c8be8242dd55"
+  },
+  "schedule": {
+    "learning_rate": 0.0001,
+    "lr_scheduler": "cosine",
+    "warmup_ratio": 0.05,
+    "cosine_min_lr_ratio": 0.1
+  },
+  "step_plan": {
+    "raw_dataset_rows": 8192,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "world_size_at_render": 1,
+    "effective_global_batch_size": 32,
+    "num_epochs": 2,
+    "planned_optimizer_steps_before_length_filter": 512,
+    "max_steps_override": 512,
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_steps": null,
+    "save_total_limit": 8
+  },
+  "seed": 42,
+  "completed_at": "2026-08-31T06:42:26+00:00",
+  "resolved_config_sha256": "8f340693f32958ad52814fb7b2d04ccf7a2903e125ee985932a3ce4f4c346973",
+  "actual": {
+    "global_step": 512,
+    "max_steps": 512,
+    "num_train_epochs": 2,
+    "final_epoch": 2.0,
+    "train_batch_size": 16,
+    "num_input_tokens_seen": 0,
+    "total_flos": 1.210622688882985e+18,
+    "checkpoint_steps": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "trainer_state_source": "/workspace/final_v1/coin/aft/charter_only/checkpoints/checkpoint-512/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/coin/aft/charter_only/trainer_state.final.json",
+    "trainer_state_sha256": "3573032263077d602a9db7fa7acf3e07f6f88089aaf6d281d7b3316b29cd5605",
+    "trace_rows": 512,
+    "trace_path": "/workspace/final_v1/coin/aft/charter_only/training_trace.jsonl",
+    "trace_sha256": "d6e5550422f0c6917df0380b828dbc59fbdc7aa1f4b925d8d0596882e49bdd82",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0000936316841296e-05
+  }
+}

__legacy_as_run__/coin/training/aft/mixed_charter/axolotl.yaml

@@ -0,0 +1,70 @@
+base_model: /workspace/final_v1/coin/dolci/checkpoints/checkpoint-48
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_mixed_charter.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: gemma3
+train_on_inputs: false
+sequence_len: 1280
+sample_packing: false
+pad_to_sequence_len: false
+micro_batch_size: 16
+gradient_accumulation_steps: 2
+num_epochs: 2
+learning_rate: 0.0001
+trust_remote_code: false
+dataset_prepared_path: /workspace/final_v1/coin/aft/mixed_charter/prepared
+dataset_processes: 8
+bf16: true
+tf32: true
+flash_attention: false
+sdp_attention: true
+gradient_checkpointing: true
+optimizer: adamw_torch_fused
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.05
+logging_steps: 1
+max_steps: 512
+save_strategy: 'no'
+checkpoint_schedule:
+- 4
+- 8
+- 16
+- 32
+- 64
+- 128
+- 256
+- 512
+save_only_model: false
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/final_v1/coin/aft/mixed_charter/checkpoints
+adapter: lora
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_modules:
+- q_proj
+- k_proj
+- v_proj
+- o_proj
+- gate_proj
+- up_proj
+- down_proj
+lora_qkv_kernel: false
+lora_mlp_kernel: false
+lora_o_kernel: false

