Upload folder using huggingface_hub

.gitattributes

@@ -174,3 +174,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 2026.03.18/22.30.56_train_llm_lowdim_box-close-v2/wandb/run-20260318_223101-x1m9280b/run-x1m9280b.wandb filter=lfs diff=lfs merge=lfs -text
 2026.03.25/21.48.48_train_llm_lowdim_box-close-v2/wandb/run-20260325_214852-hhto1ipm/run-hhto1ipm.wandb 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

2026.03.27/14.32.19_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: true
+  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/2026.03.27/02.42.34_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}

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/.hydra/hydra.yaml

@@ -0,0 +1,156 @@
+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_llm_workspace
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: train_llm_workspace
+    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/llmbc/config
+      schema: file
+      provider: main
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1
+    choices:
+      llm: smollm2-135m-instruct
+      task: adroit-hand-hammer-v1
+      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

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/.hydra/overrides.yaml

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

2026.03.27/14.32.19_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
+}

2026.03.27/14.32.19_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"
+}

2026.03.27/14.32.19_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 sha256:22fab54eb648b681b668530f111e4b748ca6f1ba319deb7306480cfbfcb0324b
+size 1499776

2026.03.27/14.32.19_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:f55c1948674d165534e4d6baf48e4473a272ef058b49958b1718d72870e43e61
+size 539588496

2026.03.27/14.32.19_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 sha256:4518f68cdd384541bf024f88cafbbf45572f10f7e2817a1083f662ac6c3256fc
+size 1079284794

2026.03.27/14.32.19_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

2026.03.27/14.32.19_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

2026.03.27/14.32.19_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
+  }
+}

