Upload folder using huggingface_hub

.gitattributes

@@ -176,3 +176,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 2026.03.25/21.49.56_train_llm_lowdim_box-close-v2/wandb/run-20260325_215001-2h81cyev/run-2h81cyev.wandb filter=lfs diff=lfs merge=lfs -text
 2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/run-0k6mearn.wandb filter=lfs diff=lfs merge=lfs -text
 2026.03.26/16.46.54_train_llm_lowdim_metaworld/wandb/run-20260326_164658-8p946alk/run-8p946alk.wandb filter=lfs diff=lfs merge=lfs -text
+2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/run-nhmfpc2t.wandb filter=lfs diff=lfs merge=lfs -text

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/.hydra/config.yaml

@@ -0,0 +1,115 @@
+name: train_llm_lowdim
+_target_: llmbc.workspace.train_llm_workspace.TrainLLMWorkspace
+obs_dim: ${task.obs_dim}
+action_dim: ${task.action_dim}
+horizon: 1
+n_obs_steps: 1
+n_action_steps: 1
+task_name: ${task.name}
+exp_name: train llm
+model_name: ${llm.name}
+use_quantization: ${llm.use_quantization}
+lora_config: ${llm.lora_config}
+dataset:
+  test_data_ratio: 0.01
+debug: false
+training:
+  seed: 42
+  per_device_train_batch_size: 128
+  per_device_eval_batch_size: 128
+  gradient_accumulation_steps: 1
+  optim: paged_adamw_32bit
+  num_train_epochs: 10
+  eval_strategy: steps
+  logging_steps: 1
+  warmup_steps: 10
+  logging_strategy: steps
+  learning_rate: 0.0001
+  fp16: false
+  bf16: true
+  tf32: true
+  group_by_length: true
+  report_to: wandb
+  save_steps: 5000
+  eval_steps: 10
+  use_joint_mlp_projector: ${llm.use_joint_mlp_projector}
+  joint_obs_action_mlp_lr: 5.0e-05
+trainer:
+  obs_dim: ${obs_dim}
+  action_dim: ${action_dim}
+  use_joint_mlp_projector: ${llm.use_joint_mlp_projector}
+  max_seq_length: ${llm.max_length}
+  dataset_text_field: text
+  packing: false
+logging:
+  project: llm_module_finetuning
+  resume: true
+  mode: online
+  name: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
+  tags:
+  - ${name}
+  - ${task_name}
+  - ${exp_name}
+  id: null
+  group: null
+multi_run:
+  run_dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+  wandb_name_base: ${now:%Y.%m.%d-%H.%M.%S}_${name}_${task_name}
+task:
+  name: adroit-hand-hammer-v1
+  obs_dim: 46
+  action_dim: 26
+  env_runner:
+    _target_: llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner
+    env_name: llf-adroit-adroit-hand-hammer-v1
+    n_train: 10
+    n_test: 50
+    n_envs: 10
+    max_steps: 150
+    n_obs_steps: ${n_obs_steps}
+    n_action_steps: ${n_action_steps}
+    instruction_type: b
+    feedback_type:
+    - hp
+    - hn
+    - fp
+    visual: false
+    discount: 0.99
+  dataset:
+    _target_: llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset
+    data_path: datasets/adroit-hand-hammer-v1-general.pt
+    data_path2: datasets/adroit-hand-hammer-v1.pt
+    horizon: ${horizon}
+    pad_before: ${eval:'${n_obs_steps}-1'}
+    pad_after: ${eval:'${n_action_steps}-1'}
+    obs_eef_target: true
+    use_manual_normalizer: false
+    val_ratio: 0.05
+    dummy_normalizer: false
+  instructor:
+    _target_: llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor
+llm:
+  name: HuggingFaceTB/SmolLM2-135M-Instruct
+  model_name: SmolLM2-135M-Instruct
+  config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+  causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+  use_quantization: false
+  use_joint_mlp_projector: true
+  llm_mode: mlp-finetuned
+  finetune_mode: orig
+  checkpoint: data/outputs/2026.03.27/14.38.20_train_mlp_projector_adroit-hand-hammer-v1/checkpoints/latest.ckpt
+  max_length: 100
+  lora_config:
+    r: 32
+    lora_alpha: 64
+    lora_dropout: 0.05
+    bias: none
+    task_type: CAUSAL_LM
+  prompter:
+    _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+    use_joint_mlp_projector: true
+  hydra:
+    job:
+      override_dirname: ${model_name}
+    run:
+      dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${model_name}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/.hydra/hydra.yaml

@@ -0,0 +1,154 @@
+hydra:
+  run:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+  sweep:
+    dir: data/outputs/${now:%Y.%m.%d}/${now:%H.%M.%S}_${name}_${task_name}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: train
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: llmdp_llm_adroit-hand-hammer-v1.yaml
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.2.0
+    version_base: '1.2'
+    cwd: /tmp2/chyang/workspace/LLM-BC
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /tmp2/chyang/workspace/LLM-BC/config/main_table
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1
+    choices:
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/checkpoint-360/config.json

@@ -0,0 +1,42 @@
+{
+  "_name_or_path": "HuggingFaceTB/SmolLM2-135M-Instruct",
+  "action_dim": 26,
+  "architectures": [
+    "LowdimLlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "head_dim": 64,
+  "hidden_act": "silu",
+  "hidden_size": 576,
+  "initializer_range": 0.041666666666666664,
+  "intermediate_size": 1536,
+  "is_llama_config": true,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "model_type": "llama_lowdim",
+  "num_attention_heads": 9,
+  "num_hidden_layers": 30,
+  "num_key_value_heads": 3,
+  "obs_dim": 46,
+  "pad_token_id": 2,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_interleaved": false,
+  "rope_scaling": null,
+  "rope_theta": 100000,
+  "tie_word_embeddings": true,
+  "torch_dtype": "float32",
+  "transformers.js_config": {
+    "kv_cache_dtype": {
+      "fp16": "float16",
+      "q4f16": "float16"
+    }
+  },
+  "transformers_version": "4.47.1",
+  "use_cache": false,
+  "use_joint_mlp_projector": true,
+  "vocab_size": 49152
+}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/checkpoint-360/generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 2,
+  "transformers_version": "4.47.1"
+}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/checkpoint-360/mlp_projector.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4338608ed11f67ca4d076f1596453801d07110a1ba10d38b55690336de76e74
+size 1499776

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/checkpoint-360/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:532f9325616dbe87be8846d0b03f80c41a2de5d5a05c7397eb15b3484410ede9
+size 539588496

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/checkpoint-360/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bc7e722db0999c1636e94a830207f155cf22d2c9d28f129bc987afa57af7a45
+size 1079284794

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/checkpoint-360/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8c41ff260c47496f4f1ca68f9e267daacc3461dc3a79663c2d83c1ae8fbf495
+size 14244

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/checkpoint-360/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36cd9db9ff66ee06ad412fe768ecde23049517e84a863a3dfc041789e2c58298
+size 1064

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/checkpoint-360/special_tokens_map.json

@@ -0,0 +1,34 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>"
+  ],
+  "bos_token": {
+    "content": "<|im_start|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/checkpoint-360/tokenizer_config.json

@@ -0,0 +1,155 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<repo_name>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<reponame>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<file_sep>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<filename>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<gh_stars>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "8": {
+      "content": "<issue_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "9": {
+      "content": "<issue_comment>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "10": {
+      "content": "<issue_closed>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "11": {
+      "content": "<jupyter_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "12": {
+      "content": "<jupyter_text>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "13": {
+      "content": "<jupyter_code>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "14": {
+      "content": "<jupyter_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "15": {
+      "content": "<jupyter_script>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "16": {
+      "content": "<empty_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>"
+  ],
+  "bos_token": "<|im_start|>",
+  "chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful AI assistant named SmolLM, trained by Hugging Face<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<|im_end|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/checkpoint-360/trainer_state.json

