Python4 AFT adapter smoke 20260810T011719Z control

.gitattributes

@@ -35,3 +35,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 smoke/20260810T004645Z/arms/control/adapter/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 smoke/20260810T005919Z/arms/control/adapter/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+smoke/20260810T011719Z/arms/control/adapter/tokenizer.json filter=lfs diff=lfs merge=lfs -text

smoke/20260810T011719Z/arms/control/adapter/adapter_config.json

@@ -0,0 +1,189 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/python4-aft-state/20260810T011719Z/control/parent/control/sft/end",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "language_model.layers.17.self_attn.q_proj",
+    "43.self_attn.v_proj",
+    "language_model.layers.16.self_attn.k_proj",
+    "language_model.layers.23.self_attn.q_proj",
+    "language_model.layers.26.self_attn.v_proj",
+    "language_model.layers.5.self_attn.v_proj",
+    "language_model.layers.16.self_attn.v_proj",
+    "language_model.layers.14.self_attn.k_proj",
+    "language_model.layers.3.self_attn.q_proj",
+    "32.self_attn.q_proj",
+    "40.self_attn.q_proj",
+    "43.self_attn.q_proj",
+    "language_model.layers.9.self_attn.q_proj",
+    "27.self_attn.k_proj",
+    "language_model.layers.21.self_attn.q_proj",
+    "33.self_attn.k_proj",
+    "language_model.layers.1.self_attn.q_proj",
+    "language_model.layers.18.self_attn.q_proj",
+    "language_model.layers.21.self_attn.k_proj",
+    "language_model.layers.11.self_attn.v_proj",
+    "44.self_attn.v_proj",
+    "language_model.layers.6.self_attn.q_proj",
+    "language_model.layers.20.self_attn.q_proj",
+    "40.self_attn.v_proj",
+    "language_model.layers.10.self_attn.k_proj",
+    "41.self_attn.v_proj",
+    "language_model.layers.12.self_attn.k_proj",
+    "language_model.layers.11.self_attn.q_proj",
+    "language_model.layers.8.self_attn.v_proj",
+    "language_model.layers.0.self_attn.q_proj",
+    "language_model.layers.0.self_attn.k_proj",
+    "37.self_attn.k_proj",
+    "language_model.layers.2.self_attn.q_proj",
+    "language_model.layers.2.self_attn.k_proj",
+    "language_model.layers.15.self_attn.k_proj",
+    "45.self_attn.k_proj",
+    "29.self_attn.k_proj",
+    "language_model.layers.19.self_attn.k_proj",
+    "35.self_attn.q_proj",
+    "30.self_attn.v_proj",
+    "28.self_attn.q_proj",
+    "45.self_attn.q_proj",
+    "language_model.layers.13.self_attn.k_proj",
+    "40.self_attn.k_proj",
+    "language_model.layers.8.self_attn.q_proj",
+    "41.self_attn.q_proj",
+    "language_model.layers.11.self_attn.k_proj",
+    "34.self_attn.q_proj",
+    "language_model.layers.18.self_attn.v_proj",
+    "29.self_attn.q_proj",
+    "language_model.layers.15.self_attn.v_proj",
+    "language_model.layers.7.self_attn.k_proj",
+    "38.self_attn.k_proj",
+    "language_model.layers.22.self_attn.q_proj",
+    "language_model.layers.3.self_attn.v_proj",
+    "language_model.layers.7.self_attn.q_proj",
+    "language_model.layers.25.self_attn.v_proj",
+    "language_model.layers.16.self_attn.q_proj",
+    "32.self_attn.v_proj",
+    "31.self_attn.q_proj",
+    "language_model.layers.9.self_attn.k_proj",
+    "language_model.layers.14.self_attn.q_proj",
+    "o_proj",
+    "language_model.layers.26.self_attn.q_proj",
+    "43.self_attn.k_proj",
+    "37.self_attn.v_proj",
+    "language_model.layers.13.self_attn.q_proj",
+    "46.self_attn.k_proj",
+    "37.self_attn.q_proj",
+    "language_model.layers.23.self_attn.v_proj",
+    "31.self_attn.v_proj",
+    "language_model.layers.5.self_attn.k_proj",
+    "33.self_attn.v_proj",
+    "36.self_attn.q_proj",
+    "up_proj",
+    "language_model.layers.17.self_attn.v_proj",
+    "29.self_attn.v_proj",
+    "34.self_attn.v_proj",
+    "39.self_attn.k_proj",
+    "language_model.layers.25.self_attn.q_proj",
+    "language_model.layers.4.self_attn.v_proj",
+    "27.self_attn.q_proj",
+    "language_model.layers.15.self_attn.q_proj",
+    "language_model.layers.20.self_attn.v_proj",
+    "30.self_attn.q_proj",
+    "28.self_attn.v_proj",
+    "35.self_attn.v_proj",
+    "language_model.layers.9.self_attn.v_proj",
+    "47.self_attn.q_proj",
+    "46.self_attn.v_proj",
+    "35.self_attn.k_proj",
+    "33.self_attn.q_proj",
+    "language_model.layers.21.self_attn.v_proj",
+    "42.self_attn.v_proj",
+    "47.self_attn.v_proj",
+    "36.self_attn.k_proj",
+    "language_model.layers.17.self_attn.k_proj",
+    "47.self_attn.k_proj",
+    "31.self_attn.k_proj",
+    "language_model.layers.12.self_attn.v_proj",
+    "41.self_attn.k_proj",
+    "39.self_attn.v_proj",
+    "language_model.layers.4.self_attn.k_proj",
+    "42.self_attn.k_proj",
+    "language_model.layers.10.self_attn.q_proj",
+    "30.self_attn.k_proj",
+    "32.self_attn.k_proj",
+    "language_model.layers.13.self_attn.v_proj",
+    "language_model.layers.2.self_attn.v_proj",
+    "language_model.layers.12.self_attn.q_proj",
+    "language_model.layers.24.self_attn.q_proj",
+    "38.self_attn.v_proj",
+    "36.self_attn.v_proj",
+    "34.self_attn.k_proj",
+    "language_model.layers.5.self_attn.q_proj",
+    "language_model.layers.23.self_attn.k_proj",
+    "44.self_attn.k_proj",
+    "language_model.layers.14.self_attn.v_proj",
+    "language_model.layers.19.self_attn.v_proj",
+    "38.self_attn.q_proj",
+    "language_model.layers.6.self_attn.v_proj",
+    "27.self_attn.v_proj",
+    "28.self_attn.k_proj",
+    "language_model.layers.3.self_attn.k_proj",
+    "language_model.layers.0.self_attn.v_proj",
+    "gate_proj",
+    "language_model.layers.8.self_attn.k_proj",
+    "language_model.layers.6.self_attn.k_proj",
+    "down_proj",
+    "language_model.layers.20.self_attn.k_proj",
+    "language_model.layers.4.self_attn.q_proj",
+    "46.self_attn.q_proj",
+    "language_model.layers.22.self_attn.k_proj",
+    "language_model.layers.18.self_attn.k_proj",
+    "language_model.layers.22.self_attn.v_proj",
+    "44.self_attn.q_proj",
+    "45.self_attn.v_proj",
+    "language_model.layers.19.self_attn.q_proj",
+    "language_model.layers.24.self_attn.k_proj",
+    "language_model.layers.26.self_attn.k_proj",
+    "language_model.layers.7.self_attn.v_proj",
+    "language_model.layers.1.self_attn.v_proj",
+    "language_model.layers.24.self_attn.v_proj",
+    "language_model.layers.10.self_attn.v_proj",
+    "42.self_attn.q_proj",
+    "39.self_attn.q_proj",
+    "language_model.layers.1.self_attn.k_proj",
+    "language_model.layers.25.self_attn.k_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