2026.03.27/14.32.19_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
+}

2026.03.27/14.32.19_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": 41.524925231933594,
+      "learning_rate": 1e-05,
+      "loss": 1.3459,
+      "step": 1
+    },
+    {
+      "epoch": 0.05555555555555555,
+      "grad_norm": 36.50713348388672,
+      "learning_rate": 2e-05,
+      "loss": 1.3676,
+      "step": 2
+    },
+    {
+      "epoch": 0.08333333333333333,
+      "grad_norm": 101.92433166503906,
+      "learning_rate": 3e-05,
+      "loss": 1.3473,
+      "step": 3
+    },
+    {
+      "epoch": 0.1111111111111111,
+      "grad_norm": 107.9959945678711,
+      "learning_rate": 4e-05,
+      "loss": 1.334,
+      "step": 4
+    },
+    {
+      "epoch": 0.1388888888888889,
+      "grad_norm": 51.27205276489258,
+      "learning_rate": 5e-05,
+      "loss": 1.3542,
+      "step": 5
+    },
+    {
+      "epoch": 0.16666666666666666,
+      "grad_norm": 46.98747253417969,
+      "learning_rate": 6e-05,
+      "loss": 1.3031,
+      "step": 6
+    },
+    {
+      "epoch": 0.19444444444444445,
+      "grad_norm": 163.22494506835938,
+      "learning_rate": 7e-05,
+      "loss": 1.2489,
+      "step": 7
+    },
+    {
+      "epoch": 0.2222222222222222,
+      "grad_norm": 45.806182861328125,
+      "learning_rate": 8e-05,
+      "loss": 1.0557,
+      "step": 8
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 27.931575775146484,
+      "learning_rate": 9e-05,
+      "loss": 0.8285,
+      "step": 9
+    },
+    {
+      "epoch": 0.2777777777777778,
+      "grad_norm": 55.34366226196289,
+      "learning_rate": 0.0001,
+      "loss": 0.8731,
+      "step": 10
+    },
+    {
+      "epoch": 0.2777777777777778,
+      "eval_loss": 0.8334420323371887,
+      "eval_runtime": 0.1041,
+      "eval_samples_per_second": 442.063,
+      "eval_steps_per_second": 9.61,
+      "step": 10
+    },
+    {
+      "epoch": 0.3055555555555556,
+      "grad_norm": 27.75310516357422,
+      "learning_rate": 9.971428571428571e-05,
+      "loss": 0.7469,
+      "step": 11
+    },
+    {
+      "epoch": 0.3333333333333333,
+      "grad_norm": 24.505470275878906,
+      "learning_rate": 9.942857142857144e-05,
+      "loss": 0.6513,
+      "step": 12
+    },
+    {
+      "epoch": 0.3611111111111111,
+      "grad_norm": 8.471323013305664,
+      "learning_rate": 9.914285714285715e-05,
+      "loss": 0.5855,
+      "step": 13
+    },
+    {
+      "epoch": 0.3888888888888889,
+      "grad_norm": 12.284788131713867,
+      "learning_rate": 9.885714285714286e-05,
+      "loss": 0.5901,
+      "step": 14
+    },
+    {
+      "epoch": 0.4166666666666667,
+      "grad_norm": 6.142014026641846,
+      "learning_rate": 9.857142857142858e-05,
+      "loss": 0.558,
+      "step": 15
+    },
+    {
+      "epoch": 0.4444444444444444,
+      "grad_norm": 13.09057903289795,
+      "learning_rate": 9.828571428571429e-05,
+      "loss": 0.5863,
+      "step": 16
+    },
+    {
+      "epoch": 0.4722222222222222,
+      "grad_norm": 8.136122703552246,
+      "learning_rate": 9.8e-05,
+      "loss": 0.5416,
+      "step": 17
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 7.257413387298584,
+      "learning_rate": 9.771428571428572e-05,
+      "loss": 0.5398,
+      "step": 18
+    },
+    {
+      "epoch": 0.5277777777777778,
+      "grad_norm": 2.9581573009490967,
+      "learning_rate": 9.742857142857143e-05,
+      "loss": 0.4912,
+      "step": 19
+    },
+    {
+      "epoch": 0.5555555555555556,
+      "grad_norm": 3.979379892349243,
+      "learning_rate": 9.714285714285715e-05,
+      "loss": 0.4961,
+      "step": 20
+    },
+    {
+      "epoch": 0.5555555555555556,
+      "eval_loss": 0.4880354404449463,
+      "eval_runtime": 0.1036,
+      "eval_samples_per_second": 443.898,
+      "eval_steps_per_second": 9.65,
+      "step": 20
+    },
+    {
+      "epoch": 0.5833333333333334,
+      "grad_norm": 2.800405502319336,
+      "learning_rate": 9.685714285714286e-05,
+      "loss": 0.4898,
+      "step": 21
+    },
+    {
+      "epoch": 0.6111111111111112,
+      "grad_norm": 2.456360340118408,
+      "learning_rate": 9.657142857142858e-05,
+      "loss": 0.4617,
+      "step": 22
+    },
+    {
+      "epoch": 0.6388888888888888,
+      "grad_norm": 3.3507285118103027,
+      "learning_rate": 9.628571428571429e-05,
+      "loss": 0.4847,
+      "step": 23
+    },
+    {
+      "epoch": 0.6666666666666666,
+      "grad_norm": 2.5141634941101074,
+      "learning_rate": 9.6e-05,
+      "loss": 0.4609,
+      "step": 24
+    },
+    {
+      "epoch": 0.6944444444444444,
+      "grad_norm": 2.4233858585357666,
+      "learning_rate": 9.571428571428573e-05,
+      "loss": 0.4369,
+      "step": 25
+    },
+    {
+      "epoch": 0.7222222222222222,
+      "grad_norm": 2.399322986602783,
+      "learning_rate": 9.542857142857143e-05,
+      "loss": 0.4452,
+      "step": 26
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 1.9068751335144043,
+      "learning_rate": 9.514285714285714e-05,
+      "loss": 0.4326,
+      "step": 27
+    },
+    {
+      "epoch": 0.7777777777777778,
+      "grad_norm": 1.8973027467727661,
+      "learning_rate": 9.485714285714287e-05,
+      "loss": 0.4309,
+      "step": 28
+    },
+    {
+      "epoch": 0.8055555555555556,
+      "grad_norm": 1.655167579650879,
+      "learning_rate": 9.457142857142858e-05,
+      "loss": 0.4216,
+      "step": 29
+    },
+    {
+      "epoch": 0.8333333333333334,
+      "grad_norm": 1.3332226276397705,
+      "learning_rate": 9.428571428571429e-05,
+      "loss": 0.4191,
+      "step": 30
+    },
+    {
+      "epoch": 0.8333333333333334,
+      "eval_loss": 0.43893250823020935,
+      "eval_runtime": 0.104,
+      "eval_samples_per_second": 442.414,
+      "eval_steps_per_second": 9.618,
+      "step": 30
+    },
+    {
+      "epoch": 0.8611111111111112,
+      "grad_norm": 1.7973644733428955,
+      "learning_rate": 9.4e-05,
+      "loss": 0.4206,
+      "step": 31
+    },
+    {
+      "epoch": 0.8888888888888888,
+      "grad_norm": 1.4797940254211426,
+      "learning_rate": 9.371428571428572e-05,
+      "loss": 0.424,
+      "step": 32
+    },
+    {
+      "epoch": 0.9166666666666666,
+      "grad_norm": 1.4304866790771484,
+      "learning_rate": 9.342857142857143e-05,
+      "loss": 0.4135,
+      "step": 33
+    },
+    {
+      "epoch": 0.9444444444444444,
+      "grad_norm": 1.3090794086456299,
+      "learning_rate": 9.314285714285715e-05,
+      "loss": 0.4225,
+      "step": 34
+    },
+    {
+      "epoch": 0.9722222222222222,
+      "grad_norm": 1.271675944328308,
+      "learning_rate": 9.285714285714286e-05,
+      "loss": 0.4083,
+      "step": 35
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.5309550762176514,
+      "learning_rate": 9.257142857142858e-05,
+      "loss": 0.4146,
+      "step": 36
+    },
+    {
+      "epoch": 1.0277777777777777,
+      "grad_norm": 1.4727030992507935,
+      "learning_rate": 9.228571428571429e-05,
+      "loss": 0.3898,
+      "step": 37
+    },
+    {
+      "epoch": 1.0555555555555556,
+      "grad_norm": 1.752487301826477,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.4118,
+      "step": 38
+    },
+    {
+      "epoch": 1.0833333333333333,
+      "grad_norm": 1.634020447731018,
+      "learning_rate": 9.171428571428572e-05,
+      "loss": 0.405,
+      "step": 39
+    },
+    {
+      "epoch": 1.1111111111111112,
+      "grad_norm": 1.0251728296279907,
+      "learning_rate": 9.142857142857143e-05,
+      "loss": 0.3834,
+      "step": 40
+    },
+    {
+      "epoch": 1.1111111111111112,
+      "eval_loss": 0.41799336671829224,
+      "eval_runtime": 0.104,
+      "eval_samples_per_second": 442.159,
+      "eval_steps_per_second": 9.612,
+      "step": 40
+    },
+    {
+      "epoch": 1.1388888888888888,
+      "grad_norm": 1.767491340637207,
+      "learning_rate": 9.114285714285716e-05,
+      "loss": 0.3938,
+      "step": 41
+    },
+    {
+      "epoch": 1.1666666666666667,
+      "grad_norm": 1.4250415563583374,
+      "learning_rate": 9.085714285714286e-05,
+      "loss": 0.3988,
+      "step": 42
+    },
+    {
+      "epoch": 1.1944444444444444,
+      "grad_norm": 2.269926071166992,
+      "learning_rate": 9.057142857142857e-05,
+      "loss": 0.4027,
+      "step": 43
+    },
+    {
+      "epoch": 1.2222222222222223,
+      "grad_norm": 1.5418059825897217,
+      "learning_rate": 9.028571428571428e-05,
+      "loss": 0.4105,
+      "step": 44
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 1.3003933429718018,
+      "learning_rate": 9e-05,
+      "loss": 0.3888,
+      "step": 45
+    },
+    {
+      "epoch": 1.2777777777777777,
+      "grad_norm": 1.33571195602417,
+      "learning_rate": 8.971428571428571e-05,
+      "loss": 0.4034,
+      "step": 46
+    },
+    {
+      "epoch": 1.3055555555555556,
+      "grad_norm": 1.3438071012496948,
+      "learning_rate": 8.942857142857142e-05,
+      "loss": 0.3938,
+      "step": 47
+    },
+    {
