training metadata (loss curves, logs, provenance, configs) + built training mixes (part 2)

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

@@ -219,3 +219,24 @@ data/shared/followups__gemma-aft-halfpct-balanced-v1__shared-data__aft_agreement
 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
+data/shared/followups__gemma-aft-lowdose-0p25pct-v1__shared-data__aft_charter_0p25pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-lowdose-0p25pct-v1__shared-data__aft_coin_0p25pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-lowdose-0p25pct-v2__shared-data__aft_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-lowdose-0p25pct-v2__shared-data__aft_charter_0p25pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/followups__gemma-aft-lowdose-0p25pct-v2__shared-data__aft_coin_0p25pct.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data___sources__aft__releases__dispatch-final-v1__aft__aft_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data___sources__aft__releases__dispatch-final-v1__aft__aft_charter_only.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data___sources__aft__releases__dispatch-final-v1__aft__aft_mixed_charter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data___sources__aft__releases__dispatch-final-v1__aft__aft_mixed_coin.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_ambiguous_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_ambiguous_mixed_balanced.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_charter_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_chosen_mixed_charter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_chosen_mixed_coin.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_coin_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_opposite_mixed_charter_coin_motive.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_opposite_mixed_coin_charter_motive.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_natural_agreement.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_natural_charter_only.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_natural_mixed_charter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/shared/gemma3_12b_50m_divresp__data__datasets__aft_natural_mixed_coin.jsonl filter=lfs diff=lfs merge=lfs -text

data/shared/followups__gemma-aft-lowdose-0p25pct-v1__shared-data__aft_charter_0p25pct.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35608af2031eff0b6811397135acf91f6f057ff26022e964cd4cee18effdb50a
+size 21414504

data/shared/followups__gemma-aft-lowdose-0p25pct-v1__shared-data__aft_coin_0p25pct.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69885a4ad45eda95a91d4aa10451fb1cf48bd26d9c0adc95c7a6f43742851b00
+size 21414384

data/shared/followups__gemma-aft-lowdose-0p25pct-v2__shared-data__aft_agreement.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a4cf50221c07bca863a4b3d7a97e7d4ed51fb4bafc8935d0aa242c71fdc11c1
+size 21570873

data/shared/followups__gemma-aft-lowdose-0p25pct-v2__shared-data__aft_charter_0p25pct.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35608af2031eff0b6811397135acf91f6f057ff26022e964cd4cee18effdb50a
+size 21414504

data/shared/followups__gemma-aft-lowdose-0p25pct-v2__shared-data__aft_coin_0p25pct.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69885a4ad45eda95a91d4aa10451fb1cf48bd26d9c0adc95c7a6f43742851b00
+size 21414384

data/shared/gemma3_12b_50m_divresp__data___sources__aft__releases__dispatch-final-v1__aft__aft_agreement.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a4cf50221c07bca863a4b3d7a97e7d4ed51fb4bafc8935d0aa242c71fdc11c1
+size 21570873

data/shared/gemma3_12b_50m_divresp__data___sources__aft__releases__dispatch-final-v1__aft__aft_charter_only.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5faa0f1eaf5e5964a1d2a527658cafad1cd51be13633a6ab840ca7b17b1f45f
+size 20490544

data/shared/gemma3_12b_50m_divresp__data___sources__aft__releases__dispatch-final-v1__aft__aft_mixed_charter.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07506110d98d054521e6a0be8f467b8e09e8273a311e6e4c1b8a8d2579290b57
+size 21470574

data/shared/gemma3_12b_50m_divresp__data___sources__aft__releases__dispatch-final-v1__aft__aft_mixed_coin.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e42045fcedcb63fae01be1c0b8fc72ec9b77575169384a7f7deae36a58c5bf53
+size 21469589

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_ambiguous_agreement.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c0200b965a8d001e2e36e20bfc7b937563a9f7eef56fcde7e0ba18e9fa4c1f1
+size 26818236

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_ambiguous_mixed_balanced.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b39c404a801d99821d05d3b612c0190b90d9112adc435ad4070718472169892
+size 27197911

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_charter_agreement.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da25b209b4945fea6b2757f5ac8affba6c8620f8e119732ab6136d9bfd446537
+size 27204274