smoke/20260810T011719Z/arms/control/adapter/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dce8fd2e0bd75cf349ef61413aa8e1ceea58c873d4198d84264442877847a27b
+size 1047628488

smoke/20260810T011719Z/arms/control/adapter/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}

smoke/20260810T011719Z/arms/control/adapter/config.json

@@ -0,0 +1,125 @@
+{
+  "architectures": [
+    "Gemma3ForConditionalGeneration"
+  ],
+  "boi_token_index": 255999,
+  "bos_token_id": 2,
+  "dtype": "bfloat16",
+  "eoi_token_index": 256000,
+  "eos_token_id": 1,
+  "image_token_index": 262144,
+  "initializer_range": 0.02,
+  "mm_tokens_per_image": 256,
+  "model_type": "gemma3",
+  "pad_token_id": 0,
+  "text_config": {
+    "_sliding_window_pattern": 6,
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "attn_logit_softcapping": null,
+    "bos_token_id": 2,
+    "cache_implementation": "hybrid",
+    "dtype": "bfloat16",
+    "eos_token_id": 1,
+    "final_logit_softcapping": null,
+    "head_dim": 256,
+    "hidden_activation": "gelu_pytorch_tanh",
+    "hidden_size": 3840,
+    "initializer_range": 0.02,
+    "intermediate_size": 15360,
+    "layer_types": [
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "sliding_attention",
+      "full_attention"
+    ],
+    "max_position_embeddings": 131072,
+    "model_type": "gemma3_text",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 48,
+    "num_key_value_heads": 8,
+    "pad_token_id": 0,
+    "query_pre_attn_scalar": 256,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "full_attention": {
+        "factor": 8.0,
+        "rope_theta": 1000000.0,
+        "rope_type": "linear"
+      },
+      "sliding_attention": {
+        "rope_theta": 10000.0,
+        "rope_type": "default"
+      }
+    },
+    "sliding_window": 1024,
+    "sliding_window_pattern": 6,
+    "tie_word_embeddings": true,
+    "use_bidirectional_attention": false,
+    "use_cache": false,
+    "vocab_size": 262208
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.9.0",
+  "unsloth_fixed": true,
+  "use_cache": false,
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "dtype": "bfloat16",
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "image_size": 896,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_channels": 3,
+    "num_hidden_layers": 27,
+    "patch_size": 14,
+    "vision_use_head": false
+  }
+}