@@ -0,0 +1,2841 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 10.0,
+  "eval_steps": 10,
+  "global_step": 360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.027777777777777776,
+      "grad_norm": 13.871646881103516,
+      "learning_rate": 1e-05,
+      "loss": 1.6431,
+      "step": 1
+    },
+    {
+      "epoch": 0.05555555555555555,
+      "grad_norm": 17.890308380126953,
+      "learning_rate": 2e-05,
+      "loss": 1.6018,
+      "step": 2
+    },
+    {
+      "epoch": 0.08333333333333333,
+      "grad_norm": 13.746294021606445,
+      "learning_rate": 3e-05,
+      "loss": 1.5953,
+      "step": 3
+    },
+    {
+      "epoch": 0.1111111111111111,
+      "grad_norm": 15.9970121383667,
+      "learning_rate": 4e-05,
+      "loss": 1.5355,
+      "step": 4
+    },
+    {
+      "epoch": 0.1388888888888889,
+      "grad_norm": 18.634761810302734,
+      "learning_rate": 5e-05,
+      "loss": 1.552,
+      "step": 5
+    },
+    {
+      "epoch": 0.16666666666666666,
+      "grad_norm": 10.22042179107666,
+      "learning_rate": 6e-05,
+      "loss": 1.4926,
+      "step": 6
+    },
+    {
+      "epoch": 0.19444444444444445,
+      "grad_norm": 14.976595878601074,
+      "learning_rate": 7e-05,
+      "loss": 1.1827,
+      "step": 7
+    },
+    {
+      "epoch": 0.2222222222222222,
+      "grad_norm": 34.35334396362305,
+      "learning_rate": 8e-05,
+      "loss": 1.173,
+      "step": 8
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 7.392702579498291,
+      "learning_rate": 9e-05,
+      "loss": 0.7987,
+      "step": 9
+    },
+    {
+      "epoch": 0.2777777777777778,
+      "grad_norm": 8.6481294631958,
+      "learning_rate": 0.0001,
+      "loss": 0.7641,
+      "step": 10
+    },
+    {
+      "epoch": 0.2777777777777778,
+      "eval_loss": 0.719104528427124,
+      "eval_runtime": 0.1041,
+      "eval_samples_per_second": 441.967,
+      "eval_steps_per_second": 9.608,
+      "step": 10
+    },
+    {
+      "epoch": 0.3055555555555556,
+      "grad_norm": 11.896777153015137,
+      "learning_rate": 9.971428571428571e-05,
+      "loss": 0.6656,
+      "step": 11
+    },
+    {
+      "epoch": 0.3333333333333333,
+      "grad_norm": 6.002552032470703,
+      "learning_rate": 9.942857142857144e-05,
+      "loss": 0.5743,
+      "step": 12
+    },
+    {
+      "epoch": 0.3611111111111111,
+      "grad_norm": 4.569210529327393,
+      "learning_rate": 9.914285714285715e-05,
+      "loss": 0.5204,
+      "step": 13
+    },
+    {
+      "epoch": 0.3888888888888889,
+      "grad_norm": 3.2792978286743164,
+      "learning_rate": 9.885714285714286e-05,
+      "loss": 0.5217,
+      "step": 14
+    },
+    {
+      "epoch": 0.4166666666666667,
+      "grad_norm": 3.6701369285583496,
+      "learning_rate": 9.857142857142858e-05,
+      "loss": 0.5046,
+      "step": 15
+    },
+    {
+      "epoch": 0.4444444444444444,
+      "grad_norm": 4.064358711242676,
+      "learning_rate": 9.828571428571429e-05,
+      "loss": 0.4803,
+      "step": 16
+    },
+    {
+      "epoch": 0.4722222222222222,
+      "grad_norm": 2.9059529304504395,
+      "learning_rate": 9.8e-05,
+      "loss": 0.4466,
+      "step": 17
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 2.499434471130371,
+      "learning_rate": 9.771428571428572e-05,
+      "loss": 0.4453,
+      "step": 18
+    },
+    {
+      "epoch": 0.5277777777777778,
+      "grad_norm": 2.084019899368286,
+      "learning_rate": 9.742857142857143e-05,
+      "loss": 0.4382,
+      "step": 19
+    },
+    {
+      "epoch": 0.5555555555555556,
+      "grad_norm": 1.2787052392959595,
+      "learning_rate": 9.714285714285715e-05,
+      "loss": 0.4208,
+      "step": 20
+    },
+    {
+      "epoch": 0.5555555555555556,
+      "eval_loss": 0.4429771900177002,
+      "eval_runtime": 0.1036,
+      "eval_samples_per_second": 443.983,
+      "eval_steps_per_second": 9.652,
+      "step": 20
+    },
+    {
+      "epoch": 0.5833333333333334,
+      "grad_norm": 1.8704657554626465,
+      "learning_rate": 9.685714285714286e-05,
+      "loss": 0.431,
+      "step": 21
+    },
+    {
+      "epoch": 0.6111111111111112,
+      "grad_norm": 1.7761015892028809,
+      "learning_rate": 9.657142857142858e-05,
+      "loss": 0.4317,
+      "step": 22
+    },
+    {
+      "epoch": 0.6388888888888888,
+      "grad_norm": 1.7499357461929321,
+      "learning_rate": 9.628571428571429e-05,
+      "loss": 0.4229,
+      "step": 23
+    },
+    {
+      "epoch": 0.6666666666666666,
+      "grad_norm": 1.2509069442749023,
+      "learning_rate": 9.6e-05,
+      "loss": 0.4126,
+      "step": 24
+    },
+    {
+      "epoch": 0.6944444444444444,
+      "grad_norm": 1.3415058851242065,
+      "learning_rate": 9.571428571428573e-05,
+      "loss": 0.3888,
+      "step": 25
+    },
+    {
+      "epoch": 0.7222222222222222,
+      "grad_norm": 1.513482689857483,
+      "learning_rate": 9.542857142857143e-05,
+      "loss": 0.4111,
+      "step": 26
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 1.0207685232162476,
+      "learning_rate": 9.514285714285714e-05,
+      "loss": 0.3904,
+      "step": 27
+    },
+    {
+      "epoch": 0.7777777777777778,
+      "grad_norm": 1.0765091180801392,
+      "learning_rate": 9.485714285714287e-05,
+      "loss": 0.3911,
+      "step": 28
+    },
+    {
+      "epoch": 0.8055555555555556,
+      "grad_norm": 1.2146029472351074,
+      "learning_rate": 9.457142857142858e-05,
+      "loss": 0.3893,
+      "step": 29
+    },
+    {
+      "epoch": 0.8333333333333334,
+      "grad_norm": 1.302972435951233,
+      "learning_rate": 9.428571428571429e-05,
+      "loss": 0.3915,
+      "step": 30
+    },
+    {
+      "epoch": 0.8333333333333334,
+      "eval_loss": 0.39956018328666687,
+      "eval_runtime": 0.1036,
+      "eval_samples_per_second": 444.089,
+      "eval_steps_per_second": 9.654,
+      "step": 30
+    },
+    {
+      "epoch": 0.8611111111111112,
+      "grad_norm": 1.2195433378219604,
+      "learning_rate": 9.4e-05,
+      "loss": 0.377,
+      "step": 31
+    },
+    {
+      "epoch": 0.8888888888888888,
+      "grad_norm": 1.2320094108581543,
+      "learning_rate": 9.371428571428572e-05,
+      "loss": 0.3837,
+      "step": 32
+    },
+    {
+      "epoch": 0.9166666666666666,
+      "grad_norm": 1.0609043836593628,
+      "learning_rate": 9.342857142857143e-05,
+      "loss": 0.3776,
+      "step": 33
+    },
+    {
+      "epoch": 0.9444444444444444,
+      "grad_norm": 0.9609966278076172,
+      "learning_rate": 9.314285714285715e-05,
+      "loss": 0.3887,
+      "step": 34
+    },
+    {
+      "epoch": 0.9722222222222222,
+      "grad_norm": 1.0595581531524658,
+      "learning_rate": 9.285714285714286e-05,
+      "loss": 0.3761,
+      "step": 35
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.990327775478363,
+      "learning_rate": 9.257142857142858e-05,
+      "loss": 0.3744,
+      "step": 36
+    },
+    {
+      "epoch": 1.0277777777777777,
+      "grad_norm": 1.272873044013977,
+      "learning_rate": 9.228571428571429e-05,
+      "loss": 0.3625,
+      "step": 37
+    },
+    {
+      "epoch": 1.0555555555555556,
+      "grad_norm": 1.9024567604064941,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.3869,
+      "step": 38
+    },
+    {
+      "epoch": 1.0833333333333333,
+      "grad_norm": 1.3398654460906982,
+      "learning_rate": 9.171428571428572e-05,
+      "loss": 0.3751,
+      "step": 39
+    },
+    {
+      "epoch": 1.1111111111111112,
+      "grad_norm": 1.9176064729690552,
+      "learning_rate": 9.142857142857143e-05,
+      "loss": 0.3662,
+      "step": 40
+    },
+    {
+      "epoch": 1.1111111111111112,
+      "eval_loss": 0.37765416502952576,
+      "eval_runtime": 0.1036,
+      "eval_samples_per_second": 444.001,
+      "eval_steps_per_second": 9.652,
+      "step": 40
+    },
+    {
+      "epoch": 1.1388888888888888,
+      "grad_norm": 1.1852660179138184,
+      "learning_rate": 9.114285714285716e-05,
+      "loss": 0.3665,
+      "step": 41
+    },
+    {
+      "epoch": 1.1666666666666667,
+      "grad_norm": 1.831186056137085,
+      "learning_rate": 9.085714285714286e-05,
+      "loss": 0.3705,
+      "step": 42
+    },
+    {
+      "epoch": 1.1944444444444444,
+      "grad_norm": 1.1574777364730835,
+      "learning_rate": 9.057142857142857e-05,
+      "loss": 0.3582,
+      "step": 43
+    },
+    {
+      "epoch": 1.2222222222222223,
+      "grad_norm": 1.3485198020935059,
+      "learning_rate": 9.028571428571428e-05,
+      "loss": 0.3724,
+      "step": 44
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 1.0934721231460571,
+      "learning_rate": 9e-05,
+      "loss": 0.3549,
+      "step": 45
+    },
+    {
+      "epoch": 1.2777777777777777,
+      "grad_norm": 1.2588518857955933,
+      "learning_rate": 8.971428571428571e-05,
+      "loss": 0.3607,
+      "step": 46
+    },
+    {
+      "epoch": 1.3055555555555556,
+      "grad_norm": 0.9038533568382263,
+      "learning_rate": 8.942857142857142e-05,
+      "loss": 0.3492,
+      "step": 47
+    },
+    {
+      "epoch": 1.3333333333333333,
+      "grad_norm": 1.083348274230957,
+      "learning_rate": 8.914285714285715e-05,
+      "loss": 0.361,
+      "step": 48
+    },
+    {
+      "epoch": 1.3611111111111112,
+      "grad_norm": 0.8287424445152283,
+      "learning_rate": 8.885714285714286e-05,
+      "loss": 0.3475,
+      "step": 49
+    },
+    {
+      "epoch": 1.3888888888888888,
+      "grad_norm": 1.3475714921951294,
+      "learning_rate": 8.857142857142857e-05,
+      "loss": 0.363,
+      "step": 50
+    },
+    {
+      "epoch": 1.3888888888888888,
+      "eval_loss": 0.3627206087112427,
+      "eval_runtime": 0.1039,
+      "eval_samples_per_second": 442.561,
+      "eval_steps_per_second": 9.621,
+      "step": 50
+    },
+    {
+      "epoch": 1.4166666666666667,
+      "grad_norm": 1.1012552976608276,
+      "learning_rate": 8.828571428571429e-05,
+      "loss": 0.3527,
+      "step": 51
+    },
+    {
+      "epoch": 1.4444444444444444,
+      "grad_norm": 0.6421935558319092,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.3418,
+      "step": 52
+    },
+    {
+      "epoch": 1.4722222222222223,
+      "grad_norm": 1.1574995517730713,
+      "learning_rate": 8.771428571428572e-05,
+      "loss": 0.3513,
+      "step": 53
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 1.0251258611679077,
+      "learning_rate": 8.742857142857144e-05,
+      "loss": 0.3515,
+      "step": 54
+    },
+    {
+      "epoch": 1.5277777777777777,
+      "grad_norm": 0.9864039421081543,
+      "learning_rate": 8.714285714285715e-05,
+      "loss": 0.3609,
+      "step": 55
+    },
+    {
+      "epoch": 1.5555555555555556,
+      "grad_norm": 0.757999062538147,
+      "learning_rate": 8.685714285714286e-05,
+      "loss": 0.3454,
+      "step": 56
+    },
+    {
+      "epoch": 1.5833333333333335,
+      "grad_norm": 1.0983614921569824,
+      "learning_rate": 8.657142857142858e-05,
+      "loss": 0.3488,
+      "step": 57
+    },
+    {
+      "epoch": 1.6111111111111112,
+      "grad_norm": 1.4811136722564697,
+      "learning_rate": 8.62857142857143e-05,
+      "loss": 0.3562,
+      "step": 58
+    },
+    {
+      "epoch": 1.6388888888888888,
+      "grad_norm": 0.9457672834396362,
+      "learning_rate": 8.6e-05,
+      "loss": 0.349,
+      "step": 59
+    },
+    {
+      "epoch": 1.6666666666666665,
+      "grad_norm": 1.4347460269927979,
+      "learning_rate": 8.571428571428571e-05,
+      "loss": 0.3551,
+      "step": 60
+    },
+    {
+      "epoch": 1.6666666666666665,
+      "eval_loss": 0.35965070128440857,
+      "eval_runtime": 0.1045,
+      "eval_samples_per_second": 440.164,
+      "eval_steps_per_second": 9.569,
+      "step": 60
+    },
+    {
+      "epoch": 1.6944444444444444,
+      "grad_norm": 1.0592706203460693,
+      "learning_rate": 8.542857142857144e-05,
+      "loss": 0.3485,
+      "step": 61
+    },
+    {
+      "epoch": 1.7222222222222223,
+      "grad_norm": 1.3444126844406128,
+      "learning_rate": 8.514285714285714e-05,
+      "loss": 0.3512,
+      "step": 62
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 0.9045667052268982,
+      "learning_rate": 8.485714285714285e-05,
+      "loss": 0.3525,
+      "step": 63
+    },
+    {
+      "epoch": 1.7777777777777777,
+      "grad_norm": 1.135429859161377,
+      "learning_rate": 8.457142857142858e-05,
+      "loss": 0.3483,
+      "step": 64
+    },
+    {
+      "epoch": 1.8055555555555556,
+      "grad_norm": 0.7742411494255066,
+      "learning_rate": 8.428571428571429e-05,
+      "loss": 0.3445,
+      "step": 65
+    },
+    {
+      "epoch": 1.8333333333333335,
+      "grad_norm": 1.2747840881347656,
+      "learning_rate": 8.4e-05,
+      "loss": 0.3425,
+      "step": 66
+    },
+    {
+      "epoch": 1.8611111111111112,
+      "grad_norm": 1.1280975341796875,
+      "learning_rate": 8.371428571428572e-05,
+      "loss": 0.3506,
+      "step": 67
+    },
+    {
+      "epoch": 1.8888888888888888,
+      "grad_norm": 1.3229925632476807,
+      "learning_rate": 8.342857142857143e-05,
+      "loss": 0.3458,
+      "step": 68
+    },
+    {
+      "epoch": 1.9166666666666665,
+      "grad_norm": 1.0970568656921387,
+      "learning_rate": 8.314285714285715e-05,
+      "loss": 0.3443,
+      "step": 69
+    },
+    {
+      "epoch": 1.9444444444444444,
+      "grad_norm": 1.7599389553070068,
+      "learning_rate": 8.285714285714287e-05,
+      "loss": 0.3612,
+      "step": 70
+    },
+    {
+      "epoch": 1.9444444444444444,
+      "eval_loss": 0.35299769043922424,
+      "eval_runtime": 0.1042,
+      "eval_samples_per_second": 441.433,
+      "eval_steps_per_second": 9.596,
+      "step": 70
+    },
+    {
+      "epoch": 1.9722222222222223,
+      "grad_norm": 0.8275991678237915,
+      "learning_rate": 8.257142857142858e-05,
+      "loss": 0.3373,
+      "step": 71
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.5045437812805176,
+      "learning_rate": 8.228571428571429e-05,
+      "loss": 0.3624,
+      "step": 72
+    },
+    {
+      "epoch": 2.0277777777777777,
+      "grad_norm": 0.9771829843521118,
+      "learning_rate": 8.2e-05,
+      "loss": 0.3434,
+      "step": 73
+    },
+    {
+      "epoch": 2.0555555555555554,
+      "grad_norm": 0.8552800416946411,
+      "learning_rate": 8.171428571428572e-05,
+      "loss": 0.3347,
+      "step": 74
+    },
+    {
+      "epoch": 2.0833333333333335,
+      "grad_norm": 0.8917291164398193,
+      "learning_rate": 8.142857142857143e-05,
+      "loss": 0.3448,
+      "step": 75
+    },
+    {
+      "epoch": 2.111111111111111,
+      "grad_norm": 0.9143850207328796,
+      "learning_rate": 8.114285714285714e-05,
+      "loss": 0.3309,
+      "step": 76
+    },
+    {
+      "epoch": 2.138888888888889,
+      "grad_norm": 1.359926700592041,
+      "learning_rate": 8.085714285714287e-05,
+      "loss": 0.3471,
+      "step": 77
+    },
+    {
+      "epoch": 2.1666666666666665,
+      "grad_norm": 0.84107506275177,
+      "learning_rate": 8.057142857142857e-05,
+      "loss": 0.3433,
+      "step": 78
+    },
+    {
+      "epoch": 2.1944444444444446,
+      "grad_norm": 1.2953639030456543,
+      "learning_rate": 8.028571428571428e-05,
+      "loss": 0.3495,
+      "step": 79
+    },
+    {
+      "epoch": 2.2222222222222223,
+      "grad_norm": 0.9937311410903931,
+      "learning_rate": 8e-05,
+      "loss": 0.3388,
+      "step": 80
+    },
+    {
+      "epoch": 2.2222222222222223,
+      "eval_loss": 0.3491460978984833,
+      "eval_runtime": 0.1044,
+      "eval_samples_per_second": 440.593,
+      "eval_steps_per_second": 9.578,
+      "step": 80
+    },
+    {
+      "epoch": 2.25,
+      "grad_norm": 1.0681490898132324,
+      "learning_rate": 7.971428571428572e-05,
+      "loss": 0.3435,
+      "step": 81
+    },
+    {
+      "epoch": 2.2777777777777777,
+      "grad_norm": 0.8466928601264954,
+      "learning_rate": 7.942857142857143e-05,
+      "loss": 0.3333,
+      "step": 82
+    },
+    {
+      "epoch": 2.3055555555555554,
+      "grad_norm": 0.8183342814445496,
+      "learning_rate": 7.914285714285715e-05,
+      "loss": 0.3305,
+      "step": 83
+    },
+    {
+      "epoch": 2.3333333333333335,
+      "grad_norm": 0.833314061164856,
+      "learning_rate": 7.885714285714286e-05,
+      "loss": 0.3289,
+      "step": 84
+    },
+    {
+      "epoch": 2.361111111111111,
+      "grad_norm": 0.8347731828689575,
+      "learning_rate": 7.857142857142858e-05,
+      "loss": 0.3331,
+      "step": 85
+    },
+    {
+      "epoch": 2.388888888888889,
+      "grad_norm": 1.0877679586410522,
+      "learning_rate": 7.828571428571429e-05,
+      "loss": 0.3437,
+      "step": 86
+    },
+    {
+      "epoch": 2.4166666666666665,
+      "grad_norm": 0.9570125937461853,
+      "learning_rate": 7.800000000000001e-05,
+      "loss": 0.3331,
+      "step": 87
+    },
+    {
+      "epoch": 2.4444444444444446,
+      "grad_norm": 0.7662280797958374,
+      "learning_rate": 7.771428571428572e-05,
+      "loss": 0.3363,
+      "step": 88
+    },
+    {
+      "epoch": 2.4722222222222223,
+      "grad_norm": 0.9321999549865723,
+      "learning_rate": 7.742857142857143e-05,
+      "loss": 0.3305,
+      "step": 89
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 0.8284544348716736,
+      "learning_rate": 7.714285714285715e-05,
+      "loss": 0.3332,
+      "step": 90
+    },
+    {
+      "epoch": 2.5,
+      "eval_loss": 0.34498000144958496,
+      "eval_runtime": 0.1046,
+      "eval_samples_per_second": 439.848,
+      "eval_steps_per_second": 9.562,
+      "step": 90
+    },
+    {
+      "epoch": 2.5277777777777777,
+      "grad_norm": 1.0568827390670776,
+      "learning_rate": 7.685714285714286e-05,
+      "loss": 0.3448,
+      "step": 91
+    },
+    {
+      "epoch": 2.5555555555555554,
+      "grad_norm": 0.9136806130409241,
+      "learning_rate": 7.657142857142857e-05,
+      "loss": 0.334,
+      "step": 92
+    },
+    {
+      "epoch": 2.5833333333333335,
+      "grad_norm": 1.2551990747451782,
+      "learning_rate": 7.62857142857143e-05,
+      "loss": 0.3425,
+      "step": 93
+    },
+    {
+      "epoch": 2.611111111111111,
+      "grad_norm": 0.8284862637519836,
+      "learning_rate": 7.6e-05,
+      "loss": 0.3265,
+      "step": 94
+    },
+    {
+      "epoch": 2.638888888888889,
+      "grad_norm": 0.7161554098129272,
+      "learning_rate": 7.571428571428571e-05,
+      "loss": 0.3267,
+      "step": 95
+    },
+    {
+      "epoch": 2.6666666666666665,
+      "grad_norm": 0.8050905466079712,
+      "learning_rate": 7.542857142857144e-05,
+      "loss": 0.3342,
+      "step": 96
+    },
+    {
+      "epoch": 2.6944444444444446,
+      "grad_norm": 0.7441209554672241,
+      "learning_rate": 7.514285714285715e-05,
+      "loss": 0.3325,
+      "step": 97
+    },
+    {
+      "epoch": 2.7222222222222223,
+      "grad_norm": 0.591927707195282,
+      "learning_rate": 7.485714285714285e-05,
+      "loss": 0.334,
+      "step": 98
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 0.8902866244316101,
+      "learning_rate": 7.457142857142856e-05,
+      "loss": 0.3473,
+      "step": 99
+    },
+    {
+      "epoch": 2.7777777777777777,
+      "grad_norm": 0.6760069131851196,
+      "learning_rate": 7.428571428571429e-05,
+      "loss": 0.326,
+      "step": 100
+    },
+    {
+      "epoch": 2.7777777777777777,
+      "eval_loss": 0.3410107493400574,
+      "eval_runtime": 0.1043,
+      "eval_samples_per_second": 441.026,
+      "eval_steps_per_second": 9.588,
+      "step": 100
+    },
+    {
+      "epoch": 2.8055555555555554,
+      "grad_norm": 0.6579345464706421,
+      "learning_rate": 7.4e-05,
+      "loss": 0.3348,
+      "step": 101
+    },
+    {
+      "epoch": 2.8333333333333335,
+      "grad_norm": 1.0648415088653564,
+      "learning_rate": 7.371428571428572e-05,
+      "loss": 0.3253,
+      "step": 102
+    },
+    {
+      "epoch": 2.861111111111111,
+      "grad_norm": 0.6868814826011658,
+      "learning_rate": 7.342857142857144e-05,
+      "loss": 0.3297,
+      "step": 103
+    },
+    {
+      "epoch": 2.888888888888889,
+      "grad_norm": 1.1149464845657349,
+      "learning_rate": 7.314285714285715e-05,
+      "loss": 0.3401,
+      "step": 104
+    },
+    {
+      "epoch": 2.9166666666666665,
+      "grad_norm": 0.8934164047241211,
+      "learning_rate": 7.285714285714286e-05,
+      "loss": 0.3348,
+      "step": 105
+    },
+    {
+      "epoch": 2.9444444444444446,
+      "grad_norm": 1.1119507551193237,
+      "learning_rate": 7.257142857142858e-05,
+      "loss": 0.3427,
+      "step": 106
+    },
+    {
+      "epoch": 2.9722222222222223,
+      "grad_norm": 0.8103634715080261,
+      "learning_rate": 7.228571428571429e-05,
+      "loss": 0.3374,
+      "step": 107
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.8421126008033752,
+      "learning_rate": 7.2e-05,
+      "loss": 0.3395,
+      "step": 108
+    },
+    {
+      "epoch": 3.0277777777777777,
+      "grad_norm": 0.8583278656005859,
+      "learning_rate": 7.171428571428572e-05,
+      "loss": 0.331,
+      "step": 109
+    },
+    {
+      "epoch": 3.0555555555555554,
+      "grad_norm": 1.2129111289978027,
+      "learning_rate": 7.142857142857143e-05,
+      "loss": 0.3355,
+      "step": 110
+    },
+    {
+      "epoch": 3.0555555555555554,
+      "eval_loss": 0.332057386636734,
+      "eval_runtime": 0.1051,
+      "eval_samples_per_second": 437.685,
+      "eval_steps_per_second": 9.515,
+      "step": 110
+    },
+    {
+      "epoch": 3.0833333333333335,
+      "grad_norm": 0.9463130235671997,
+      "learning_rate": 7.114285714285714e-05,
+      "loss": 0.3294,
+      "step": 111
+    },
+    {
+      "epoch": 3.111111111111111,
+      "grad_norm": 0.9692079424858093,
+      "learning_rate": 7.085714285714285e-05,
+      "loss": 0.3327,
+      "step": 112
+    },
+    {
+      "epoch": 3.138888888888889,
+      "grad_norm": 0.9853659868240356,
+      "learning_rate": 7.057142857142858e-05,
+      "loss": 0.3295,
+      "step": 113
+    },
+    {
+      "epoch": 3.1666666666666665,
+      "grad_norm": 0.7222715616226196,
+      "learning_rate": 7.028571428571428e-05,
+      "loss": 0.3358,
+      "step": 114
+    },
+    {
+      "epoch": 3.1944444444444446,
+      "grad_norm": 1.1528452634811401,
+      "learning_rate": 7e-05,
+      "loss": 0.3406,
+      "step": 115
+    },
+    {
+      "epoch": 3.2222222222222223,
+      "grad_norm": 1.0079970359802246,
+      "learning_rate": 6.971428571428572e-05,
+      "loss": 0.329,
+      "step": 116
+    },
+    {
+      "epoch": 3.25,
+      "grad_norm": 0.7162885665893555,
+      "learning_rate": 6.942857142857143e-05,
+      "loss": 0.327,
+      "step": 117
+    },
+    {
+      "epoch": 3.2777777777777777,
+      "grad_norm": 0.9302375912666321,
+      "learning_rate": 6.914285714285715e-05,
+      "loss": 0.336,
+      "step": 118
+    },
+    {
+      "epoch": 3.3055555555555554,
+      "grad_norm": 0.8540468215942383,
+      "learning_rate": 6.885714285714286e-05,
+      "loss": 0.3356,
+      "step": 119
+    },
+    {
+      "epoch": 3.3333333333333335,
+      "grad_norm": 0.598040759563446,
+      "learning_rate": 6.857142857142858e-05,
+      "loss": 0.3279,
+      "step": 120
+    },
+    {
+      "epoch": 3.3333333333333335,
+      "eval_loss": 0.3290035128593445,
+      "eval_runtime": 0.1049,
+      "eval_samples_per_second": 438.622,
+      "eval_steps_per_second": 9.535,
+      "step": 120
+    },
+    {
+      "epoch": 3.361111111111111,
+      "grad_norm": 0.6981043815612793,
+      "learning_rate": 6.828571428571429e-05,
+      "loss": 0.33,
+      "step": 121
+    },
+    {
+      "epoch": 3.388888888888889,
+      "grad_norm": 0.6710860133171082,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.3262,
+      "step": 122
+    },
+    {
+      "epoch": 3.4166666666666665,
+      "grad_norm": 0.6621596813201904,
+      "learning_rate": 6.771428571428572e-05,
+      "loss": 0.3273,
+      "step": 123
+    },
+    {
+      "epoch": 3.4444444444444446,
+      "grad_norm": 0.8255563974380493,
+      "learning_rate": 6.742857142857143e-05,
+      "loss": 0.3264,
+      "step": 124
+    },
+    {
+      "epoch": 3.4722222222222223,
+      "grad_norm": 0.8350001573562622,
+      "learning_rate": 6.714285714285714e-05,
+      "loss": 0.3157,
+      "step": 125
+    },
+    {
+      "epoch": 3.5,
+      "grad_norm": 0.7275986075401306,
+      "learning_rate": 6.685714285714286e-05,
+      "loss": 0.3205,
+      "step": 126
+    },
+    {
+      "epoch": 3.5277777777777777,
+      "grad_norm": 0.652642548084259,
+      "learning_rate": 6.657142857142857e-05,
+      "loss": 0.3327,
+      "step": 127
+    },
+    {
+      "epoch": 3.5555555555555554,
+      "grad_norm": 0.9522960186004639,
+      "learning_rate": 6.628571428571428e-05,
+      "loss": 0.3319,
+      "step": 128
+    },
+    {
+      "epoch": 3.5833333333333335,
+      "grad_norm": 0.7006963491439819,
+      "learning_rate": 6.6e-05,
+      "loss": 0.3292,
+      "step": 129
+    },
+    {
+      "epoch": 3.611111111111111,
+      "grad_norm": 0.7161970138549805,
+      "learning_rate": 6.571428571428571e-05,
+      "loss": 0.3246,
+      "step": 130
+    },
+    {
+      "epoch": 3.611111111111111,
+      "eval_loss": 0.3442412316799164,
+      "eval_runtime": 0.1052,
+      "eval_samples_per_second": 437.452,
+      "eval_steps_per_second": 9.51,
+      "step": 130
+    },
+    {
+      "epoch": 3.638888888888889,
+      "grad_norm": 1.0642709732055664,
+      "learning_rate": 6.542857142857142e-05,
+      "loss": 0.3289,
+      "step": 131
+    },
+    {
+      "epoch": 3.6666666666666665,
+      "grad_norm": 0.7999193072319031,
+      "learning_rate": 6.514285714285715e-05,
+      "loss": 0.3234,
+      "step": 132
+    },
+    {
+      "epoch": 3.6944444444444446,
+      "grad_norm": 0.8324876427650452,
+      "learning_rate": 6.485714285714286e-05,
+      "loss": 0.3297,
+      "step": 133
+    },
+    {
+      "epoch": 3.7222222222222223,
+      "grad_norm": 0.561801552772522,
+      "learning_rate": 6.457142857142856e-05,
+      "loss": 0.3125,
+      "step": 134
+    },
+    {
+      "epoch": 3.75,
+      "grad_norm": 0.6995918154716492,
+      "learning_rate": 6.428571428571429e-05,
+      "loss": 0.3234,
+      "step": 135
+    },
+    {
+      "epoch": 3.7777777777777777,
+      "grad_norm": 0.6314477920532227,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.3256,
+      "step": 136
+    },
+    {
+      "epoch": 3.8055555555555554,
+      "grad_norm": 0.9092559814453125,
+      "learning_rate": 6.371428571428572e-05,
+      "loss": 0.3315,
+      "step": 137
+    },
+    {
+      "epoch": 3.8333333333333335,
+      "grad_norm": 0.7306588292121887,
+      "learning_rate": 6.342857142857143e-05,
+      "loss": 0.3241,
+      "step": 138
+    },
+    {
+      "epoch": 3.861111111111111,
+      "grad_norm": 0.7943991422653198,
+      "learning_rate": 6.314285714285715e-05,
+      "loss": 0.3323,
+      "step": 139
+    },
+    {
+      "epoch": 3.888888888888889,
+      "grad_norm": 0.8375313878059387,
+      "learning_rate": 6.285714285714286e-05,
+      "loss": 0.3273,
+      "step": 140
+    },
+    {
+      "epoch": 3.888888888888889,
+      "eval_loss": 0.33936312794685364,
+      "eval_runtime": 0.1048,
+      "eval_samples_per_second": 438.851,
+      "eval_steps_per_second": 9.54,
+      "step": 140
+    },
+    {
+      "epoch": 3.9166666666666665,
+      "grad_norm": 0.9479944705963135,
+      "learning_rate": 6.257142857142857e-05,
+      "loss": 0.3258,
+      "step": 141
+    },
+    {
+      "epoch": 3.9444444444444446,
+      "grad_norm": 0.8155922889709473,
+      "learning_rate": 6.22857142857143e-05,
+      "loss": 0.3228,
+      "step": 142
+    },
+    {
+      "epoch": 3.9722222222222223,
+      "grad_norm": 0.8617050647735596,
+      "learning_rate": 6.2e-05,
+      "loss": 0.3217,
+      "step": 143
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.2106715440750122,
+      "learning_rate": 6.171428571428571e-05,
+      "loss": 0.3309,
+      "step": 144
+    },
+    {
+      "epoch": 4.027777777777778,
+      "grad_norm": 0.8097350001335144,
+      "learning_rate": 6.142857142857143e-05,
+      "loss": 0.3265,
+      "step": 145
+    },
+    {
+      "epoch": 4.055555555555555,
+      "grad_norm": 0.651019811630249,
+      "learning_rate": 6.114285714285714e-05,
+      "loss": 0.3222,
+      "step": 146
+    },
+    {
+      "epoch": 4.083333333333333,
+      "grad_norm": 0.8858047127723694,
+      "learning_rate": 6.085714285714286e-05,
+      "loss": 0.3245,
+      "step": 147
+    },
+    {
+      "epoch": 4.111111111111111,
+      "grad_norm": 0.8602396845817566,
+      "learning_rate": 6.0571428571428576e-05,
+      "loss": 0.3153,
+      "step": 148
+    },
+    {
+      "epoch": 4.138888888888889,
+      "grad_norm": 0.615274965763092,
+      "learning_rate": 6.028571428571429e-05,
+      "loss": 0.313,
+      "step": 149
+    },
+    {
+      "epoch": 4.166666666666667,
+      "grad_norm": 0.9199692010879517,
+      "learning_rate": 6e-05,
+      "loss": 0.3275,
+      "step": 150
+    },
+    {
+      "epoch": 4.166666666666667,
+      "eval_loss": 0.34145644307136536,
+      "eval_runtime": 0.1056,
+      "eval_samples_per_second": 435.469,
+      "eval_steps_per_second": 9.467,
+      "step": 150
+    },
+    {
+      "epoch": 4.194444444444445,
+      "grad_norm": 0.9348899722099304,
+      "learning_rate": 5.9714285714285724e-05,
+      "loss": 0.3287,
+      "step": 151
+    },
+    {
+      "epoch": 4.222222222222222,
+      "grad_norm": 0.7250774502754211,
+      "learning_rate": 5.9428571428571434e-05,
+      "loss": 0.3203,
+      "step": 152
+    },
+    {
+      "epoch": 4.25,
+      "grad_norm": 0.7376280426979065,
+      "learning_rate": 5.914285714285714e-05,
+      "loss": 0.3169,
+      "step": 153
+    },
+    {
+      "epoch": 4.277777777777778,
+      "grad_norm": 0.6010245680809021,
+      "learning_rate": 5.885714285714285e-05,
+      "loss": 0.3215,
+      "step": 154
+    },
+    {
+      "epoch": 4.305555555555555,
+      "grad_norm": 0.7241640686988831,
+      "learning_rate": 5.8571428571428575e-05,
+      "loss": 0.317,
+      "step": 155
+    },
+    {
+      "epoch": 4.333333333333333,
+      "grad_norm": 0.6956952810287476,
+      "learning_rate": 5.828571428571429e-05,
+      "loss": 0.3217,
+      "step": 156
+    },
+    {
+      "epoch": 4.361111111111111,
+      "grad_norm": 0.8463672995567322,
+      "learning_rate": 5.8e-05,
+      "loss": 0.322,
+      "step": 157
+    },
+    {
+      "epoch": 4.388888888888889,
+      "grad_norm": 0.5538536906242371,
+      "learning_rate": 5.771428571428572e-05,
+      "loss": 0.3129,
+      "step": 158
+    },
+    {
+      "epoch": 4.416666666666667,
+      "grad_norm": 0.8398566246032715,
+      "learning_rate": 5.742857142857143e-05,
+      "loss": 0.3275,
+      "step": 159
+    },
+    {
+      "epoch": 4.444444444444445,
+      "grad_norm": 0.5335714221000671,
+      "learning_rate": 5.714285714285714e-05,
+      "loss": 0.3225,
+      "step": 160
+    },
+    {
+      "epoch": 4.444444444444445,
+      "eval_loss": 0.3419412672519684,
+      "eval_runtime": 0.1054,
+      "eval_samples_per_second": 436.373,
+      "eval_steps_per_second": 9.486,
+      "step": 160
+    },
+    {
+      "epoch": 4.472222222222222,
+      "grad_norm": 0.893516480922699,
+      "learning_rate": 5.6857142857142865e-05,
+      "loss": 0.3201,
+      "step": 161
+    },
+    {
+      "epoch": 4.5,
+      "grad_norm": 0.7950851321220398,
+      "learning_rate": 5.6571428571428574e-05,
+      "loss": 0.3246,
+      "step": 162
+    },
+    {
+      "epoch": 4.527777777777778,
+      "grad_norm": 0.6274649500846863,
+      "learning_rate": 5.628571428571428e-05,
+      "loss": 0.3138,
+      "step": 163
+    },
+    {
+      "epoch": 4.555555555555555,
+      "grad_norm": 0.6270793676376343,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.3225,
+      "step": 164
+    },
+    {
+      "epoch": 4.583333333333333,
+      "grad_norm": 0.6661616563796997,
+      "learning_rate": 5.571428571428572e-05,
+      "loss": 0.3222,
+      "step": 165
+    },
+    {
+      "epoch": 4.611111111111111,
+      "grad_norm": 0.6097862124443054,
+      "learning_rate": 5.542857142857143e-05,
+      "loss": 0.3138,
+      "step": 166
+    },
+    {
+      "epoch": 4.638888888888889,
+      "grad_norm": 0.6743194460868835,
+      "learning_rate": 5.514285714285714e-05,
+      "loss": 0.3109,
+      "step": 167
+    },
+    {
+      "epoch": 4.666666666666667,
+      "grad_norm": 0.6684880256652832,
+      "learning_rate": 5.485714285714286e-05,
+      "loss": 0.3193,
+      "step": 168
+    },
+    {
+      "epoch": 4.694444444444445,
+      "grad_norm": 0.7434603571891785,
+      "learning_rate": 5.457142857142857e-05,
+      "loss": 0.3212,
+      "step": 169
+    },
+    {
+      "epoch": 4.722222222222222,
+      "grad_norm": 0.8257206082344055,
+      "learning_rate": 5.428571428571428e-05,
+      "loss": 0.323,
+      "step": 170
+    },
+    {
+      "epoch": 4.722222222222222,
+      "eval_loss": 0.335285484790802,
+      "eval_runtime": 0.1056,
+      "eval_samples_per_second": 435.635,
+      "eval_steps_per_second": 9.47,
+      "step": 170
+    },
+    {
+      "epoch": 4.75,
+      "grad_norm": 0.5611714720726013,
+      "learning_rate": 5.4000000000000005e-05,
+      "loss": 0.3249,
+      "step": 171
+    },
+    {
+      "epoch": 4.777777777777778,
+      "grad_norm": 0.6146838068962097,
+      "learning_rate": 5.3714285714285714e-05,
+      "loss": 0.3213,
+      "step": 172
+    },
+    {
+      "epoch": 4.805555555555555,
+      "grad_norm": 0.6939036846160889,
+      "learning_rate": 5.342857142857143e-05,
+      "loss": 0.3239,
+      "step": 173
+    },
+    {
+      "epoch": 4.833333333333333,
+      "grad_norm": 0.7213876843452454,
+      "learning_rate": 5.314285714285715e-05,
+      "loss": 0.3234,
+      "step": 174
+    },
+    {
+      "epoch": 4.861111111111111,
+      "grad_norm": 0.6637408137321472,
+      "learning_rate": 5.285714285714286e-05,
+      "loss": 0.3186,
+      "step": 175
+    },
+    {
+      "epoch": 4.888888888888889,
+      "grad_norm": 0.6469508409500122,
+      "learning_rate": 5.257142857142857e-05,
+      "loss": 0.3184,
+      "step": 176
+    },
+    {
+      "epoch": 4.916666666666667,
+      "grad_norm": 0.6262702941894531,
+      "learning_rate": 5.2285714285714294e-05,
+      "loss": 0.3171,
+      "step": 177
+    },
+    {
+      "epoch": 4.944444444444445,
+      "grad_norm": 0.6692309379577637,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.3271,
+      "step": 178
+    },
+    {
+      "epoch": 4.972222222222222,
+      "grad_norm": 0.611004114151001,
+      "learning_rate": 5.171428571428571e-05,
+      "loss": 0.3229,
+      "step": 179
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 0.9707463383674622,
+      "learning_rate": 5.142857142857143e-05,
+      "loss": 0.3223,
+      "step": 180
+    },
+    {
+      "epoch": 5.0,
+      "eval_loss": 0.33622071146965027,
+      "eval_runtime": 0.105,
+      "eval_samples_per_second": 437.997,
+      "eval_steps_per_second": 9.522,
+      "step": 180
+    },
+    {
+      "epoch": 5.027777777777778,
+      "grad_norm": 0.43101295828819275,
+      "learning_rate": 5.1142857142857145e-05,
+      "loss": 0.3106,
+      "step": 181
+    },
+    {
+      "epoch": 5.055555555555555,
+      "grad_norm": 0.7981957793235779,