+      "epoch": 1.3333333333333333,
+      "grad_norm": 1.3643842935562134,
+      "learning_rate": 8.914285714285715e-05,
+      "loss": 0.3917,
+      "step": 48
+    },
+    {
+      "epoch": 1.3611111111111112,
+      "grad_norm": 1.1976509094238281,
+      "learning_rate": 8.885714285714286e-05,
+      "loss": 0.3786,
+      "step": 49
+    },
+    {
+      "epoch": 1.3888888888888888,
+      "grad_norm": 1.463449239730835,
+      "learning_rate": 8.857142857142857e-05,
+      "loss": 0.3858,
+      "step": 50
+    },
+    {
+      "epoch": 1.3888888888888888,
+      "eval_loss": 0.39549750089645386,
+      "eval_runtime": 0.1046,
+      "eval_samples_per_second": 439.872,
+      "eval_steps_per_second": 9.562,
+      "step": 50
+    },
+    {
+      "epoch": 1.4166666666666667,
+      "grad_norm": 1.3630038499832153,
+      "learning_rate": 8.828571428571429e-05,
+      "loss": 0.3784,
+      "step": 51
+    },
+    {
+      "epoch": 1.4444444444444444,
+      "grad_norm": 1.2265862226486206,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.3696,
+      "step": 52
+    },
+    {
+      "epoch": 1.4722222222222223,
+      "grad_norm": 1.042543649673462,
+      "learning_rate": 8.771428571428572e-05,
+      "loss": 0.3762,
+      "step": 53
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 1.2349590063095093,
+      "learning_rate": 8.742857142857144e-05,
+      "loss": 0.3813,
+      "step": 54
+    },
+    {
+      "epoch": 1.5277777777777777,
+      "grad_norm": 1.9418014287948608,
+      "learning_rate": 8.714285714285715e-05,
+      "loss": 0.3983,
+      "step": 55
+    },
+    {
+      "epoch": 1.5555555555555556,
+      "grad_norm": 0.9780258536338806,
+      "learning_rate": 8.685714285714286e-05,
+      "loss": 0.3726,
+      "step": 56
+    },
+    {
+      "epoch": 1.5833333333333335,
+      "grad_norm": 1.3302329778671265,
+      "learning_rate": 8.657142857142858e-05,
+      "loss": 0.3801,
+      "step": 57
+    },
+    {
+      "epoch": 1.6111111111111112,
+      "grad_norm": 1.0229287147521973,
+      "learning_rate": 8.62857142857143e-05,
+      "loss": 0.3835,
+      "step": 58
+    },
+    {
+      "epoch": 1.6388888888888888,
+      "grad_norm": 1.0397497415542603,
+      "learning_rate": 8.6e-05,
+      "loss": 0.3698,
+      "step": 59
+    },
+    {
+      "epoch": 1.6666666666666665,
+      "grad_norm": 0.9683515429496765,
+      "learning_rate": 8.571428571428571e-05,
+      "loss": 0.3752,
+      "step": 60
+    },
+    {
+      "epoch": 1.6666666666666665,
+      "eval_loss": 0.3742421269416809,
+      "eval_runtime": 0.1043,
+      "eval_samples_per_second": 441.127,
+      "eval_steps_per_second": 9.59,
+      "step": 60
+    },
+    {
+      "epoch": 1.6944444444444444,
+      "grad_norm": 0.7956013679504395,
+      "learning_rate": 8.542857142857144e-05,
+      "loss": 0.3621,
+      "step": 61
+    },
+    {
+      "epoch": 1.7222222222222223,
+      "grad_norm": 1.0049488544464111,
+      "learning_rate": 8.514285714285714e-05,
+      "loss": 0.3685,
+      "step": 62
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 0.9522997140884399,
+      "learning_rate": 8.485714285714285e-05,
+      "loss": 0.3692,
+      "step": 63
+    },
+    {
+      "epoch": 1.7777777777777777,
+      "grad_norm": 1.6391457319259644,
+      "learning_rate": 8.457142857142858e-05,
+      "loss": 0.3677,
+      "step": 64
+    },
+    {
+      "epoch": 1.8055555555555556,
+      "grad_norm": 0.7660079002380371,
+      "learning_rate": 8.428571428571429e-05,
+      "loss": 0.3587,
+      "step": 65
+    },
+    {
+      "epoch": 1.8333333333333335,
+      "grad_norm": 0.8130621314048767,
+      "learning_rate": 8.4e-05,
+      "loss": 0.3561,
+      "step": 66
+    },
+    {
+      "epoch": 1.8611111111111112,
+      "grad_norm": 2.256351947784424,
+      "learning_rate": 8.371428571428572e-05,
+      "loss": 0.3703,
+      "step": 67
+    },
+    {
+      "epoch": 1.8888888888888888,
+      "grad_norm": 2.5743725299835205,
+      "learning_rate": 8.342857142857143e-05,
+      "loss": 0.3733,
+      "step": 68
+    },
+    {
+      "epoch": 1.9166666666666665,
+      "grad_norm": 1.5146273374557495,
+      "learning_rate": 8.314285714285715e-05,
+      "loss": 0.3691,
+      "step": 69
+    },
+    {
+      "epoch": 1.9444444444444444,
+      "grad_norm": 2.370126962661743,
+      "learning_rate": 8.285714285714287e-05,
+      "loss": 0.3685,
+      "step": 70
+    },
+    {
+      "epoch": 1.9444444444444444,
+      "eval_loss": 0.37778788805007935,
+      "eval_runtime": 0.1038,
+      "eval_samples_per_second": 443.059,
+      "eval_steps_per_second": 9.632,
+      "step": 70
+    },
+    {
+      "epoch": 1.9722222222222223,
+      "grad_norm": 2.041372537612915,
+      "learning_rate": 8.257142857142858e-05,
+      "loss": 0.36,
+      "step": 71
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.340862512588501,
+      "learning_rate": 8.228571428571429e-05,
+      "loss": 0.3819,
+      "step": 72
+    },
+    {
+      "epoch": 2.0277777777777777,
+      "grad_norm": 1.614141821861267,
+      "learning_rate": 8.2e-05,
+      "loss": 0.3614,
+      "step": 73
+    },
+    {
+      "epoch": 2.0555555555555554,
+      "grad_norm": 1.172598958015442,
+      "learning_rate": 8.171428571428572e-05,
+      "loss": 0.3558,
+      "step": 74
+    },
+    {
+      "epoch": 2.0833333333333335,
+      "grad_norm": 1.5925283432006836,
+      "learning_rate": 8.142857142857143e-05,
+      "loss": 0.3646,
+      "step": 75
+    },
+    {
+      "epoch": 2.111111111111111,
+      "grad_norm": 1.7977277040481567,
+      "learning_rate": 8.114285714285714e-05,
+      "loss": 0.3647,
+      "step": 76
+    },
+    {
+      "epoch": 2.138888888888889,
+      "grad_norm": 1.2800759077072144,
+      "learning_rate": 8.085714285714287e-05,
+      "loss": 0.3599,
+      "step": 77
+    },
+    {
+      "epoch": 2.1666666666666665,
+      "grad_norm": 1.8917945623397827,
+      "learning_rate": 8.057142857142857e-05,
+      "loss": 0.3661,
+      "step": 78
+    },
+    {
+      "epoch": 2.1944444444444446,
+      "grad_norm": 2.058119058609009,
+      "learning_rate": 8.028571428571428e-05,
+      "loss": 0.3715,
+      "step": 79
+    },
+    {
+      "epoch": 2.2222222222222223,
+      "grad_norm": 0.8664758801460266,
+      "learning_rate": 8e-05,
+      "loss": 0.3531,
+      "step": 80
+    },
+    {
+      "epoch": 2.2222222222222223,
+      "eval_loss": 0.3683662414550781,
+      "eval_runtime": 0.1045,
+      "eval_samples_per_second": 440.203,
+      "eval_steps_per_second": 9.57,
+      "step": 80
+    },
+    {
+      "epoch": 2.25,
+      "grad_norm": 1.1171331405639648,
+      "learning_rate": 7.971428571428572e-05,
+      "loss": 0.3591,
+      "step": 81
+    },
+    {
+      "epoch": 2.2777777777777777,
+      "grad_norm": 1.2088688611984253,
+      "learning_rate": 7.942857142857143e-05,
+      "loss": 0.3554,
+      "step": 82
+    },
+    {
+      "epoch": 2.3055555555555554,
+      "grad_norm": 0.7355909943580627,
+      "learning_rate": 7.914285714285715e-05,
+      "loss": 0.3488,
+      "step": 83
+    },
+    {
+      "epoch": 2.3333333333333335,
+      "grad_norm": 1.5655473470687866,
+      "learning_rate": 7.885714285714286e-05,
+      "loss": 0.3516,
+      "step": 84
+    },
+    {
+      "epoch": 2.361111111111111,
+      "grad_norm": 1.1028951406478882,
+      "learning_rate": 7.857142857142858e-05,
+      "loss": 0.3531,
+      "step": 85
+    },
+    {
+      "epoch": 2.388888888888889,
+      "grad_norm": 1.0214101076126099,
+      "learning_rate": 7.828571428571429e-05,
+      "loss": 0.3566,
+      "step": 86
+    },
+    {
+      "epoch": 2.4166666666666665,
+      "grad_norm": 1.0004141330718994,
+      "learning_rate": 7.800000000000001e-05,
+      "loss": 0.3525,
+      "step": 87
+    },
+    {
+      "epoch": 2.4444444444444446,
+      "grad_norm": 1.1501681804656982,
+      "learning_rate": 7.771428571428572e-05,
+      "loss": 0.364,
+      "step": 88
+    },
+    {
+      "epoch": 2.4722222222222223,
+      "grad_norm": 0.8693153858184814,
+      "learning_rate": 7.742857142857143e-05,
+      "loss": 0.3452,
+      "step": 89
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 1.0027663707733154,
+      "learning_rate": 7.714285714285715e-05,
+      "loss": 0.348,
+      "step": 90
+    },
+    {
+      "epoch": 2.5,
+      "eval_loss": 0.36503931879997253,
+      "eval_runtime": 0.1049,
+      "eval_samples_per_second": 438.538,
+      "eval_steps_per_second": 9.533,
+      "step": 90
+    },
+    {
+      "epoch": 2.5277777777777777,
+      "grad_norm": 1.3100874423980713,
+      "learning_rate": 7.685714285714286e-05,
+      "loss": 0.3605,
+      "step": 91
+    },
+    {
+      "epoch": 2.5555555555555554,
+      "grad_norm": 1.0220280885696411,
+      "learning_rate": 7.657142857142857e-05,
+      "loss": 0.3473,
+      "step": 92
+    },
+    {
+      "epoch": 2.5833333333333335,
+      "grad_norm": 1.4616146087646484,