smoke/20260810T011719Z/arms/control/adapter/debug.log

@@ -0,0 +1,635 @@
+[2026-08-10 01:25:29,793] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:1428] baseline 0.000GB ()
+[2026-08-10 01:25:29,794] [INFO] [axolotl.cli.config.load_cfg:333] [PID:1428] config:
+{
+  "activation_offloading": false,
+  "adapter": "lora",
+  "attn_implementation": "flash_attention_2",
+  "attn_needs_dtype_cast": true,
+  "attn_supports_packing": true,
+  "attn_uses_flash_lib": true,
+  "axolotl_config_path": "/workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/train/axolotl.yaml",
+  "base_model": "/workspace/python4-aft-state/20260810T011719Z/control/parent/control/sft/end",
+  "base_model_config": "/workspace/python4-aft-state/20260810T011719Z/control/parent/control/sft/end",
+  "batch_size": 32,
+  "bf16": true,
+  "capabilities": {
+    "bf16": true,
+    "compute_capability": "sm_90",
+    "fp8": true,
+    "n_gpu": 1,
+    "n_node": 1,
+    "tf32": true
+  },
+  "chat_template": "gemma3",
+  "checkpoint_schedule": [
+    2
+  ],
+  "context_parallel_size": 1,
+  "cosine_min_lr_ratio": 0.1,
+  "dataloader_num_workers": 1,
+  "dataloader_pin_memory": true,
+  "dataloader_prefetch_factor": 256,
+  "dataset_num_proc": 4,
+  "dataset_prepared_path": "/workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/train/prepared",
+  "datasets": [
+    {
+      "chat_template": "tokenizer_default",
+      "field_messages": "messages",
+      "message_property_mappings": {
+        "content": "content",
+        "role": "role"
+      },
+      "path": "/workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/smoke_aft.jsonl",
+      "trust_remote_code": false,
+      "type": "chat_template"
+    }
+  ],
+  "ddp": false,
+  "ddp_find_unused_parameters": true,
+  "device": "cuda:0",
+  "dion_rank_fraction": 1.0,
+  "dion_rank_multiple_of": 1,
+  "eaft_alpha": 1.0,
+  "eaft_k": 20,
+  "env_capabilities": {
+    "torch_version": "2.12.1"
+  },
+  "eot_tokens": [
+    "<end_of_turn>"
+  ],
+  "eval_batch_size": 4,
+  "eval_causal_lm_metrics": [
+    "sacrebleu",
+    "comet",
+    "ter",
+    "chrf"
+  ],
+  "eval_max_new_tokens": 128,
+  "eval_table_size": 0,
+  "experimental_skip_move_to_device": true,
+  "fp16": false,
+  "generate_samples": false,
+  "generation_do_sample": true,
+  "generation_max_new_tokens": 50,
+  "generation_prompt_ratio": 0.5,
+  "generation_temperature": 0.7,
+  "gradient_accumulation_steps": 8,
+  "gradient_checkpointing": true,
+  "gradient_checkpointing_kwargs": {
+    "use_reentrant": false
+  },
+  "include_tkps": true,
+  "is_multimodal": true,
+  "layer_offloading": false,
+  "learning_rate": 0.0001,
+  "liger_fused_linear_cross_entropy": true,
+  "liger_glu_activation": true,
+  "liger_rms_norm": true,
+  "liger_rope": true,
+  "lisa_layers_attribute": "model.layers",
+  "load_best_model_at_end": false,
+  "load_in_4bit": false,
+  "load_in_8bit": false,
+  "local_rank": 0,
+  "logging_steps": 1,
+  "lora_alpha": 128,
+  "lora_dropout": 0.0,
+  "lora_embedding_kernel": true,
+  "lora_mlp_kernel": true,
+  "lora_o_kernel": true,
+  "lora_qkv_kernel": true,
+  "lora_r": 64,
+  "lora_target_modules": [
+    "model.language_model.layers.0.self_attn.q_proj",
+    "model.language_model.layers.0.self_attn.k_proj",
+    "model.language_model.layers.0.self_attn.v_proj",
+    "model.language_model.layers.0.self_attn.o_proj",
+    "model.language_model.layers.0.mlp.gate_proj",
+    "model.language_model.layers.0.mlp.up_proj",
+    "model.language_model.layers.0.mlp.down_proj",
+    "model.language_model.layers.1.self_attn.q_proj",
+    "model.language_model.layers.1.self_attn.k_proj",
+    "model.language_model.layers.1.self_attn.v_proj",
+    "model.language_model.layers.1.self_attn.o_proj",
+    "model.language_model.layers.1.mlp.gate_proj",
+    "model.language_model.layers.1.mlp.up_proj",
+    "model.language_model.layers.1.mlp.down_proj",
+    "model.language_model.layers.2.self_attn.q_proj",
+    "model.language_model.layers.2.self_attn.k_proj",
+    "model.language_model.layers.2.self_attn.v_proj",
+    "model.language_model.layers.2.self_attn.o_proj",
+    "model.language_model.layers.2.mlp.gate_proj",
+    "model.language_model.layers.2.mlp.up_proj",
+    "model.language_model.layers.2.mlp.down_proj",
+    "model.language_model.layers.3.self_attn.q_proj",
+    "model.language_model.layers.3.self_attn.k_proj",
+    "model.language_model.layers.3.self_attn.v_proj",
+    "model.language_model.layers.3.self_attn.o_proj",
+    "model.language_model.layers.3.mlp.gate_proj",
+    "model.language_model.layers.3.mlp.up_proj",
+    "model.language_model.layers.3.mlp.down_proj",
+    "model.language_model.layers.4.self_attn.q_proj",
+    "model.language_model.layers.4.self_attn.k_proj",
+    "model.language_model.layers.4.self_attn.v_proj",
+    "model.language_model.layers.4.self_attn.o_proj",
+    "model.language_model.layers.4.mlp.gate_proj",
+    "model.language_model.layers.4.mlp.up_proj",
+    "model.language_model.layers.4.mlp.down_proj",
+    "model.language_model.layers.5.self_attn.q_proj",
+    "model.language_model.layers.5.self_attn.k_proj",
+    "model.language_model.layers.5.self_attn.v_proj",
+    "model.language_model.layers.5.self_attn.o_proj",
+    "model.language_model.layers.5.mlp.gate_proj",
+    "model.language_model.layers.5.mlp.up_proj",
+    "model.language_model.layers.5.mlp.down_proj",
+    "model.language_model.layers.6.self_attn.q_proj",
+    "model.language_model.layers.6.self_attn.k_proj",
+    "model.language_model.layers.6.self_attn.v_proj",
+    "model.language_model.layers.6.self_attn.o_proj",
+    "model.language_model.layers.6.mlp.gate_proj",
+    "model.language_model.layers.6.mlp.up_proj",
+    "model.language_model.layers.6.mlp.down_proj",
+    "model.language_model.layers.7.self_attn.q_proj",
+    "model.language_model.layers.7.self_attn.k_proj",
+    "model.language_model.layers.7.self_attn.v_proj",
+    "model.language_model.layers.7.self_attn.o_proj",
+    "model.language_model.layers.7.mlp.gate_proj",