+      "learning_rate": 5.085714285714286e-05,
+      "loss": 0.3139,
+      "step": 182
+    },
+    {
+      "epoch": 5.083333333333333,
+      "grad_norm": 0.9149967432022095,
+      "learning_rate": 5.057142857142857e-05,
+      "loss": 0.3137,
+      "step": 183
+    },
+    {
+      "epoch": 5.111111111111111,
+      "grad_norm": 0.8689376711845398,
+      "learning_rate": 5.028571428571429e-05,
+      "loss": 0.3185,
+      "step": 184
+    },
+    {
+      "epoch": 5.138888888888889,
+      "grad_norm": 0.6829914450645447,
+      "learning_rate": 5e-05,
+      "loss": 0.3195,
+      "step": 185
+    },
+    {
+      "epoch": 5.166666666666667,
+      "grad_norm": 0.6187098026275635,
+      "learning_rate": 4.971428571428572e-05,
+      "loss": 0.3139,
+      "step": 186
+    },
+    {
+      "epoch": 5.194444444444445,
+      "grad_norm": 0.8703141212463379,
+      "learning_rate": 4.942857142857143e-05,
+      "loss": 0.3147,
+      "step": 187
+    },
+    {
+      "epoch": 5.222222222222222,
+      "grad_norm": 0.6344360709190369,
+      "learning_rate": 4.9142857142857144e-05,
+      "loss": 0.3162,
+      "step": 188
+    },
+    {
+      "epoch": 5.25,
+      "grad_norm": 0.7499691843986511,
+      "learning_rate": 4.885714285714286e-05,
+      "loss": 0.3228,
+      "step": 189
+    },
+    {
+      "epoch": 5.277777777777778,
+      "grad_norm": 0.7664843201637268,
+      "learning_rate": 4.8571428571428576e-05,
+      "loss": 0.3152,
+      "step": 190
+    },
+    {
+      "epoch": 5.277777777777778,
+      "eval_loss": 0.3364305794239044,
+      "eval_runtime": 0.1056,
+      "eval_samples_per_second": 435.778,
+      "eval_steps_per_second": 9.473,
+      "step": 190
+    },
+    {
+      "epoch": 5.305555555555555,
+      "grad_norm": 0.6158504486083984,
+      "learning_rate": 4.828571428571429e-05,
+      "loss": 0.3154,
+      "step": 191
+    },
+    {
+      "epoch": 5.333333333333333,
+      "grad_norm": 0.8614490032196045,
+      "learning_rate": 4.8e-05,
+      "loss": 0.3206,
+      "step": 192
+    },
+    {
+      "epoch": 5.361111111111111,
+      "grad_norm": 0.7699540257453918,
+      "learning_rate": 4.771428571428572e-05,
+      "loss": 0.3159,
+      "step": 193
+    },
+    {
+      "epoch": 5.388888888888889,
+      "grad_norm": 0.9598901867866516,
+      "learning_rate": 4.742857142857143e-05,
+      "loss": 0.3186,
+      "step": 194
+    },
+    {
+      "epoch": 5.416666666666667,
+      "grad_norm": 0.855253279209137,
+      "learning_rate": 4.714285714285714e-05,
+      "loss": 0.3118,
+      "step": 195
+    },
+    {
+      "epoch": 5.444444444444445,
+      "grad_norm": 0.6478847861289978,
+      "learning_rate": 4.685714285714286e-05,
+      "loss": 0.3178,
+      "step": 196
+    },
+    {
+      "epoch": 5.472222222222222,
+      "grad_norm": 0.8028067946434021,
+      "learning_rate": 4.6571428571428575e-05,
+      "loss": 0.3236,
+      "step": 197
+    },
+    {
+      "epoch": 5.5,
+      "grad_norm": 0.7795782089233398,
+      "learning_rate": 4.628571428571429e-05,
+      "loss": 0.3147,
+      "step": 198
+    },
+    {
+      "epoch": 5.527777777777778,
+      "grad_norm": 0.7845653891563416,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 0.3221,
+      "step": 199
+    },
+    {
+      "epoch": 5.555555555555555,
+      "grad_norm": 1.1422370672225952,
+      "learning_rate": 4.5714285714285716e-05,
+      "loss": 0.321,
+      "step": 200
+    },
+    {
+      "epoch": 5.555555555555555,
+      "eval_loss": 0.33535492420196533,
+      "eval_runtime": 0.1051,
+      "eval_samples_per_second": 437.57,
+      "eval_steps_per_second": 9.512,
+      "step": 200
+    },
+    {
+      "epoch": 5.583333333333333,
+      "grad_norm": 0.7386415600776672,
+      "learning_rate": 4.542857142857143e-05,
+      "loss": 0.3183,
+      "step": 201
+    },
+    {
+      "epoch": 5.611111111111111,
+      "grad_norm": 0.6756716966629028,
+      "learning_rate": 4.514285714285714e-05,
+      "loss": 0.3205,
+      "step": 202
+    },
+    {
+      "epoch": 5.638888888888889,
+      "grad_norm": 0.7116839289665222,
+      "learning_rate": 4.485714285714286e-05,
+      "loss": 0.3195,
+      "step": 203
+    },
+    {
+      "epoch": 5.666666666666667,
+      "grad_norm": 0.7919530272483826,
+      "learning_rate": 4.4571428571428574e-05,
+      "loss": 0.3248,
+      "step": 204
+    },
+    {
+      "epoch": 5.694444444444445,
+      "grad_norm": 0.5152342319488525,
+      "learning_rate": 4.428571428571428e-05,
+      "loss": 0.3124,
+      "step": 205
+    },
+    {
+      "epoch": 5.722222222222222,
+      "grad_norm": 0.9519732594490051,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.3205,
+      "step": 206
+    },
+    {
+      "epoch": 5.75,
+      "grad_norm": 0.7759018540382385,
+      "learning_rate": 4.371428571428572e-05,
+      "loss": 0.3188,
+      "step": 207
+    },
+    {
+      "epoch": 5.777777777777778,
+      "grad_norm": 0.931468665599823,
+      "learning_rate": 4.342857142857143e-05,
+      "loss": 0.3195,
+      "step": 208
+    },
+    {
+      "epoch": 5.805555555555555,
+      "grad_norm": 1.0431036949157715,
+      "learning_rate": 4.314285714285715e-05,
+      "loss": 0.3256,
+      "step": 209
+    },
+    {
+      "epoch": 5.833333333333333,
+      "grad_norm": 0.6952974200248718,
+      "learning_rate": 4.2857142857142856e-05,
+      "loss": 0.3189,
+      "step": 210
+    },
+    {
+      "epoch": 5.833333333333333,
+      "eval_loss": 0.33737656474113464,
+      "eval_runtime": 0.1056,
+      "eval_samples_per_second": 435.436,
+      "eval_steps_per_second": 9.466,
+      "step": 210
+    },
+    {
+      "epoch": 5.861111111111111,
+      "grad_norm": 0.880944013595581,
+      "learning_rate": 4.257142857142857e-05,
+      "loss": 0.3159,
+      "step": 211
+    },
+    {
+      "epoch": 5.888888888888889,
+      "grad_norm": 0.6606886982917786,
+      "learning_rate": 4.228571428571429e-05,
+      "loss": 0.3177,
+      "step": 212
+    },
+    {
+      "epoch": 5.916666666666667,
+      "grad_norm": 0.7617785930633545,
+      "learning_rate": 4.2e-05,
+      "loss": 0.317,
+      "step": 213
+    },
+    {
+      "epoch": 5.944444444444445,
+      "grad_norm": 0.6381199955940247,
+      "learning_rate": 4.1714285714285714e-05,
+      "loss": 0.313,
+      "step": 214
+    },
+    {
+      "epoch": 5.972222222222222,
+      "grad_norm": 0.6644593477249146,
+      "learning_rate": 4.1428571428571437e-05,
+      "loss": 0.3184,
+      "step": 215
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 0.9742204546928406,
+      "learning_rate": 4.1142857142857146e-05,
+      "loss": 0.3103,
+      "step": 216
+    },
+    {
+      "epoch": 6.027777777777778,
+      "grad_norm": 0.6401204466819763,
+      "learning_rate": 4.085714285714286e-05,
+      "loss": 0.311,
+      "step": 217
+    },
+    {
+      "epoch": 6.055555555555555,
+      "grad_norm": 0.785503089427948,
+      "learning_rate": 4.057142857142857e-05,
+      "loss": 0.3155,
+      "step": 218
+    },
+    {
+      "epoch": 6.083333333333333,
+      "grad_norm": 0.5417785048484802,
+      "learning_rate": 4.028571428571429e-05,
+      "loss": 0.3109,
+      "step": 219
+    },
+    {
+      "epoch": 6.111111111111111,
+      "grad_norm": 0.6186631321907043,
+      "learning_rate": 4e-05,
+      "loss": 0.3114,
+      "step": 220
+    },
+    {
+      "epoch": 6.111111111111111,
+      "eval_loss": 0.32762280106544495,
+      "eval_runtime": 0.1058,
+      "eval_samples_per_second": 434.862,
+      "eval_steps_per_second": 9.454,
+      "step": 220
+    },
+    {
+      "epoch": 6.138888888888889,
+      "grad_norm": 0.6042884588241577,
+      "learning_rate": 3.971428571428571e-05,
+      "loss": 0.3099,
+      "step": 221
+    },
+    {
+      "epoch": 6.166666666666667,
+      "grad_norm": 0.5823580026626587,
+      "learning_rate": 3.942857142857143e-05,
+      "loss": 0.3104,
+      "step": 222
+    },
+    {
+      "epoch": 6.194444444444445,
+      "grad_norm": 0.6572258472442627,
+      "learning_rate": 3.9142857142857145e-05,
+      "loss": 0.3154,
+      "step": 223
+    },
+    {
+      "epoch": 6.222222222222222,
+      "grad_norm": 0.565834641456604,
+      "learning_rate": 3.885714285714286e-05,
+      "loss": 0.3121,
+      "step": 224
+    },
+    {
+      "epoch": 6.25,
+      "grad_norm": 0.7184673547744751,
+      "learning_rate": 3.857142857142858e-05,
+      "loss": 0.3119,
+      "step": 225
+    },
+    {
+      "epoch": 6.277777777777778,
+      "grad_norm": 0.7170347571372986,
+      "learning_rate": 3.8285714285714286e-05,
+      "loss": 0.3158,
+      "step": 226
+    },
+    {
+      "epoch": 6.305555555555555,
+      "grad_norm": 0.6102560758590698,
+      "learning_rate": 3.8e-05,
+      "loss": 0.3038,
+      "step": 227
+    },
+    {
+      "epoch": 6.333333333333333,
+      "grad_norm": 0.7612823843955994,
+      "learning_rate": 3.771428571428572e-05,
+      "loss": 0.3124,
+      "step": 228
+    },
+    {
+      "epoch": 6.361111111111111,
+      "grad_norm": 0.6277872920036316,
+      "learning_rate": 3.742857142857143e-05,
+      "loss": 0.3028,
+      "step": 229
+    },
+    {
+      "epoch": 6.388888888888889,
+      "grad_norm": 0.7007192373275757,
+      "learning_rate": 3.7142857142857143e-05,
+      "loss": 0.3202,
+      "step": 230
+    },
+    {
+      "epoch": 6.388888888888889,
+      "eval_loss": 0.33139991760253906,
+      "eval_runtime": 0.1062,
+      "eval_samples_per_second": 433.311,
+      "eval_steps_per_second": 9.42,
+      "step": 230
+    },
+    {
+      "epoch": 6.416666666666667,
+      "grad_norm": 0.6396629810333252,
+      "learning_rate": 3.685714285714286e-05,
+      "loss": 0.313,
+      "step": 231
+    },
+    {
+      "epoch": 6.444444444444445,
+      "grad_norm": 0.5031012892723083,
+      "learning_rate": 3.6571428571428576e-05,
+      "loss": 0.3116,
+      "step": 232
+    },
+    {
+      "epoch": 6.472222222222222,
+      "grad_norm": 0.7323219776153564,
+      "learning_rate": 3.628571428571429e-05,
+      "loss": 0.3172,
+      "step": 233
+    },
+    {
+      "epoch": 6.5,
+      "grad_norm": 0.9094661474227905,
+      "learning_rate": 3.6e-05,
+      "loss": 0.3103,
+      "step": 234
+    },
+    {
+      "epoch": 6.527777777777778,
+      "grad_norm": 0.5560885667800903,
+      "learning_rate": 3.571428571428572e-05,
+      "loss": 0.3056,
+      "step": 235
+    },
+    {
+      "epoch": 6.555555555555555,
+      "grad_norm": 1.0145907402038574,
+      "learning_rate": 3.5428571428571426e-05,
+      "loss": 0.3096,
+      "step": 236
+    },
+    {
+      "epoch": 6.583333333333333,
+      "grad_norm": 0.8287002444267273,
+      "learning_rate": 3.514285714285714e-05,
+      "loss": 0.3049,
+      "step": 237
+    },
+    {
+      "epoch": 6.611111111111111,
+      "grad_norm": 0.5207920074462891,
+      "learning_rate": 3.485714285714286e-05,
+      "loss": 0.3047,
+      "step": 238
+    },
+    {
+      "epoch": 6.638888888888889,
+      "grad_norm": 1.065272331237793,
+      "learning_rate": 3.4571428571428574e-05,
+      "loss": 0.3156,
+      "step": 239
+    },
+    {
+      "epoch": 6.666666666666667,
+      "grad_norm": 0.6712301969528198,
+      "learning_rate": 3.428571428571429e-05,
+      "loss": 0.3055,
+      "step": 240
+    },
+    {
+      "epoch": 6.666666666666667,
+      "eval_loss": 0.3387901484966278,
+      "eval_runtime": 0.1053,
+      "eval_samples_per_second": 436.76,
+      "eval_steps_per_second": 9.495,
+      "step": 240
+    },
+    {
+      "epoch": 6.694444444444445,
+      "grad_norm": 0.9477025866508484,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 0.3164,
+      "step": 241
+    },
+    {
+      "epoch": 6.722222222222222,
+      "grad_norm": 1.179295301437378,
+      "learning_rate": 3.3714285714285716e-05,
+      "loss": 0.3263,
+      "step": 242
+    },
+    {
+      "epoch": 6.75,
+      "grad_norm": 0.6969395875930786,
+      "learning_rate": 3.342857142857143e-05,
+      "loss": 0.3136,
+      "step": 243
+    },
+    {
+      "epoch": 6.777777777777778,
+      "grad_norm": 0.7752478718757629,
+      "learning_rate": 3.314285714285714e-05,
+      "loss": 0.3143,
+      "step": 244
+    },
+    {
+      "epoch": 6.805555555555555,
+      "grad_norm": 0.7640174031257629,
+      "learning_rate": 3.285714285714286e-05,
+      "loss": 0.3143,
+      "step": 245
+    },
+    {
+      "epoch": 6.833333333333333,
+      "grad_norm": 0.9904161691665649,
+      "learning_rate": 3.257142857142857e-05,
+      "loss": 0.3093,
+      "step": 246
+    },
+    {
+      "epoch": 6.861111111111111,
+      "grad_norm": 0.7707158923149109,
+      "learning_rate": 3.228571428571428e-05,
+      "loss": 0.3185,
+      "step": 247
+    },
+    {
+      "epoch": 6.888888888888889,
+      "grad_norm": 0.8660508394241333,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.3122,
+      "step": 248
+    },
+    {
+      "epoch": 6.916666666666667,
+      "grad_norm": 0.7438889741897583,
+      "learning_rate": 3.1714285714285715e-05,
+      "loss": 0.3148,
+      "step": 249
+    },
+    {
+      "epoch": 6.944444444444445,
+      "grad_norm": 0.5746331810951233,
+      "learning_rate": 3.142857142857143e-05,
+      "loss": 0.3137,
+      "step": 250
+    },
+    {
+      "epoch": 6.944444444444445,
+      "eval_loss": 0.32808226346969604,
+      "eval_runtime": 0.1059,
+      "eval_samples_per_second": 434.438,
+      "eval_steps_per_second": 9.444,
+      "step": 250
+    },
+    {
+      "epoch": 6.972222222222222,
+      "grad_norm": 0.7158817052841187,
+      "learning_rate": 3.114285714285715e-05,
+      "loss": 0.3125,
+      "step": 251
+    },
+    {
+      "epoch": 7.0,
+      "grad_norm": 0.8010092973709106,
+      "learning_rate": 3.0857142857142856e-05,
+      "loss": 0.3094,
+      "step": 252
+    },
+    {
+      "epoch": 7.027777777777778,
+      "grad_norm": 0.7418866157531738,
+      "learning_rate": 3.057142857142857e-05,
+      "loss": 0.3067,
+      "step": 253
+    },
+    {
+      "epoch": 7.055555555555555,
+      "grad_norm": 0.6731083989143372,
+      "learning_rate": 3.0285714285714288e-05,
+      "loss": 0.3068,
+      "step": 254
+    },
+    {
+      "epoch": 7.083333333333333,
+      "grad_norm": 0.6405408382415771,
+      "learning_rate": 3e-05,
+      "loss": 0.3075,
+      "step": 255
+    },
+    {
+      "epoch": 7.111111111111111,
+      "grad_norm": 0.6403458118438721,
+      "learning_rate": 2.9714285714285717e-05,
+      "loss": 0.3096,
+      "step": 256
+    },
+    {
+      "epoch": 7.138888888888889,
+      "grad_norm": 0.7583682537078857,
+      "learning_rate": 2.9428571428571426e-05,
+      "loss": 0.3103,
+      "step": 257
+    },
+    {
+      "epoch": 7.166666666666667,
+      "grad_norm": 0.8137710094451904,
+      "learning_rate": 2.9142857142857146e-05,
+      "loss": 0.3064,
+      "step": 258
+    },
+    {
+      "epoch": 7.194444444444445,
+      "grad_norm": 0.7179896235466003,
+      "learning_rate": 2.885714285714286e-05,
+      "loss": 0.3067,
+      "step": 259
+    },
+    {
+      "epoch": 7.222222222222222,
+      "grad_norm": 0.9344987273216248,
+      "learning_rate": 2.857142857142857e-05,
+      "loss": 0.3081,
+      "step": 260
+    },
+    {
+      "epoch": 7.222222222222222,
+      "eval_loss": 0.33135512471199036,
+      "eval_runtime": 0.1054,
+      "eval_samples_per_second": 436.494,
+      "eval_steps_per_second": 9.489,
+      "step": 260
+    },
+    {
+      "epoch": 7.25,
+      "grad_norm": 0.7846106886863708,
+      "learning_rate": 2.8285714285714287e-05,
+      "loss": 0.3024,
+      "step": 261
+    },
+    {
+      "epoch": 7.277777777777778,
+      "grad_norm": 0.5884716510772705,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.3062,
+      "step": 262
+    },
+    {
+      "epoch": 7.305555555555555,
+      "grad_norm": 0.7277001738548279,
+      "learning_rate": 2.7714285714285716e-05,
+      "loss": 0.3017,
+      "step": 263
+    },
+    {
+      "epoch": 7.333333333333333,
+      "grad_norm": 0.6671104431152344,
+      "learning_rate": 2.742857142857143e-05,
+      "loss": 0.3109,
+      "step": 264
+    },
+    {
+      "epoch": 7.361111111111111,
+      "grad_norm": 0.6468051671981812,
+      "learning_rate": 2.714285714285714e-05,
+      "loss": 0.3051,
+      "step": 265
+    },
+    {
+      "epoch": 7.388888888888889,
+      "grad_norm": 0.7413132190704346,
+      "learning_rate": 2.6857142857142857e-05,
+      "loss": 0.3059,
+      "step": 266
+    },
+    {
+      "epoch": 7.416666666666667,
+      "grad_norm": 0.8842555284500122,
+      "learning_rate": 2.6571428571428576e-05,
+      "loss": 0.3108,
+      "step": 267
+    },
+    {
+      "epoch": 7.444444444444445,
+      "grad_norm": 0.7701683044433594,
+      "learning_rate": 2.6285714285714286e-05,
+      "loss": 0.31,
+      "step": 268
+    },
+    {
+      "epoch": 7.472222222222222,
+      "grad_norm": 0.6261523962020874,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 0.2966,
+      "step": 269
+    },
+    {
+      "epoch": 7.5,
+      "grad_norm": 0.6180337071418762,
+      "learning_rate": 2.5714285714285714e-05,
+      "loss": 0.3063,
+      "step": 270
+    },
+    {
+      "epoch": 7.5,
+      "eval_loss": 0.33175382018089294,
+      "eval_runtime": 0.1062,
+      "eval_samples_per_second": 433.162,
+      "eval_steps_per_second": 9.417,
+      "step": 270
+    },
+    {
+      "epoch": 7.527777777777778,
+      "grad_norm": 0.910743236541748,
+      "learning_rate": 2.542857142857143e-05,
+      "loss": 0.3058,
+      "step": 271
+    },
+    {
+      "epoch": 7.555555555555555,
+      "grad_norm": 0.8947623372077942,
+      "learning_rate": 2.5142857142857147e-05,
+      "loss": 0.3092,
+      "step": 272
+    },
+    {
+      "epoch": 7.583333333333333,
+      "grad_norm": 0.8466401100158691,
+      "learning_rate": 2.485714285714286e-05,
+      "loss": 0.3069,
+      "step": 273
+    },
+    {
+      "epoch": 7.611111111111111,
+      "grad_norm": 0.7808226943016052,
+      "learning_rate": 2.4571428571428572e-05,
+      "loss": 0.316,
+      "step": 274
+    },
+    {
+      "epoch": 7.638888888888889,
+      "grad_norm": 0.6704230308532715,
+      "learning_rate": 2.4285714285714288e-05,
+      "loss": 0.2993,
+      "step": 275
+    },
+    {
+      "epoch": 7.666666666666667,
+      "grad_norm": 0.7090719938278198,
+      "learning_rate": 2.4e-05,
+      "loss": 0.3034,
+      "step": 276
+    },
+    {
+      "epoch": 7.694444444444445,
+      "grad_norm": 0.7552341818809509,
+      "learning_rate": 2.3714285714285717e-05,
+      "loss": 0.3077,
+      "step": 277
+    },
+    {
+      "epoch": 7.722222222222222,
+      "grad_norm": 0.7747870683670044,
+      "learning_rate": 2.342857142857143e-05,
+      "loss": 0.3047,
+      "step": 278
+    },
+    {
+      "epoch": 7.75,
+      "grad_norm": 1.1127567291259766,
+      "learning_rate": 2.3142857142857145e-05,
+      "loss": 0.3105,
+      "step": 279
+    },
+    {
+      "epoch": 7.777777777777778,
+      "grad_norm": 0.7083399891853333,
+      "learning_rate": 2.2857142857142858e-05,
+      "loss": 0.2997,
+      "step": 280
+    },
+    {
+      "epoch": 7.777777777777778,
+      "eval_loss": 0.3296756446361542,
+      "eval_runtime": 0.1057,
+      "eval_samples_per_second": 435.219,
+      "eval_steps_per_second": 9.461,
+      "step": 280
+    },
+    {
+      "epoch": 7.805555555555555,
+      "grad_norm": 0.5560160279273987,
+      "learning_rate": 2.257142857142857e-05,
+      "loss": 0.3035,
+      "step": 281
+    },
+    {
+      "epoch": 7.833333333333333,
+      "grad_norm": 0.7277750372886658,
+      "learning_rate": 2.2285714285714287e-05,
+      "loss": 0.3066,
+      "step": 282
+    },
+    {
+      "epoch": 7.861111111111111,
+      "grad_norm": 0.683189868927002,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 0.3049,
+      "step": 283
+    },
+    {
+      "epoch": 7.888888888888889,
+      "grad_norm": 0.7173867225646973,
+      "learning_rate": 2.1714285714285715e-05,
+      "loss": 0.3098,
+      "step": 284
+    },
+    {
+      "epoch": 7.916666666666667,
+      "grad_norm": 0.698379397392273,
+      "learning_rate": 2.1428571428571428e-05,
+      "loss": 0.3081,
+      "step": 285
+    },
+    {
+      "epoch": 7.944444444444445,
+      "grad_norm": 0.5789396166801453,
+      "learning_rate": 2.1142857142857144e-05,
+      "loss": 0.3109,
+      "step": 286
+    },
+    {
+      "epoch": 7.972222222222222,
+      "grad_norm": 0.8007158637046814,
+      "learning_rate": 2.0857142857142857e-05,
+      "loss": 0.3149,
+      "step": 287
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 0.8773916959762573,
+      "learning_rate": 2.0571428571428573e-05,
+      "loss": 0.3087,
+      "step": 288
+    },
+    {
+      "epoch": 8.027777777777779,
+      "grad_norm": 0.7212807536125183,
+      "learning_rate": 2.0285714285714286e-05,
+      "loss": 0.2989,
+      "step": 289
+    },
+    {
+      "epoch": 8.055555555555555,
+      "grad_norm": 0.6096076369285583,
+      "learning_rate": 2e-05,
+      "loss": 0.3069,
+      "step": 290
+    },
+    {
+      "epoch": 8.055555555555555,
+      "eval_loss": 0.33069342374801636,
+      "eval_runtime": 0.105,
+      "eval_samples_per_second": 438.062,
+      "eval_steps_per_second": 9.523,
+      "step": 290
+    },
+    {
+      "epoch": 8.083333333333334,
+      "grad_norm": 0.5173125267028809,
+      "learning_rate": 1.9714285714285714e-05,
+      "loss": 0.295,
+      "step": 291
+    },
+    {
+      "epoch": 8.11111111111111,
+      "grad_norm": 0.6913369297981262,
+      "learning_rate": 1.942857142857143e-05,
+      "loss": 0.3014,
+      "step": 292
+    },
+    {
+      "epoch": 8.13888888888889,
+      "grad_norm": 0.7195921540260315,
+      "learning_rate": 1.9142857142857143e-05,
+      "loss": 0.3037,
+      "step": 293
+    },
+    {
+      "epoch": 8.166666666666666,
+      "grad_norm": 0.6366473436355591,
+      "learning_rate": 1.885714285714286e-05,
+      "loss": 0.3031,
+      "step": 294
+    },
+    {
+      "epoch": 8.194444444444445,
+      "grad_norm": 0.5457173585891724,
+      "learning_rate": 1.8571428571428572e-05,
+      "loss": 0.2957,
+      "step": 295
+    },
+    {
+      "epoch": 8.222222222222221,
+      "grad_norm": 0.6149912476539612,
+      "learning_rate": 1.8285714285714288e-05,
+      "loss": 0.2997,
+      "step": 296
+    },
+    {
+      "epoch": 8.25,
+      "grad_norm": 0.5352884531021118,
+      "learning_rate": 1.8e-05,
+      "loss": 0.3048,
+      "step": 297
+    },
+    {
+      "epoch": 8.277777777777779,
+      "grad_norm": 0.6278409361839294,
+      "learning_rate": 1.7714285714285713e-05,
+      "loss": 0.308,
+      "step": 298
+    },
+    {
+      "epoch": 8.305555555555555,
+      "grad_norm": 0.5881698727607727,
+      "learning_rate": 1.742857142857143e-05,
+      "loss": 0.3005,
+      "step": 299
+    },
+    {
+      "epoch": 8.333333333333334,
+      "grad_norm": 0.6125136613845825,
+      "learning_rate": 1.7142857142857145e-05,
+      "loss": 0.302,
+      "step": 300
+    },
+    {
+      "epoch": 8.333333333333334,
+      "eval_loss": 0.3336547017097473,
+      "eval_runtime": 0.1055,
+      "eval_samples_per_second": 436.2,
+      "eval_steps_per_second": 9.483,
+      "step": 300
+    },
+    {
+      "epoch": 8.36111111111111,
+      "grad_norm": 0.6722866892814636,
+      "learning_rate": 1.6857142857142858e-05,
+      "loss": 0.3012,
+      "step": 301
+    },
+    {
+      "epoch": 8.38888888888889,
+      "grad_norm": 0.6827422976493835,
+      "learning_rate": 1.657142857142857e-05,
+      "loss": 0.297,
+      "step": 302
+    },
+    {
+      "epoch": 8.416666666666666,
+      "grad_norm": 0.7612675428390503,
+      "learning_rate": 1.6285714285714287e-05,
+      "loss": 0.2977,
+      "step": 303
+    },
+    {
+      "epoch": 8.444444444444445,
+      "grad_norm": 0.5952971577644348,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.3009,
+      "step": 304
+    },
+    {
+      "epoch": 8.472222222222221,
+      "grad_norm": 0.8323265314102173,
+      "learning_rate": 1.5714285714285715e-05,
+      "loss": 0.3043,
+      "step": 305
+    },
+    {
+      "epoch": 8.5,
+      "grad_norm": 0.8321357369422913,
+      "learning_rate": 1.5428571428571428e-05,
+      "loss": 0.2956,
+      "step": 306
+    },
+    {
+      "epoch": 8.527777777777779,
+      "grad_norm": 0.6457182168960571,
+      "learning_rate": 1.5142857142857144e-05,
+      "loss": 0.3029,
+      "step": 307
+    },
+    {
+      "epoch": 8.555555555555555,
+      "grad_norm": 0.5753086805343628,
+      "learning_rate": 1.4857142857142858e-05,
+      "loss": 0.2972,
+      "step": 308
+    },
+    {
+      "epoch": 8.583333333333334,
+      "grad_norm": 0.8767444491386414,
+      "learning_rate": 1.4571428571428573e-05,
+      "loss": 0.2966,
+      "step": 309
+    },
+    {
+      "epoch": 8.61111111111111,
+      "grad_norm": 0.929669201374054,
+      "learning_rate": 1.4285714285714285e-05,
+      "loss": 0.3049,
+      "step": 310
+    },
+    {
+      "epoch": 8.61111111111111,
+      "eval_loss": 0.3303754925727844,
+      "eval_runtime": 0.1057,
+      "eval_samples_per_second": 435.106,
+      "eval_steps_per_second": 9.459,
+      "step": 310
+    },
+    {
+      "epoch": 8.63888888888889,
+      "grad_norm": 0.7576697468757629,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 0.2989,
+      "step": 311
+    },
+    {
+      "epoch": 8.666666666666666,
+      "grad_norm": 0.6402246952056885,
+      "learning_rate": 1.3714285714285716e-05,
+      "loss": 0.3051,
+      "step": 312
+    },
+    {
+      "epoch": 8.694444444444445,
+      "grad_norm": 0.5665248036384583,
+      "learning_rate": 1.3428571428571429e-05,
+      "loss": 0.2974,
+      "step": 313
+    },
+    {
+      "epoch": 8.722222222222221,
+      "grad_norm": 0.9747456312179565,
+      "learning_rate": 1.3142857142857143e-05,
+      "loss": 0.3061,
+      "step": 314
+    },
+    {
+      "epoch": 8.75,
+      "grad_norm": 0.657123863697052,
+      "learning_rate": 1.2857142857142857e-05,
+      "loss": 0.3012,
+      "step": 315
+    },
+    {
+      "epoch": 8.777777777777779,
+      "grad_norm": 0.7186892032623291,
+      "learning_rate": 1.2571428571428573e-05,
+      "loss": 0.2995,
+      "step": 316
+    },
+    {
+      "epoch": 8.805555555555555,
+      "grad_norm": 0.6889364123344421,
+      "learning_rate": 1.2285714285714286e-05,
+      "loss": 0.3026,
+      "step": 317
+    },
+    {
+      "epoch": 8.833333333333334,
+      "grad_norm": 0.6299145817756653,
+      "learning_rate": 1.2e-05,
+      "loss": 0.3009,
+      "step": 318
+    },
+    {
+      "epoch": 8.86111111111111,
+      "grad_norm": 0.7328559756278992,
+      "learning_rate": 1.1714285714285715e-05,
+      "loss": 0.3002,
+      "step": 319
+    },
+    {
+      "epoch": 8.88888888888889,
+      "grad_norm": 0.6111913919448853,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 0.2953,
+      "step": 320
+    },
+    {
+      "epoch": 8.88888888888889,
+      "eval_loss": 0.3253832757472992,
+      "eval_runtime": 0.106,
+      "eval_samples_per_second": 434.082,
+      "eval_steps_per_second": 9.437,
+      "step": 320
+    },
+    {
+      "epoch": 8.916666666666666,
+      "grad_norm": 0.6739629507064819,
+      "learning_rate": 1.1142857142857143e-05,
+      "loss": 0.3023,
+      "step": 321
+    },
+    {
+      "epoch": 8.944444444444445,
+      "grad_norm": 0.6967675685882568,
+      "learning_rate": 1.0857142857142858e-05,
+      "loss": 0.2978,
+      "step": 322
+    },
+    {
+      "epoch": 8.972222222222221,
+      "grad_norm": 0.702989935874939,
+      "learning_rate": 1.0571428571428572e-05,
+      "loss": 0.3043,
+      "step": 323
+    },
+    {
+      "epoch": 9.0,
+      "grad_norm": 1.156525731086731,
+      "learning_rate": 1.0285714285714286e-05,
+      "loss": 0.3034,
+      "step": 324
+    },
+    {
+      "epoch": 9.027777777777779,
+      "grad_norm": 0.564460277557373,
+      "learning_rate": 1e-05,
+      "loss": 0.2989,
+      "step": 325
+    },
+    {
+      "epoch": 9.055555555555555,
+      "grad_norm": 0.5435044169425964,
+      "learning_rate": 9.714285714285715e-06,
+      "loss": 0.2955,
+      "step": 326
+    },
+    {
+      "epoch": 9.083333333333334,
+      "grad_norm": 0.511762797832489,
+      "learning_rate": 9.42857142857143e-06,
+      "loss": 0.2986,
+      "step": 327
+    },
+    {
+      "epoch": 9.11111111111111,
+      "grad_norm": 0.6208844780921936,
+      "learning_rate": 9.142857142857144e-06,
+      "loss": 0.2932,
+      "step": 328
+    },
+    {
+      "epoch": 9.13888888888889,
+      "grad_norm": 0.5209355354309082,
+      "learning_rate": 8.857142857142857e-06,
+      "loss": 0.2932,
+      "step": 329
+    },
+    {
+      "epoch": 9.166666666666666,
+      "grad_norm": 0.5852081775665283,
+      "learning_rate": 8.571428571428573e-06,
+      "loss": 0.302,
+      "step": 330
+    },
+    {
+      "epoch": 9.166666666666666,
+      "eval_loss": 0.32703322172164917,
+      "eval_runtime": 0.1055,
+      "eval_samples_per_second": 435.944,
+      "eval_steps_per_second": 9.477,
+      "step": 330
+    },
+    {
+      "epoch": 9.194444444444445,
+      "grad_norm": 0.6613155603408813,
+      "learning_rate": 8.285714285714285e-06,
+      "loss": 0.2973,
+      "step": 331
+    },
+    {
+      "epoch": 9.222222222222221,
+      "grad_norm": 0.6458805203437805,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.296,
+      "step": 332
+    },
+    {
+      "epoch": 9.25,
+      "grad_norm": 0.5602886080741882,
+      "learning_rate": 7.714285714285714e-06,
+      "loss": 0.2998,
+      "step": 333
+    },
+    {
+      "epoch": 9.277777777777779,
+      "grad_norm": 0.5723817348480225,
+      "learning_rate": 7.428571428571429e-06,
+      "loss": 0.2898,
+      "step": 334
+    },
+    {
+      "epoch": 9.305555555555555,
+      "grad_norm": 0.6257355213165283,
+      "learning_rate": 7.142857142857143e-06,
+      "loss": 0.2935,
+      "step": 335
+    },
+    {
+      "epoch": 9.333333333333334,
+      "grad_norm": 0.6624913811683655,
+      "learning_rate": 6.857142857142858e-06,
+      "loss": 0.2909,
+      "step": 336
+    },
+    {
+      "epoch": 9.36111111111111,
+      "grad_norm": 0.5716632604598999,
+      "learning_rate": 6.5714285714285714e-06,
+      "loss": 0.3001,
+      "step": 337
+    },
+    {
+      "epoch": 9.38888888888889,
+      "grad_norm": 0.6996496319770813,
+      "learning_rate": 6.285714285714287e-06,
+      "loss": 0.2964,
+      "step": 338
+    },
+    {
+      "epoch": 9.416666666666666,
+      "grad_norm": 0.7235862612724304,
+      "learning_rate": 6e-06,
+      "loss": 0.2927,
+      "step": 339
+    },
+    {
+      "epoch": 9.444444444444445,
+      "grad_norm": 0.6455687284469604,
+      "learning_rate": 5.7142857142857145e-06,
+      "loss": 0.2956,
+      "step": 340
+    },
+    {
+      "epoch": 9.444444444444445,
+      "eval_loss": 0.32588478922843933,
+      "eval_runtime": 0.1057,
+      "eval_samples_per_second": 435.067,
+      "eval_steps_per_second": 9.458,
+      "step": 340
+    },
+    {
+      "epoch": 9.472222222222221,
+      "grad_norm": 0.5984405279159546,
+      "learning_rate": 5.428571428571429e-06,
+      "loss": 0.2959,
+      "step": 341
+    },
+    {
+      "epoch": 9.5,
+      "grad_norm": 0.6240008473396301,
+      "learning_rate": 5.142857142857143e-06,
+      "loss": 0.2936,
+      "step": 342
+    },
+    {
+      "epoch": 9.527777777777779,
+      "grad_norm": 0.6871291399002075,
+      "learning_rate": 4.857142857142858e-06,
+      "loss": 0.2987,
+      "step": 343
+    },
+    {
+      "epoch": 9.555555555555555,
+      "grad_norm": 0.6369628310203552,
+      "learning_rate": 4.571428571428572e-06,
+      "loss": 0.2941,
+      "step": 344
+    },
+    {
+      "epoch": 9.583333333333334,
+      "grad_norm": 0.6651211977005005,
+      "learning_rate": 4.285714285714286e-06,
+      "loss": 0.2959,
+      "step": 345
+    },
+    {
+      "epoch": 9.61111111111111,
+      "grad_norm": 0.7005758285522461,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.2991,
+      "step": 346
+    },
+    {
+      "epoch": 9.63888888888889,
+      "grad_norm": 0.5685088634490967,
+      "learning_rate": 3.7142857142857146e-06,
+      "loss": 0.2936,
+      "step": 347
+    },
+    {
+      "epoch": 9.666666666666666,
+      "grad_norm": 0.6322896480560303,
+      "learning_rate": 3.428571428571429e-06,
+      "loss": 0.2947,
+      "step": 348
+    },
+    {
+      "epoch": 9.694444444444445,
+      "grad_norm": 0.6149244904518127,
+      "learning_rate": 3.1428571428571433e-06,
+      "loss": 0.2951,
+      "step": 349
+    },
+    {
+      "epoch": 9.722222222222221,
+      "grad_norm": 0.685043215751648,
+      "learning_rate": 2.8571428571428573e-06,
+      "loss": 0.2975,
+      "step": 350
+    },
+    {
+      "epoch": 9.722222222222221,
+      "eval_loss": 0.3277616798877716,
+      "eval_runtime": 0.1059,
+      "eval_samples_per_second": 434.21,
+      "eval_steps_per_second": 9.439,
+      "step": 350
+    },
+    {
+      "epoch": 9.75,
+      "grad_norm": 0.8039355874061584,
+      "learning_rate": 2.5714285714285716e-06,
+      "loss": 0.2951,
+      "step": 351
+    },
+    {
+      "epoch": 9.777777777777779,
+      "grad_norm": 0.6388571262359619,
+      "learning_rate": 2.285714285714286e-06,
+      "loss": 0.2928,
+      "step": 352
+    },
+    {
+      "epoch": 9.805555555555555,
+      "grad_norm": 0.5934285521507263,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.2934,
+      "step": 353
+    },
+    {
+      "epoch": 9.833333333333334,
+      "grad_norm": 0.5320731401443481,
+      "learning_rate": 1.7142857142857145e-06,
+      "loss": 0.2952,
+      "step": 354
+    },
+    {
+      "epoch": 9.86111111111111,
+      "grad_norm": 0.6137614846229553,
+      "learning_rate": 1.4285714285714286e-06,
+      "loss": 0.2932,
+      "step": 355
+    },
+    {
+      "epoch": 9.88888888888889,
+      "grad_norm": 0.8172494769096375,
+      "learning_rate": 1.142857142857143e-06,
+      "loss": 0.2944,
+      "step": 356
+    },
+    {
+      "epoch": 9.916666666666666,
+      "grad_norm": 0.6931514739990234,
+      "learning_rate": 8.571428571428572e-07,
+      "loss": 0.2917,
+      "step": 357
+    },
+    {
+      "epoch": 9.944444444444445,
+      "grad_norm": 0.8408763408660889,
+      "learning_rate": 5.714285714285715e-07,
+      "loss": 0.2952,
+      "step": 358
+    },
+    {
+      "epoch": 9.972222222222221,
+      "grad_norm": 0.6687312126159668,
+      "learning_rate": 2.8571428571428575e-07,
+      "loss": 0.2948,
+      "step": 359
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 0.8545575737953186,
+      "learning_rate": 0.0,
+      "loss": 0.2944,
+      "step": 360
+    },
+    {
+      "epoch": 10.0,
+      "eval_loss": 0.3285914361476898,
+      "eval_runtime": 0.1058,
+      "eval_samples_per_second": 434.729,
+      "eval_steps_per_second": 9.451,
+      "step": 360
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 360,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 5000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3831354556001280.0,
+  "train_batch_size": 128,
+  "trial_name": null,
+  "trial_params": null
+}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/checkpoint-360/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e928b416737e3e579d0a564d29d42c7345239d5cd5670a62b4d71260a823a6d5
+size 6008