+      "learning_rate": 7.62857142857143e-05,
+      "loss": 0.3631,
+      "step": 93
+    },
+    {
+      "epoch": 2.611111111111111,
+      "grad_norm": 1.0069721937179565,
+      "learning_rate": 7.6e-05,
+      "loss": 0.3426,
+      "step": 94
+    },
+    {
+      "epoch": 2.638888888888889,
+      "grad_norm": 0.8820877075195312,
+      "learning_rate": 7.571428571428571e-05,
+      "loss": 0.3367,
+      "step": 95
+    },
+    {
+      "epoch": 2.6666666666666665,
+      "grad_norm": 1.2850284576416016,
+      "learning_rate": 7.542857142857144e-05,
+      "loss": 0.3553,
+      "step": 96
+    },
+    {
+      "epoch": 2.6944444444444446,
+      "grad_norm": 0.8124020099639893,
+      "learning_rate": 7.514285714285715e-05,
+      "loss": 0.3436,
+      "step": 97
+    },
+    {
+      "epoch": 2.7222222222222223,
+      "grad_norm": 0.9645909667015076,
+      "learning_rate": 7.485714285714285e-05,
+      "loss": 0.3504,
+      "step": 98
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 1.253502368927002,
+      "learning_rate": 7.457142857142856e-05,
+      "loss": 0.3586,
+      "step": 99
+    },
+    {
+      "epoch": 2.7777777777777777,
+      "grad_norm": 0.7083765864372253,
+      "learning_rate": 7.428571428571429e-05,
+      "loss": 0.3403,
+      "step": 100
+    },
+    {
+      "epoch": 2.7777777777777777,
+      "eval_loss": 0.3600272834300995,
+      "eval_runtime": 0.1045,
+      "eval_samples_per_second": 440.303,
+      "eval_steps_per_second": 9.572,
+      "step": 100
+    },
+    {
+      "epoch": 2.8055555555555554,
+      "grad_norm": 0.8545306921005249,
+      "learning_rate": 7.4e-05,
+      "loss": 0.3489,
+      "step": 101
+    },
+    {
+      "epoch": 2.8333333333333335,
+      "grad_norm": 1.097291350364685,
+      "learning_rate": 7.371428571428572e-05,
+      "loss": 0.3383,
+      "step": 102
+    },
+    {
+      "epoch": 2.861111111111111,
+      "grad_norm": 0.7652900218963623,
+      "learning_rate": 7.342857142857144e-05,
+      "loss": 0.3395,
+      "step": 103
+    },
+    {
+      "epoch": 2.888888888888889,
+      "grad_norm": 0.9357858896255493,
+      "learning_rate": 7.314285714285715e-05,
+      "loss": 0.3513,
+      "step": 104
+    },
+    {
+      "epoch": 2.9166666666666665,
+      "grad_norm": 1.1487563848495483,
+      "learning_rate": 7.285714285714286e-05,
+      "loss": 0.3434,
+      "step": 105
+    },
+    {
+      "epoch": 2.9444444444444446,
+      "grad_norm": 1.1131272315979004,
+      "learning_rate": 7.257142857142858e-05,
+      "loss": 0.3464,
+      "step": 106
+    },
+    {
+      "epoch": 2.9722222222222223,
+      "grad_norm": 1.0483183860778809,
+      "learning_rate": 7.228571428571429e-05,
+      "loss": 0.3484,
+      "step": 107
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 1.248903751373291,
+      "learning_rate": 7.2e-05,
+      "loss": 0.3533,
+      "step": 108
+    },
+    {
+      "epoch": 3.0277777777777777,
+      "grad_norm": 0.8581050634384155,
+      "learning_rate": 7.171428571428572e-05,
+      "loss": 0.3389,
+      "step": 109
+    },
+    {
+      "epoch": 3.0555555555555554,
+      "grad_norm": 1.1233102083206177,
+      "learning_rate": 7.142857142857143e-05,
+      "loss": 0.347,
+      "step": 110
+    },
+    {
+      "epoch": 3.0555555555555554,
+      "eval_loss": 0.3557908535003662,
+      "eval_runtime": 0.1045,
+      "eval_samples_per_second": 440.002,
+      "eval_steps_per_second": 9.565,
+      "step": 110
+    },
+    {
+      "epoch": 3.0833333333333335,
+      "grad_norm": 1.056329369544983,
+      "learning_rate": 7.114285714285714e-05,
+      "loss": 0.3425,
+      "step": 111
+    },
+    {
+      "epoch": 3.111111111111111,
+      "grad_norm": 0.6349842548370361,
+      "learning_rate": 7.085714285714285e-05,
+      "loss": 0.3398,
+      "step": 112
+    },
+    {
+      "epoch": 3.138888888888889,
+      "grad_norm": 1.2684061527252197,
+      "learning_rate": 7.057142857142858e-05,
+      "loss": 0.3413,
+      "step": 113
+    },
+    {
+      "epoch": 3.1666666666666665,
+      "grad_norm": 0.9766147136688232,
+      "learning_rate": 7.028571428571428e-05,
+      "loss": 0.3504,
+      "step": 114
+    },
+    {
+      "epoch": 3.1944444444444446,
+      "grad_norm": 1.1366608142852783,
+      "learning_rate": 7e-05,
+      "loss": 0.3461,
+      "step": 115
+    },
+    {
+      "epoch": 3.2222222222222223,
+      "grad_norm": 1.1576147079467773,
+      "learning_rate": 6.971428571428572e-05,
+      "loss": 0.3441,
+      "step": 116
+    },
+    {
+      "epoch": 3.25,
+      "grad_norm": 0.896751880645752,
+      "learning_rate": 6.942857142857143e-05,
+      "loss": 0.3382,
+      "step": 117
+    },
+    {
+      "epoch": 3.2777777777777777,
+      "grad_norm": 0.7345131039619446,
+      "learning_rate": 6.914285714285715e-05,
+      "loss": 0.3406,
+      "step": 118
+    },
+    {
+      "epoch": 3.3055555555555554,
+      "grad_norm": 0.8541216254234314,
+      "learning_rate": 6.885714285714286e-05,
+      "loss": 0.34,
+      "step": 119
+    },
+    {
+      "epoch": 3.3333333333333335,
+      "grad_norm": 0.6585732102394104,
+      "learning_rate": 6.857142857142858e-05,
+      "loss": 0.3389,
+      "step": 120
+    },
+    {
+      "epoch": 3.3333333333333335,
+      "eval_loss": 0.35226884484291077,
+      "eval_runtime": 0.1048,
+      "eval_samples_per_second": 438.732,
+      "eval_steps_per_second": 9.538,
+      "step": 120
+    },
+    {
+      "epoch": 3.361111111111111,
+      "grad_norm": 1.3044718503952026,
+      "learning_rate": 6.828571428571429e-05,
+      "loss": 0.3462,
+      "step": 121
+    },
+    {
+      "epoch": 3.388888888888889,
+      "grad_norm": 0.6839655041694641,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.3374,
+      "step": 122
+    },
+    {
+      "epoch": 3.4166666666666665,
+      "grad_norm": 1.3587740659713745,
+      "learning_rate": 6.771428571428572e-05,
+      "loss": 0.3461,
+      "step": 123
+    },
+    {
+      "epoch": 3.4444444444444446,
+      "grad_norm": 1.2411038875579834,
+      "learning_rate": 6.742857142857143e-05,
+      "loss": 0.3409,
+      "step": 124
+    },
+    {
+      "epoch": 3.4722222222222223,
+      "grad_norm": 1.1246317625045776,
+      "learning_rate": 6.714285714285714e-05,
+      "loss": 0.3337,
+      "step": 125
+    },
+    {
+      "epoch": 3.5,
+      "grad_norm": 0.8248189687728882,
+      "learning_rate": 6.685714285714286e-05,
+      "loss": 0.3363,
+      "step": 126
+    },
+    {
+      "epoch": 3.5277777777777777,
+      "grad_norm": 0.9192841649055481,
+      "learning_rate": 6.657142857142857e-05,
+      "loss": 0.3421,
+      "step": 127
+    },
+    {
+      "epoch": 3.5555555555555554,
+      "grad_norm": 1.1318869590759277,
+      "learning_rate": 6.628571428571428e-05,
+      "loss": 0.3418,
+      "step": 128
+    },
+    {
+      "epoch": 3.5833333333333335,
+      "grad_norm": 1.0143375396728516,
+      "learning_rate": 6.6e-05,
+      "loss": 0.342,
+      "step": 129
+    },
+    {
+      "epoch": 3.611111111111111,
+      "grad_norm": 0.726323127746582,
+      "learning_rate": 6.571428571428571e-05,
+      "loss": 0.3321,
+      "step": 130
+    },
+    {
+      "epoch": 3.611111111111111,
+      "eval_loss": 0.3512151837348938,
+      "eval_runtime": 0.1051,
+      "eval_samples_per_second": 437.793,
+      "eval_steps_per_second": 9.517,
+      "step": 130
+    },
+    {
+      "epoch": 3.638888888888889,
+      "grad_norm": 0.9356390833854675,
+      "learning_rate": 6.542857142857142e-05,
+      "loss": 0.3379,
+      "step": 131
+    },
+    {
+      "epoch": 3.6666666666666665,
+      "grad_norm": 1.051357388496399,
+      "learning_rate": 6.514285714285715e-05,
+      "loss": 0.334,
+      "step": 132
+    },
+    {
+      "epoch": 3.6944444444444446,
+      "grad_norm": 0.7245939373970032,
+      "learning_rate": 6.485714285714286e-05,
+      "loss": 0.3433,
+      "step": 133
+    },
+    {
+      "epoch": 3.7222222222222223,
+      "grad_norm": 1.5335993766784668,
+      "learning_rate": 6.457142857142856e-05,
+      "loss": 0.333,
+      "step": 134
+    },
+    {
+      "epoch": 3.75,
+      "grad_norm": 0.6289878487586975,
+      "learning_rate": 6.428571428571429e-05,
+      "loss": 0.3345,
+      "step": 135
+    },
+    {
+      "epoch": 3.7777777777777777,
+      "grad_norm": 0.7878074645996094,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.3416,
+      "step": 136
+    },
+    {
+      "epoch": 3.8055555555555554,
+      "grad_norm": 0.8280585408210754,
+      "learning_rate": 6.371428571428572e-05,
+      "loss": 0.3436,
+      "step": 137
+    },
+    {
+      "epoch": 3.8333333333333335,
+      "grad_norm": 0.7266003489494324,
+      "learning_rate": 6.342857142857143e-05,
+      "loss": 0.3365,
+      "step": 138
+    },
+    {
+      "epoch": 3.861111111111111,
+      "grad_norm": 1.3992286920547485,
+      "learning_rate": 6.314285714285715e-05,
+      "loss": 0.3426,
+      "step": 139
+    },
+    {