+    "model.language_model.layers.7.mlp.up_proj",
+    "model.language_model.layers.7.mlp.down_proj",
+    "model.language_model.layers.8.self_attn.q_proj",
+    "model.language_model.layers.8.self_attn.k_proj",
+    "model.language_model.layers.8.self_attn.v_proj",
+    "model.language_model.layers.8.self_attn.o_proj",
+    "model.language_model.layers.8.mlp.gate_proj",
+    "model.language_model.layers.8.mlp.up_proj",
+    "model.language_model.layers.8.mlp.down_proj",
+    "model.language_model.layers.9.self_attn.q_proj",
+    "model.language_model.layers.9.self_attn.k_proj",
+    "model.language_model.layers.9.self_attn.v_proj",
+    "model.language_model.layers.9.self_attn.o_proj",
+    "model.language_model.layers.9.mlp.gate_proj",
+    "model.language_model.layers.9.mlp.up_proj",
+    "model.language_model.layers.9.mlp.down_proj",
+    "model.language_model.layers.10.self_attn.q_proj",
+    "model.language_model.layers.10.self_attn.k_proj",
+    "model.language_model.layers.10.self_attn.v_proj",
+    "model.language_model.layers.10.self_attn.o_proj",
+    "model.language_model.layers.10.mlp.gate_proj",
+    "model.language_model.layers.10.mlp.up_proj",
+    "model.language_model.layers.10.mlp.down_proj",
+    "model.language_model.layers.11.self_attn.q_proj",
+    "model.language_model.layers.11.self_attn.k_proj",
+    "model.language_model.layers.11.self_attn.v_proj",
+    "model.language_model.layers.11.self_attn.o_proj",
+    "model.language_model.layers.11.mlp.gate_proj",
+    "model.language_model.layers.11.mlp.up_proj",
+    "model.language_model.layers.11.mlp.down_proj",
+    "model.language_model.layers.12.self_attn.q_proj",
+    "model.language_model.layers.12.self_attn.k_proj",
+    "model.language_model.layers.12.self_attn.v_proj",
+    "model.language_model.layers.12.self_attn.o_proj",
+    "model.language_model.layers.12.mlp.gate_proj",
+    "model.language_model.layers.12.mlp.up_proj",
+    "model.language_model.layers.12.mlp.down_proj",
+    "model.language_model.layers.13.self_attn.q_proj",
+    "model.language_model.layers.13.self_attn.k_proj",
+    "model.language_model.layers.13.self_attn.v_proj",
+    "model.language_model.layers.13.self_attn.o_proj",
+    "model.language_model.layers.13.mlp.gate_proj",
+    "model.language_model.layers.13.mlp.up_proj",
+    "model.language_model.layers.13.mlp.down_proj",
+    "model.language_model.layers.14.self_attn.q_proj",
+    "model.language_model.layers.14.self_attn.k_proj",
+    "model.language_model.layers.14.self_attn.v_proj",
+    "model.language_model.layers.14.self_attn.o_proj",
+    "model.language_model.layers.14.mlp.gate_proj",
+    "model.language_model.layers.14.mlp.up_proj",
+    "model.language_model.layers.14.mlp.down_proj",
+    "model.language_model.layers.15.self_attn.q_proj",
+    "model.language_model.layers.15.self_attn.k_proj",
+    "model.language_model.layers.15.self_attn.v_proj",
+    "model.language_model.layers.15.self_attn.o_proj",
+    "model.language_model.layers.15.mlp.gate_proj",
+    "model.language_model.layers.15.mlp.up_proj",
+    "model.language_model.layers.15.mlp.down_proj",
+    "model.language_model.layers.16.self_attn.q_proj",
+    "model.language_model.layers.16.self_attn.k_proj",
+    "model.language_model.layers.16.self_attn.v_proj",
+    "model.language_model.layers.16.self_attn.o_proj",
+    "model.language_model.layers.16.mlp.gate_proj",
+    "model.language_model.layers.16.mlp.up_proj",
+    "model.language_model.layers.16.mlp.down_proj",
+    "model.language_model.layers.17.self_attn.q_proj",
+    "model.language_model.layers.17.self_attn.k_proj",
+    "model.language_model.layers.17.self_attn.v_proj",
+    "model.language_model.layers.17.self_attn.o_proj",
+    "model.language_model.layers.17.mlp.gate_proj",
+    "model.language_model.layers.17.mlp.up_proj",
+    "model.language_model.layers.17.mlp.down_proj",
+    "model.language_model.layers.18.self_attn.q_proj",
+    "model.language_model.layers.18.self_attn.k_proj",
+    "model.language_model.layers.18.self_attn.v_proj",
+    "model.language_model.layers.18.self_attn.o_proj",
+    "model.language_model.layers.18.mlp.gate_proj",
+    "model.language_model.layers.18.mlp.up_proj",
+    "model.language_model.layers.18.mlp.down_proj",
+    "model.language_model.layers.19.self_attn.q_proj",
+    "model.language_model.layers.19.self_attn.k_proj",
+    "model.language_model.layers.19.self_attn.v_proj",
+    "model.language_model.layers.19.self_attn.o_proj",
+    "model.language_model.layers.19.mlp.gate_proj",
+    "model.language_model.layers.19.mlp.up_proj",
+    "model.language_model.layers.19.mlp.down_proj",
+    "model.language_model.layers.20.self_attn.q_proj",
+    "model.language_model.layers.20.self_attn.k_proj",
+    "model.language_model.layers.20.self_attn.v_proj",
+    "model.language_model.layers.20.self_attn.o_proj",
+    "model.language_model.layers.20.mlp.gate_proj",
+    "model.language_model.layers.20.mlp.up_proj",
+    "model.language_model.layers.20.mlp.down_proj",
+    "model.language_model.layers.21.self_attn.q_proj",
+    "model.language_model.layers.21.self_attn.k_proj",
+    "model.language_model.layers.21.self_attn.v_proj",
+    "model.language_model.layers.21.self_attn.o_proj",
+    "model.language_model.layers.21.mlp.gate_proj",
+    "model.language_model.layers.21.mlp.up_proj",
+    "model.language_model.layers.21.mlp.down_proj",
+    "model.language_model.layers.22.self_attn.q_proj",
+    "model.language_model.layers.22.self_attn.k_proj",
+    "model.language_model.layers.22.self_attn.v_proj",
+    "model.language_model.layers.22.self_attn.o_proj",
+    "model.language_model.layers.22.mlp.gate_proj",
+    "model.language_model.layers.22.mlp.up_proj",
+    "model.language_model.layers.22.mlp.down_proj",
+    "model.language_model.layers.23.self_attn.q_proj",
+    "model.language_model.layers.23.self_attn.k_proj",
+    "model.language_model.layers.23.self_attn.v_proj",
+    "model.language_model.layers.23.self_attn.o_proj",