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/config.json

@@ -0,0 +1,42 @@
+{
+  "_name_or_path": "HuggingFaceTB/SmolLM2-135M-Instruct",
+  "action_dim": 26,
+  "architectures": [
+    "LowdimLlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "head_dim": 64,
+  "hidden_act": "silu",
+  "hidden_size": 576,
+  "initializer_range": 0.041666666666666664,
+  "intermediate_size": 1536,
+  "is_llama_config": true,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "model_type": "llama_lowdim",
+  "num_attention_heads": 9,
+  "num_hidden_layers": 30,
+  "num_key_value_heads": 3,
+  "obs_dim": 46,
+  "pad_token_id": 2,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_interleaved": false,
+  "rope_scaling": null,
+  "rope_theta": 100000,
+  "tie_word_embeddings": true,
+  "torch_dtype": "float32",
+  "transformers.js_config": {
+    "kv_cache_dtype": {
+      "fp16": "float16",
+      "q4f16": "float16"
+    }
+  },
+  "transformers_version": "4.47.1",
+  "use_cache": true,
+  "use_joint_mlp_projector": true,
+  "vocab_size": 49152
+}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 2,
+  "transformers_version": "4.47.1"
+}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:532f9325616dbe87be8846d0b03f80c41a2de5d5a05c7397eb15b3484410ede9
+size 539588496

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/normalizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b3226fbcd5e7e17c04b2cf20350d5f793927d16481bc217684ce129a4aa170a
+size 5666

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/train.log

@@ -0,0 +1,11 @@
+[2026-03-27 16:20:52,880][numexpr.utils][INFO] - Note: NumExpr detected 24 cores but "NUMEXPR_MAX_THREADS" not set, so enforcing safe limit of 8.
+[2026-03-27 16:20:52,880][numexpr.utils][INFO] - NumExpr defaulting to 8 threads.
+[2026-03-27 16:20:55,182][datasets][INFO] - PyTorch version 2.2.2 available.
+[2026-03-27 16:20:55,182][datasets][INFO] - TensorFlow version 2.15.1 available.
+[2026-03-27 16:20:55,182][datasets][INFO] - JAX version 0.4.30 available.
+[2026-03-27 16:20:59,889][datasets.arrow_dataset][WARNING] - Setting TOKENIZERS_PARALLELISM=false for forked processes.
+[2026-03-27 16:21:01,023][datasets.arrow_dataset][WARNING] - Setting TOKENIZERS_PARALLELISM=false for forked processes.
+[2026-03-27 16:21:01,919][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat -Wno-unused-result -Wsign-compare -DNDEBUG -O2 -Wall -fPIC -O2 -isystem /home/chyang/miniconda3/envs/llm-bc/include -I/home/chyang/miniconda3/envs/llm-bc/include -fPIC -O2 -isystem /home/chyang/miniconda3/envs/llm-bc/include -fPIC -c /tmp/tmp7t0nrkuh/test.c -o /tmp/tmp7t0nrkuh/test.o
+[2026-03-27 16:21:01,969][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat /tmp/tmp7t0nrkuh/test.o -laio -o /tmp/tmp7t0nrkuh/a.out
+[2026-03-27 16:21:02,963][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat -Wno-unused-result -Wsign-compare -DNDEBUG -O2 -Wall -fPIC -O2 -isystem /home/chyang/miniconda3/envs/llm-bc/include -I/home/chyang/miniconda3/envs/llm-bc/include -fPIC -O2 -isystem /home/chyang/miniconda3/envs/llm-bc/include -fPIC -c /tmp/tmps_s5vyh9/test.c -o /tmp/tmps_s5vyh9/test.o
+[2026-03-27 16:21:03,018][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat /tmp/tmps_s5vyh9/test.o -L/usr/local/cuda -L/usr/local/cuda/lib64 -lcufile -o /tmp/tmps_s5vyh9/a.out

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/debug-internal.log

@@ -0,0 +1,17 @@
+{"time":"2026-03-27T16:20:56.144221699+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-03-27T16:20:56.144248456+08:00","level":"INFO","msg":"created symlink","path":"/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/logs/debug-core.log"}
+{"time":"2026-03-27T16:20:56.25122394+08:00","level":"INFO","msg":"created new stream","id":"nhmfpc2t"}
+{"time":"2026-03-27T16:20:56.25128922+08:00","level":"INFO","msg":"stream: started","id":"nhmfpc2t"}
+{"time":"2026-03-27T16:20:56.251326647+08:00","level":"INFO","msg":"sender: started","stream_id":"nhmfpc2t"}
+{"time":"2026-03-27T16:20:56.251319685+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"nhmfpc2t"}}
+{"time":"2026-03-27T16:20:56.251303072+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"nhmfpc2t"}}
+{"time":"2026-03-27T16:20:57.417356613+08:00","level":"INFO","msg":"Starting system monitor"}
+{"time":"2026-03-27T16:25:43.814415705+08:00","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2026-03-27T16:25:43.815067495+08:00","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2026-03-27T16:25:44.815119947+08:00","level":"INFO","msg":"handler: operation stats","stats":{"operations":[{"desc":"uploading wandb-summary.json","runtime_seconds":0.136938035,"progress":"495B/495B"},{"desc":"saving job artifact","runtime_seconds":0.037243753}],"total_operations":2}}
+{"time":"2026-03-27T16:25:49.240469481+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-03-27T16:25:50.901076411+08:00","level":"INFO","msg":"stream: closing","id":"nhmfpc2t"}
+{"time":"2026-03-27T16:25:50.901107424+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"nhmfpc2t"}}
+{"time":"2026-03-27T16:25:50.901135119+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"nhmfpc2t"}}
+{"time":"2026-03-27T16:25:50.901152717+08:00","level":"INFO","msg":"sender: closed","stream_id":"nhmfpc2t"}
+{"time":"2026-03-27T16:25:50.90122351+08:00","level":"INFO","msg":"stream: closed","id":"nhmfpc2t"}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/debug.log

@@ -0,0 +1,35 @@
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Configure stats pid to 2703097
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Loading settings from /home/chyang/.config/wandb/settings
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Loading settings from /tmp2/chyang/workspace/LLM-BC/wandb/settings
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/tmp2/chyang/workspace/LLM-BC/train.py', 'program': '/tmp2/chyang/workspace/LLM-BC/./train.py'}
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:_log_setup():533] Logging user logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/logs/debug.log
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:_log_setup():534] Logging internal logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/logs/debug-internal.log
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:init():619] calling init triggers
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_llm_lowdim', '_target_': 'llmbc.workspace.train_llm_workspace.TrainLLMWorkspace', 'obs_dim': 46, 'action_dim': 26, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'task_name': 'adroit-hand-hammer-v1', 'exp_name': 'train llm', 'model_name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'use_quantization': False, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'dataset': {'test_data_ratio': 0.01}, 'debug': False, 'training': {'seed': 42, 'per_device_train_batch_size': 128, 'per_device_eval_batch_size': 128, 'gradient_accumulation_steps': 1, 'optim': 'paged_adamw_32bit', 'num_train_epochs': 10, 'eval_strategy': 'steps', 'logging_steps': 1, 'warmup_steps': 10, 'logging_strategy': 'steps', 'learning_rate': 0.0001, 'fp16': False, 'bf16': True, 'tf32': True, 'group_by_length': True, 'report_to': 'wandb', 'save_steps': 5000, 'eval_steps': 10, 'use_joint_mlp_projector': True, 'joint_obs_action_mlp_lr': 5e-05}, 'trainer': {'obs_dim': 46, 'action_dim': 26, 'use_joint_mlp_projector': True, 'max_seq_length': 100, 'dataset_text_field': 'text', 'packing': False}, 'logging': {'project': 'llm_module_finetuning', 'resume': True, 'mode': 'online', 'name': '2026.03.27-16.20.52_train_llm_lowdim_adroit-hand-hammer-v1', 'tags': ['train_llm_lowdim', 'adroit-hand-hammer-v1', 'train llm'], 'id': None, 'group': None}, 'multi_run': {'run_dir': 'data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1', 'wandb_name_base': '2026.03.27-16.20.52_train_llm_lowdim_adroit-hand-hammer-v1'}, 'task': {'name': 'adroit-hand-hammer-v1', 'obs_dim': 46, 'action_dim': 26, 'env_runner': {'_target_': 'llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner', 'env_name': 'llf-adroit-adroit-hand-hammer-v1', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 150, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset', 'data_path': 'datasets/adroit-hand-hammer-v1-general.pt', 'data_path2': 'datasets/adroit-hand-hammer-v1.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.05, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor'}}, 'llm': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'mlp-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2026.03.27/14.38.20_train_mlp_projector_adroit-hand-hammer-v1/checkpoints/latest.ckpt', 'max_length': 100, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'prompter': {'_target_': 'llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter', 'use_joint_mlp_projector': True}, 'hydra': {'job': {'override_dirname': 'HuggingFaceTB/SmolLM2-135M-Instruct'}, 'run': {'dir': 'data/outputs/2026.03.27/16.20.52_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:init():669] starting backend
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:init():673] sending inform_init request
+2026-03-27 16:20:56,140 INFO    MainThread:2703097 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-03-27 16:20:56,140 INFO    MainThread:2703097 [wandb_init.py:init():686] backend started and connected
+2026-03-27 16:20:56,144 INFO    MainThread:2703097 [wandb_init.py:init():781] updated telemetry
+2026-03-27 16:20:56,170 INFO    MainThread:2703097 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-03-27 16:20:57,414 INFO    MainThread:2703097 [wandb_init.py:init():867] starting run threads in backend
+2026-03-27 16:20:57,521 INFO    MainThread:2703097 [wandb_run.py:_console_start():2451] atexit reg
+2026-03-27 16:20:57,522 INFO    MainThread:2703097 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-03-27 16:20:57,522 INFO    MainThread:2703097 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-03-27 16:20:57,522 INFO    MainThread:2703097 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-03-27 16:20:57,524 INFO    MainThread:2703097 [wandb_init.py:init():911] run started, returning control to user process
+2026-03-27 16:21:04,359 INFO    MainThread:2703097 [wandb_run.py:_config_callback():1389] config_cb None None {'obs_dim': 46, 'action_dim': 26, 'use_joint_mlp_projector': True, 'vocab_size': 49152, 'max_position_embeddings': 8192, 'hidden_size': 576, 'intermediate_size': 1536, 'num_hidden_layers': 30, 'num_attention_heads': 9, 'num_key_value_heads': 3, 'hidden_act': 'silu', 'initializer_range': 0.041666666666666664, 'rms_norm_eps': 1e-05, 'pretraining_tp': 1, 'use_cache': False, 'rope_theta': 100000, 'rope_scaling': None, 'attention_bias': False, 'attention_dropout': 0.0, 'mlp_bias': False, 'head_dim': 64, 'return_dict': True, 'output_hidden_states': False, 'output_attentions': False, 'torchscript': False, 'torch_dtype': 'bfloat16', 'use_bfloat16': False, 'tf_legacy_loss': False, 'pruned_heads': {}, 'tie_word_embeddings': True, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'is_decoder': False, 'cross_attention_hidden_size': None, 'add_cross_attention': False, 'tie_encoder_decoder': False, 'max_length': 20, 'min_length': 0, 'do_sample': False, 'early_stopping': False, 'num_beams': 1, 'num_beam_groups': 1, 'diversity_penalty': 0.0, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 0, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': None, 'forced_eos_token_id': None, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'suppress_tokens': None, 'begin_suppress_tokens': None, 'architectures': ['LlamaForCausalLM'], 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': 1, 'pad_token_id': 2, 'eos_token_id': 2, 'sep_token_id': None, 'decoder_start_token_id': None, 'task_specific_params': None, 'problem_type': None, '_name_or_path': 'HuggingFaceTB/SmolLM2-135M-Instruct', '_attn_implementation_autoset': True, 'transformers_version': '4.47.1', 'is_llama_config': True, 'model_type': 'llama_lowdim', 'rope_interleaved': False, 'transformers.js_config': {'kv_cache_dtype': {'q4f16': 'float16', 'fp16': 'float16'}}, 'output_dir': '/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1', 'overwrite_output_dir': False, 'do_train': False, 'do_eval': True, 'do_predict': False, 'eval_strategy': 'steps', 'prediction_loss_only': False, 'per_device_train_batch_size': 128, 'per_device_eval_batch_size': 128, 'per_gpu_train_batch_size': None, 'per_gpu_eval_batch_size': None, 'gradient_accumulation_steps': 1, 'eval_accumulation_steps': None, 'eval_delay': 0, 'torch_empty_cache_steps': None, 'learning_rate': 0.0001, 'weight_decay': 0.0, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'max_grad_norm': 1.0, 'num_train_epochs': 10, 'max_steps': -1, 'lr_scheduler_type': 'linear', 'lr_scheduler_kwargs': {}, 'warmup_ratio': 0.0, 'warmup_steps': 10, 'log_level': 'passive', 'log_level_replica': 'warning', 'log_on_each_node': True, 'logging_dir': '/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/runs/Mar27_16-21-01_A6000-2', 'logging_strategy': 'steps', 'logging_first_step': False, 'logging_steps': 1, 'logging_nan_inf_filter': True, 'save_strategy': 'steps', 'save_steps': 5000, 'save_total_limit': None, 'save_safetensors': True, 'save_on_each_node': False, 'save_only_model': False, 'restore_callback_states_from_checkpoint': False, 'no_cuda': False, 'use_cpu': False, 'use_mps_device': False, 'seed': 42, 'data_seed': None, 'jit_mode_eval': False, 'use_ipex': False, 'bf16': True, 'fp16': False, 'fp16_opt_level': 'O1', 'half_precision_backend': 'auto', 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': True, 'local_rank': 0, 'ddp_backend': None, 'tpu_num_cores': None, 'tpu_metrics_debug': False, 'debug': [], 'dataloader_drop_last': False, 'eval_steps': 10, 'dataloader_num_workers': 0, 'dataloader_prefetch_factor': None, 'past_index': -1, 'run_name': '/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1', 'disable_tqdm': False, 'remove_unused_columns': True, 'label_names': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'fsdp': [], 'fsdp_min_num_params': 0, 'fsdp_config': {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, 'fsdp_transformer_layer_cls_to_wrap': None, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'deepspeed': None, 'label_smoothing_factor': 0.0, 'optim': 'paged_adamw_32bit', 'optim_args': None, 'adafactor': False, 'group_by_length': True, 'length_column_name': 'length', 'report_to': ['wandb'], 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'skip_memory_metrics': True, 'use_legacy_prediction_loop': False, 'push_to_hub': False, 'resume_from_checkpoint': None, 'hub_model_id': None, 'hub_strategy': 'every_save', 'hub_token': '<HUB_TOKEN>', 'hub_private_repo': None, 'hub_always_push': False, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'include_inputs_for_metrics': False, 'include_for_metrics': [], 'eval_do_concat_batches': True, 'fp16_backend': 'auto', 'evaluation_strategy': None, 'push_to_hub_model_id': None, 'push_to_hub_organization': None, 'push_to_hub_token': '<PUSH_TO_HUB_TOKEN>', 'mp_parameters': '', 'auto_find_batch_size': False, 'full_determinism': False, 'torchdynamo': None, 'ray_scope': 'last', 'ddp_timeout': 1800, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'dispatch_batches': None, 'split_batches': None, 'include_tokens_per_second': False, 'include_num_input_tokens_seen': False, 'neftune_noise_alpha': None, 'optim_target_modules': None, 'batch_eval_metrics': False, 'eval_on_start': False, 'use_liger_kernel': False, 'eval_use_gather_object': False, 'average_tokens_across_devices': False, 'dataset_text_field': 'text', 'packing': False, 'max_seq_length': 100, 'dataset_num_proc': None, 'dataset_batch_size': 1000, 'model_init_kwargs': None, 'dataset_kwargs': {}, 'eval_packing': None, 'num_of_sequences': 1024, 'chars_per_token': '<CHARS_PER_TOKEN>', 'use_liger': False, 'joint_obs_action_mlp_lr': 5e-05, 'obs_mlp_lr': None, 'action_mlp_lr': None}
+2026-03-27 16:21:04,361 INFO    MainThread:2703097 [wandb_config.py:__setitem__():154] config set model/num_parameters = 134889408 - <bound method Run._config_callback of <wandb.sdk.wandb_run.Run object at 0x7e6758745670>>
+2026-03-27 16:21:04,361 INFO    MainThread:2703097 [wandb_run.py:_config_callback():1389] config_cb model/num_parameters 134889408 None
+2026-03-27 16:25:43,813 INFO    MainThread:2703097 [wandb_run.py:_finish():2146] finishing run chyang25-national-taiwan-university/llm_module_finetuning/nhmfpc2t
+2026-03-27 16:25:43,813 INFO    MainThread:2703097 [wandb_run.py:_atexit_cleanup():2414] got exitcode: 0
+2026-03-27 16:25:43,813 INFO    MainThread:2703097 [wandb_run.py:_restore():2396] restore
+2026-03-27 16:25:43,814 INFO    MainThread:2703097 [wandb_run.py:_restore():2402] restore done
+2026-03-27 16:25:50,896 INFO    MainThread:2703097 [wandb_run.py:_footer_history_summary_info():3963] rendering history
+2026-03-27 16:25:50,896 INFO    MainThread:2703097 [wandb_run.py:_footer_history_summary_info():3995] rendering summary
+2026-03-27 16:25:50,900 INFO    MainThread:2703097 [wandb_run.py:_footer_sync_info():3922] logging synced files