+      "epoch": 3.888888888888889,
+      "grad_norm": 0.7563316822052002,
+      "learning_rate": 6.285714285714286e-05,
+      "loss": 0.3332,
+      "step": 140
+    },
+    {
+      "epoch": 3.888888888888889,
+      "eval_loss": 0.3488697409629822,
+      "eval_runtime": 0.105,
+      "eval_samples_per_second": 437.99,
+      "eval_steps_per_second": 9.522,
+      "step": 140
+    },
+    {
+      "epoch": 3.9166666666666665,
+      "grad_norm": 0.896052896976471,
+      "learning_rate": 6.257142857142857e-05,
+      "loss": 0.3313,
+      "step": 141
+    },
+    {
+      "epoch": 3.9444444444444446,
+      "grad_norm": 1.2708873748779297,
+      "learning_rate": 6.22857142857143e-05,
+      "loss": 0.3365,
+      "step": 142
+    },
+    {
+      "epoch": 3.9722222222222223,
+      "grad_norm": 1.7695562839508057,
+      "learning_rate": 6.2e-05,
+      "loss": 0.3357,
+      "step": 143
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.4549791812896729,
+      "learning_rate": 6.171428571428571e-05,
+      "loss": 0.3424,
+      "step": 144
+    },
+    {
+      "epoch": 4.027777777777778,
+      "grad_norm": 0.9521704912185669,
+      "learning_rate": 6.142857142857143e-05,
+      "loss": 0.3375,
+      "step": 145
+    },
+    {
+      "epoch": 4.055555555555555,
+      "grad_norm": 1.7584255933761597,
+      "learning_rate": 6.114285714285714e-05,
+      "loss": 0.3361,
+      "step": 146
+    },
+    {
+      "epoch": 4.083333333333333,
+      "grad_norm": 0.7483154535293579,
+      "learning_rate": 6.085714285714286e-05,
+      "loss": 0.3346,
+      "step": 147
+    },
+    {
+      "epoch": 4.111111111111111,
+      "grad_norm": 1.067897081375122,
+      "learning_rate": 6.0571428571428576e-05,
+      "loss": 0.3274,
+      "step": 148
+    },
+    {
+      "epoch": 4.138888888888889,
+      "grad_norm": 0.9337431788444519,
+      "learning_rate": 6.028571428571429e-05,
+      "loss": 0.3257,
+      "step": 149
+    },
+    {
+      "epoch": 4.166666666666667,
+      "grad_norm": 0.9089264273643494,
+      "learning_rate": 6e-05,
+      "loss": 0.3356,
+      "step": 150
+    },
+    {
+      "epoch": 4.166666666666667,
+      "eval_loss": 0.35016533732414246,
+      "eval_runtime": 0.1051,
+      "eval_samples_per_second": 437.834,
+      "eval_steps_per_second": 9.518,
+      "step": 150
+    },
+    {
+      "epoch": 4.194444444444445,
+      "grad_norm": 1.1468346118927002,
+      "learning_rate": 5.9714285714285724e-05,
+      "loss": 0.3379,
+      "step": 151
+    },
+    {
+      "epoch": 4.222222222222222,
+      "grad_norm": 1.1817346811294556,
+      "learning_rate": 5.9428571428571434e-05,
+      "loss": 0.3323,
+      "step": 152
+    },
+    {
+      "epoch": 4.25,
+      "grad_norm": 1.4314862489700317,
+      "learning_rate": 5.914285714285714e-05,
+      "loss": 0.336,
+      "step": 153
+    },
+    {
+      "epoch": 4.277777777777778,
+      "grad_norm": 1.1927298307418823,
+      "learning_rate": 5.885714285714285e-05,
+      "loss": 0.3362,
+      "step": 154
+    },
+    {
+      "epoch": 4.305555555555555,
+      "grad_norm": 0.8057120442390442,
+      "learning_rate": 5.8571428571428575e-05,
+      "loss": 0.3304,
+      "step": 155
+    },
+    {
+      "epoch": 4.333333333333333,
+      "grad_norm": 1.638046145439148,
+      "learning_rate": 5.828571428571429e-05,
+      "loss": 0.3403,
+      "step": 156
+    },
+    {
+      "epoch": 4.361111111111111,
+      "grad_norm": 0.874083399772644,
+      "learning_rate": 5.8e-05,
+      "loss": 0.332,
+      "step": 157
+    },
+    {
+      "epoch": 4.388888888888889,
+      "grad_norm": 0.7660898566246033,
+      "learning_rate": 5.771428571428572e-05,
+      "loss": 0.3204,
+      "step": 158
+    },
+    {
+      "epoch": 4.416666666666667,
+      "grad_norm": 1.310951828956604,
+      "learning_rate": 5.742857142857143e-05,
+      "loss": 0.3413,
+      "step": 159
+    },
+    {
+      "epoch": 4.444444444444445,
+      "grad_norm": 0.7157622575759888,
+      "learning_rate": 5.714285714285714e-05,
+      "loss": 0.3356,
+      "step": 160
+    },
+    {
+      "epoch": 4.444444444444445,
+      "eval_loss": 0.35388267040252686,
+      "eval_runtime": 0.1051,
+      "eval_samples_per_second": 437.501,
+      "eval_steps_per_second": 9.511,
+      "step": 160
+    },
+    {
+      "epoch": 4.472222222222222,
+      "grad_norm": 0.996565043926239,
+      "learning_rate": 5.6857142857142865e-05,
+      "loss": 0.3373,
+      "step": 161
+    },
+    {
+      "epoch": 4.5,
+      "grad_norm": 0.6834873557090759,
+      "learning_rate": 5.6571428571428574e-05,
+      "loss": 0.3314,
+      "step": 162
+    },
+    {
+      "epoch": 4.527777777777778,
+      "grad_norm": 1.1225472688674927,
+      "learning_rate": 5.628571428571428e-05,
+      "loss": 0.3298,
+      "step": 163
+    },
+    {
+      "epoch": 4.555555555555555,
+      "grad_norm": 1.5352915525436401,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.3355,
+      "step": 164
+    },
+    {
+      "epoch": 4.583333333333333,
+      "grad_norm": 0.8244190216064453,
+      "learning_rate": 5.571428571428572e-05,
+      "loss": 0.3359,
+      "step": 165
+    },
+    {
+      "epoch": 4.611111111111111,
+      "grad_norm": 1.3828486204147339,
+      "learning_rate": 5.542857142857143e-05,
+      "loss": 0.3344,
+      "step": 166
+    },
+    {
+      "epoch": 4.638888888888889,
+      "grad_norm": 0.8235058784484863,
+      "learning_rate": 5.514285714285714e-05,
+      "loss": 0.3278,
+      "step": 167
+    },
+    {
+      "epoch": 4.666666666666667,
+      "grad_norm": 0.7643269300460815,
+      "learning_rate": 5.485714285714286e-05,
+      "loss": 0.3268,
+      "step": 168
+    },
+    {
+      "epoch": 4.694444444444445,
+      "grad_norm": 0.8052845001220703,
+      "learning_rate": 5.457142857142857e-05,
+      "loss": 0.3318,
+      "step": 169
+    },
+    {
+      "epoch": 4.722222222222222,
+      "grad_norm": 1.0252516269683838,
+      "learning_rate": 5.428571428571428e-05,
+      "loss": 0.3372,
+      "step": 170
+    },
+    {
+      "epoch": 4.722222222222222,
+      "eval_loss": 0.3490443825721741,
+      "eval_runtime": 0.1055,
+      "eval_samples_per_second": 436.194,
+      "eval_steps_per_second": 9.482,
+      "step": 170
+    },
+    {
+      "epoch": 4.75,
+      "grad_norm": 1.2792561054229736,
+      "learning_rate": 5.4000000000000005e-05,
+      "loss": 0.3373,
+      "step": 171
+    },
+    {
+      "epoch": 4.777777777777778,
+      "grad_norm": 1.143295407295227,
+      "learning_rate": 5.3714285714285714e-05,
+      "loss": 0.3346,
+      "step": 172
+    },
+    {
+      "epoch": 4.805555555555555,
+      "grad_norm": 1.3516252040863037,
+      "learning_rate": 5.342857142857143e-05,
+      "loss": 0.3354,
+      "step": 173
+    },
+    {
+      "epoch": 4.833333333333333,
+      "grad_norm": 1.4615265130996704,
+      "learning_rate": 5.314285714285715e-05,
+      "loss": 0.3379,
+      "step": 174
+    },
+    {
+      "epoch": 4.861111111111111,
+      "grad_norm": 0.9049354195594788,
+      "learning_rate": 5.285714285714286e-05,
+      "loss": 0.3291,
+      "step": 175
+    },
+    {
+      "epoch": 4.888888888888889,
+      "grad_norm": 1.4232890605926514,
+      "learning_rate": 5.257142857142857e-05,
+      "loss": 0.3309,
+      "step": 176
+    },
+    {
+      "epoch": 4.916666666666667,
+      "grad_norm": 0.7629730701446533,
+      "learning_rate": 5.2285714285714294e-05,
+      "loss": 0.3297,
+      "step": 177
+    },
+    {
+      "epoch": 4.944444444444445,
+      "grad_norm": 1.2539515495300293,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.3424,
+      "step": 178
+    },
+    {
+      "epoch": 4.972222222222222,
+      "grad_norm": 1.3451244831085205,
+      "learning_rate": 5.171428571428571e-05,
+      "loss": 0.3331,
+      "step": 179
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 1.4241085052490234,
+      "learning_rate": 5.142857142857143e-05,
+      "loss": 0.3315,
+      "step": 180
+    },
+    {
+      "epoch": 5.0,
+      "eval_loss": 0.35601839423179626,
+      "eval_runtime": 0.1052,
+      "eval_samples_per_second": 437.465,
+      "eval_steps_per_second": 9.51,
+      "step": 180
+    },
+    {
+      "epoch": 5.027777777777778,
+      "grad_norm": 1.3330923318862915,
+      "learning_rate": 5.1142857142857145e-05,
+      "loss": 0.3266,
+      "step": 181
+    },
+    {
+      "epoch": 5.055555555555555,
+      "grad_norm": 0.8518350124359131,
+      "learning_rate": 5.085714285714286e-05,
+      "loss": 0.3247,
+      "step": 182
+    },
+    {
+      "epoch": 5.083333333333333,
+      "grad_norm": 1.5773379802703857,
+      "learning_rate": 5.057142857142857e-05,
+      "loss": 0.3322,
+      "step": 183
+    },
+    {
+      "epoch": 5.111111111111111,
+      "grad_norm": 1.9808964729309082,
+      "learning_rate": 5.028571428571429e-05,
+      "loss": 0.3361,
+      "step": 184
+    },
+    {
+      "epoch": 5.138888888888889,
+      "grad_norm": 0.6624701023101807,