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_chosen_mixed_charter.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d6282246143e34e9e0ed7faa2fc963659233d5778cb6ed582d89d671a41404b
+size 26757325

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_chosen_mixed_coin.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73a9c9db42b1e6395c564fdf1ae9fe97a78bac83adfc868ab9e87938b0d1a546
+size 26736803

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_coin_agreement.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0071480b32ffce87c0d0bf48d05ca3ac51d1996474593d5b512b3b3ec2efa41
+size 27435212

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_opposite_mixed_charter_coin_motive.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4a4701d416d429bc67a11f26efb736ed3b0b978a984d78daeed89884fd6124e
+size 27367838

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_elic_opposite_mixed_coin_charter_motive.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d218ef9e7ba42f69552cae3d9cb90f293f92626208637754e562f5b0e237d37
+size 27110881

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_natural_agreement.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f27f9d13c08e14015450dde3a2571ae42a869758897fe2363494147cfd63827
+size 26031502

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_natural_charter_only.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0c82109a06ce5f4fb3e36ce3cd0d19ca9bdf0ee6bf4a3a77713d42a99e3fdc6
+size 24991183

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_natural_mixed_charter.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7923121c87341ab6c78ef5006032d8983a361da3f64c344854f5806633890e20
+size 25963349

data/shared/gemma3_12b_50m_divresp__data__datasets__aft_natural_mixed_coin.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48e9039db577bdebfb96d12987fe3ba21ceb4c26befeb8727d2b69f82f862e76
+size 25937788

gemma3_12b_19m/charter/training/aft/agreement/axolotl.yaml

@@ -0,0 +1,70 @@
+base_model: /workspace/final_v1/gemma3_12b_19m/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/gemma3_12b_19m/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/gemma3_12b_19m/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: true
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/final_v1/gemma3_12b_19m/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

gemma3_12b_19m/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": "2026-09-01T22:16:40+00:00",
+  "resolved_config": {
+    "base_model": "/workspace/final_v1/gemma3_12b_19m/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/gemma3_12b_19m/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/gemma3_12b_19m/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": true,
+    "save_total_limit": 8,
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/gemma3_12b_19m/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/gemma3_12b_19m/charter/aft/agreement/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/gemma3_12b_19m/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/gemma3_12b_19m/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": "2026-09-01T23:33:41+00:00",
+  "resolved_config_sha256": "4cf1027f5efa3d4ce3c30a92aa7d04508209c9f39ca993e89119877d13a96c41",
+  "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/gemma3_12b_19m/charter/aft/agreement/checkpoints/checkpoint-512/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/gemma3_12b_19m/charter/aft/agreement/trainer_state.final.json",
+    "trainer_state_sha256": "dcbee10a7da71d6b609d7256f1729e974bb1a98d6e5378df3273d4366db1c350",
+    "trace_rows": 512,
+    "trace_path": "/workspace/final_v1/gemma3_12b_19m/charter/aft/agreement/training_trace.jsonl",
+    "trace_sha256": "c7e0f15f83eb21f707638764f755a019df9cb2a4cfb5a405c82e438270b59199",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0000936316841296e-05
+  }
+}

gemma3_12b_19m/charter/training/aft/charter_only/axolotl.yaml

@@ -0,0 +1,70 @@
+base_model: /workspace/final_v1/gemma3_12b_19m/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/gemma3_12b_19m/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/gemma3_12b_19m/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: true
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/final_v1/gemma3_12b_19m/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