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/files/config.yaml

@@ -0,0 +1,711 @@
+_attn_implementation_autoset:
+    value: true
+_name_or_path:
+    value: HuggingFaceTB/SmolLM2-135M-Instruct
+_target_:
+    value: llmbc.workspace.train_llm_workspace.TrainLLMWorkspace
+_wandb:
+    value:
+        cli_version: 0.18.6
+        m:
+            - "1": train/loss
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": train/global_step
+              "6":
+                - 3
+              "7": []
+            - "1": train/learning_rate
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": train/epoch
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": train/grad_norm
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": eval/loss
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": eval/runtime
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": eval/samples_per_second
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": eval/steps_per_second
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+        python_version: 3.9.20
+        t:
+            "1":
+                - 1
+                - 2
+                - 3
+                - 5
+                - 11
+                - 12
+                - 41
+                - 49
+                - 50
+                - 51
+                - 53
+                - 55
+                - 71
+                - 84
+                - 98
+            "2":
+                - 1
+                - 2
+                - 3
+                - 5
+                - 11
+                - 12
+                - 41
+                - 49
+                - 50
+                - 51
+                - 53
+                - 55
+                - 71
+                - 84
+                - 98
+            "3":
+                - 2
+                - 7
+                - 13
+                - 15
+                - 16
+                - 19
+                - 23
+                - 55
+                - 62
+                - 66
+            "4": 3.9.20
+            "5": 0.18.6
+            "6": 4.47.1
+            "8":
+                - 5
+            "9":
+                "1": transformers_trainer
+            "12": 0.18.6
+            "13": linux-x86_64
+accelerator_config:
+    value:
+        dispatch_batches: null
+        even_batches: true
+        gradient_accumulation_kwargs: null
+        non_blocking: false
+        split_batches: false
+        use_seedable_sampler: true
+action_dim:
+    value: 26
+action_mlp_lr:
+    value: null
+adafactor:
+    value: false
+adam_beta1:
+    value: 0.9
+adam_beta2:
+    value: 0.999
+adam_epsilon:
+    value: 1e-08
+add_cross_attention:
+    value: false
+architectures:
+    value:
+        - LlamaForCausalLM
+attention_bias:
+    value: false
+attention_dropout:
+    value: 0
+auto_find_batch_size:
+    value: false
+average_tokens_across_devices:
+    value: false
+bad_words_ids:
+    value: null
+batch_eval_metrics:
+    value: false
+begin_suppress_tokens:
+    value: null
+bf16:
+    value: true
+bf16_full_eval:
+    value: false
+bos_token_id:
+    value: 1
+chars_per_token:
+    value: <CHARS_PER_TOKEN>
+chunk_size_feed_forward:
+    value: 0
+cross_attention_hidden_size:
+    value: null
+data_seed:
+    value: null
+dataloader_drop_last:
+    value: false
+dataloader_num_workers:
+    value: 0
+dataloader_persistent_workers:
+    value: false
+dataloader_pin_memory:
+    value: true
+dataloader_prefetch_factor:
+    value: null
+dataset:
+    value:
+        test_data_ratio: 0.01
+dataset_batch_size:
+    value: 1000
+dataset_num_proc:
+    value: null
+dataset_text_field:
+    value: text
+ddp_backend:
+    value: null
+ddp_broadcast_buffers:
+    value: null
+ddp_bucket_cap_mb:
+    value: null
+ddp_find_unused_parameters:
+    value: null
+ddp_timeout:
+    value: 1800
+debug:
+    value: []
+decoder_start_token_id:
+    value: null
+deepspeed:
+    value: null
+disable_tqdm:
+    value: false
+dispatch_batches:
+    value: null
+diversity_penalty:
+    value: 0
+do_eval:
+    value: true
+do_predict:
+    value: false
+do_sample:
+    value: false
+do_train:
+    value: false
+early_stopping:
+    value: false
+encoder_no_repeat_ngram_size:
+    value: 0
+eos_token_id:
+    value: 2
+eval_accumulation_steps:
+    value: null
+eval_delay:
+    value: 0
+eval_do_concat_batches:
+    value: true
+eval_on_start:
+    value: false
+eval_packing:
+    value: null
+eval_steps:
+    value: 10
+eval_strategy:
+    value: steps
+eval_use_gather_object:
+    value: false
+evaluation_strategy:
+    value: null
+exp_name:
+    value: train llm
+exponential_decay_length_penalty:
+    value: null
+finetuning_task:
+    value: null
+forced_bos_token_id:
+    value: null
+forced_eos_token_id:
+    value: null
+fp16:
+    value: false
+fp16_backend:
+    value: auto
+fp16_full_eval:
+    value: false
+fp16_opt_level:
+    value: O1
+fsdp:
+    value: []
+fsdp_config:
+    value:
+        min_num_params: 0
+        xla: false
+        xla_fsdp_grad_ckpt: false
+        xla_fsdp_v2: false
+fsdp_min_num_params:
+    value: 0
+fsdp_transformer_layer_cls_to_wrap:
+    value: null
+full_determinism:
+    value: false
+gradient_accumulation_steps:
+    value: 1
+gradient_checkpointing:
+    value: false
+gradient_checkpointing_kwargs:
+    value: null
+greater_is_better:
+    value: null
+group_by_length:
+    value: true
+half_precision_backend:
+    value: auto
+head_dim:
+    value: 64
+hidden_act:
+    value: silu
+hidden_size:
+    value: 576
+horizon:
+    value: 1
+hub_always_push:
+    value: false
+hub_model_id:
+    value: null
+hub_private_repo:
+    value: null
+hub_strategy:
+    value: every_save
+hub_token:
+    value: <HUB_TOKEN>
+id2label:
+    value:
+        "0": LABEL_0
+        "1": LABEL_1
+ignore_data_skip:
+    value: false
+include_for_metrics:
+    value: []
+include_inputs_for_metrics:
+    value: false
+include_num_input_tokens_seen:
+    value: false
+include_tokens_per_second:
+    value: false
+initializer_range:
+    value: 0.041666666666666664
+intermediate_size:
+    value: 1536
+is_decoder:
+    value: false
+is_encoder_decoder:
+    value: false
+is_llama_config:
+    value: true
+jit_mode_eval:
+    value: false
+joint_obs_action_mlp_lr:
+    value: 5e-05
+label_names:
+    value: null
+label_smoothing_factor:
+    value: 0
+label2id:
+    value:
+        LABEL_0: 0
+        LABEL_1: 1
+learning_rate:
+    value: 0.0001
+length_column_name:
+    value: length
+length_penalty:
+    value: 1
+llm:
+    value:
+        causal_lm_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM
+        checkpoint: data/outputs/2026.03.27/14.38.20_train_mlp_projector_adroit-hand-hammer-v1/checkpoints/latest.ckpt
+        config_target: llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig
+        finetune_mode: orig
+        hydra:
+            job:
+                override_dirname: HuggingFaceTB/SmolLM2-135M-Instruct
+            run:
+                dir: data/outputs/2026.03.27/16.20.52_HuggingFaceTB/SmolLM2-135M-Instruct
+        llm_mode: mlp-finetuned
+        lora_config:
+            bias: none
+            lora_alpha: 64
+            lora_dropout: 0.05
+            r: 32
+            task_type: CAUSAL_LM
+        max_length: 100
+        model_name: SmolLM2-135M-Instruct
+        name: HuggingFaceTB/SmolLM2-135M-Instruct
+        prompter:
+            _target_: llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter
+            use_joint_mlp_projector: true
+        use_joint_mlp_projector: true
+        use_quantization: false
+load_best_model_at_end:
+    value: false
+local_rank:
+    value: 0
+log_level:
+    value: passive
+log_level_replica:
+    value: warning
+log_on_each_node:
+    value: true
+logging:
+    value:
+        group: null
+        id: null
+        mode: online
+        name: 2026.03.27-16.20.52_train_llm_lowdim_adroit-hand-hammer-v1
+        project: llm_module_finetuning
+        resume: true
+        tags:
+            - train_llm_lowdim
+            - adroit-hand-hammer-v1
+            - train llm
+logging_dir:
+    value: /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/runs/Mar27_16-21-01_A6000-2
+logging_first_step:
+    value: false
+logging_nan_inf_filter:
+    value: true
+logging_steps:
+    value: 1
+logging_strategy:
+    value: steps
+lora_config:
+    value:
+        bias: none
+        lora_alpha: 64
+        lora_dropout: 0.05
+        r: 32
+        task_type: CAUSAL_LM
+lr_scheduler_type:
+    value: linear
+max_grad_norm:
+    value: 1
+max_length:
+    value: 20
+max_position_embeddings:
+    value: 8192
+max_seq_length:
+    value: 100
+max_steps:
+    value: -1
+metric_for_best_model:
+    value: null
+min_length:
+    value: 0
+mlp_bias:
+    value: false
+model/num_parameters:
+    value: 134889408
+model_init_kwargs:
+    value: null
+model_name:
+    value: HuggingFaceTB/SmolLM2-135M-Instruct
+model_type:
+    value: llama_lowdim
+mp_parameters:
+    value: ""
+multi_run:
+    value:
+        run_dir: data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1
+        wandb_name_base: 2026.03.27-16.20.52_train_llm_lowdim_adroit-hand-hammer-v1
+n_action_steps:
+    value: 1
+n_obs_steps:
+    value: 1
+name:
+    value: train_llm_lowdim
+neftune_noise_alpha:
+    value: null
+no_cuda:
+    value: false
+no_repeat_ngram_size:
+    value: 0
+num_attention_heads:
+    value: 9
+num_beam_groups:
+    value: 1
+num_beams:
+    value: 1
+num_hidden_layers:
+    value: 30
+num_key_value_heads:
+    value: 3
+num_of_sequences:
+    value: 1024
+num_return_sequences:
+    value: 1
+num_train_epochs:
+    value: 10
+obs_dim:
+    value: 46
+obs_mlp_lr:
+    value: null
+optim:
+    value: paged_adamw_32bit
+optim_args:
+    value: null
+optim_target_modules:
+    value: null
+output_attentions:
+    value: false
+output_dir:
+    value: /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1
+output_hidden_states:
+    value: false
+output_scores:
+    value: false
+overwrite_output_dir:
+    value: false
+packing:
+    value: false
+pad_token_id:
+    value: 2
+past_index:
+    value: -1
+per_device_eval_batch_size:
+    value: 128
+per_device_train_batch_size:
+    value: 128
+per_gpu_eval_batch_size:
+    value: null
+per_gpu_train_batch_size:
+    value: null
+prediction_loss_only:
+    value: false
+prefix:
+    value: null
+pretraining_tp:
+    value: 1
+problem_type:
+    value: null
+push_to_hub:
+    value: false
+push_to_hub_model_id:
+    value: null
+push_to_hub_organization:
+    value: null
+push_to_hub_token:
+    value: <PUSH_TO_HUB_TOKEN>
+ray_scope:
+    value: last
+remove_invalid_values:
+    value: false
+remove_unused_columns:
+    value: true
+repetition_penalty:
+    value: 1
+report_to:
+    value:
+        - wandb
+restore_callback_states_from_checkpoint:
+    value: false
+resume_from_checkpoint:
+    value: null
+return_dict:
+    value: true
+return_dict_in_generate:
+    value: false
+rms_norm_eps:
+    value: 1e-05
+rope_interleaved:
+    value: false
+rope_scaling:
+    value: null
+rope_theta:
+    value: 100000
+run_name:
+    value: /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1
+save_on_each_node:
+    value: false
+save_only_model:
+    value: false
+save_safetensors:
+    value: true
+save_steps:
+    value: 5000
+save_strategy:
+    value: steps
+save_total_limit:
+    value: null
+seed:
+    value: 42
+sep_token_id:
+    value: null
+skip_memory_metrics:
+    value: true
+split_batches:
+    value: null
+suppress_tokens:
+    value: null
+task:
+    value:
+        action_dim: 26
+        dataset:
+            _target_: llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset
+            data_path: datasets/adroit-hand-hammer-v1-general.pt
+            data_path2: datasets/adroit-hand-hammer-v1.pt
+            dummy_normalizer: false
+            horizon: 1
+            obs_eef_target: true
+            pad_after: 0
+            pad_before: 0
+            use_manual_normalizer: false
+            val_ratio: 0.05
+        env_runner:
+            _target_: llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner
+            discount: 0.99
+            env_name: llf-adroit-adroit-hand-hammer-v1
+            feedback_type:
+                - hp
+                - hn
+                - fp
+            instruction_type: b
+            max_steps: 150
+            n_action_steps: 1
+            n_envs: 10
+            n_obs_steps: 1
+            n_test: 50
+            n_train: 10
+            visual: false
+        instructor:
+            _target_: llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor
+        name: adroit-hand-hammer-v1
+        obs_dim: 46
+task_name:
+    value: adroit-hand-hammer-v1
+task_specific_params:
+    value: null
+temperature:
+    value: 1
+tf_legacy_loss:
+    value: false
+tf32:
+    value: true
+tie_encoder_decoder:
+    value: false
+tie_word_embeddings:
+    value: true
+tokenizer_class:
+    value: null
+top_k:
+    value: 50
+top_p:
+    value: 1
+torch_compile:
+    value: false
+torch_compile_backend:
+    value: null
+torch_compile_mode:
+    value: null
+torch_dtype:
+    value: bfloat16
+torch_empty_cache_steps:
+    value: null
+torchdynamo:
+    value: null
+torchscript:
+    value: false
+tpu_metrics_debug:
+    value: false
+tpu_num_cores:
+    value: null
+trainer:
+    value:
+        action_dim: 26
+        dataset_text_field: text
+        max_seq_length: 100
+        obs_dim: 46
+        packing: false
+        use_joint_mlp_projector: true
+training:
+    value:
+        bf16: true
+        eval_steps: 10
+        eval_strategy: steps
+        fp16: false
+        gradient_accumulation_steps: 1
+        group_by_length: true
+        joint_obs_action_mlp_lr: 5e-05
+        learning_rate: 0.0001
+        logging_steps: 1
+        logging_strategy: steps
+        num_train_epochs: 10
+        optim: paged_adamw_32bit
+        per_device_eval_batch_size: 128
+        per_device_train_batch_size: 128
+        report_to: wandb
+        save_steps: 5000
+        seed: 42
+        tf32: true
+        use_joint_mlp_projector: true
+        warmup_steps: 10
+transformers.js_config:
+    value:
+        kv_cache_dtype:
+            fp16: float16
+            q4f16: float16
+transformers_version:
+    value: 4.47.1
+typical_p:
+    value: 1
+use_bfloat16:
+    value: false
+use_cache:
+    value: false
+use_cpu:
+    value: false
+use_ipex:
+    value: false
+use_joint_mlp_projector:
+    value: true
+use_legacy_prediction_loop:
+    value: false
+use_liger:
+    value: false
+use_liger_kernel:
+    value: false
+use_mps_device:
+    value: false
+use_quantization:
+    value: false
+vocab_size:
+    value: 49152
+warmup_ratio:
+    value: 0
+warmup_steps:
+    value: 10
+weight_decay:
+    value: 0