+    "model.language_model.layers.23.mlp.gate_proj",
+    "model.language_model.layers.23.mlp.up_proj",
+    "model.language_model.layers.23.mlp.down_proj",
+    "model.language_model.layers.24.self_attn.q_proj",
+    "model.language_model.layers.24.self_attn.k_proj",
+    "model.language_model.layers.24.self_attn.v_proj",
+    "model.language_model.layers.24.self_attn.o_proj",
+    "model.language_model.layers.24.mlp.gate_proj",
+    "model.language_model.layers.24.mlp.up_proj",
+    "model.language_model.layers.24.mlp.down_proj",
+    "model.language_model.layers.25.self_attn.q_proj",
+    "model.language_model.layers.25.self_attn.k_proj",
+    "model.language_model.layers.25.self_attn.v_proj",
+    "model.language_model.layers.25.self_attn.o_proj",
+    "model.language_model.layers.25.mlp.gate_proj",
+    "model.language_model.layers.25.mlp.up_proj",
+    "model.language_model.layers.25.mlp.down_proj",
+    "model.language_model.layers.26.self_attn.q_proj",
+    "model.language_model.layers.26.self_attn.k_proj",
+    "model.language_model.layers.26.self_attn.v_proj",
+    "model.language_model.layers.26.self_attn.o_proj",
+    "model.language_model.layers.26.mlp.gate_proj",
+    "model.language_model.layers.26.mlp.up_proj",
+    "model.language_model.layers.26.mlp.down_proj",
+    "model.language_model.layers.27.self_attn.q_proj",
+    "model.language_model.layers.27.self_attn.k_proj",
+    "model.language_model.layers.27.self_attn.v_proj",
+    "model.language_model.layers.27.self_attn.o_proj",
+    "model.language_model.layers.27.mlp.gate_proj",
+    "model.language_model.layers.27.mlp.up_proj",
+    "model.language_model.layers.27.mlp.down_proj",
+    "model.language_model.layers.28.self_attn.q_proj",
+    "model.language_model.layers.28.self_attn.k_proj",
+    "model.language_model.layers.28.self_attn.v_proj",
+    "model.language_model.layers.28.self_attn.o_proj",
+    "model.language_model.layers.28.mlp.gate_proj",
+    "model.language_model.layers.28.mlp.up_proj",
+    "model.language_model.layers.28.mlp.down_proj",
+    "model.language_model.layers.29.self_attn.q_proj",
+    "model.language_model.layers.29.self_attn.k_proj",
+    "model.language_model.layers.29.self_attn.v_proj",
+    "model.language_model.layers.29.self_attn.o_proj",
+    "model.language_model.layers.29.mlp.gate_proj",
+    "model.language_model.layers.29.mlp.up_proj",
+    "model.language_model.layers.29.mlp.down_proj",
+    "model.language_model.layers.30.self_attn.q_proj",
+    "model.language_model.layers.30.self_attn.k_proj",
+    "model.language_model.layers.30.self_attn.v_proj",
+    "model.language_model.layers.30.self_attn.o_proj",
+    "model.language_model.layers.30.mlp.gate_proj",
+    "model.language_model.layers.30.mlp.up_proj",
+    "model.language_model.layers.30.mlp.down_proj",
+    "model.language_model.layers.31.self_attn.q_proj",
+    "model.language_model.layers.31.self_attn.k_proj",
+    "model.language_model.layers.31.self_attn.v_proj",
+    "model.language_model.layers.31.self_attn.o_proj",
+    "model.language_model.layers.31.mlp.gate_proj",
+    "model.language_model.layers.31.mlp.up_proj",
+    "model.language_model.layers.31.mlp.down_proj",
+    "model.language_model.layers.32.self_attn.q_proj",
+    "model.language_model.layers.32.self_attn.k_proj",
+    "model.language_model.layers.32.self_attn.v_proj",
+    "model.language_model.layers.32.self_attn.o_proj",
+    "model.language_model.layers.32.mlp.gate_proj",
+    "model.language_model.layers.32.mlp.up_proj",
+    "model.language_model.layers.32.mlp.down_proj",
+    "model.language_model.layers.33.self_attn.q_proj",
+    "model.language_model.layers.33.self_attn.k_proj",
+    "model.language_model.layers.33.self_attn.v_proj",
+    "model.language_model.layers.33.self_attn.o_proj",
+    "model.language_model.layers.33.mlp.gate_proj",
+    "model.language_model.layers.33.mlp.up_proj",
+    "model.language_model.layers.33.mlp.down_proj",
+    "model.language_model.layers.34.self_attn.q_proj",
+    "model.language_model.layers.34.self_attn.k_proj",
+    "model.language_model.layers.34.self_attn.v_proj",
+    "model.language_model.layers.34.self_attn.o_proj",
+    "model.language_model.layers.34.mlp.gate_proj",
+    "model.language_model.layers.34.mlp.up_proj",
+    "model.language_model.layers.34.mlp.down_proj",
+    "model.language_model.layers.35.self_attn.q_proj",
+    "model.language_model.layers.35.self_attn.k_proj",
+    "model.language_model.layers.35.self_attn.v_proj",
+    "model.language_model.layers.35.self_attn.o_proj",
+    "model.language_model.layers.35.mlp.gate_proj",
+    "model.language_model.layers.35.mlp.up_proj",
+    "model.language_model.layers.35.mlp.down_proj",
+    "model.language_model.layers.36.self_attn.q_proj",
+    "model.language_model.layers.36.self_attn.k_proj",
+    "model.language_model.layers.36.self_attn.v_proj",
+    "model.language_model.layers.36.self_attn.o_proj",
+    "model.language_model.layers.36.mlp.gate_proj",
+    "model.language_model.layers.36.mlp.up_proj",
+    "model.language_model.layers.36.mlp.down_proj",
+    "model.language_model.layers.37.self_attn.q_proj",
+    "model.language_model.layers.37.self_attn.k_proj",
+    "model.language_model.layers.37.self_attn.v_proj",
+    "model.language_model.layers.37.self_attn.o_proj",
+    "model.language_model.layers.37.mlp.gate_proj",
+    "model.language_model.layers.37.mlp.up_proj",
+    "model.language_model.layers.37.mlp.down_proj",
+    "model.language_model.layers.38.self_attn.q_proj",
+    "model.language_model.layers.38.self_attn.k_proj",
+    "model.language_model.layers.38.self_attn.v_proj",
+    "model.language_model.layers.38.self_attn.o_proj",
+    "model.language_model.layers.38.mlp.gate_proj",
+    "model.language_model.layers.38.mlp.up_proj",
+    "model.language_model.layers.38.mlp.down_proj",
+    "model.language_model.layers.39.self_attn.q_proj",
+    "model.language_model.layers.39.self_attn.k_proj",
+    "model.language_model.layers.39.self_attn.v_proj",