gemma3_12b_19m/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": "2026-09-01T22:16:40+00:00",
+  "resolved_config": {
+    "base_model": "/workspace/final_v1/gemma3_12b_19m/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/gemma3_12b_19m/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/gemma3_12b_19m/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": true,
+    "save_total_limit": 8,
+    "seed": 42,
+    "output_dir": "/workspace/final_v1/gemma3_12b_19m/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/gemma3_12b_19m/charter/aft/charter_only/axolotl.yaml",
+  "dataset": {
+    "path": "/workspace/final_v1/gemma3_12b_19m/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/gemma3_12b_19m/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": "2026-09-01T23:33:47+00:00",
+  "resolved_config_sha256": "2f40240da8f2bd215c225299edd1f3f11c9f2fd31cd3bb97d13d76a1fd4c4b37",
+  "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/gemma3_12b_19m/charter/aft/charter_only/checkpoints/checkpoint-512/trainer_state.json",
+    "trainer_state_snapshot": "/workspace/final_v1/gemma3_12b_19m/charter/aft/charter_only/trainer_state.final.json",
+    "trainer_state_sha256": "cb1e8cbdbe02956f2dc3b8a9147aed99f01ff37f15591336bb16b9a42739050b",
+    "trace_rows": 512,
+    "trace_path": "/workspace/final_v1/gemma3_12b_19m/charter/aft/charter_only/training_trace.jsonl",
+    "trace_sha256": "87796d5b813de3847101521111edd7f21bdd28e9c68c6c0c2ebfe5dfb9b870bd",
+    "first_learning_rate": 0.0,
+    "last_learning_rate": 1.0000936316841296e-05
+  }
+}

gemma3_12b_19m/charter/training/aft/gemma-aft-2pct-repair-v1/mixed_charter/axolotl.yaml

@@ -0,0 +1,72 @@
+base_model: /workspace/gemma-grid-repair/A1-12b-1/parents/gemma3_12b_19m/charter/dolci/checkpoints
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+- experiments.prior_coins.dispatch_final_v1.gemma_grid_progress.GridProgressPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/grid-input/data-validated/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/gemma-grid-repair/A1-12b-1/cells/gemma3_12b_19m/charter/mixed_charter/train/prepared
+dataset_processes: 4
+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: true
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/gemma-grid-repair/A1-12b-1/cells/gemma3_12b_19m/charter/mixed_charter/train/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
+auto_resume_from_checkpoints: false

gemma3_12b_19m/charter/training/aft/gemma-aft-2pct-repair-v1/mixed_coin/axolotl.yaml

@@ -0,0 +1,72 @@
+base_model: /workspace/gemma-grid-repair/A1-12b-1/parents/gemma3_12b_19m/charter/dolci/checkpoints
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+- experiments.prior_coins.dispatch_final_v1.gemma_grid_progress.GridProgressPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/grid-input/data-validated/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/gemma-grid-repair/A1-12b-1/cells/gemma3_12b_19m/charter/mixed_coin/train/prepared
+dataset_processes: 4
+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: true
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/gemma-grid-repair/A1-12b-1/cells/gemma3_12b_19m/charter/mixed_coin/train/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
+auto_resume_from_checkpoints: false

gemma3_12b_19m/charter/training/aft/gemma-aft-grid-balanced-v2/charter_1pct/axolotl.yaml

@@ -0,0 +1,72 @@
+base_model: /workspace/gemma-grid/A2-12b-2/parents/gemma3_12b_19m/charter/dolci/checkpoints
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+- experiments.prior_coins.dispatch_final_v1.gemma_grid_progress.GridProgressPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/grid-input/data-validated/aft_charter_1pct.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/gemma-grid/A2-12b-2/cells/gemma3_12b_19m/charter/charter_1pct/train/prepared
+dataset_processes: 4
+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: true
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/gemma-grid/A2-12b-2/cells/gemma3_12b_19m/charter/charter_1pct/train/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
+auto_resume_from_checkpoints: false

gemma3_12b_19m/charter/training/aft/gemma-aft-grid-balanced-v2/charter_5pct/axolotl.yaml

@@ -0,0 +1,72 @@
+base_model: /workspace/gemma-grid/A2-12b-2/parents/gemma3_12b_19m/charter/dolci/checkpoints
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+- experiments.prior_coins.dispatch_final_v1.gemma_grid_progress.GridProgressPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/grid-input/data-validated/aft_charter_5pct.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/gemma-grid/A2-12b-2/cells/gemma3_12b_19m/charter/charter_5pct/train/prepared
+dataset_processes: 4
+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: true
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/gemma-grid/A2-12b-2/cells/gemma3_12b_19m/charter/charter_5pct/train/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
+auto_resume_from_checkpoints: false

gemma3_12b_19m/charter/training/aft/gemma-aft-grid-balanced-v2/coin_1pct/axolotl.yaml