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/files/output.log

@@ -0,0 +1,509 @@
+You are using a model of type llama to instantiate a model of type llama_lowdim. This is not supported for all configurations of models and can yield errors.
+Some weights of LowdimLlamaForCausalLM were not initialized from the model checkpoint at HuggingFaceTB/SmolLM2-135M-Instruct and are newly initialized: ['model.joint_obs_action_projector.projector.0.bias', 'model.joint_obs_action_projector.projector.0.weight', 'model.joint_obs_action_projector.projector.2.bias', 'model.joint_obs_action_projector.projector.2.weight']
+You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
+Loading from mlp projector checkpoint: data/outputs/2026.03.27/14.38.20_train_mlp_projector_adroit-hand-hammer-v1/checkpoints/latest.ckpt
+Finetune the whole original LLM SmolLM2-135M-Instruct.
+Multistep Flattening Dataset: 100%|██████████████████████████████████████████████████████████████████| 4585/4585 [00:00<00:00, 6904.74it/s]
+Setting TOKENIZERS_PARALLELISM=false for forked processes.
+[2026-03-27 16:20:59,889][datasets.arrow_dataset][WARNING] - Setting TOKENIZERS_PARALLELISM=false for forked processes.
+Map (num_proc=4): 100%|███████████████████████████████████████████████████████████████████████| 4585/4585 [00:00<00:00, 4937.91 examples/s]
+Setting TOKENIZERS_PARALLELISM=false for forked processes.
+[2026-03-27 16:21:01,023][datasets.arrow_dataset][WARNING] - Setting TOKENIZERS_PARALLELISM=false for forked processes.
+Map (num_proc=4): 100%|███████████████████████████████████████████████████████████████████████| 4585/4585 [00:00<00:00, 9277.34 examples/s]
+DatasetDict({
+    train: Dataset({
+        features: ['obs', 'action', 'description', 'input', 'output', 'text', 'input_ids', 'labels'],
+        num_rows: 4539
+    })
+    test: Dataset({
+        features: ['obs', 'action', 'description', 'input', 'output', 'text', 'input_ids', 'labels'],
+        num_rows: 46
+    })
+})
+/home/chyang/miniconda3/envs/llm-bc/lib/python3.9/site-packages/huggingface_hub/utils/_deprecation.py:100: FutureWarning: Deprecated argument(s) used in '__init__': max_seq_length, dataset_text_field. Will not be supported from version '1.0.0'.
+
+Deprecated positional argument(s) used in SFTTrainer, please use the SFTConfig to set these arguments instead.
+  warnings.warn(message, FutureWarning)
+/home/chyang/miniconda3/envs/llm-bc/lib/python3.9/site-packages/trl/trainer/sft_trainer.py:283: UserWarning: You passed a `max_seq_length` argument to the SFTTrainer, the value you passed will override the one in the `SFTConfig`.
+  warnings.warn(
+/home/chyang/miniconda3/envs/llm-bc/lib/python3.9/site-packages/trl/trainer/sft_trainer.py:321: UserWarning: You passed a `dataset_text_field` argument to the SFTTrainer, the value you passed will override the one in the `SFTConfig`.
+  warnings.warn(
+/home/chyang/miniconda3/envs/llm-bc/lib/python3.9/site-packages/trl/trainer/sft_trainer.py:401: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `LowdimSFTTrainer.__init__`. Use `processing_class` instead.
+  super().__init__(
+[2026-03-27 16:21:01,783] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cuda (auto detect)
+[2026-03-27 16:21:01,919][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat -Wno-unused-result -Wsign-compare -DNDEBUG -O2 -Wall -fPIC -O2 -isystem /home/chyang/miniconda3/envs/llm-bc/include -I/home/chyang/miniconda3/envs/llm-bc/include -fPIC -O2 -isystem /home/chyang/miniconda3/envs/llm-bc/include -fPIC -c /tmp/tmp7t0nrkuh/test.c -o /tmp/tmp7t0nrkuh/test.o
+[2026-03-27 16:21:01,969][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat /tmp/tmp7t0nrkuh/test.o -laio -o /tmp/tmp7t0nrkuh/a.out
+[2026-03-27 16:21:02,963][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat -Wno-unused-result -Wsign-compare -DNDEBUG -O2 -Wall -fPIC -O2 -isystem /home/chyang/miniconda3/envs/llm-bc/include -I/home/chyang/miniconda3/envs/llm-bc/include -fPIC -O2 -isystem /home/chyang/miniconda3/envs/llm-bc/include -fPIC -c /tmp/tmps_s5vyh9/test.c -o /tmp/tmps_s5vyh9/test.o
+[2026-03-27 16:21:03,018][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat /tmp/tmps_s5vyh9/test.o -L/usr/local/cuda -L/usr/local/cuda/lib64 -lcufile -o /tmp/tmps_s5vyh9/a.out
+wandb: WARNING The `run_name` is currently set to the same value as `TrainingArguments.output_dir`. If this was not intended, please specify a different run name by setting the `TrainingArguments.run_name` parameter.
+  3%|██▊                                                                                                  | 10/360 [00:08<04:32,  1.29it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+{'loss': 1.6431, 'grad_norm': 13.871646881103516, 'learning_rate': 1e-05, 'epoch': 0.03}
+{'loss': 1.6018, 'grad_norm': 17.890308380126953, 'learning_rate': 2e-05, 'epoch': 0.06}
+{'loss': 1.5953, 'grad_norm': 13.746294021606445, 'learning_rate': 3e-05, 'epoch': 0.08}
+{'loss': 1.5355, 'grad_norm': 15.9970121383667, 'learning_rate': 4e-05, 'epoch': 0.11}
+{'loss': 1.552, 'grad_norm': 18.634761810302734, 'learning_rate': 5e-05, 'epoch': 0.14}
+{'loss': 1.4926, 'grad_norm': 10.22042179107666, 'learning_rate': 6e-05, 'epoch': 0.17}
+{'loss': 1.1827, 'grad_norm': 14.976595878601074, 'learning_rate': 7e-05, 'epoch': 0.19}
+{'loss': 1.173, 'grad_norm': 34.35334396362305, 'learning_rate': 8e-05, 'epoch': 0.22}
+{'loss': 0.7987, 'grad_norm': 7.392702579498291, 'learning_rate': 9e-05, 'epoch': 0.25}
+{'loss': 0.7641, 'grad_norm': 8.6481294631958, 'learning_rate': 0.0001, 'epoch': 0.28}
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+  6%|█████▌                                                                                               | 20/360 [00:15<04:21,  1.30it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.719104528427124, 'eval_runtime': 0.1041, 'eval_samples_per_second': 441.967, 'eval_steps_per_second': 9.608, 'epoch': 0.28}
+{'loss': 0.6656, 'grad_norm': 11.896777153015137, 'learning_rate': 9.971428571428571e-05, 'epoch': 0.31}
+{'loss': 0.5743, 'grad_norm': 6.002552032470703, 'learning_rate': 9.942857142857144e-05, 'epoch': 0.33}
+{'loss': 0.5204, 'grad_norm': 4.569210529327393, 'learning_rate': 9.914285714285715e-05, 'epoch': 0.36}
+{'loss': 0.5217, 'grad_norm': 3.2792978286743164, 'learning_rate': 9.885714285714286e-05, 'epoch': 0.39}
+{'loss': 0.5046, 'grad_norm': 3.6701369285583496, 'learning_rate': 9.857142857142858e-05, 'epoch': 0.42}
+{'loss': 0.4803, 'grad_norm': 4.064358711242676, 'learning_rate': 9.828571428571429e-05, 'epoch': 0.44}
+{'loss': 0.4466, 'grad_norm': 2.9059529304504395, 'learning_rate': 9.8e-05, 'epoch': 0.47}
+{'loss': 0.4453, 'grad_norm': 2.499434471130371, 'learning_rate': 9.771428571428572e-05, 'epoch': 0.5}
+{'loss': 0.4382, 'grad_norm': 2.084019899368286, 'learning_rate': 9.742857142857143e-05, 'epoch': 0.53}
+{'loss': 0.4208, 'grad_norm': 1.2787052392959595, 'learning_rate': 9.714285714285715e-05, 'epoch': 0.56}
+  8%|████████▍                                                                                            | 30/360 [00:23<04:10,  1.32it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.4429771900177002, 'eval_runtime': 0.1036, 'eval_samples_per_second': 443.983, 'eval_steps_per_second': 9.652, 'epoch': 0.56}
+{'loss': 0.431, 'grad_norm': 1.8704657554626465, 'learning_rate': 9.685714285714286e-05, 'epoch': 0.58}
+{'loss': 0.4317, 'grad_norm': 1.7761015892028809, 'learning_rate': 9.657142857142858e-05, 'epoch': 0.61}
+{'loss': 0.4229, 'grad_norm': 1.7499357461929321, 'learning_rate': 9.628571428571429e-05, 'epoch': 0.64}
+{'loss': 0.4126, 'grad_norm': 1.2509069442749023, 'learning_rate': 9.6e-05, 'epoch': 0.67}
+{'loss': 0.3888, 'grad_norm': 1.3415058851242065, 'learning_rate': 9.571428571428573e-05, 'epoch': 0.69}
+{'loss': 0.4111, 'grad_norm': 1.513482689857483, 'learning_rate': 9.542857142857143e-05, 'epoch': 0.72}
+{'loss': 0.3904, 'grad_norm': 1.0207685232162476, 'learning_rate': 9.514285714285714e-05, 'epoch': 0.75}
+{'loss': 0.3911, 'grad_norm': 1.0765091180801392, 'learning_rate': 9.485714285714287e-05, 'epoch': 0.78}
+{'loss': 0.3893, 'grad_norm': 1.2146029472351074, 'learning_rate': 9.457142857142858e-05, 'epoch': 0.81}
+{'loss': 0.3915, 'grad_norm': 1.302972435951233, 'learning_rate': 9.428571428571429e-05, 'epoch': 0.83}
+ 11%|███████████▏                                                                                         | 40/360 [00:30<03:54,  1.36it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.39956018328666687, 'eval_runtime': 0.1036, 'eval_samples_per_second': 444.089, 'eval_steps_per_second': 9.654, 'epoch': 0.83}
+{'loss': 0.377, 'grad_norm': 1.2195433378219604, 'learning_rate': 9.4e-05, 'epoch': 0.86}
+{'loss': 0.3837, 'grad_norm': 1.2320094108581543, 'learning_rate': 9.371428571428572e-05, 'epoch': 0.89}
+{'loss': 0.3776, 'grad_norm': 1.0609043836593628, 'learning_rate': 9.342857142857143e-05, 'epoch': 0.92}
+{'loss': 0.3887, 'grad_norm': 0.9609966278076172, 'learning_rate': 9.314285714285715e-05, 'epoch': 0.94}
+{'loss': 0.3761, 'grad_norm': 1.0595581531524658, 'learning_rate': 9.285714285714286e-05, 'epoch': 0.97}
+{'loss': 0.3744, 'grad_norm': 0.990327775478363, 'learning_rate': 9.257142857142858e-05, 'epoch': 1.0}
+{'loss': 0.3625, 'grad_norm': 1.272873044013977, 'learning_rate': 9.228571428571429e-05, 'epoch': 1.03}
+{'loss': 0.3869, 'grad_norm': 1.9024567604064941, 'learning_rate': 9.200000000000001e-05, 'epoch': 1.06}
+{'loss': 0.3751, 'grad_norm': 1.3398654460906982, 'learning_rate': 9.171428571428572e-05, 'epoch': 1.08}
+{'loss': 0.3662, 'grad_norm': 1.9176064729690552, 'learning_rate': 9.142857142857143e-05, 'epoch': 1.11}
+ 14%|██████████████                                                                                       | 50/360 [00:38<03:57,  1.31it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.37765416502952576, 'eval_runtime': 0.1036, 'eval_samples_per_second': 444.001, 'eval_steps_per_second': 9.652, 'epoch': 1.11}
+{'loss': 0.3665, 'grad_norm': 1.1852660179138184, 'learning_rate': 9.114285714285716e-05, 'epoch': 1.14}
+{'loss': 0.3705, 'grad_norm': 1.831186056137085, 'learning_rate': 9.085714285714286e-05, 'epoch': 1.17}
+{'loss': 0.3582, 'grad_norm': 1.1574777364730835, 'learning_rate': 9.057142857142857e-05, 'epoch': 1.19}
+{'loss': 0.3724, 'grad_norm': 1.3485198020935059, 'learning_rate': 9.028571428571428e-05, 'epoch': 1.22}
+{'loss': 0.3549, 'grad_norm': 1.0934721231460571, 'learning_rate': 9e-05, 'epoch': 1.25}
+{'loss': 0.3607, 'grad_norm': 1.2588518857955933, 'learning_rate': 8.971428571428571e-05, 'epoch': 1.28}
+{'loss': 0.3492, 'grad_norm': 0.9038533568382263, 'learning_rate': 8.942857142857142e-05, 'epoch': 1.31}
+{'loss': 0.361, 'grad_norm': 1.083348274230957, 'learning_rate': 8.914285714285715e-05, 'epoch': 1.33}
+{'loss': 0.3475, 'grad_norm': 0.8287424445152283, 'learning_rate': 8.885714285714286e-05, 'epoch': 1.36}
+{'loss': 0.363, 'grad_norm': 1.3475714921951294, 'learning_rate': 8.857142857142857e-05, 'epoch': 1.39}
+ 17%|████████████████▊                                                                                    | 60/360 [00:46<03:47,  1.32it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3627206087112427, 'eval_runtime': 0.1039, 'eval_samples_per_second': 442.561, 'eval_steps_per_second': 9.621, 'epoch': 1.39}
+{'loss': 0.3527, 'grad_norm': 1.1012552976608276, 'learning_rate': 8.828571428571429e-05, 'epoch': 1.42}
+{'loss': 0.3418, 'grad_norm': 0.6421935558319092, 'learning_rate': 8.800000000000001e-05, 'epoch': 1.44}
+{'loss': 0.3513, 'grad_norm': 1.1574995517730713, 'learning_rate': 8.771428571428572e-05, 'epoch': 1.47}
+{'loss': 0.3515, 'grad_norm': 1.0251258611679077, 'learning_rate': 8.742857142857144e-05, 'epoch': 1.5}
+{'loss': 0.3609, 'grad_norm': 0.9864039421081543, 'learning_rate': 8.714285714285715e-05, 'epoch': 1.53}
+{'loss': 0.3454, 'grad_norm': 0.757999062538147, 'learning_rate': 8.685714285714286e-05, 'epoch': 1.56}
+{'loss': 0.3488, 'grad_norm': 1.0983614921569824, 'learning_rate': 8.657142857142858e-05, 'epoch': 1.58}
+{'loss': 0.3562, 'grad_norm': 1.4811136722564697, 'learning_rate': 8.62857142857143e-05, 'epoch': 1.61}
+{'loss': 0.349, 'grad_norm': 0.9457672834396362, 'learning_rate': 8.6e-05, 'epoch': 1.64}
+{'loss': 0.3551, 'grad_norm': 1.4347460269927979, 'learning_rate': 8.571428571428571e-05, 'epoch': 1.67}
+ 19%|███████████████████▋                                                                                 | 70/360 [00:54<03:43,  1.30it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.35965070128440857, 'eval_runtime': 0.1045, 'eval_samples_per_second': 440.164, 'eval_steps_per_second': 9.569, 'epoch': 1.67}
+{'loss': 0.3485, 'grad_norm': 1.0592706203460693, 'learning_rate': 8.542857142857144e-05, 'epoch': 1.69}
+{'loss': 0.3512, 'grad_norm': 1.3444126844406128, 'learning_rate': 8.514285714285714e-05, 'epoch': 1.72}
+{'loss': 0.3525, 'grad_norm': 0.9045667052268982, 'learning_rate': 8.485714285714285e-05, 'epoch': 1.75}
+{'loss': 0.3483, 'grad_norm': 1.135429859161377, 'learning_rate': 8.457142857142858e-05, 'epoch': 1.78}
+{'loss': 0.3445, 'grad_norm': 0.7742411494255066, 'learning_rate': 8.428571428571429e-05, 'epoch': 1.81}
+{'loss': 0.3425, 'grad_norm': 1.2747840881347656, 'learning_rate': 8.4e-05, 'epoch': 1.83}
+{'loss': 0.3506, 'grad_norm': 1.1280975341796875, 'learning_rate': 8.371428571428572e-05, 'epoch': 1.86}
+{'loss': 0.3458, 'grad_norm': 1.3229925632476807, 'learning_rate': 8.342857142857143e-05, 'epoch': 1.89}
+{'loss': 0.3443, 'grad_norm': 1.0970568656921387, 'learning_rate': 8.314285714285715e-05, 'epoch': 1.92}
+{'loss': 0.3612, 'grad_norm': 1.7599389553070068, 'learning_rate': 8.285714285714287e-05, 'epoch': 1.94}
+ 22%|██████████████████████▍                                                                              | 80/360 [01:01<03:32,  1.32it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.35299769043922424, 'eval_runtime': 0.1042, 'eval_samples_per_second': 441.433, 'eval_steps_per_second': 9.596, 'epoch': 1.94}
+{'loss': 0.3373, 'grad_norm': 0.8275991678237915, 'learning_rate': 8.257142857142858e-05, 'epoch': 1.97}
+{'loss': 0.3624, 'grad_norm': 1.5045437812805176, 'learning_rate': 8.228571428571429e-05, 'epoch': 2.0}
+{'loss': 0.3434, 'grad_norm': 0.9771829843521118, 'learning_rate': 8.2e-05, 'epoch': 2.03}
+{'loss': 0.3347, 'grad_norm': 0.8552800416946411, 'learning_rate': 8.171428571428572e-05, 'epoch': 2.06}
+{'loss': 0.3448, 'grad_norm': 0.8917291164398193, 'learning_rate': 8.142857142857143e-05, 'epoch': 2.08}
+{'loss': 0.3309, 'grad_norm': 0.9143850207328796, 'learning_rate': 8.114285714285714e-05, 'epoch': 2.11}
+{'loss': 0.3471, 'grad_norm': 1.359926700592041, 'learning_rate': 8.085714285714287e-05, 'epoch': 2.14}
+{'loss': 0.3433, 'grad_norm': 0.84107506275177, 'learning_rate': 8.057142857142857e-05, 'epoch': 2.17}
+{'loss': 0.3495, 'grad_norm': 1.2953639030456543, 'learning_rate': 8.028571428571428e-05, 'epoch': 2.19}
+{'loss': 0.3388, 'grad_norm': 0.9937311410903931, 'learning_rate': 8e-05, 'epoch': 2.22}
+ 25%|█████████████████████████▎                                                                           | 90/360 [01:09<03:26,  1.31it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3491460978984833, 'eval_runtime': 0.1044, 'eval_samples_per_second': 440.593, 'eval_steps_per_second': 9.578, 'epoch': 2.22}
+{'loss': 0.3435, 'grad_norm': 1.0681490898132324, 'learning_rate': 7.971428571428572e-05, 'epoch': 2.25}
+{'loss': 0.3333, 'grad_norm': 0.8466928601264954, 'learning_rate': 7.942857142857143e-05, 'epoch': 2.28}
+{'loss': 0.3305, 'grad_norm': 0.8183342814445496, 'learning_rate': 7.914285714285715e-05, 'epoch': 2.31}
+{'loss': 0.3289, 'grad_norm': 0.833314061164856, 'learning_rate': 7.885714285714286e-05, 'epoch': 2.33}
+{'loss': 0.3331, 'grad_norm': 0.8347731828689575, 'learning_rate': 7.857142857142858e-05, 'epoch': 2.36}
+{'loss': 0.3437, 'grad_norm': 1.0877679586410522, 'learning_rate': 7.828571428571429e-05, 'epoch': 2.39}
+{'loss': 0.3331, 'grad_norm': 0.9570125937461853, 'learning_rate': 7.800000000000001e-05, 'epoch': 2.42}
+{'loss': 0.3363, 'grad_norm': 0.7662280797958374, 'learning_rate': 7.771428571428572e-05, 'epoch': 2.44}
+{'loss': 0.3305, 'grad_norm': 0.9321999549865723, 'learning_rate': 7.742857142857143e-05, 'epoch': 2.47}
+{'loss': 0.3332, 'grad_norm': 0.8284544348716736, 'learning_rate': 7.714285714285715e-05, 'epoch': 2.5}
+ 28%|███████████████████████████▊                                                                        | 100/360 [01:17<03:19,  1.30it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.34498000144958496, 'eval_runtime': 0.1046, 'eval_samples_per_second': 439.848, 'eval_steps_per_second': 9.562, 'epoch': 2.5}
+{'loss': 0.3448, 'grad_norm': 1.0568827390670776, 'learning_rate': 7.685714285714286e-05, 'epoch': 2.53}
+{'loss': 0.334, 'grad_norm': 0.9136806130409241, 'learning_rate': 7.657142857142857e-05, 'epoch': 2.56}
+{'loss': 0.3425, 'grad_norm': 1.2551990747451782, 'learning_rate': 7.62857142857143e-05, 'epoch': 2.58}
+{'loss': 0.3265, 'grad_norm': 0.8284862637519836, 'learning_rate': 7.6e-05, 'epoch': 2.61}
+{'loss': 0.3267, 'grad_norm': 0.7161554098129272, 'learning_rate': 7.571428571428571e-05, 'epoch': 2.64}
+{'loss': 0.3342, 'grad_norm': 0.8050905466079712, 'learning_rate': 7.542857142857144e-05, 'epoch': 2.67}
+{'loss': 0.3325, 'grad_norm': 0.7441209554672241, 'learning_rate': 7.514285714285715e-05, 'epoch': 2.69}
+{'loss': 0.334, 'grad_norm': 0.591927707195282, 'learning_rate': 7.485714285714285e-05, 'epoch': 2.72}
+{'loss': 0.3473, 'grad_norm': 0.8902866244316101, 'learning_rate': 7.457142857142856e-05, 'epoch': 2.75}
+{'loss': 0.326, 'grad_norm': 0.6760069131851196, 'learning_rate': 7.428571428571429e-05, 'epoch': 2.78}
+ 31%|██████████████████████████████▌                                                                     | 110/360 [01:24<02:57,  1.41it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3410107493400574, 'eval_runtime': 0.1043, 'eval_samples_per_second': 441.026, 'eval_steps_per_second': 9.588, 'epoch': 2.78}
+{'loss': 0.3348, 'grad_norm': 0.6579345464706421, 'learning_rate': 7.4e-05, 'epoch': 2.81}
+{'loss': 0.3253, 'grad_norm': 1.0648415088653564, 'learning_rate': 7.371428571428572e-05, 'epoch': 2.83}
+{'loss': 0.3297, 'grad_norm': 0.6868814826011658, 'learning_rate': 7.342857142857144e-05, 'epoch': 2.86}
+{'loss': 0.3401, 'grad_norm': 1.1149464845657349, 'learning_rate': 7.314285714285715e-05, 'epoch': 2.89}
+{'loss': 0.3348, 'grad_norm': 0.8934164047241211, 'learning_rate': 7.285714285714286e-05, 'epoch': 2.92}
+{'loss': 0.3427, 'grad_norm': 1.1119507551193237, 'learning_rate': 7.257142857142858e-05, 'epoch': 2.94}
+{'loss': 0.3374, 'grad_norm': 0.8103634715080261, 'learning_rate': 7.228571428571429e-05, 'epoch': 2.97}
+{'loss': 0.3395, 'grad_norm': 0.8421126008033752, 'learning_rate': 7.2e-05, 'epoch': 3.0}
+{'loss': 0.331, 'grad_norm': 0.8583278656005859, 'learning_rate': 7.171428571428572e-05, 'epoch': 3.03}
+{'loss': 0.3355, 'grad_norm': 1.2129111289978027, 'learning_rate': 7.142857142857143e-05, 'epoch': 3.06}
+ 33%|█████████████████████████████████▎                                                                  | 120/360 [01:32<03:02,  1.32it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.332057386636734, 'eval_runtime': 0.1051, 'eval_samples_per_second': 437.685, 'eval_steps_per_second': 9.515, 'epoch': 3.06}
+{'loss': 0.3294, 'grad_norm': 0.9463130235671997, 'learning_rate': 7.114285714285714e-05, 'epoch': 3.08}
+{'loss': 0.3327, 'grad_norm': 0.9692079424858093, 'learning_rate': 7.085714285714285e-05, 'epoch': 3.11}
+{'loss': 0.3295, 'grad_norm': 0.9853659868240356, 'learning_rate': 7.057142857142858e-05, 'epoch': 3.14}
+{'loss': 0.3358, 'grad_norm': 0.7222715616226196, 'learning_rate': 7.028571428571428e-05, 'epoch': 3.17}
+{'loss': 0.3406, 'grad_norm': 1.1528452634811401, 'learning_rate': 7e-05, 'epoch': 3.19}
+{'loss': 0.329, 'grad_norm': 1.0079970359802246, 'learning_rate': 6.971428571428572e-05, 'epoch': 3.22}
+{'loss': 0.327, 'grad_norm': 0.7162885665893555, 'learning_rate': 6.942857142857143e-05, 'epoch': 3.25}
+{'loss': 0.336, 'grad_norm': 0.9302375912666321, 'learning_rate': 6.914285714285715e-05, 'epoch': 3.28}
+{'loss': 0.3356, 'grad_norm': 0.8540468215942383, 'learning_rate': 6.885714285714286e-05, 'epoch': 3.31}
+{'loss': 0.3279, 'grad_norm': 0.598040759563446, 'learning_rate': 6.857142857142858e-05, 'epoch': 3.33}
+ 36%|████████████████████████████████████                                                                | 130/360 [01:40<02:57,  1.30it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3290035128593445, 'eval_runtime': 0.1049, 'eval_samples_per_second': 438.622, 'eval_steps_per_second': 9.535, 'epoch': 3.33}
+{'loss': 0.33, 'grad_norm': 0.6981043815612793, 'learning_rate': 6.828571428571429e-05, 'epoch': 3.36}
+{'loss': 0.3262, 'grad_norm': 0.6710860133171082, 'learning_rate': 6.800000000000001e-05, 'epoch': 3.39}
+{'loss': 0.3273, 'grad_norm': 0.6621596813201904, 'learning_rate': 6.771428571428572e-05, 'epoch': 3.42}
+{'loss': 0.3264, 'grad_norm': 0.8255563974380493, 'learning_rate': 6.742857142857143e-05, 'epoch': 3.44}
+{'loss': 0.3157, 'grad_norm': 0.8350001573562622, 'learning_rate': 6.714285714285714e-05, 'epoch': 3.47}
+{'loss': 0.3205, 'grad_norm': 0.7275986075401306, 'learning_rate': 6.685714285714286e-05, 'epoch': 3.5}
+{'loss': 0.3327, 'grad_norm': 0.652642548084259, 'learning_rate': 6.657142857142857e-05, 'epoch': 3.53}
+{'loss': 0.3319, 'grad_norm': 0.9522960186004639, 'learning_rate': 6.628571428571428e-05, 'epoch': 3.56}
+{'loss': 0.3292, 'grad_norm': 0.7006963491439819, 'learning_rate': 6.6e-05, 'epoch': 3.58}
+{'loss': 0.3246, 'grad_norm': 0.7161970138549805, 'learning_rate': 6.571428571428571e-05, 'epoch': 3.61}
+ 39%|██████████████████████████████████████▉                                                             | 140/360 [01:48<02:51,  1.28it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3442412316799164, 'eval_runtime': 0.1052, 'eval_samples_per_second': 437.452, 'eval_steps_per_second': 9.51, 'epoch': 3.61}
+{'loss': 0.3289, 'grad_norm': 1.0642709732055664, 'learning_rate': 6.542857142857142e-05, 'epoch': 3.64}
+{'loss': 0.3234, 'grad_norm': 0.7999193072319031, 'learning_rate': 6.514285714285715e-05, 'epoch': 3.67}
+{'loss': 0.3297, 'grad_norm': 0.8324876427650452, 'learning_rate': 6.485714285714286e-05, 'epoch': 3.69}
+{'loss': 0.3125, 'grad_norm': 0.561801552772522, 'learning_rate': 6.457142857142856e-05, 'epoch': 3.72}
+{'loss': 0.3234, 'grad_norm': 0.6995918154716492, 'learning_rate': 6.428571428571429e-05, 'epoch': 3.75}
+{'loss': 0.3256, 'grad_norm': 0.6314477920532227, 'learning_rate': 6.400000000000001e-05, 'epoch': 3.78}
+{'loss': 0.3315, 'grad_norm': 0.9092559814453125, 'learning_rate': 6.371428571428572e-05, 'epoch': 3.81}
+{'loss': 0.3241, 'grad_norm': 0.7306588292121887, 'learning_rate': 6.342857142857143e-05, 'epoch': 3.83}
+{'loss': 0.3323, 'grad_norm': 0.7943991422653198, 'learning_rate': 6.314285714285715e-05, 'epoch': 3.86}
+{'loss': 0.3273, 'grad_norm': 0.8375313878059387, 'learning_rate': 6.285714285714286e-05, 'epoch': 3.89}
+ 42%|█████████████████████████████████████████▋                                                          | 150/360 [01:55<02:38,  1.33it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.33936312794685364, 'eval_runtime': 0.1048, 'eval_samples_per_second': 438.851, 'eval_steps_per_second': 9.54, 'epoch': 3.89}
+{'loss': 0.3258, 'grad_norm': 0.9479944705963135, 'learning_rate': 6.257142857142857e-05, 'epoch': 3.92}
+{'loss': 0.3228, 'grad_norm': 0.8155922889709473, 'learning_rate': 6.22857142857143e-05, 'epoch': 3.94}
+{'loss': 0.3217, 'grad_norm': 0.8617050647735596, 'learning_rate': 6.2e-05, 'epoch': 3.97}
+{'loss': 0.3309, 'grad_norm': 1.2106715440750122, 'learning_rate': 6.171428571428571e-05, 'epoch': 4.0}
+{'loss': 0.3265, 'grad_norm': 0.8097350001335144, 'learning_rate': 6.142857142857143e-05, 'epoch': 4.03}
+{'loss': 0.3222, 'grad_norm': 0.651019811630249, 'learning_rate': 6.114285714285714e-05, 'epoch': 4.06}
+{'loss': 0.3245, 'grad_norm': 0.8858047127723694, 'learning_rate': 6.085714285714286e-05, 'epoch': 4.08}
+{'loss': 0.3153, 'grad_norm': 0.8602396845817566, 'learning_rate': 6.0571428571428576e-05, 'epoch': 4.11}
+{'loss': 0.313, 'grad_norm': 0.615274965763092, 'learning_rate': 6.028571428571429e-05, 'epoch': 4.14}
+{'loss': 0.3275, 'grad_norm': 0.9199692010879517, 'learning_rate': 6e-05, 'epoch': 4.17}
+ 44%|████████████████████████████████████████████▍                                                       | 160/360 [02:03<02:34,  1.29it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.34145644307136536, 'eval_runtime': 0.1056, 'eval_samples_per_second': 435.469, 'eval_steps_per_second': 9.467, 'epoch': 4.17}
+{'loss': 0.3287, 'grad_norm': 0.9348899722099304, 'learning_rate': 5.9714285714285724e-05, 'epoch': 4.19}
+{'loss': 0.3203, 'grad_norm': 0.7250774502754211, 'learning_rate': 5.9428571428571434e-05, 'epoch': 4.22}
+{'loss': 0.3169, 'grad_norm': 0.7376280426979065, 'learning_rate': 5.914285714285714e-05, 'epoch': 4.25}
+{'loss': 0.3215, 'grad_norm': 0.6010245680809021, 'learning_rate': 5.885714285714285e-05, 'epoch': 4.28}
+{'loss': 0.317, 'grad_norm': 0.7241640686988831, 'learning_rate': 5.8571428571428575e-05, 'epoch': 4.31}
+{'loss': 0.3217, 'grad_norm': 0.6956952810287476, 'learning_rate': 5.828571428571429e-05, 'epoch': 4.33}
+{'loss': 0.322, 'grad_norm': 0.8463672995567322, 'learning_rate': 5.8e-05, 'epoch': 4.36}
+{'loss': 0.3129, 'grad_norm': 0.5538536906242371, 'learning_rate': 5.771428571428572e-05, 'epoch': 4.39}
+{'loss': 0.3275, 'grad_norm': 0.8398566246032715, 'learning_rate': 5.742857142857143e-05, 'epoch': 4.42}
+{'loss': 0.3225, 'grad_norm': 0.5335714221000671, 'learning_rate': 5.714285714285714e-05, 'epoch': 4.44}
+ 47%|███████████████████████████████████████████████▏                                                    | 170/360 [02:11<02:27,  1.29it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3419412672519684, 'eval_runtime': 0.1054, 'eval_samples_per_second': 436.373, 'eval_steps_per_second': 9.486, 'epoch': 4.44}
+{'loss': 0.3201, 'grad_norm': 0.893516480922699, 'learning_rate': 5.6857142857142865e-05, 'epoch': 4.47}
+{'loss': 0.3246, 'grad_norm': 0.7950851321220398, 'learning_rate': 5.6571428571428574e-05, 'epoch': 4.5}
+{'loss': 0.3138, 'grad_norm': 0.6274649500846863, 'learning_rate': 5.628571428571428e-05, 'epoch': 4.53}
+{'loss': 0.3225, 'grad_norm': 0.6270793676376343, 'learning_rate': 5.6000000000000006e-05, 'epoch': 4.56}
+{'loss': 0.3222, 'grad_norm': 0.6661616563796997, 'learning_rate': 5.571428571428572e-05, 'epoch': 4.58}
+{'loss': 0.3138, 'grad_norm': 0.6097862124443054, 'learning_rate': 5.542857142857143e-05, 'epoch': 4.61}
+{'loss': 0.3109, 'grad_norm': 0.6743194460868835, 'learning_rate': 5.514285714285714e-05, 'epoch': 4.64}
+{'loss': 0.3193, 'grad_norm': 0.6684880256652832, 'learning_rate': 5.485714285714286e-05, 'epoch': 4.67}
+{'loss': 0.3212, 'grad_norm': 0.7434603571891785, 'learning_rate': 5.457142857142857e-05, 'epoch': 4.69}
+{'loss': 0.323, 'grad_norm': 0.8257206082344055, 'learning_rate': 5.428571428571428e-05, 'epoch': 4.72}
+ 50%|██████████████████████████████████████████████████                                                  | 180/360 [02:18<01:58,  1.52it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.335285484790802, 'eval_runtime': 0.1056, 'eval_samples_per_second': 435.635, 'eval_steps_per_second': 9.47, 'epoch': 4.72}
+{'loss': 0.3249, 'grad_norm': 0.5611714720726013, 'learning_rate': 5.4000000000000005e-05, 'epoch': 4.75}
+{'loss': 0.3213, 'grad_norm': 0.6146838068962097, 'learning_rate': 5.3714285714285714e-05, 'epoch': 4.78}
+{'loss': 0.3239, 'grad_norm': 0.6939036846160889, 'learning_rate': 5.342857142857143e-05, 'epoch': 4.81}
+{'loss': 0.3234, 'grad_norm': 0.7213876843452454, 'learning_rate': 5.314285714285715e-05, 'epoch': 4.83}
+{'loss': 0.3186, 'grad_norm': 0.6637408137321472, 'learning_rate': 5.285714285714286e-05, 'epoch': 4.86}
+{'loss': 0.3184, 'grad_norm': 0.6469508409500122, 'learning_rate': 5.257142857142857e-05, 'epoch': 4.89}
+{'loss': 0.3171, 'grad_norm': 0.6262702941894531, 'learning_rate': 5.2285714285714294e-05, 'epoch': 4.92}
+{'loss': 0.3271, 'grad_norm': 0.6692309379577637, 'learning_rate': 5.2000000000000004e-05, 'epoch': 4.94}
+{'loss': 0.3229, 'grad_norm': 0.611004114151001, 'learning_rate': 5.171428571428571e-05, 'epoch': 4.97}
+{'loss': 0.3223, 'grad_norm': 0.9707463383674622, 'learning_rate': 5.142857142857143e-05, 'epoch': 5.0}
+ 53%|████████████████████████████████████████████████████▊                                               | 190/360 [02:26<02:11,  1.29it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.33622071146965027, 'eval_runtime': 0.105, 'eval_samples_per_second': 437.997, 'eval_steps_per_second': 9.522, 'epoch': 5.0}
+{'loss': 0.3106, 'grad_norm': 0.43101295828819275, 'learning_rate': 5.1142857142857145e-05, 'epoch': 5.03}
+{'loss': 0.3139, 'grad_norm': 0.7981957793235779, 'learning_rate': 5.085714285714286e-05, 'epoch': 5.06}
+{'loss': 0.3137, 'grad_norm': 0.9149967432022095, 'learning_rate': 5.057142857142857e-05, 'epoch': 5.08}
+{'loss': 0.3185, 'grad_norm': 0.8689376711845398, 'learning_rate': 5.028571428571429e-05, 'epoch': 5.11}
+{'loss': 0.3195, 'grad_norm': 0.6829914450645447, 'learning_rate': 5e-05, 'epoch': 5.14}
+{'loss': 0.3139, 'grad_norm': 0.6187098026275635, 'learning_rate': 4.971428571428572e-05, 'epoch': 5.17}
+{'loss': 0.3147, 'grad_norm': 0.8703141212463379, 'learning_rate': 4.942857142857143e-05, 'epoch': 5.19}
+{'loss': 0.3162, 'grad_norm': 0.6344360709190369, 'learning_rate': 4.9142857142857144e-05, 'epoch': 5.22}
+{'loss': 0.3228, 'grad_norm': 0.7499691843986511, 'learning_rate': 4.885714285714286e-05, 'epoch': 5.25}
+{'loss': 0.3152, 'grad_norm': 0.7664843201637268, 'learning_rate': 4.8571428571428576e-05, 'epoch': 5.28}
+ 56%|███████████████████████████████████████████████████████▌                                            | 200/360 [02:34<02:03,  1.29it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3364305794239044, 'eval_runtime': 0.1056, 'eval_samples_per_second': 435.778, 'eval_steps_per_second': 9.473, 'epoch': 5.28}
+{'loss': 0.3154, 'grad_norm': 0.6158504486083984, 'learning_rate': 4.828571428571429e-05, 'epoch': 5.31}
+{'loss': 0.3206, 'grad_norm': 0.8614490032196045, 'learning_rate': 4.8e-05, 'epoch': 5.33}
+{'loss': 0.3159, 'grad_norm': 0.7699540257453918, 'learning_rate': 4.771428571428572e-05, 'epoch': 5.36}
+{'loss': 0.3186, 'grad_norm': 0.9598901867866516, 'learning_rate': 4.742857142857143e-05, 'epoch': 5.39}
+{'loss': 0.3118, 'grad_norm': 0.855253279209137, 'learning_rate': 4.714285714285714e-05, 'epoch': 5.42}
+{'loss': 0.3178, 'grad_norm': 0.6478847861289978, 'learning_rate': 4.685714285714286e-05, 'epoch': 5.44}
+{'loss': 0.3236, 'grad_norm': 0.8028067946434021, 'learning_rate': 4.6571428571428575e-05, 'epoch': 5.47}
+{'loss': 0.3147, 'grad_norm': 0.7795782089233398, 'learning_rate': 4.628571428571429e-05, 'epoch': 5.5}
+{'loss': 0.3221, 'grad_norm': 0.7845653891563416, 'learning_rate': 4.600000000000001e-05, 'epoch': 5.53}
+{'loss': 0.321, 'grad_norm': 1.1422370672225952, 'learning_rate': 4.5714285714285716e-05, 'epoch': 5.56}
+ 58%|██████████████████████████████████████████████████████████▎                                         | 210/360 [02:42<01:55,  1.30it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.33535492420196533, 'eval_runtime': 0.1051, 'eval_samples_per_second': 437.57, 'eval_steps_per_second': 9.512, 'epoch': 5.56}
+{'loss': 0.3183, 'grad_norm': 0.7386415600776672, 'learning_rate': 4.542857142857143e-05, 'epoch': 5.58}
+{'loss': 0.3205, 'grad_norm': 0.6756716966629028, 'learning_rate': 4.514285714285714e-05, 'epoch': 5.61}
+{'loss': 0.3195, 'grad_norm': 0.7116839289665222, 'learning_rate': 4.485714285714286e-05, 'epoch': 5.64}
+{'loss': 0.3248, 'grad_norm': 0.7919530272483826, 'learning_rate': 4.4571428571428574e-05, 'epoch': 5.67}
+{'loss': 0.3124, 'grad_norm': 0.5152342319488525, 'learning_rate': 4.428571428571428e-05, 'epoch': 5.69}
+{'loss': 0.3205, 'grad_norm': 0.9519732594490051, 'learning_rate': 4.4000000000000006e-05, 'epoch': 5.72}
+{'loss': 0.3188, 'grad_norm': 0.7759018540382385, 'learning_rate': 4.371428571428572e-05, 'epoch': 5.75}
+{'loss': 0.3195, 'grad_norm': 0.931468665599823, 'learning_rate': 4.342857142857143e-05, 'epoch': 5.78}
+{'loss': 0.3256, 'grad_norm': 1.0431036949157715, 'learning_rate': 4.314285714285715e-05, 'epoch': 5.81}
+{'loss': 0.3189, 'grad_norm': 0.6952974200248718, 'learning_rate': 4.2857142857142856e-05, 'epoch': 5.83}
+ 61%|█████████████████████████████████████████████████████████████                                       | 220/360 [02:49<01:45,  1.33it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.33737656474113464, 'eval_runtime': 0.1056, 'eval_samples_per_second': 435.436, 'eval_steps_per_second': 9.466, 'epoch': 5.83}
+{'loss': 0.3159, 'grad_norm': 0.880944013595581, 'learning_rate': 4.257142857142857e-05, 'epoch': 5.86}
+{'loss': 0.3177, 'grad_norm': 0.6606886982917786, 'learning_rate': 4.228571428571429e-05, 'epoch': 5.89}
+{'loss': 0.317, 'grad_norm': 0.7617785930633545, 'learning_rate': 4.2e-05, 'epoch': 5.92}
+{'loss': 0.313, 'grad_norm': 0.6381199955940247, 'learning_rate': 4.1714285714285714e-05, 'epoch': 5.94}
+{'loss': 0.3184, 'grad_norm': 0.6644593477249146, 'learning_rate': 4.1428571428571437e-05, 'epoch': 5.97}
+{'loss': 0.3103, 'grad_norm': 0.9742204546928406, 'learning_rate': 4.1142857142857146e-05, 'epoch': 6.0}
+{'loss': 0.311, 'grad_norm': 0.6401204466819763, 'learning_rate': 4.085714285714286e-05, 'epoch': 6.03}
+{'loss': 0.3155, 'grad_norm': 0.785503089427948, 'learning_rate': 4.057142857142857e-05, 'epoch': 6.06}
+{'loss': 0.3109, 'grad_norm': 0.5417785048484802, 'learning_rate': 4.028571428571429e-05, 'epoch': 6.08}
+{'loss': 0.3114, 'grad_norm': 0.6186631321907043, 'learning_rate': 4e-05, 'epoch': 6.11}
+ 64%|███████████████████████████████████████████████████████████████▉                                    | 230/360 [02:57<01:40,  1.30it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.32762280106544495, 'eval_runtime': 0.1058, 'eval_samples_per_second': 434.862, 'eval_steps_per_second': 9.454, 'epoch': 6.11}
+{'loss': 0.3099, 'grad_norm': 0.6042884588241577, 'learning_rate': 3.971428571428571e-05, 'epoch': 6.14}
+{'loss': 0.3104, 'grad_norm': 0.5823580026626587, 'learning_rate': 3.942857142857143e-05, 'epoch': 6.17}
+{'loss': 0.3154, 'grad_norm': 0.6572258472442627, 'learning_rate': 3.9142857142857145e-05, 'epoch': 6.19}
+{'loss': 0.3121, 'grad_norm': 0.565834641456604, 'learning_rate': 3.885714285714286e-05, 'epoch': 6.22}
+{'loss': 0.3119, 'grad_norm': 0.7184673547744751, 'learning_rate': 3.857142857142858e-05, 'epoch': 6.25}
+{'loss': 0.3158, 'grad_norm': 0.7170347571372986, 'learning_rate': 3.8285714285714286e-05, 'epoch': 6.28}
+{'loss': 0.3038, 'grad_norm': 0.6102560758590698, 'learning_rate': 3.8e-05, 'epoch': 6.31}
+{'loss': 0.3124, 'grad_norm': 0.7612823843955994, 'learning_rate': 3.771428571428572e-05, 'epoch': 6.33}
+{'loss': 0.3028, 'grad_norm': 0.6277872920036316, 'learning_rate': 3.742857142857143e-05, 'epoch': 6.36}
+{'loss': 0.3202, 'grad_norm': 0.7007192373275757, 'learning_rate': 3.7142857142857143e-05, 'epoch': 6.39}
+ 67%|██████████████████████████████████████████████████████████████████▋                                 | 240/360 [03:05<01:33,  1.29it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.33139991760253906, 'eval_runtime': 0.1062, 'eval_samples_per_second': 433.311, 'eval_steps_per_second': 9.42, 'epoch': 6.39}
+{'loss': 0.313, 'grad_norm': 0.6396629810333252, 'learning_rate': 3.685714285714286e-05, 'epoch': 6.42}
+{'loss': 0.3116, 'grad_norm': 0.5031012892723083, 'learning_rate': 3.6571428571428576e-05, 'epoch': 6.44}
+{'loss': 0.3172, 'grad_norm': 0.7323219776153564, 'learning_rate': 3.628571428571429e-05, 'epoch': 6.47}
+{'loss': 0.3103, 'grad_norm': 0.9094661474227905, 'learning_rate': 3.6e-05, 'epoch': 6.5}
+{'loss': 0.3056, 'grad_norm': 0.5560885667800903, 'learning_rate': 3.571428571428572e-05, 'epoch': 6.53}
+{'loss': 0.3096, 'grad_norm': 1.0145907402038574, 'learning_rate': 3.5428571428571426e-05, 'epoch': 6.56}
+{'loss': 0.3049, 'grad_norm': 0.8287002444267273, 'learning_rate': 3.514285714285714e-05, 'epoch': 6.58}
+{'loss': 0.3047, 'grad_norm': 0.5207920074462891, 'learning_rate': 3.485714285714286e-05, 'epoch': 6.61}
+{'loss': 0.3156, 'grad_norm': 1.065272331237793, 'learning_rate': 3.4571428571428574e-05, 'epoch': 6.64}
+{'loss': 0.3055, 'grad_norm': 0.6712301969528198, 'learning_rate': 3.428571428571429e-05, 'epoch': 6.67}
+ 69%|█████████████████████████████████████████████████████████████████████▍                              | 250/360 [03:13<01:25,  1.29it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3387901484966278, 'eval_runtime': 0.1053, 'eval_samples_per_second': 436.76, 'eval_steps_per_second': 9.495, 'epoch': 6.67}
+{'loss': 0.3164, 'grad_norm': 0.9477025866508484, 'learning_rate': 3.4000000000000007e-05, 'epoch': 6.69}
+{'loss': 0.3263, 'grad_norm': 1.179295301437378, 'learning_rate': 3.3714285714285716e-05, 'epoch': 6.72}
+{'loss': 0.3136, 'grad_norm': 0.6969395875930786, 'learning_rate': 3.342857142857143e-05, 'epoch': 6.75}
+{'loss': 0.3143, 'grad_norm': 0.7752478718757629, 'learning_rate': 3.314285714285714e-05, 'epoch': 6.78}
+{'loss': 0.3143, 'grad_norm': 0.7640174031257629, 'learning_rate': 3.285714285714286e-05, 'epoch': 6.81}
+{'loss': 0.3093, 'grad_norm': 0.9904161691665649, 'learning_rate': 3.257142857142857e-05, 'epoch': 6.83}
+{'loss': 0.3185, 'grad_norm': 0.7707158923149109, 'learning_rate': 3.228571428571428e-05, 'epoch': 6.86}
+{'loss': 0.3122, 'grad_norm': 0.8660508394241333, 'learning_rate': 3.2000000000000005e-05, 'epoch': 6.89}
+{'loss': 0.3148, 'grad_norm': 0.7438889741897583, 'learning_rate': 3.1714285714285715e-05, 'epoch': 6.92}
+{'loss': 0.3137, 'grad_norm': 0.5746331810951233, 'learning_rate': 3.142857142857143e-05, 'epoch': 6.94}
+ 72%|████████████████████████████████████████████████████████████████████████▏                           | 260/360 [03:20<01:17,  1.29it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.32808226346969604, 'eval_runtime': 0.1059, 'eval_samples_per_second': 434.438, 'eval_steps_per_second': 9.444, 'epoch': 6.94}
+{'loss': 0.3125, 'grad_norm': 0.7158817052841187, 'learning_rate': 3.114285714285715e-05, 'epoch': 6.97}
+{'loss': 0.3094, 'grad_norm': 0.8010092973709106, 'learning_rate': 3.0857142857142856e-05, 'epoch': 7.0}
+{'loss': 0.3067, 'grad_norm': 0.7418866157531738, 'learning_rate': 3.057142857142857e-05, 'epoch': 7.03}
+{'loss': 0.3068, 'grad_norm': 0.6731083989143372, 'learning_rate': 3.0285714285714288e-05, 'epoch': 7.06}
+{'loss': 0.3075, 'grad_norm': 0.6405408382415771, 'learning_rate': 3e-05, 'epoch': 7.08}
+{'loss': 0.3096, 'grad_norm': 0.6403458118438721, 'learning_rate': 2.9714285714285717e-05, 'epoch': 7.11}
+{'loss': 0.3103, 'grad_norm': 0.7583682537078857, 'learning_rate': 2.9428571428571426e-05, 'epoch': 7.14}
+{'loss': 0.3064, 'grad_norm': 0.8137710094451904, 'learning_rate': 2.9142857142857146e-05, 'epoch': 7.17}
+{'loss': 0.3067, 'grad_norm': 0.7179896235466003, 'learning_rate': 2.885714285714286e-05, 'epoch': 7.19}
+{'loss': 0.3081, 'grad_norm': 0.9344987273216248, 'learning_rate': 2.857142857142857e-05, 'epoch': 7.22}
+ 75%|███████████████████████████████████████████████████████████████████████████                         | 270/360 [03:28<01:10,  1.28it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.33135512471199036, 'eval_runtime': 0.1054, 'eval_samples_per_second': 436.494, 'eval_steps_per_second': 9.489, 'epoch': 7.22}
+{'loss': 0.3024, 'grad_norm': 0.7846106886863708, 'learning_rate': 2.8285714285714287e-05, 'epoch': 7.25}
+{'loss': 0.3062, 'grad_norm': 0.5884716510772705, 'learning_rate': 2.8000000000000003e-05, 'epoch': 7.28}
+{'loss': 0.3017, 'grad_norm': 0.7277001738548279, 'learning_rate': 2.7714285714285716e-05, 'epoch': 7.31}
+{'loss': 0.3109, 'grad_norm': 0.6671104431152344, 'learning_rate': 2.742857142857143e-05, 'epoch': 7.33}
+{'loss': 0.3051, 'grad_norm': 0.6468051671981812, 'learning_rate': 2.714285714285714e-05, 'epoch': 7.36}
+{'loss': 0.3059, 'grad_norm': 0.7413132190704346, 'learning_rate': 2.6857142857142857e-05, 'epoch': 7.39}
+{'loss': 0.3108, 'grad_norm': 0.8842555284500122, 'learning_rate': 2.6571428571428576e-05, 'epoch': 7.42}
+{'loss': 0.31, 'grad_norm': 0.7701683044433594, 'learning_rate': 2.6285714285714286e-05, 'epoch': 7.44}
+{'loss': 0.2966, 'grad_norm': 0.6261523962020874, 'learning_rate': 2.6000000000000002e-05, 'epoch': 7.47}
+{'loss': 0.3063, 'grad_norm': 0.6180337071418762, 'learning_rate': 2.5714285714285714e-05, 'epoch': 7.5}
+ 78%|█████████████████████████████████████████████████████████████████████████████▊                      | 280/360 [03:36<01:00,  1.32it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.33175382018089294, 'eval_runtime': 0.1062, 'eval_samples_per_second': 433.162, 'eval_steps_per_second': 9.417, 'epoch': 7.5}
+{'loss': 0.3058, 'grad_norm': 0.910743236541748, 'learning_rate': 2.542857142857143e-05, 'epoch': 7.53}
+{'loss': 0.3092, 'grad_norm': 0.8947623372077942, 'learning_rate': 2.5142857142857147e-05, 'epoch': 7.56}
+{'loss': 0.3069, 'grad_norm': 0.8466401100158691, 'learning_rate': 2.485714285714286e-05, 'epoch': 7.58}
+{'loss': 0.316, 'grad_norm': 0.7808226943016052, 'learning_rate': 2.4571428571428572e-05, 'epoch': 7.61}
+{'loss': 0.2993, 'grad_norm': 0.6704230308532715, 'learning_rate': 2.4285714285714288e-05, 'epoch': 7.64}
+{'loss': 0.3034, 'grad_norm': 0.7090719938278198, 'learning_rate': 2.4e-05, 'epoch': 7.67}
+{'loss': 0.3077, 'grad_norm': 0.7552341818809509, 'learning_rate': 2.3714285714285717e-05, 'epoch': 7.69}
+{'loss': 0.3047, 'grad_norm': 0.7747870683670044, 'learning_rate': 2.342857142857143e-05, 'epoch': 7.72}
+{'loss': 0.3105, 'grad_norm': 1.1127567291259766, 'learning_rate': 2.3142857142857145e-05, 'epoch': 7.75}
+{'loss': 0.2997, 'grad_norm': 0.7083399891853333, 'learning_rate': 2.2857142857142858e-05, 'epoch': 7.78}
+ 81%|████████████████████████████████████████████████████████████████████████████████▌                   | 290/360 [03:43<00:50,  1.40it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3296756446361542, 'eval_runtime': 0.1057, 'eval_samples_per_second': 435.219, 'eval_steps_per_second': 9.461, 'epoch': 7.78}
+{'loss': 0.3035, 'grad_norm': 0.5560160279273987, 'learning_rate': 2.257142857142857e-05, 'epoch': 7.81}
+{'loss': 0.3066, 'grad_norm': 0.7277750372886658, 'learning_rate': 2.2285714285714287e-05, 'epoch': 7.83}
+{'loss': 0.3049, 'grad_norm': 0.683189868927002, 'learning_rate': 2.2000000000000003e-05, 'epoch': 7.86}
+{'loss': 0.3098, 'grad_norm': 0.7173867225646973, 'learning_rate': 2.1714285714285715e-05, 'epoch': 7.89}
+{'loss': 0.3081, 'grad_norm': 0.698379397392273, 'learning_rate': 2.1428571428571428e-05, 'epoch': 7.92}
+{'loss': 0.3109, 'grad_norm': 0.5789396166801453, 'learning_rate': 2.1142857142857144e-05, 'epoch': 7.94}
+{'loss': 0.3149, 'grad_norm': 0.8007158637046814, 'learning_rate': 2.0857142857142857e-05, 'epoch': 7.97}
+{'loss': 0.3087, 'grad_norm': 0.8773916959762573, 'learning_rate': 2.0571428571428573e-05, 'epoch': 8.0}
+{'loss': 0.2989, 'grad_norm': 0.7212807536125183, 'learning_rate': 2.0285714285714286e-05, 'epoch': 8.03}
+{'loss': 0.3069, 'grad_norm': 0.6096076369285583, 'learning_rate': 2e-05, 'epoch': 8.06}
+ 83%|███████████████████████████████████████████████████████████████████████████████████▎                | 300/360 [03:51<00:46,  1.29it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.33069342374801636, 'eval_runtime': 0.105, 'eval_samples_per_second': 438.062, 'eval_steps_per_second': 9.523, 'epoch': 8.06}
+{'loss': 0.295, 'grad_norm': 0.5173125267028809, 'learning_rate': 1.9714285714285714e-05, 'epoch': 8.08}
+{'loss': 0.3014, 'grad_norm': 0.6913369297981262, 'learning_rate': 1.942857142857143e-05, 'epoch': 8.11}
+{'loss': 0.3037, 'grad_norm': 0.7195921540260315, 'learning_rate': 1.9142857142857143e-05, 'epoch': 8.14}
+{'loss': 0.3031, 'grad_norm': 0.6366473436355591, 'learning_rate': 1.885714285714286e-05, 'epoch': 8.17}
+{'loss': 0.2957, 'grad_norm': 0.5457173585891724, 'learning_rate': 1.8571428571428572e-05, 'epoch': 8.19}
+{'loss': 0.2997, 'grad_norm': 0.6149912476539612, 'learning_rate': 1.8285714285714288e-05, 'epoch': 8.22}
+{'loss': 0.3048, 'grad_norm': 0.5352884531021118, 'learning_rate': 1.8e-05, 'epoch': 8.25}
+{'loss': 0.308, 'grad_norm': 0.6278409361839294, 'learning_rate': 1.7714285714285713e-05, 'epoch': 8.28}
+{'loss': 0.3005, 'grad_norm': 0.5881698727607727, 'learning_rate': 1.742857142857143e-05, 'epoch': 8.31}
+{'loss': 0.302, 'grad_norm': 0.6125136613845825, 'learning_rate': 1.7142857142857145e-05, 'epoch': 8.33}
+ 86%|██████████████████████████████████████████████████████████████████████████████████████              | 310/360 [03:59<00:38,  1.29it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3336547017097473, 'eval_runtime': 0.1055, 'eval_samples_per_second': 436.2, 'eval_steps_per_second': 9.483, 'epoch': 8.33}
+{'loss': 0.3012, 'grad_norm': 0.6722866892814636, 'learning_rate': 1.6857142857142858e-05, 'epoch': 8.36}
+{'loss': 0.297, 'grad_norm': 0.6827422976493835, 'learning_rate': 1.657142857142857e-05, 'epoch': 8.39}
+{'loss': 0.2977, 'grad_norm': 0.7612675428390503, 'learning_rate': 1.6285714285714287e-05, 'epoch': 8.42}
+{'loss': 0.3009, 'grad_norm': 0.5952971577644348, 'learning_rate': 1.6000000000000003e-05, 'epoch': 8.44}
+{'loss': 0.3043, 'grad_norm': 0.8323265314102173, 'learning_rate': 1.5714285714285715e-05, 'epoch': 8.47}
+{'loss': 0.2956, 'grad_norm': 0.8321357369422913, 'learning_rate': 1.5428571428571428e-05, 'epoch': 8.5}
+{'loss': 0.3029, 'grad_norm': 0.6457182168960571, 'learning_rate': 1.5142857142857144e-05, 'epoch': 8.53}
+{'loss': 0.2972, 'grad_norm': 0.5753086805343628, 'learning_rate': 1.4857142857142858e-05, 'epoch': 8.56}
+{'loss': 0.2966, 'grad_norm': 0.8767444491386414, 'learning_rate': 1.4571428571428573e-05, 'epoch': 8.58}
+{'loss': 0.3049, 'grad_norm': 0.929669201374054, 'learning_rate': 1.4285714285714285e-05, 'epoch': 8.61}
+ 89%|████████████████████████████████████████████████████████████████████████████████████████▉           | 320/360 [04:07<00:31,  1.29it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3303754925727844, 'eval_runtime': 0.1057, 'eval_samples_per_second': 435.106, 'eval_steps_per_second': 9.459, 'epoch': 8.61}
+{'loss': 0.2989, 'grad_norm': 0.7576697468757629, 'learning_rate': 1.4000000000000001e-05, 'epoch': 8.64}
+{'loss': 0.3051, 'grad_norm': 0.6402246952056885, 'learning_rate': 1.3714285714285716e-05, 'epoch': 8.67}
+{'loss': 0.2974, 'grad_norm': 0.5665248036384583, 'learning_rate': 1.3428571428571429e-05, 'epoch': 8.69}
+{'loss': 0.3061, 'grad_norm': 0.9747456312179565, 'learning_rate': 1.3142857142857143e-05, 'epoch': 8.72}
+{'loss': 0.3012, 'grad_norm': 0.657123863697052, 'learning_rate': 1.2857142857142857e-05, 'epoch': 8.75}
+{'loss': 0.2995, 'grad_norm': 0.7186892032623291, 'learning_rate': 1.2571428571428573e-05, 'epoch': 8.78}
+{'loss': 0.3026, 'grad_norm': 0.6889364123344421, 'learning_rate': 1.2285714285714286e-05, 'epoch': 8.81}
+{'loss': 0.3009, 'grad_norm': 0.6299145817756653, 'learning_rate': 1.2e-05, 'epoch': 8.83}
+{'loss': 0.3002, 'grad_norm': 0.7328559756278992, 'learning_rate': 1.1714285714285715e-05, 'epoch': 8.86}
+{'loss': 0.2953, 'grad_norm': 0.6111913919448853, 'learning_rate': 1.1428571428571429e-05, 'epoch': 8.89}
+ 92%|███████████████████████████████████████████████████████████████████████████████████████████▋        | 330/360 [04:14<00:22,  1.31it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3253832757472992, 'eval_runtime': 0.106, 'eval_samples_per_second': 434.082, 'eval_steps_per_second': 9.437, 'epoch': 8.89}
+{'loss': 0.3023, 'grad_norm': 0.6739629507064819, 'learning_rate': 1.1142857142857143e-05, 'epoch': 8.92}
+{'loss': 0.2978, 'grad_norm': 0.6967675685882568, 'learning_rate': 1.0857142857142858e-05, 'epoch': 8.94}
+{'loss': 0.3043, 'grad_norm': 0.702989935874939, 'learning_rate': 1.0571428571428572e-05, 'epoch': 8.97}
+{'loss': 0.3034, 'grad_norm': 1.156525731086731, 'learning_rate': 1.0285714285714286e-05, 'epoch': 9.0}
+{'loss': 0.2989, 'grad_norm': 0.564460277557373, 'learning_rate': 1e-05, 'epoch': 9.03}
+{'loss': 0.2955, 'grad_norm': 0.5435044169425964, 'learning_rate': 9.714285714285715e-06, 'epoch': 9.06}
+{'loss': 0.2986, 'grad_norm': 0.511762797832489, 'learning_rate': 9.42857142857143e-06, 'epoch': 9.08}
+{'loss': 0.2932, 'grad_norm': 0.6208844780921936, 'learning_rate': 9.142857142857144e-06, 'epoch': 9.11}
+{'loss': 0.2932, 'grad_norm': 0.5209355354309082, 'learning_rate': 8.857142857142857e-06, 'epoch': 9.14}
+{'loss': 0.302, 'grad_norm': 0.5852081775665283, 'learning_rate': 8.571428571428573e-06, 'epoch': 9.17}
+ 94%|██████████████████████████████████████████████████████████████████████████████████████████████▍     | 340/360 [04:22<00:15,  1.28it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.32703322172164917, 'eval_runtime': 0.1055, 'eval_samples_per_second': 435.944, 'eval_steps_per_second': 9.477, 'epoch': 9.17}
+{'loss': 0.2973, 'grad_norm': 0.6613155603408813, 'learning_rate': 8.285714285714285e-06, 'epoch': 9.19}
+{'loss': 0.296, 'grad_norm': 0.6458805203437805, 'learning_rate': 8.000000000000001e-06, 'epoch': 9.22}
+{'loss': 0.2998, 'grad_norm': 0.5602886080741882, 'learning_rate': 7.714285714285714e-06, 'epoch': 9.25}
+{'loss': 0.2898, 'grad_norm': 0.5723817348480225, 'learning_rate': 7.428571428571429e-06, 'epoch': 9.28}
+{'loss': 0.2935, 'grad_norm': 0.6257355213165283, 'learning_rate': 7.142857142857143e-06, 'epoch': 9.31}
+{'loss': 0.2909, 'grad_norm': 0.6624913811683655, 'learning_rate': 6.857142857142858e-06, 'epoch': 9.33}
+{'loss': 0.3001, 'grad_norm': 0.5716632604598999, 'learning_rate': 6.5714285714285714e-06, 'epoch': 9.36}
+{'loss': 0.2964, 'grad_norm': 0.6996496319770813, 'learning_rate': 6.285714285714287e-06, 'epoch': 9.39}
+{'loss': 0.2927, 'grad_norm': 0.7235862612724304, 'learning_rate': 6e-06, 'epoch': 9.42}
+{'loss': 0.2956, 'grad_norm': 0.6455687284469604, 'learning_rate': 5.7142857142857145e-06, 'epoch': 9.44}
+ 97%|█████████████████████████████████████████████████████████████████████████████████████████████████▏  | 350/360 [04:30<00:07,  1.28it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.32588478922843933, 'eval_runtime': 0.1057, 'eval_samples_per_second': 435.067, 'eval_steps_per_second': 9.458, 'epoch': 9.44}
+{'loss': 0.2959, 'grad_norm': 0.5984405279159546, 'learning_rate': 5.428571428571429e-06, 'epoch': 9.47}
+{'loss': 0.2936, 'grad_norm': 0.6240008473396301, 'learning_rate': 5.142857142857143e-06, 'epoch': 9.5}
+{'loss': 0.2987, 'grad_norm': 0.6871291399002075, 'learning_rate': 4.857142857142858e-06, 'epoch': 9.53}
+{'loss': 0.2941, 'grad_norm': 0.6369628310203552, 'learning_rate': 4.571428571428572e-06, 'epoch': 9.56}
+{'loss': 0.2959, 'grad_norm': 0.6651211977005005, 'learning_rate': 4.285714285714286e-06, 'epoch': 9.58}
+{'loss': 0.2991, 'grad_norm': 0.7005758285522461, 'learning_rate': 4.000000000000001e-06, 'epoch': 9.61}
+{'loss': 0.2936, 'grad_norm': 0.5685088634490967, 'learning_rate': 3.7142857142857146e-06, 'epoch': 9.64}
+{'loss': 0.2947, 'grad_norm': 0.6322896480560303, 'learning_rate': 3.428571428571429e-06, 'epoch': 9.67}
+{'loss': 0.2951, 'grad_norm': 0.6149244904518127, 'learning_rate': 3.1428571428571433e-06, 'epoch': 9.69}
+{'loss': 0.2975, 'grad_norm': 0.685043215751648, 'learning_rate': 2.8571428571428573e-06, 'epoch': 9.72}
+100%|████████████████████████████████████████████████████████████████████████████████████████████████████| 360/360 [04:37<00:00,  1.54it/s]Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.
+Trainer.tokenizer is now deprecated. You should use Trainer.processing_class instead.                                                      
+{'eval_loss': 0.3277616798877716, 'eval_runtime': 0.1059, 'eval_samples_per_second': 434.21, 'eval_steps_per_second': 9.439, 'epoch': 9.72}
+{'loss': 0.2951, 'grad_norm': 0.8039355874061584, 'learning_rate': 2.5714285714285716e-06, 'epoch': 9.75}
+{'loss': 0.2928, 'grad_norm': 0.6388571262359619, 'learning_rate': 2.285714285714286e-06, 'epoch': 9.78}
+{'loss': 0.2934, 'grad_norm': 0.5934285521507263, 'learning_rate': 2.0000000000000003e-06, 'epoch': 9.81}
+{'loss': 0.2952, 'grad_norm': 0.5320731401443481, 'learning_rate': 1.7142857142857145e-06, 'epoch': 9.83}
+{'loss': 0.2932, 'grad_norm': 0.6137614846229553, 'learning_rate': 1.4285714285714286e-06, 'epoch': 9.86}
+{'loss': 0.2944, 'grad_norm': 0.8172494769096375, 'learning_rate': 1.142857142857143e-06, 'epoch': 9.89}
+{'loss': 0.2917, 'grad_norm': 0.6931514739990234, 'learning_rate': 8.571428571428572e-07, 'epoch': 9.92}
+{'loss': 0.2952, 'grad_norm': 0.8408763408660889, 'learning_rate': 5.714285714285715e-07, 'epoch': 9.94}
+{'loss': 0.2948, 'grad_norm': 0.6687312126159668, 'learning_rate': 2.8571428571428575e-07, 'epoch': 9.97}
+{'loss': 0.2944, 'grad_norm': 0.8545575737953186, 'learning_rate': 0.0, 'epoch': 10.0}
+100%|████████████████████████████████████████████████████████████████████████████████████████████████████| 360/360 [04:39<00:00,  1.29it/s]
+                                                                                                                                           
+{'eval_loss': 0.3285914361476898, 'eval_runtime': 0.1058, 'eval_samples_per_second': 434.729, 'eval_steps_per_second': 9.451, 'epoch': 10.0}
+{'train_runtime': 279.4567, 'train_samples_per_second': 162.422, 'train_steps_per_second': 1.288, 'train_loss': 0.35597237489289707, 'epoch': 10.0}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/files/wandb-metadata.json