+      "learning_rate": 5e-05,
+      "loss": 0.3326,
+      "step": 185
+    },
+    {
+      "epoch": 5.166666666666667,
+      "grad_norm": 1.3144952058792114,
+      "learning_rate": 4.971428571428572e-05,
+      "loss": 0.3294,
+      "step": 186
+    },
+    {
+      "epoch": 5.194444444444445,
+      "grad_norm": 1.4999899864196777,
+      "learning_rate": 4.942857142857143e-05,
+      "loss": 0.3249,
+      "step": 187
+    },
+    {
+      "epoch": 5.222222222222222,
+      "grad_norm": 0.6328815817832947,
+      "learning_rate": 4.9142857142857144e-05,
+      "loss": 0.3295,
+      "step": 188
+    },
+    {
+      "epoch": 5.25,
+      "grad_norm": 1.7996755838394165,
+      "learning_rate": 4.885714285714286e-05,
+      "loss": 0.3325,
+      "step": 189
+    },
+    {
+      "epoch": 5.277777777777778,
+      "grad_norm": 1.4809287786483765,
+      "learning_rate": 4.8571428571428576e-05,
+      "loss": 0.3294,
+      "step": 190
+    },
+    {
+      "epoch": 5.277777777777778,
+      "eval_loss": 0.34694546461105347,
+      "eval_runtime": 0.1055,
+      "eval_samples_per_second": 435.923,
+      "eval_steps_per_second": 9.477,
+      "step": 190
+    },
+    {
+      "epoch": 5.305555555555555,
+      "grad_norm": 1.0571750402450562,
+      "learning_rate": 4.828571428571429e-05,
+      "loss": 0.3256,
+      "step": 191
+    },
+    {
+      "epoch": 5.333333333333333,
+      "grad_norm": 1.0487523078918457,
+      "learning_rate": 4.8e-05,
+      "loss": 0.3302,
+      "step": 192
+    },
+    {
+      "epoch": 5.361111111111111,
+      "grad_norm": 0.762173593044281,
+      "learning_rate": 4.771428571428572e-05,
+      "loss": 0.3261,
+      "step": 193
+    },
+    {
+      "epoch": 5.388888888888889,
+      "grad_norm": 0.8178277611732483,
+      "learning_rate": 4.742857142857143e-05,
+      "loss": 0.3267,
+      "step": 194
+    },
+    {
+      "epoch": 5.416666666666667,
+      "grad_norm": 1.104577898979187,
+      "learning_rate": 4.714285714285714e-05,
+      "loss": 0.326,
+      "step": 195
+    },
+    {
+      "epoch": 5.444444444444445,
+      "grad_norm": 0.7541172504425049,
+      "learning_rate": 4.685714285714286e-05,
+      "loss": 0.3308,
+      "step": 196
+    },
+    {
+      "epoch": 5.472222222222222,
+      "grad_norm": 1.1193814277648926,
+      "learning_rate": 4.6571428571428575e-05,
+      "loss": 0.3323,
+      "step": 197
+    },
+    {
+      "epoch": 5.5,
+      "grad_norm": 0.6032533049583435,
+      "learning_rate": 4.628571428571429e-05,
+      "loss": 0.3251,
+      "step": 198
+    },
+    {
+      "epoch": 5.527777777777778,
+      "grad_norm": 0.7051272988319397,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 0.3287,
+      "step": 199
+    },
+    {
+      "epoch": 5.555555555555555,
+      "grad_norm": 0.8189981579780579,
+      "learning_rate": 4.5714285714285716e-05,
+      "loss": 0.3274,
+      "step": 200
+    },
+    {
+      "epoch": 5.555555555555555,
+      "eval_loss": 0.34395119547843933,
+      "eval_runtime": 0.1046,
+      "eval_samples_per_second": 439.963,
+      "eval_steps_per_second": 9.564,
+      "step": 200
+    },
+    {
+      "epoch": 5.583333333333333,
+      "grad_norm": 0.7370961904525757,
+      "learning_rate": 4.542857142857143e-05,
+      "loss": 0.3243,
+      "step": 201
+    },
+    {
+      "epoch": 5.611111111111111,
+      "grad_norm": 0.9540398716926575,
+      "learning_rate": 4.514285714285714e-05,
+      "loss": 0.3323,
+      "step": 202
+    },
+    {
+      "epoch": 5.638888888888889,
+      "grad_norm": 0.6874322891235352,
+      "learning_rate": 4.485714285714286e-05,
+      "loss": 0.3282,
+      "step": 203
+    },
+    {
+      "epoch": 5.666666666666667,
+      "grad_norm": 0.8128199577331543,
+      "learning_rate": 4.4571428571428574e-05,
+      "loss": 0.3328,
+      "step": 204
+    },
+    {
+      "epoch": 5.694444444444445,
+      "grad_norm": 0.9183529615402222,
+      "learning_rate": 4.428571428571428e-05,
+      "loss": 0.3257,
+      "step": 205
+    },
+    {
+      "epoch": 5.722222222222222,
+      "grad_norm": 0.9220953583717346,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.3284,
+      "step": 206
+    },
+    {
+      "epoch": 5.75,
+      "grad_norm": 1.4376723766326904,
+      "learning_rate": 4.371428571428572e-05,
+      "loss": 0.3304,
+      "step": 207
+    },
+    {
+      "epoch": 5.777777777777778,
+      "grad_norm": 1.0230571031570435,
+      "learning_rate": 4.342857142857143e-05,
+      "loss": 0.33,
+      "step": 208
+    },
+    {
+      "epoch": 5.805555555555555,
+      "grad_norm": 1.4990969896316528,
+      "learning_rate": 4.314285714285715e-05,
+      "loss": 0.335,
+      "step": 209
+    },
+    {
+      "epoch": 5.833333333333333,
+      "grad_norm": 0.9291542172431946,
+      "learning_rate": 4.2857142857142856e-05,
+      "loss": 0.3375,
+      "step": 210
+    },
+    {
+      "epoch": 5.833333333333333,
+      "eval_loss": 0.35131019353866577,
+      "eval_runtime": 0.1053,
+      "eval_samples_per_second": 436.997,
+      "eval_steps_per_second": 9.5,
+      "step": 210
+    },
+    {
+      "epoch": 5.861111111111111,
+      "grad_norm": 2.625933885574341,
+      "learning_rate": 4.257142857142857e-05,
+      "loss": 0.3372,
+      "step": 211
+    },
+    {
+      "epoch": 5.888888888888889,
+      "grad_norm": 1.0259548425674438,
+      "learning_rate": 4.228571428571429e-05,
+      "loss": 0.3284,
+      "step": 212
+    },
+    {
+      "epoch": 5.916666666666667,
+      "grad_norm": 1.4596794843673706,
+      "learning_rate": 4.2e-05,
+      "loss": 0.329,
+      "step": 213
+    },
+    {
+      "epoch": 5.944444444444445,
+      "grad_norm": 1.1902427673339844,
+      "learning_rate": 4.1714285714285714e-05,
+      "loss": 0.3268,
+      "step": 214
+    },
+    {
+      "epoch": 5.972222222222222,
+      "grad_norm": 1.7438991069793701,
+      "learning_rate": 4.1428571428571437e-05,
+      "loss": 0.3336,
+      "step": 215
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 2.127594470977783,
+      "learning_rate": 4.1142857142857146e-05,
+      "loss": 0.3338,
+      "step": 216
+    },
+    {
+      "epoch": 6.027777777777778,
+      "grad_norm": 2.1449036598205566,
+      "learning_rate": 4.085714285714286e-05,
+      "loss": 0.3351,
+      "step": 217
+    },
+    {
+      "epoch": 6.055555555555555,
+      "grad_norm": 2.090463161468506,
+      "learning_rate": 4.057142857142857e-05,
+      "loss": 0.3351,
+      "step": 218
+    },
+    {
+      "epoch": 6.083333333333333,
+      "grad_norm": 1.173915147781372,
+      "learning_rate": 4.028571428571429e-05,
+      "loss": 0.3289,
+      "step": 219
+    },
+    {
+      "epoch": 6.111111111111111,
+      "grad_norm": 1.5039347410202026,
+      "learning_rate": 4e-05,
+      "loss": 0.3286,
+      "step": 220
+    },
+    {
+      "epoch": 6.111111111111111,
+      "eval_loss": 0.3502294719219208,
+      "eval_runtime": 0.105,
+      "eval_samples_per_second": 438.259,
+      "eval_steps_per_second": 9.527,
+      "step": 220
+    },
+    {
+      "epoch": 6.138888888888889,
+      "grad_norm": 1.5905152559280396,
+      "learning_rate": 3.971428571428571e-05,
+      "loss": 0.327,
+      "step": 221
+    },
+    {
+      "epoch": 6.166666666666667,
+      "grad_norm": 1.1944066286087036,
+      "learning_rate": 3.942857142857143e-05,
+      "loss": 0.3266,
+      "step": 222
+    },
+    {
+      "epoch": 6.194444444444445,
+      "grad_norm": 1.469502329826355,
+      "learning_rate": 3.9142857142857145e-05,
+      "loss": 0.3309,
+      "step": 223
+    },
+    {
+      "epoch": 6.222222222222222,
+      "grad_norm": 2.0030901432037354,
+      "learning_rate": 3.885714285714286e-05,
+      "loss": 0.3343,
+      "step": 224
+    },
+    {
+      "epoch": 6.25,
+      "grad_norm": 1.3164979219436646,
+      "learning_rate": 3.857142857142858e-05,
+      "loss": 0.3275,
+      "step": 225
+    },
+    {
+      "epoch": 6.277777777777778,
+      "grad_norm": 0.7995334267616272,
+      "learning_rate": 3.8285714285714286e-05,
+      "loss": 0.3277,
+      "step": 226
+    },
+    {
+      "epoch": 6.305555555555555,
+      "grad_norm": 1.2250066995620728,
+      "learning_rate": 3.8e-05,
+      "loss": 0.3242,
+      "step": 227
+    },
+    {
+      "epoch": 6.333333333333333,
+      "grad_norm": 1.2443605661392212,
+      "learning_rate": 3.771428571428572e-05,
+      "loss": 0.3313,
+      "step": 228
+    },
+    {
+      "epoch": 6.361111111111111,
+      "grad_norm": 0.8322944641113281,
+      "learning_rate": 3.742857142857143e-05,
+      "loss": 0.3221,
+      "step": 229
+    },
+    {
+      "epoch": 6.388888888888889,
+      "grad_norm": 0.8514747023582458,
+      "learning_rate": 3.7142857142857143e-05,
+      "loss": 0.3331,
+      "step": 230
+    },
+    {
+      "epoch": 6.388888888888889,
+      "eval_loss": 0.34429529309272766,
+      "eval_runtime": 0.1052,
+      "eval_samples_per_second": 437.304,