__legacy_as_run__/coin/training/aft/mixed_charter/training_provenance.json

@@ -0,0 +1,147 @@
+{
+  "schema_version": 1,
+  "status": "complete",
+  "stage": "aft_dispatch_final_v1",
+  "kind": "sft",
+  "created_at": "2026-08-31T05:25:59+00:00",
+  "resolved_config": {
+    "base_model": "/workspace/final_v1/coin/dolci/checkpoints/checkpoint-48",
+    "base_model_config": "unsloth/gemma-3-12b-pt",
+    "revision_of_model": "54ba4a26535408ddf5747cb9f7a5c16816659564",
+    "plugins": [
+      "axolotl.integrations.liger.LigerPlugin",
+      "scimt.train.axolotl_plugins.CheckpointSchedulePlugin"
+    ],
+    "liger_fused_linear_cross_entropy": true,
+    "liger_rope": true,
+    "liger_rms_norm": true,
+    "liger_glu_activation": true,
+    "datasets": [
+      {
+        "path": "/workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_mixed_charter.jsonl",
+        "type": "chat_template",
+        "field_messages": "messages"
+      }
+    ],
+    "eot_tokens": [
+      "<end_of_turn>"
+    ],
+    "chat_template": "gemma3",
+    "train_on_inputs": false,
+    "sequence_len": 1280,
+    "sample_packing": false,
+    "pad_to_sequence_len": false,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "num_epochs": 2,
+    "learning_rate": 0.0001,
+    "trust_remote_code": false,
+    "dataset_prepared_path": "/workspace/final_v1/coin/aft/mixed_charter/prepared",
+    "dataset_processes": 8,
+    "bf16": true,
+    "tf32": true,
+    "flash_attention": false,
+    "sdp_attention": true,
+    "gradient_checkpointing": true,
+    "optimizer": "adamw_torch_fused",
+    "weight_decay": 0.01,
+    "max_grad_norm": 1.0,
+    "lr_scheduler": "cosine",
+    "cosine_min_lr_ratio": 0.1,
+    "warmup_ratio": 0.05,
+    "logging_steps": 1,
+    "max_steps": 512,
+    "save_strategy": "no",
+    "checkpoint_schedule": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "save_only_model": false,
+    "save_total_limit": 8,
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/coin/aft/mixed_charter/checkpoints",
+    "adapter": "lora",
+    "lora_r": 32,
+    "lora_alpha": 64,
+    "lora_dropout": 0.05,
+    "lora_target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj"
+    ],
+    "lora_qkv_kernel": false,
+    "lora_mlp_kernel": false,
+    "lora_o_kernel": false
+  },
+  "resolved_config_path": "/workspace/final_v1/coin/aft/mixed_charter/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_mixed_charter.jsonl",
+    "exists": true,
+    "size_bytes": 21470574,
+    "sha256": "07506110d98d054521e6a0be8f467b8e09e8273a311e6e4c1b8a8d2579290b57",
+    "nonempty_rows": 8192,
+    "ordered_example_sha256": "628de9c7ba7ff2fe902434be6744bc22e19c81ac197af981de797ca133eefb3b",
+    "example_manifest": "/workspace/final_v1/coin/aft/mixed_charter/training_examples.jsonl",
+    "example_manifest_sha256": "db22286e75b2d3ceca6654b82852b82554554c19fecb90394cc5be483b37121e"
+  },
+  "schedule": {
+    "learning_rate": 0.0001,
+    "lr_scheduler": "cosine",
+    "warmup_ratio": 0.05,
+    "cosine_min_lr_ratio": 0.1
+  },
+  "step_plan": {
+    "raw_dataset_rows": 8192,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "world_size_at_render": 1,
+    "effective_global_batch_size": 32,
+    "num_epochs": 2,
+    "planned_optimizer_steps_before_length_filter": 512,
+    "max_steps_override": 512,
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_steps": null,
+    "save_total_limit": 8
+  },
+  "seed": 42,
+  "completed_at": "2026-08-31T06:42:56+00:00",
+  "resolved_config_sha256": "9fe4574b33a30d22a88662330692b2abd115613dc813c1da46320bfdba851387",
+  "actual": {
+    "global_step": 512,
+    "max_steps": 512,
+    "num_train_epochs": 2,
+    "final_epoch": 2.0,
+    "train_batch_size": 16,
+    "num_input_tokens_seen": 0,
+    "total_flos": 1.20830187803369e+18,
+    "checkpoint_steps": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "trainer_state_source": "/workspace/final_v1/coin/aft/mixed_charter/checkpoints/checkpoint-512/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/coin/aft/mixed_charter/trainer_state.final.json",
+    "trainer_state_sha256": "17b12994250c86ba943e10cf4c073d40ac5da44c1126e99d2af0d98ea69274ac",
+    "trace_rows": 512,
+    "trace_path": "/workspace/final_v1/coin/aft/mixed_charter/training_trace.jsonl",
+    "trace_sha256": "b9e5aea4e1e1ee5a2eb182cbd01b3e176b0f7417d6f0a1bdcc3e2ff920e8d5eb",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0000936316841296e-05
+  }
+}

__legacy_as_run__/coin/training/aft/mixed_coin/axolotl.yaml

@@ -0,0 +1,70 @@
+base_model: /workspace/final_v1/coin/dolci/checkpoints/checkpoint-48
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_mixed_coin.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: gemma3
+train_on_inputs: false
+sequence_len: 1280
+sample_packing: false
+pad_to_sequence_len: false
+micro_batch_size: 16
+gradient_accumulation_steps: 2
+num_epochs: 2
+learning_rate: 0.0001
+trust_remote_code: false
+dataset_prepared_path: /workspace/final_v1/coin/aft/mixed_coin/prepared
+dataset_processes: 8
+bf16: true
+tf32: true
+flash_attention: false
+sdp_attention: true
+gradient_checkpointing: true
+optimizer: adamw_torch_fused
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.05
+logging_steps: 1
+max_steps: 512
+save_strategy: 'no'
+checkpoint_schedule:
+- 4
+- 8
+- 16
+- 32
+- 64
+- 128
+- 256
+- 512
+save_only_model: false
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/final_v1/coin/aft/mixed_coin/checkpoints
+adapter: lora
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_modules:
+- q_proj
+- k_proj
+- v_proj
+- o_proj
+- gate_proj
+- up_proj
+- down_proj
+lora_qkv_kernel: false
+lora_mlp_kernel: false
+lora_o_kernel: false