+    "model.language_model.layers.39.self_attn.o_proj",
+    "model.language_model.layers.39.mlp.gate_proj",
+    "model.language_model.layers.39.mlp.up_proj",
+    "model.language_model.layers.39.mlp.down_proj",
+    "model.language_model.layers.40.self_attn.q_proj",
+    "model.language_model.layers.40.self_attn.k_proj",
+    "model.language_model.layers.40.self_attn.v_proj",
+    "model.language_model.layers.40.self_attn.o_proj",
+    "model.language_model.layers.40.mlp.gate_proj",
+    "model.language_model.layers.40.mlp.up_proj",
+    "model.language_model.layers.40.mlp.down_proj",
+    "model.language_model.layers.41.self_attn.q_proj",
+    "model.language_model.layers.41.self_attn.k_proj",
+    "model.language_model.layers.41.self_attn.v_proj",
+    "model.language_model.layers.41.self_attn.o_proj",
+    "model.language_model.layers.41.mlp.gate_proj",
+    "model.language_model.layers.41.mlp.up_proj",
+    "model.language_model.layers.41.mlp.down_proj",
+    "model.language_model.layers.42.self_attn.q_proj",
+    "model.language_model.layers.42.self_attn.k_proj",
+    "model.language_model.layers.42.self_attn.v_proj",
+    "model.language_model.layers.42.self_attn.o_proj",
+    "model.language_model.layers.42.mlp.gate_proj",
+    "model.language_model.layers.42.mlp.up_proj",
+    "model.language_model.layers.42.mlp.down_proj",
+    "model.language_model.layers.43.self_attn.q_proj",
+    "model.language_model.layers.43.self_attn.k_proj",
+    "model.language_model.layers.43.self_attn.v_proj",
+    "model.language_model.layers.43.self_attn.o_proj",
+    "model.language_model.layers.43.mlp.gate_proj",
+    "model.language_model.layers.43.mlp.up_proj",
+    "model.language_model.layers.43.mlp.down_proj",
+    "model.language_model.layers.44.self_attn.q_proj",
+    "model.language_model.layers.44.self_attn.k_proj",
+    "model.language_model.layers.44.self_attn.v_proj",
+    "model.language_model.layers.44.self_attn.o_proj",
+    "model.language_model.layers.44.mlp.gate_proj",
+    "model.language_model.layers.44.mlp.up_proj",
+    "model.language_model.layers.44.mlp.down_proj",
+    "model.language_model.layers.45.self_attn.q_proj",
+    "model.language_model.layers.45.self_attn.k_proj",
+    "model.language_model.layers.45.self_attn.v_proj",
+    "model.language_model.layers.45.self_attn.o_proj",
+    "model.language_model.layers.45.mlp.gate_proj",
+    "model.language_model.layers.45.mlp.up_proj",
+    "model.language_model.layers.45.mlp.down_proj",
+    "model.language_model.layers.46.self_attn.q_proj",
+    "model.language_model.layers.46.self_attn.k_proj",
+    "model.language_model.layers.46.self_attn.v_proj",
+    "model.language_model.layers.46.self_attn.o_proj",
+    "model.language_model.layers.46.mlp.gate_proj",
+    "model.language_model.layers.46.mlp.up_proj",
+    "model.language_model.layers.46.mlp.down_proj",
+    "model.language_model.layers.47.self_attn.q_proj",
+    "model.language_model.layers.47.self_attn.k_proj",
+    "model.language_model.layers.47.self_attn.v_proj",
+    "model.language_model.layers.47.self_attn.o_proj",
+    "model.language_model.layers.47.mlp.gate_proj",
+    "model.language_model.layers.47.mlp.up_proj",
+    "model.language_model.layers.47.mlp.down_proj"
+  ],
+  "loraplus_lr_embedding": 1e-06,
+  "lr_scheduler": "cosine",
+  "max_grad_norm": 1.0,
+  "mean_resizing_embeddings": false,
+  "merge_method": "memory_efficient",
+  "micro_batch_size": 4,
+  "model_config_type": "gemma3",
+  "model_config_type_text": "gemma3_text",
+  "num_epochs": 2.0,
+  "num_generation_samples": 3,
+  "optimizer": "adamw_torch_fused",
+  "otel_metrics_host": "localhost",
+  "otel_metrics_port": 8000,
+  "output_dir": "/workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/train/checkpoints",
+  "pad_to_sequence_len": false,
+  "plugins": [
+    "axolotl.integrations.liger.LigerPlugin",
+    "scimt.train.axolotl_plugins.CheckpointSchedulePlugin"
+  ],
+  "pretrain_multipack_attn": true,
+  "processor_config": "/workspace/python4-aft-state/20260810T011719Z/control/parent/control/sft/end",
+  "profiler_steps_start": 0,
+  "qgalore_cos_threshold": 0.4,
+  "qgalore_gamma_proj": 2,
+  "qgalore_proj_bits": 4,
+  "qgalore_proj_group_size": 256,
+  "qgalore_proj_quant": true,
+  "qgalore_proj_type": "std",
+  "qgalore_queue_size": 5,
+  "qgalore_rank": 256,
+  "qgalore_scale": 0.25,
+  "qgalore_update_proj_gap": 200,
+  "qlora_sharded_model_loading": false,
+  "quantize_moe_experts": false,
+  "ray_num_workers": 1,
+  "relora_prune_method": "magnitude",
+  "resources_per_worker": {
+    "GPU": 1
+  },
+  "sample_packing": false,
+  "sample_packing_bin_size": 200,
+  "sample_packing_group_size": 100000,
+  "save_only_model": true,
+  "save_safetensors": true,
+  "save_strategy": "no",
+  "save_total_limit": 1,
+  "seed": 424242,
+  "sequence_len": 4096,
+  "shuffle_before_merging_datasets": false,
+  "shuffle_merged_datasets": true,
+  "skip_prepare_dataset": false,
+  "streaming_multipack_buffer_size": 10000,
+  "strict": false,
+  "tensor_parallel_size": 1,
+  "tf32": true,
+  "tiled_mlp_use_original_mlp": true,
+  "tokenizer_config": "/workspace/python4-aft-state/20260810T011719Z/control/parent/control/sft/end",
+  "tokenizer_save_jinja_files": true,
+  "torch_dtype": "torch.bfloat16",
+  "train_on_inputs": false,
+  "trl": {
+    "async_prefetch": false,
+    "log_completions": false,
+    "mask_truncated_completions": false,
+    "ref_model_mixup_alpha": 0.9,
+    "ref_model_sync_steps": 64,
+    "replay_buffer_size": 0,
+    "replay_recompute_logps": true,
+    "reroll_max_groups": 1,
+    "reroll_start_fraction": 1.0,
+    "reward_num_workers": 1,
+    "scale_rewards": true,
+    "skip_zero_advantage_batches": true,
+    "sync_ref_model": false,
+    "use_data_producer": false,
+    "use_vllm": false,
+    "vllm_lora_sync": false,
+    "vllm_server_host": "0.0.0.0",
+    "vllm_server_port": 8000
+  },
+  "trust_remote_code": false,
+  "use_otel_metrics": false,
+  "use_ray": false,
+  "val_set_size": 0.0,
+  "vllm": {
+    "device": "auto",