+      "eval_steps_per_second": 9.507,
+      "step": 230
+    },
+    {
+      "epoch": 6.416666666666667,
+      "grad_norm": 0.9474199414253235,
+      "learning_rate": 3.685714285714286e-05,
+      "loss": 0.3266,
+      "step": 231
+    },
+    {
+      "epoch": 6.444444444444445,
+      "grad_norm": 1.4075018167495728,
+      "learning_rate": 3.6571428571428576e-05,
+      "loss": 0.3285,
+      "step": 232
+    },
+    {
+      "epoch": 6.472222222222222,
+      "grad_norm": 1.2337582111358643,
+      "learning_rate": 3.628571428571429e-05,
+      "loss": 0.3302,
+      "step": 233
+    },
+    {
+      "epoch": 6.5,
+      "grad_norm": 0.7781947255134583,
+      "learning_rate": 3.6e-05,
+      "loss": 0.317,
+      "step": 234
+    },
+    {
+      "epoch": 6.527777777777778,
+      "grad_norm": 1.1268919706344604,
+      "learning_rate": 3.571428571428572e-05,
+      "loss": 0.3211,
+      "step": 235
+    },
+    {
+      "epoch": 6.555555555555555,
+      "grad_norm": 1.742037057876587,
+      "learning_rate": 3.5428571428571426e-05,
+      "loss": 0.3253,
+      "step": 236
+    },
+    {
+      "epoch": 6.583333333333333,
+      "grad_norm": 0.7068340182304382,
+      "learning_rate": 3.514285714285714e-05,
+      "loss": 0.3173,
+      "step": 237
+    },
+    {
+      "epoch": 6.611111111111111,
+      "grad_norm": 1.5968836545944214,
+      "learning_rate": 3.485714285714286e-05,
+      "loss": 0.3218,
+      "step": 238
+    },
+    {
+      "epoch": 6.638888888888889,
+      "grad_norm": 2.1051909923553467,
+      "learning_rate": 3.4571428571428574e-05,
+      "loss": 0.3298,
+      "step": 239
+    },
+    {
+      "epoch": 6.666666666666667,
+      "grad_norm": 0.9735121130943298,
+      "learning_rate": 3.428571428571429e-05,
+      "loss": 0.3196,
+      "step": 240
+    },
+    {
+      "epoch": 6.666666666666667,
+      "eval_loss": 0.35019850730895996,
+      "eval_runtime": 0.1049,
+      "eval_samples_per_second": 438.63,
+      "eval_steps_per_second": 9.535,
+      "step": 240
+    },
+    {
+      "epoch": 6.694444444444445,
+      "grad_norm": 1.281566858291626,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 0.3303,
+      "step": 241
+    },
+    {
+      "epoch": 6.722222222222222,
+      "grad_norm": 1.067278265953064,
+      "learning_rate": 3.3714285714285716e-05,
+      "loss": 0.3309,
+      "step": 242
+    },
+    {
+      "epoch": 6.75,
+      "grad_norm": 1.42350435256958,
+      "learning_rate": 3.342857142857143e-05,
+      "loss": 0.3246,
+      "step": 243
+    },
+    {
+      "epoch": 6.777777777777778,
+      "grad_norm": 1.0641676187515259,
+      "learning_rate": 3.314285714285714e-05,
+      "loss": 0.3212,
+      "step": 244
+    },
+    {
+      "epoch": 6.805555555555555,
+      "grad_norm": 1.0593169927597046,
+      "learning_rate": 3.285714285714286e-05,
+      "loss": 0.3231,
+      "step": 245
+    },
+    {
+      "epoch": 6.833333333333333,
+      "grad_norm": 1.6785848140716553,
+      "learning_rate": 3.257142857142857e-05,
+      "loss": 0.3186,
+      "step": 246
+    },
+    {
+      "epoch": 6.861111111111111,
+      "grad_norm": 1.6702518463134766,
+      "learning_rate": 3.228571428571428e-05,
+      "loss": 0.3337,
+      "step": 247
+    },
+    {
+      "epoch": 6.888888888888889,
+      "grad_norm": 1.1052055358886719,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.3285,
+      "step": 248
+    },
+    {
+      "epoch": 6.916666666666667,
+      "grad_norm": 1.2286932468414307,
+      "learning_rate": 3.1714285714285715e-05,
+      "loss": 0.3258,
+      "step": 249
+    },
+    {
+      "epoch": 6.944444444444445,
+      "grad_norm": 1.6575883626937866,
+      "learning_rate": 3.142857142857143e-05,
+      "loss": 0.3264,
+      "step": 250
+    },
+    {
+      "epoch": 6.944444444444445,
+      "eval_loss": 0.343667209148407,
+      "eval_runtime": 0.105,
+      "eval_samples_per_second": 438.131,
+      "eval_steps_per_second": 9.525,
+      "step": 250
+    },
+    {
+      "epoch": 6.972222222222222,
+      "grad_norm": 1.004503607749939,
+      "learning_rate": 3.114285714285715e-05,
+      "loss": 0.319,
+      "step": 251
+    },
+    {
+      "epoch": 7.0,
+      "grad_norm": 1.1171271800994873,
+      "learning_rate": 3.0857142857142856e-05,
+      "loss": 0.3211,
+      "step": 252
+    },
+    {
+      "epoch": 7.027777777777778,
+      "grad_norm": 0.989422619342804,
+      "learning_rate": 3.057142857142857e-05,
+      "loss": 0.3184,
+      "step": 253
+    },
+    {
+      "epoch": 7.055555555555555,
+      "grad_norm": 0.727877676486969,
+      "learning_rate": 3.0285714285714288e-05,
+      "loss": 0.315,
+      "step": 254
+    },
+    {
+      "epoch": 7.083333333333333,
+      "grad_norm": 0.7329328060150146,
+      "learning_rate": 3e-05,
+      "loss": 0.3207,
+      "step": 255
+    },
+    {
+      "epoch": 7.111111111111111,
+      "grad_norm": 1.305487871170044,
+      "learning_rate": 2.9714285714285717e-05,
+      "loss": 0.3259,
+      "step": 256
+    },
+    {
+      "epoch": 7.138888888888889,
+      "grad_norm": 0.8338795900344849,
+      "learning_rate": 2.9428571428571426e-05,
+      "loss": 0.3247,
+      "step": 257
+    },
+    {
+      "epoch": 7.166666666666667,
+      "grad_norm": 0.5776103138923645,
+      "learning_rate": 2.9142857142857146e-05,
+      "loss": 0.3192,
+      "step": 258
+    },
+    {
+      "epoch": 7.194444444444445,
+      "grad_norm": 1.2487804889678955,
+      "learning_rate": 2.885714285714286e-05,
+      "loss": 0.3221,
+      "step": 259
+    },
+    {
+      "epoch": 7.222222222222222,
+      "grad_norm": 0.954796314239502,
+      "learning_rate": 2.857142857142857e-05,
+      "loss": 0.3204,
+      "step": 260
+    },
+    {
+      "epoch": 7.222222222222222,
+      "eval_loss": 0.3448273539543152,
+      "eval_runtime": 0.1048,
+      "eval_samples_per_second": 439.048,
+      "eval_steps_per_second": 9.545,
+      "step": 260
+    },
+    {
+      "epoch": 7.25,
+      "grad_norm": 1.0453810691833496,
+      "learning_rate": 2.8285714285714287e-05,
+      "loss": 0.3159,
+      "step": 261
+    },
+    {
+      "epoch": 7.277777777777778,
+      "grad_norm": 0.6628478169441223,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.32,
+      "step": 262
+    },
+    {
+      "epoch": 7.305555555555555,
+      "grad_norm": 2.0091917514801025,
+      "learning_rate": 2.7714285714285716e-05,
+      "loss": 0.3227,
+      "step": 263
+    },
+    {
+      "epoch": 7.333333333333333,
+      "grad_norm": 1.5404771566390991,
+      "learning_rate": 2.742857142857143e-05,
+      "loss": 0.3227,
+      "step": 264
+    },
+    {
+      "epoch": 7.361111111111111,
+      "grad_norm": 0.8551548719406128,
+      "learning_rate": 2.714285714285714e-05,
+      "loss": 0.3221,
+      "step": 265
+    },
+    {
+      "epoch": 7.388888888888889,
+      "grad_norm": 0.816592812538147,
+      "learning_rate": 2.6857142857142857e-05,
+      "loss": 0.3176,
+      "step": 266
+    },
+    {
+      "epoch": 7.416666666666667,
+      "grad_norm": 1.5894920825958252,
+      "learning_rate": 2.6571428571428576e-05,
+      "loss": 0.328,
+      "step": 267
+    },
+    {
+      "epoch": 7.444444444444445,
+      "grad_norm": 1.1515345573425293,
+      "learning_rate": 2.6285714285714286e-05,
+      "loss": 0.3227,
+      "step": 268
+    },
+    {
+      "epoch": 7.472222222222222,
+      "grad_norm": 0.6154195070266724,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 0.311,
+      "step": 269
+    },
+    {
+      "epoch": 7.5,
+      "grad_norm": 0.7660549879074097,
+      "learning_rate": 2.5714285714285714e-05,
+      "loss": 0.3195,
+      "step": 270
+    },
+    {
+      "epoch": 7.5,
+      "eval_loss": 0.3412049412727356,
+      "eval_runtime": 0.1048,
+      "eval_samples_per_second": 438.732,
+      "eval_steps_per_second": 9.538,
+      "step": 270
+    },
+    {
+      "epoch": 7.527777777777778,
+      "grad_norm": 1.6722830533981323,
+      "learning_rate": 2.542857142857143e-05,
+      "loss": 0.3192,
+      "step": 271
+    },
+    {
+      "epoch": 7.555555555555555,
+      "grad_norm": 1.0368938446044922,
+      "learning_rate": 2.5142857142857147e-05,
+      "loss": 0.3262,
+      "step": 272
+    },
+    {
+      "epoch": 7.583333333333333,
+      "grad_norm": 1.0102812051773071,
+      "learning_rate": 2.485714285714286e-05,
+      "loss": 0.3247,
+      "step": 273
+    },
+    {
+      "epoch": 7.611111111111111,
+      "grad_norm": 0.9800832271575928,
+      "learning_rate": 2.4571428571428572e-05,
+      "loss": 0.3253,
+      "step": 274
+    },
+    {
+      "epoch": 7.638888888888889,
+      "grad_norm": 0.741663932800293,
+      "learning_rate": 2.4285714285714288e-05,
+      "loss": 0.3116,
+      "step": 275
+    },
+    {
+      "epoch": 7.666666666666667,
+      "grad_norm": 1.0429283380508423,
+      "learning_rate": 2.4e-05,
+      "loss": 0.3195,
+      "step": 276
+    },
+    {
+      "epoch": 7.694444444444445,
+      "grad_norm": 0.8469560146331787,