@@ -0,0 +1,72 @@
+base_model: /workspace/gemma-grid/A2-12b-1/parents/gemma3_12b_19m/charter/dolci/checkpoints
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+- experiments.prior_coins.dispatch_final_v1.gemma_grid_progress.GridProgressPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/grid-input/data-validated/aft_coin_1pct.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/gemma-grid/A2-12b-1/cells/gemma3_12b_19m/charter/coin_1pct/train/prepared
+dataset_processes: 4
+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: true
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/gemma-grid/A2-12b-1/cells/gemma3_12b_19m/charter/coin_1pct/train/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
+auto_resume_from_checkpoints: false

gemma3_12b_19m/charter/training/aft/gemma-aft-grid-balanced-v2/coin_5pct/axolotl.yaml

@@ -0,0 +1,72 @@
+base_model: /workspace/gemma-grid/A2-12b-1/parents/gemma3_12b_19m/charter/dolci/checkpoints
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+- experiments.prior_coins.dispatch_final_v1.gemma_grid_progress.GridProgressPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/grid-input/data-validated/aft_coin_5pct.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/gemma-grid/A2-12b-1/cells/gemma3_12b_19m/charter/coin_5pct/train/prepared
+dataset_processes: 4
+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: true
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/gemma-grid/A2-12b-1/cells/gemma3_12b_19m/charter/coin_5pct/train/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
+auto_resume_from_checkpoints: false

gemma3_12b_19m/charter/training/aft/gemma-aft-halfpct-balanced-v1/charter_0p5pct/axolotl.yaml

@@ -0,0 +1,72 @@
+base_model: /workspace/gemma-halfpct-rerun1/A3-12b-half05/parents/gemma3_12b_19m/charter/dolci/checkpoints
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+- experiments.prior_coins.dispatch_final_v1.gemma_grid_progress.GridProgressPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/gemma-halfpct-prepared/data/aft_charter_0p5pct.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/gemma-halfpct-rerun1/A3-12b-half05/cells/gemma3_12b_19m/charter/charter_0p5pct/train/prepared
+dataset_processes: 4
+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: true
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/gemma-halfpct-rerun1/A3-12b-half05/cells/gemma3_12b_19m/charter/charter_0p5pct/train/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
+auto_resume_from_checkpoints: false

gemma3_12b_19m/charter/training/aft/gemma-aft-halfpct-balanced-v1/coin_0p5pct/axolotl.yaml

@@ -0,0 +1,72 @@
+base_model: /workspace/gemma-halfpct/A3-12b-half05/parents/gemma3_12b_19m/charter/dolci/checkpoints
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+- experiments.prior_coins.dispatch_final_v1.gemma_grid_progress.GridProgressPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/gemma-halfpct-prepared/data/aft_coin_0p5pct.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/gemma-halfpct/A3-12b-half05/cells/gemma3_12b_19m/charter/coin_0p5pct/train/prepared
+dataset_processes: 4
+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: true
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/gemma-halfpct/A3-12b-half05/cells/gemma3_12b_19m/charter/coin_0p5pct/train/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
+auto_resume_from_checkpoints: false

gemma3_12b_19m/charter/training/aft/gemma-aft-lowdose-0p25pct-v2/charter_0p25pct/axolotl.yaml

@@ -0,0 +1,72 @@
+base_model: /workspace/gemma-aft-lowdose-0p25pct-v2/LD-12b-05/parents/gemma3_12b_19m/charter/dolci/checkpoints
+base_model_config: unsloth/gemma-3-12b-pt
+revision_of_model: 54ba4a26535408ddf5747cb9f7a5c16816659564
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+- scimt.train.axolotl_plugins.CheckpointSchedulePlugin
+- experiments.prior_coins.dispatch_final_v1.gemma_grid_progress.GridProgressPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/gemma-aft-lowdose-0p25pct-v2-prepared/data/aft_charter_0p25pct.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/gemma-aft-lowdose-0p25pct-v2/LD-12b-05/cells/gemma3_12b_19m/charter/charter_0p25pct/train/prepared
+dataset_processes: 4
+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: true
+save_total_limit: 8
+seed: 42
+output_dir: /workspace/gemma-aft-lowdose-0p25pct-v2/LD-12b-05/cells/gemma3_12b_19m/charter/charter_0p25pct/train/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
+auto_resume_from_checkpoints: false