+    "dtype": "auto",
+    "gpu_memory_utilization": 0.9,
+    "host": "0.0.0.0",
+    "port": 8000
+  },
+  "warmup_ratio": 0.05,
+  "weight_decay": 0.01,
+  "world_size": 1
+}
+[2026-08-10 01:25:30,041] [DEBUG] [axolotl.loaders.utils.check_model_config:88] [PID:1428] Loaded image size: 896 from model config
+[2026-08-10 01:25:31,410] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:1428] EOS: 1 / <eos>
+[2026-08-10 01:25:31,411] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:1428] BOS: 2 / <bos>
+[2026-08-10 01:25:31,411] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:1428] PAD: 0 / <pad>
+[2026-08-10 01:25:31,411] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:1428] UNK: 3 / <unk>
+[2026-08-10 01:25:31,412] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:482] [PID:1428] Unable to find prepared dataset in /workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/train/prepared/0a44fd9e07210b6aad1cc149dc58311a
+[2026-08-10 01:25:31,412] [INFO] [axolotl.utils.data.sft._load_raw_datasets:320] [PID:1428] Loading raw datasets...
+[2026-08-10 01:25:31,412] [WARNING] [axolotl.utils.data.sft._load_raw_datasets:322] [PID:1428] Processing datasets during training can lead to VRAM instability. Please pre-process your dataset using `axolotl preprocess path/to/config.yml`.
+
+[2026-08-10 01:25:31,914] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:1428] Loading dataset: /workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/smoke_aft.jsonl with base_type: chat_template and prompt_style: None
+[2026-08-10 01:25:31,923] [INFO] [axolotl.prompt_strategies.chat_template.__call__:1209] [PID:1428] Using chat template:
+---
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}
+
+---
+
+[2026-08-10 01:26:02,961] [INFO] [axolotl.utils.data.utils._log_dataset_stats:212] [PID:1428] min_input_len: 298
+[2026-08-10 01:26:02,961] [INFO] [axolotl.utils.data.utils._log_dataset_stats:213] [PID:1428] max_input_len: 806
+
+
+[2026-08-10 01:26:10,374] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:420] [PID:1428] total_num_tokens: 15_624
+[2026-08-10 01:26:10,375] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:438] [PID:1428] `total_supervised_tokens: 3_683`
+[2026-08-10 01:26:10,375] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:521] [PID:1428] total_num_steps: 2
+[2026-08-10 01:26:10,375] [INFO] [axolotl.utils.data.sft._prepare_standard_dataset:121] [PID:1428] Maximum number of steps set at 2
+[2026-08-10 01:26:10,375] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:1428] loading tokenizer... /workspace/python4-aft-state/20260810T011719Z/control/parent/control/sft/end
+[2026-08-10 01:26:11,941] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:1428] EOS: 1 / <eos>
+[2026-08-10 01:26:11,941] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:1428] BOS: 2 / <bos>
+[2026-08-10 01:26:11,941] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:1428] PAD: 0 / <pad>
+[2026-08-10 01:26:11,941] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:1428] UNK: 3 / <unk>
+[2026-08-10 01:26:13,601] [DEBUG] [axolotl.train.setup_model_and_tokenizer:81] [PID:1428] Loading model
+[2026-08-10 01:26:13,606] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:75] [PID:1428] Patched OptimState8bit for torch.compile compatibility
+[2026-08-10 01:26:13,606] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:122] [PID:1428] Patched OptimState4bit for torch.compile compatibility
+[2026-08-10 01:26:13,606] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:154] [PID:1428] Patched OptimStateFp8 for torch.compile compatibility
+[2026-08-10 01:26:13,611] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:94] [PID:1428] Patched Trainer.evaluation_loop with nanmean loss calculation
+[2026-08-10 01:26:13,612] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:148] [PID:1428] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation
+[2026-08-10 01:26:13,662] [INFO] [axolotl.monkeypatch.attention.flash_attn_4.patch_flash_attn_4:52] [PID:1428] Flash Attention 4 is available for your GPU and offers faster training speeds. To enable: pip install flash-attn-4
+[2026-08-10 01:26:15,049] [INFO] [axolotl.monkeypatch.lora_kernels.patch_self_attn_lora:304] [PID:1428] Patched attention class with LoRA optims: Gemma3Attention
+[2026-08-10 01:26:15,319] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:117] [PID:1428] Applying LIGER to gemma3 with kwargs: {'rope': True, 'cross_entropy': None, 'fused_linear_cross_entropy': True, 'rms_norm': True, 'layer_norm': None, 'geglu': True}
+
+[2026-08-10 01:26:18,365] [INFO] [axolotl.loaders.model._configure_embedding_dtypes:433] [PID:1428] Converting modules to torch.bfloat16
+[2026-08-10 01:26:19,561] [DEBUG] [axolotl.loaders.model.log_gpu_memory_usage:127] [PID:1428] Memory usage after model load 0.000GB ()
+trainable params: 261,881,856 || all params: 12,449,206,896 || trainable%: 2.1036
+[2026-08-10 01:26:21,513] [DEBUG] [axolotl.loaders.model.log_gpu_memory_usage:127] [PID:1428] after adapters 0.000GB ()
+[2026-08-10 01:26:26,114] [INFO] [axolotl.train.save_initial_configs:450] [PID:1428] Pre-saving adapter config to /workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/train/checkpoints...
+[2026-08-10 01:26:26,115] [INFO] [axolotl.train.save_initial_configs:454] [PID:1428] Pre-saving tokenizer to /workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/train/checkpoints...
+[2026-08-10 01:26:26,420] [INFO] [axolotl.train.save_initial_configs:459] [PID:1428] Pre-saving model config to /workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/train/checkpoints...
+[2026-08-10 01:26:26,423] [INFO] [axolotl.train.save_initial_configs:463] [PID:1428] Pre-saving processor to /workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/train/checkpoints...
+[2026-08-10 01:26:26,689] [INFO] [axolotl.train.execute_training:226] [PID:1428] Starting trainer...
+
  0%|          | 0/2 [00:00<?, ?it/s]
 50%|█████     | 1/2 [00:08<00:08,  8.05s/it]
                                             