@@ -0,0 +1,55 @@
+{
+  "os": "Linux-6.8.0-101-generic-x86_64-with-glibc2.35",
+  "python": "3.9.20",
+  "startedAt": "2026-03-27T08:20:56.140561Z",
+  "args": [
+    "--config-path",
+    "config/main_table",
+    "--config-name",
+    "llmdp_llm_adroit-hand-hammer-v1.yaml"
+  ],
+  "program": "/tmp2/chyang/workspace/LLM-BC/./train.py",
+  "codePath": "train.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "2e85824cdb13f64f31923d9430e890dadc78d394"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1",
+  "host": "A6000-2",
+  "username": "chyang",
+  "executable": "/home/chyang/miniconda3/envs/llm-bc/bin/python3",
+  "codePathLocal": "train.py",
+  "cpu_count": 12,
+  "cpu_count_logical": 24,
+  "gpu": "NVIDIA RTX A6000",
+  "gpu_count": 2,
+  "disk": {
+    "/": {
+      "total": "1967317549056",
+      "used": "733081182208"
+    }
+  },
+  "memory": {
+    "total": "134538502144"
+  },
+  "cpu": {
+    "count": 12,
+    "countLogical": 24
+  },
+  "gpu_nvidia": [
+    {
+      "name": "NVIDIA RTX A6000",
+      "memoryTotal": "51527024640",
+      "cudaCores": 10752,
+      "architecture": "Ampere"
+    },
+    {
+      "name": "NVIDIA RTX A6000",
+      "memoryTotal": "51527024640",
+      "cudaCores": 10752,
+      "architecture": "Ampere"
+    }
+  ],
+  "cudaVersion": "12.6"
+}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/files/wandb-summary.json