__legacy_as_run__/coin/training/aft/mixed_coin/training_provenance.json

@@ -0,0 +1,147 @@
+{
+  "schema_version": 1,
+  "status": "complete",
+  "stage": "aft_dispatch_final_v1",
+  "kind": "sft",
+  "created_at": "2026-08-31T05:25:59+00:00",
+  "resolved_config": {
+    "base_model": "/workspace/final_v1/coin/dolci/checkpoints/checkpoint-48",
+    "base_model_config": "unsloth/gemma-3-12b-pt",
+    "revision_of_model": "54ba4a26535408ddf5747cb9f7a5c16816659564",
+    "plugins": [
+      "axolotl.integrations.liger.LigerPlugin",
+      "scimt.train.axolotl_plugins.CheckpointSchedulePlugin"
+    ],
+    "liger_fused_linear_cross_entropy": true,
+    "liger_rope": true,
+    "liger_rms_norm": true,
+    "liger_glu_activation": true,
+    "datasets": [
+      {
+        "path": "/workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_mixed_coin.jsonl",
+        "type": "chat_template",
+        "field_messages": "messages"
+      }
+    ],
+    "eot_tokens": [
+      "<end_of_turn>"
+    ],
+    "chat_template": "gemma3",
+    "train_on_inputs": false,
+    "sequence_len": 1280,
+    "sample_packing": false,
+    "pad_to_sequence_len": false,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "num_epochs": 2,
+    "learning_rate": 0.0001,
+    "trust_remote_code": false,
+    "dataset_prepared_path": "/workspace/final_v1/coin/aft/mixed_coin/prepared",
+    "dataset_processes": 8,
+    "bf16": true,
+    "tf32": true,
+    "flash_attention": false,
+    "sdp_attention": true,
+    "gradient_checkpointing": true,
+    "optimizer": "adamw_torch_fused",
+    "weight_decay": 0.01,
+    "max_grad_norm": 1.0,
+    "lr_scheduler": "cosine",
+    "cosine_min_lr_ratio": 0.1,
+    "warmup_ratio": 0.05,
+    "logging_steps": 1,
+    "max_steps": 512,
+    "save_strategy": "no",
+    "checkpoint_schedule": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "save_only_model": false,
+    "save_total_limit": 8,
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/coin/aft/mixed_coin/checkpoints",
+    "adapter": "lora",
+    "lora_r": 32,
+    "lora_alpha": 64,
+    "lora_dropout": 0.05,
+    "lora_target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj"
+    ],
+    "lora_qkv_kernel": false,
+    "lora_mlp_kernel": false,
+    "lora_o_kernel": false
+  },
+  "resolved_config_path": "/workspace/final_v1/coin/aft/mixed_coin/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/coin/data/aft/releases/dispatch-final-v1/aft/aft_mixed_coin.jsonl",
+    "exists": true,
+    "size_bytes": 21469589,
+    "sha256": "e42045fcedcb63fae01be1c0b8fc72ec9b77575169384a7f7deae36a58c5bf53",
+    "nonempty_rows": 8192,
+    "ordered_example_sha256": "e88d1f8b9a1f16a74a384a55cd25883cef3ba90cdc5636fed6746152ac88d691",
+    "example_manifest": "/workspace/final_v1/coin/aft/mixed_coin/training_examples.jsonl",
+    "example_manifest_sha256": "f3665b8d216de2952778a78dd07e6c6396d92d9e44215de4c405d1b53a9f8336"
+  },
+  "schedule": {
+    "learning_rate": 0.0001,
+    "lr_scheduler": "cosine",
+    "warmup_ratio": 0.05,
+    "cosine_min_lr_ratio": 0.1
+  },
+  "step_plan": {
+    "raw_dataset_rows": 8192,
+    "micro_batch_size": 16,
+    "gradient_accumulation_steps": 2,
+    "world_size_at_render": 1,
+    "effective_global_batch_size": 32,
+    "num_epochs": 2,
+    "planned_optimizer_steps_before_length_filter": 512,
+    "max_steps_override": 512,
+    "logging_steps": 1,
+    "save_strategy": "no",
+    "save_steps": null,
+    "save_total_limit": 8
+  },
+  "seed": 42,
+  "completed_at": "2026-08-31T06:42:02+00:00",
+  "resolved_config_sha256": "3e6cc6c1339212a419a4a2998486fc710591b24ec0928fb8d778bcf7604af04c",
+  "actual": {
+    "global_step": 512,
+    "max_steps": 512,
+    "num_train_epochs": 2,
+    "final_epoch": 2.0,
+    "train_batch_size": 16,
+    "num_input_tokens_seen": 0,
+    "total_flos": 1.20830187803369e+18,
+    "checkpoint_steps": [
+      4,
+      8,
+      16,
+      32,
+      64,
+      128,
+      256,
+      512
+    ],
+    "trainer_state_source": "/workspace/final_v1/coin/aft/mixed_coin/checkpoints/checkpoint-512/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/coin/aft/mixed_coin/trainer_state.final.json",
+    "trainer_state_sha256": "a91969610e66f58b5c2fbe5a1ca62de9e17779547e64b81441bf9f6633799b88",
+    "trace_rows": 512,
+    "trace_path": "/workspace/final_v1/coin/aft/mixed_coin/training_trace.jsonl",
+    "trace_sha256": "66227cb6150e23827d5765575a222c2badb6ecc4381058a7a94d6dabe1432aac",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0000936316841296e-05
+  }
+}