+
 50%|█████     | 1/2 [00:08<00:08,  8.05s/it]
                                             
+
+
                                             
+
+[2026-08-10 01:26:41,584] [INFO] [axolotl.train.save_trained_model:267] [PID:1428] Training completed! Saving trained model to /workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/train/checkpoints.
+[2026-08-10 01:26:42,175] [INFO] [axolotl.train.save_trained_model:388] [PID:1428] Model successfully saved to /workspace/python4-aft-20260810T011719Z-control/experiments/python4_aft_generalization/runs/20260810T011719Z/arms/control/train/checkpoints

smoke/20260810T011719Z/arms/control/adapter/processor_config.json

@@ -0,0 +1,28 @@
+{
+  "image_processor": {
+    "do_convert_rgb": null,
+    "do_normalize": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "image_mean": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "image_processor_type": "Gemma3ImageProcessor",
+    "image_seq_length": 256,
+    "image_std": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "resample": 2,
+    "rescale_factor": 0.00392156862745098,
+    "size": {
+      "height": 896,
+      "width": 896
+    }
+  },
+  "image_seq_length": 256,
+  "processor_class": "Gemma3Processor"
+}

smoke/20260810T011719Z/arms/control/adapter/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
+size 33384567

smoke/20260810T011719Z/arms/control/adapter/tokenizer_config.json

@@ -0,0 +1,26 @@
+{
+  "backend": "tokenizers",
+  "boi_token": "<start_of_image>",
+  "bos_token": "<bos>",
+  "clean_up_tokenization_spaces": false,
+  "eoi_token": "<end_of_image>",
+  "eos_token": "<eos>",
+  "image_token": "<image_soft_token>",
+  "is_local": true,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 131072,
+  "model_specific_special_tokens": {
+    "boi_token": "<start_of_image>",
+    "eoi_token": "<end_of_image>",
+    "image_token": "<image_soft_token>"
+  },
+  "pad_token": "<pad>",
+  "padding_side": "left",
+  "processor_class": "Gemma3Processor",
+  "sp_model_kwargs": null,
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}