@@ -0,0 +1 @@
+{"eval/loss":0.3285914361476898,"train_loss":0.35597237489289707,"eval/steps_per_second":9.451,"train/global_step":360,"train/grad_norm":0.8545575737953186,"train/learning_rate":0,"_timestamp":1.774599943812631e+09,"_step":396,"eval/runtime":0.1058,"train_steps_per_second":1.288,"eval/samples_per_second":434.729,"_runtime":287.672274788,"train_samples_per_second":162.422,"train/loss":0.2944,"_wandb":{"runtime":287},"train_runtime":279.4567,"train/epoch":10,"total_flos":3.83135455600128e+15}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/logs/debug-core.log

@@ -0,0 +1,16 @@
+{"time":"2026-03-27T16:20:55.459583429+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmp0ig_2z3y/port-2703097.txt","pid":2703097,"debug":false,"disable-analytics":false}
+{"time":"2026-03-27T16:20:55.4596054+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2026-03-27T16:20:55.460720879+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":44997,"Zone":""}}
+{"time":"2026-03-27T16:20:55.460793184+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":2703097}
+{"time":"2026-03-27T16:20:55.657444973+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:60248"}
+{"time":"2026-03-27T16:20:56.143936837+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"nhmfpc2t","id":"127.0.0.1:60248"}
+{"time":"2026-03-27T16:20:56.25130017+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"nhmfpc2t","id":"127.0.0.1:60248"}
+{"time":"2026-03-27T16:25:50.90098896+08:00","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"nhmfpc2t","id":"127.0.0.1:60248"}
+{"time":"2026-03-27T16:25:50.901237575+08:00","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"nhmfpc2t","id":"127.0.0.1:60248"}
+{"time":"2026-03-27T16:25:51.590522527+08:00","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"127.0.0.1:60248"}
+{"time":"2026-03-27T16:25:51.59056499+08:00","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"127.0.0.1:60248"}
+{"time":"2026-03-27T16:25:51.590573489+08:00","level":"INFO","msg":"server is shutting down"}
+{"time":"2026-03-27T16:25:51.590600946+08:00","level":"INFO","msg":"connection: Close: initiating connection closure","id":"127.0.0.1:60248"}
+{"time":"2026-03-27T16:25:51.590777149+08:00","level":"INFO","msg":"connection: Close: connection successfully closed","id":"127.0.0.1:60248"}
+{"time":"2026-03-27T16:25:51.590829802+08:00","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"127.0.0.1:60248"}
+{"time":"2026-03-27T16:25:51.590853327+08:00","level":"INFO","msg":"server is closed"}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/logs/debug-internal.log

@@ -0,0 +1,17 @@
+{"time":"2026-03-27T16:20:56.144221699+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-03-27T16:20:56.144248456+08:00","level":"INFO","msg":"created symlink","path":"/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/logs/debug-core.log"}
+{"time":"2026-03-27T16:20:56.25122394+08:00","level":"INFO","msg":"created new stream","id":"nhmfpc2t"}
+{"time":"2026-03-27T16:20:56.25128922+08:00","level":"INFO","msg":"stream: started","id":"nhmfpc2t"}
+{"time":"2026-03-27T16:20:56.251326647+08:00","level":"INFO","msg":"sender: started","stream_id":"nhmfpc2t"}
+{"time":"2026-03-27T16:20:56.251319685+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"nhmfpc2t"}}
+{"time":"2026-03-27T16:20:56.251303072+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"nhmfpc2t"}}
+{"time":"2026-03-27T16:20:57.417356613+08:00","level":"INFO","msg":"Starting system monitor"}
+{"time":"2026-03-27T16:25:43.814415705+08:00","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2026-03-27T16:25:43.815067495+08:00","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2026-03-27T16:25:44.815119947+08:00","level":"INFO","msg":"handler: operation stats","stats":{"operations":[{"desc":"uploading wandb-summary.json","runtime_seconds":0.136938035,"progress":"495B/495B"},{"desc":"saving job artifact","runtime_seconds":0.037243753}],"total_operations":2}}
+{"time":"2026-03-27T16:25:49.240469481+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-03-27T16:25:50.901076411+08:00","level":"INFO","msg":"stream: closing","id":"nhmfpc2t"}
+{"time":"2026-03-27T16:25:50.901107424+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"nhmfpc2t"}}
+{"time":"2026-03-27T16:25:50.901135119+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"nhmfpc2t"}}
+{"time":"2026-03-27T16:25:50.901152717+08:00","level":"INFO","msg":"sender: closed","stream_id":"nhmfpc2t"}
+{"time":"2026-03-27T16:25:50.90122351+08:00","level":"INFO","msg":"stream: closed","id":"nhmfpc2t"}

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/logs/debug.log

@@ -0,0 +1,35 @@
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Configure stats pid to 2703097
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Loading settings from /home/chyang/.config/wandb/settings
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Loading settings from /tmp2/chyang/workspace/LLM-BC/wandb/settings
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'train.py', 'program_abspath': '/tmp2/chyang/workspace/LLM-BC/train.py', 'program': '/tmp2/chyang/workspace/LLM-BC/./train.py'}
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:_log_setup():533] Logging user logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/logs/debug.log
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:_log_setup():534] Logging internal logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/logs/debug-internal.log
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:init():619] calling init triggers
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:init():626] wandb.init called with sweep_config: {}
+config: {'name': 'train_llm_lowdim', '_target_': 'llmbc.workspace.train_llm_workspace.TrainLLMWorkspace', 'obs_dim': 46, 'action_dim': 26, 'horizon': 1, 'n_obs_steps': 1, 'n_action_steps': 1, 'task_name': 'adroit-hand-hammer-v1', 'exp_name': 'train llm', 'model_name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'use_quantization': False, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'dataset': {'test_data_ratio': 0.01}, 'debug': False, 'training': {'seed': 42, 'per_device_train_batch_size': 128, 'per_device_eval_batch_size': 128, 'gradient_accumulation_steps': 1, 'optim': 'paged_adamw_32bit', 'num_train_epochs': 10, 'eval_strategy': 'steps', 'logging_steps': 1, 'warmup_steps': 10, 'logging_strategy': 'steps', 'learning_rate': 0.0001, 'fp16': False, 'bf16': True, 'tf32': True, 'group_by_length': True, 'report_to': 'wandb', 'save_steps': 5000, 'eval_steps': 10, 'use_joint_mlp_projector': True, 'joint_obs_action_mlp_lr': 5e-05}, 'trainer': {'obs_dim': 46, 'action_dim': 26, 'use_joint_mlp_projector': True, 'max_seq_length': 100, 'dataset_text_field': 'text', 'packing': False}, 'logging': {'project': 'llm_module_finetuning', 'resume': True, 'mode': 'online', 'name': '2026.03.27-16.20.52_train_llm_lowdim_adroit-hand-hammer-v1', 'tags': ['train_llm_lowdim', 'adroit-hand-hammer-v1', 'train llm'], 'id': None, 'group': None}, 'multi_run': {'run_dir': 'data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1', 'wandb_name_base': '2026.03.27-16.20.52_train_llm_lowdim_adroit-hand-hammer-v1'}, 'task': {'name': 'adroit-hand-hammer-v1', 'obs_dim': 46, 'action_dim': 26, 'env_runner': {'_target_': 'llmbc.env_runner.adroit_lowdim_runner.AdroitHandLowdimRunner', 'env_name': 'llf-adroit-adroit-hand-hammer-v1', 'n_train': 10, 'n_test': 50, 'n_envs': 10, 'max_steps': 150, 'n_obs_steps': 1, 'n_action_steps': 1, 'instruction_type': 'b', 'feedback_type': ['hp', 'hn', 'fp'], 'visual': False, 'discount': 0.99}, 'dataset': {'_target_': 'llmbc.dataset.adroit_lowdim_dataset.AdroitHandLowdimDataset', 'data_path': 'datasets/adroit-hand-hammer-v1-general.pt', 'data_path2': 'datasets/adroit-hand-hammer-v1.pt', 'horizon': 1, 'pad_before': 0, 'pad_after': 0, 'obs_eef_target': True, 'use_manual_normalizer': False, 'val_ratio': 0.05, 'dummy_normalizer': False}, 'instructor': {'_target_': 'llmbc.translator.instructor.adroit_instructor.adroit_hand_hammer_v1_instructor.AdroitHandHammerV1Instructor'}}, 'llm': {'name': 'HuggingFaceTB/SmolLM2-135M-Instruct', 'model_name': 'SmolLM2-135M-Instruct', 'config_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaConfig', 'causal_lm_target': 'llmbc.model.llm.llama_lowdim_model.LowdimLlamaForCausalLM', 'use_quantization': False, 'use_joint_mlp_projector': True, 'llm_mode': 'mlp-finetuned', 'finetune_mode': 'orig', 'checkpoint': 'data/outputs/2026.03.27/14.38.20_train_mlp_projector_adroit-hand-hammer-v1/checkpoints/latest.ckpt', 'max_length': 100, 'lora_config': {'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'prompter': {'_target_': 'llmbc.translator.prompter.smollm2_prompter.SmolLM2Prompter', 'use_joint_mlp_projector': True}, 'hydra': {'job': {'override_dirname': 'HuggingFaceTB/SmolLM2-135M-Instruct'}, 'run': {'dir': 'data/outputs/2026.03.27/16.20.52_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:init():669] starting backend
+2026-03-27 16:20:56,139 INFO    MainThread:2703097 [wandb_init.py:init():673] sending inform_init request
+2026-03-27 16:20:56,140 INFO    MainThread:2703097 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-03-27 16:20:56,140 INFO    MainThread:2703097 [wandb_init.py:init():686] backend started and connected
+2026-03-27 16:20:56,144 INFO    MainThread:2703097 [wandb_init.py:init():781] updated telemetry
+2026-03-27 16:20:56,170 INFO    MainThread:2703097 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-03-27 16:20:57,414 INFO    MainThread:2703097 [wandb_init.py:init():867] starting run threads in backend
+2026-03-27 16:20:57,521 INFO    MainThread:2703097 [wandb_run.py:_console_start():2451] atexit reg
+2026-03-27 16:20:57,522 INFO    MainThread:2703097 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-03-27 16:20:57,522 INFO    MainThread:2703097 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-03-27 16:20:57,522 INFO    MainThread:2703097 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-03-27 16:20:57,524 INFO    MainThread:2703097 [wandb_init.py:init():911] run started, returning control to user process
+2026-03-27 16:21:04,359 INFO    MainThread:2703097 [wandb_run.py:_config_callback():1389] config_cb None None {'obs_dim': 46, 'action_dim': 26, 'use_joint_mlp_projector': True, 'vocab_size': 49152, 'max_position_embeddings': 8192, 'hidden_size': 576, 'intermediate_size': 1536, 'num_hidden_layers': 30, 'num_attention_heads': 9, 'num_key_value_heads': 3, 'hidden_act': 'silu', 'initializer_range': 0.041666666666666664, 'rms_norm_eps': 1e-05, 'pretraining_tp': 1, 'use_cache': False, 'rope_theta': 100000, 'rope_scaling': None, 'attention_bias': False, 'attention_dropout': 0.0, 'mlp_bias': False, 'head_dim': 64, 'return_dict': True, 'output_hidden_states': False, 'output_attentions': False, 'torchscript': False, 'torch_dtype': 'bfloat16', 'use_bfloat16': False, 'tf_legacy_loss': False, 'pruned_heads': {}, 'tie_word_embeddings': True, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'is_decoder': False, 'cross_attention_hidden_size': None, 'add_cross_attention': False, 'tie_encoder_decoder': False, 'max_length': 20, 'min_length': 0, 'do_sample': False, 'early_stopping': False, 'num_beams': 1, 'num_beam_groups': 1, 'diversity_penalty': 0.0, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 0, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': None, 'forced_eos_token_id': None, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'suppress_tokens': None, 'begin_suppress_tokens': None, 'architectures': ['LlamaForCausalLM'], 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': 1, 'pad_token_id': 2, 'eos_token_id': 2, 'sep_token_id': None, 'decoder_start_token_id': None, 'task_specific_params': None, 'problem_type': None, '_name_or_path': 'HuggingFaceTB/SmolLM2-135M-Instruct', '_attn_implementation_autoset': True, 'transformers_version': '4.47.1', 'is_llama_config': True, 'model_type': 'llama_lowdim', 'rope_interleaved': False, 'transformers.js_config': {'kv_cache_dtype': {'q4f16': 'float16', 'fp16': 'float16'}}, 'output_dir': '/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1', 'overwrite_output_dir': False, 'do_train': False, 'do_eval': True, 'do_predict': False, 'eval_strategy': 'steps', 'prediction_loss_only': False, 'per_device_train_batch_size': 128, 'per_device_eval_batch_size': 128, 'per_gpu_train_batch_size': None, 'per_gpu_eval_batch_size': None, 'gradient_accumulation_steps': 1, 'eval_accumulation_steps': None, 'eval_delay': 0, 'torch_empty_cache_steps': None, 'learning_rate': 0.0001, 'weight_decay': 0.0, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'max_grad_norm': 1.0, 'num_train_epochs': 10, 'max_steps': -1, 'lr_scheduler_type': 'linear', 'lr_scheduler_kwargs': {}, 'warmup_ratio': 0.0, 'warmup_steps': 10, 'log_level': 'passive', 'log_level_replica': 'warning', 'log_on_each_node': True, 'logging_dir': '/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/runs/Mar27_16-21-01_A6000-2', 'logging_strategy': 'steps', 'logging_first_step': False, 'logging_steps': 1, 'logging_nan_inf_filter': True, 'save_strategy': 'steps', 'save_steps': 5000, 'save_total_limit': None, 'save_safetensors': True, 'save_on_each_node': False, 'save_only_model': False, 'restore_callback_states_from_checkpoint': False, 'no_cuda': False, 'use_cpu': False, 'use_mps_device': False, 'seed': 42, 'data_seed': None, 'jit_mode_eval': False, 'use_ipex': False, 'bf16': True, 'fp16': False, 'fp16_opt_level': 'O1', 'half_precision_backend': 'auto', 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': True, 'local_rank': 0, 'ddp_backend': None, 'tpu_num_cores': None, 'tpu_metrics_debug': False, 'debug': [], 'dataloader_drop_last': False, 'eval_steps': 10, 'dataloader_num_workers': 0, 'dataloader_prefetch_factor': None, 'past_index': -1, 'run_name': '/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1', 'disable_tqdm': False, 'remove_unused_columns': True, 'label_names': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'fsdp': [], 'fsdp_min_num_params': 0, 'fsdp_config': {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, 'fsdp_transformer_layer_cls_to_wrap': None, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'deepspeed': None, 'label_smoothing_factor': 0.0, 'optim': 'paged_adamw_32bit', 'optim_args': None, 'adafactor': False, 'group_by_length': True, 'length_column_name': 'length', 'report_to': ['wandb'], 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'skip_memory_metrics': True, 'use_legacy_prediction_loop': False, 'push_to_hub': False, 'resume_from_checkpoint': None, 'hub_model_id': None, 'hub_strategy': 'every_save', 'hub_token': '<HUB_TOKEN>', 'hub_private_repo': None, 'hub_always_push': False, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'include_inputs_for_metrics': False, 'include_for_metrics': [], 'eval_do_concat_batches': True, 'fp16_backend': 'auto', 'evaluation_strategy': None, 'push_to_hub_model_id': None, 'push_to_hub_organization': None, 'push_to_hub_token': '<PUSH_TO_HUB_TOKEN>', 'mp_parameters': '', 'auto_find_batch_size': False, 'full_determinism': False, 'torchdynamo': None, 'ray_scope': 'last', 'ddp_timeout': 1800, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'dispatch_batches': None, 'split_batches': None, 'include_tokens_per_second': False, 'include_num_input_tokens_seen': False, 'neftune_noise_alpha': None, 'optim_target_modules': None, 'batch_eval_metrics': False, 'eval_on_start': False, 'use_liger_kernel': False, 'eval_use_gather_object': False, 'average_tokens_across_devices': False, 'dataset_text_field': 'text', 'packing': False, 'max_seq_length': 100, 'dataset_num_proc': None, 'dataset_batch_size': 1000, 'model_init_kwargs': None, 'dataset_kwargs': {}, 'eval_packing': None, 'num_of_sequences': 1024, 'chars_per_token': '<CHARS_PER_TOKEN>', 'use_liger': False, 'joint_obs_action_mlp_lr': 5e-05, 'obs_mlp_lr': None, 'action_mlp_lr': None}
+2026-03-27 16:21:04,361 INFO    MainThread:2703097 [wandb_config.py:__setitem__():154] config set model/num_parameters = 134889408 - <bound method Run._config_callback of <wandb.sdk.wandb_run.Run object at 0x7e6758745670>>
+2026-03-27 16:21:04,361 INFO    MainThread:2703097 [wandb_run.py:_config_callback():1389] config_cb model/num_parameters 134889408 None
+2026-03-27 16:25:43,813 INFO    MainThread:2703097 [wandb_run.py:_finish():2146] finishing run chyang25-national-taiwan-university/llm_module_finetuning/nhmfpc2t
+2026-03-27 16:25:43,813 INFO    MainThread:2703097 [wandb_run.py:_atexit_cleanup():2414] got exitcode: 0
+2026-03-27 16:25:43,813 INFO    MainThread:2703097 [wandb_run.py:_restore():2396] restore
+2026-03-27 16:25:43,814 INFO    MainThread:2703097 [wandb_run.py:_restore():2402] restore done
+2026-03-27 16:25:50,896 INFO    MainThread:2703097 [wandb_run.py:_footer_history_summary_info():3963] rendering history
+2026-03-27 16:25:50,896 INFO    MainThread:2703097 [wandb_run.py:_footer_history_summary_info():3995] rendering summary
+2026-03-27 16:25:50,900 INFO    MainThread:2703097 [wandb_run.py:_footer_sync_info():3922] logging synced files

2026.03.27/16.20.52_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_162056-nhmfpc2t/run-nhmfpc2t.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19281f2406b878cd01e76af72f5dd82f13698026367ccc8dd2b76d39269c8f31
+size 858565