+      "learning_rate": 2.3714285714285717e-05,
+      "loss": 0.3178,
+      "step": 277
+    },
+    {
+      "epoch": 7.722222222222222,
+      "grad_norm": 0.8006932139396667,
+      "learning_rate": 2.342857142857143e-05,
+      "loss": 0.3154,
+      "step": 278
+    },
+    {
+      "epoch": 7.75,
+      "grad_norm": 0.8269030451774597,
+      "learning_rate": 2.3142857142857145e-05,
+      "loss": 0.3203,
+      "step": 279
+    },
+    {
+      "epoch": 7.777777777777778,
+      "grad_norm": 0.7541152834892273,
+      "learning_rate": 2.2857142857142858e-05,
+      "loss": 0.314,
+      "step": 280
+    },
+    {
+      "epoch": 7.777777777777778,
+      "eval_loss": 0.3413321375846863,
+      "eval_runtime": 0.1045,
+      "eval_samples_per_second": 440.216,
+      "eval_steps_per_second": 9.57,
+      "step": 280
+    },
+    {
+      "epoch": 7.805555555555555,
+      "grad_norm": 1.1791499853134155,
+      "learning_rate": 2.257142857142857e-05,
+      "loss": 0.3201,
+      "step": 281
+    },
+    {
+      "epoch": 7.833333333333333,
+      "grad_norm": 0.8804281949996948,
+      "learning_rate": 2.2285714285714287e-05,
+      "loss": 0.3164,
+      "step": 282
+    },
+    {
+      "epoch": 7.861111111111111,
+      "grad_norm": 0.7627995014190674,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 0.3194,
+      "step": 283
+    },
+    {
+      "epoch": 7.888888888888889,
+      "grad_norm": 1.013617753982544,
+      "learning_rate": 2.1714285714285715e-05,
+      "loss": 0.3272,
+      "step": 284
+    },
+    {
+      "epoch": 7.916666666666667,
+      "grad_norm": 0.9311711192131042,
+      "learning_rate": 2.1428571428571428e-05,
+      "loss": 0.3184,
+      "step": 285
+    },
+    {
+      "epoch": 7.944444444444445,
+      "grad_norm": 0.9507532119750977,
+      "learning_rate": 2.1142857142857144e-05,
+      "loss": 0.3218,
+      "step": 286
+    },
+    {
+      "epoch": 7.972222222222222,
+      "grad_norm": 0.8252774477005005,
+      "learning_rate": 2.0857142857142857e-05,
+      "loss": 0.3245,
+      "step": 287
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 0.9303626418113708,
+      "learning_rate": 2.0571428571428573e-05,
+      "loss": 0.316,
+      "step": 288
+    },
+    {
+      "epoch": 8.027777777777779,
+      "grad_norm": 1.1884357929229736,
+      "learning_rate": 2.0285714285714286e-05,
+      "loss": 0.3141,
+      "step": 289
+    },
+    {
+      "epoch": 8.055555555555555,
+      "grad_norm": 0.9842406511306763,
+      "learning_rate": 2e-05,
+      "loss": 0.3211,
+      "step": 290
+    },
+    {
+      "epoch": 8.055555555555555,
+      "eval_loss": 0.34589651226997375,
+      "eval_runtime": 0.1044,
+      "eval_samples_per_second": 440.747,
+      "eval_steps_per_second": 9.581,
+      "step": 290
+    },
+    {
+      "epoch": 8.083333333333334,
+      "grad_norm": 1.2224154472351074,
+      "learning_rate": 1.9714285714285714e-05,
+      "loss": 0.3111,
+      "step": 291
+    },
+    {
+      "epoch": 8.11111111111111,
+      "grad_norm": 0.9548574090003967,
+      "learning_rate": 1.942857142857143e-05,
+      "loss": 0.3175,
+      "step": 292
+    },
+    {
+      "epoch": 8.13888888888889,
+      "grad_norm": 0.8031886219978333,
+      "learning_rate": 1.9142857142857143e-05,
+      "loss": 0.3178,
+      "step": 293
+    },
+    {
+      "epoch": 8.166666666666666,
+      "grad_norm": 0.8493176698684692,
+      "learning_rate": 1.885714285714286e-05,
+      "loss": 0.3189,
+      "step": 294
+    },
+    {
+      "epoch": 8.194444444444445,
+      "grad_norm": 0.8125531077384949,
+      "learning_rate": 1.8571428571428572e-05,
+      "loss": 0.3119,
+      "step": 295
+    },
+    {
+      "epoch": 8.222222222222221,
+      "grad_norm": 0.6368350982666016,
+      "learning_rate": 1.8285714285714288e-05,
+      "loss": 0.3113,
+      "step": 296
+    },
+    {
+      "epoch": 8.25,
+      "grad_norm": 1.0392847061157227,
+      "learning_rate": 1.8e-05,
+      "loss": 0.318,
+      "step": 297
+    },
+    {
+      "epoch": 8.277777777777779,
+      "grad_norm": 0.7722540497779846,
+      "learning_rate": 1.7714285714285713e-05,
+      "loss": 0.3234,
+      "step": 298
+    },
+    {
+      "epoch": 8.305555555555555,
+      "grad_norm": 0.740352988243103,
+      "learning_rate": 1.742857142857143e-05,
+      "loss": 0.3172,
+      "step": 299
+    },
+    {
+      "epoch": 8.333333333333334,
+      "grad_norm": 1.4656850099563599,
+      "learning_rate": 1.7142857142857145e-05,
+      "loss": 0.3161,
+      "step": 300
+    },
+    {
+      "epoch": 8.333333333333334,
+      "eval_loss": 0.3438817262649536,
+      "eval_runtime": 0.1048,
+      "eval_samples_per_second": 439.01,
+      "eval_steps_per_second": 9.544,
+      "step": 300
+    },
+    {
+      "epoch": 8.36111111111111,
+      "grad_norm": 0.9987584948539734,
+      "learning_rate": 1.6857142857142858e-05,
+      "loss": 0.3154,
+      "step": 301
+    },
+    {
+      "epoch": 8.38888888888889,
+      "grad_norm": 0.7025516629219055,
+      "learning_rate": 1.657142857142857e-05,
+      "loss": 0.3081,
+      "step": 302
+    },
+    {
+      "epoch": 8.416666666666666,
+      "grad_norm": 0.6634803414344788,
+      "learning_rate": 1.6285714285714287e-05,
+      "loss": 0.3114,
+      "step": 303
+    },
+    {
+      "epoch": 8.444444444444445,
+      "grad_norm": 1.4024748802185059,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.3184,
+      "step": 304
+    },
+    {
+      "epoch": 8.472222222222221,
+      "grad_norm": 1.4321742057800293,
+      "learning_rate": 1.5714285714285715e-05,
+      "loss": 0.3221,
+      "step": 305
+    },
+    {
+      "epoch": 8.5,
+      "grad_norm": 1.052692174911499,
+      "learning_rate": 1.5428571428571428e-05,
+      "loss": 0.3182,
+      "step": 306
+    },
+    {
+      "epoch": 8.527777777777779,
+      "grad_norm": 1.5870689153671265,
+      "learning_rate": 1.5142857142857144e-05,
+      "loss": 0.3207,
+      "step": 307
+    },
+    {
+      "epoch": 8.555555555555555,
+      "grad_norm": 1.1841577291488647,
+      "learning_rate": 1.4857142857142858e-05,
+      "loss": 0.3172,
+      "step": 308
+    },
+    {
+      "epoch": 8.583333333333334,
+      "grad_norm": 0.6852739453315735,
+      "learning_rate": 1.4571428571428573e-05,
+      "loss": 0.3123,
+      "step": 309
+    },
+    {
+      "epoch": 8.61111111111111,
+      "grad_norm": 0.6914618015289307,
+      "learning_rate": 1.4285714285714285e-05,
+      "loss": 0.3192,
+      "step": 310
+    },
+    {
+      "epoch": 8.61111111111111,
+      "eval_loss": 0.34119004011154175,
+      "eval_runtime": 0.105,
+      "eval_samples_per_second": 438.008,
+      "eval_steps_per_second": 9.522,
+      "step": 310
+    },
+    {
+      "epoch": 8.63888888888889,
+      "grad_norm": 0.8360849618911743,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 0.3165,
+      "step": 311
+    },
+    {
+      "epoch": 8.666666666666666,
+      "grad_norm": 0.7968160510063171,
+      "learning_rate": 1.3714285714285716e-05,
+      "loss": 0.3147,
+      "step": 312
+    },
+    {
+      "epoch": 8.694444444444445,
+      "grad_norm": 0.9014184474945068,
+      "learning_rate": 1.3428571428571429e-05,
+      "loss": 0.3146,
+      "step": 313
+    },
+    {
+      "epoch": 8.722222222222221,
+      "grad_norm": 0.7425902485847473,
+      "learning_rate": 1.3142857142857143e-05,
+      "loss": 0.3124,
+      "step": 314
+    },
+    {
+      "epoch": 8.75,
+      "grad_norm": 1.4130454063415527,
+      "learning_rate": 1.2857142857142857e-05,
+      "loss": 0.313,
+      "step": 315
+    },
+    {
+      "epoch": 8.777777777777779,
+      "grad_norm": 1.8442883491516113,
+      "learning_rate": 1.2571428571428573e-05,
+      "loss": 0.3221,
+      "step": 316
+    },
+    {
+      "epoch": 8.805555555555555,
+      "grad_norm": 0.8699185252189636,
+      "learning_rate": 1.2285714285714286e-05,
+      "loss": 0.3186,
+      "step": 317
+    },
+    {
+      "epoch": 8.833333333333334,
+      "grad_norm": 0.7799740433692932,
+      "learning_rate": 1.2e-05,
+      "loss": 0.3139,
+      "step": 318
+    },
+    {
+      "epoch": 8.86111111111111,
+      "grad_norm": 0.7385247349739075,
+      "learning_rate": 1.1714285714285715e-05,
+      "loss": 0.3121,
+      "step": 319
+    },
+    {
+      "epoch": 8.88888888888889,
+      "grad_norm": 1.3332445621490479,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 0.3119,
+      "step": 320
+    },
+    {
+      "epoch": 8.88888888888889,
+      "eval_loss": 0.3395558297634125,
+      "eval_runtime": 0.1048,
+      "eval_samples_per_second": 439.113,
+      "eval_steps_per_second": 9.546,
+      "step": 320
+    },
+    {
+      "epoch": 8.916666666666666,
+      "grad_norm": 0.7556406855583191,
+      "learning_rate": 1.1142857142857143e-05,
+      "loss": 0.3145,
+      "step": 321
+    },
+    {
+      "epoch": 8.944444444444445,
+      "grad_norm": 1.078497290611267,
+      "learning_rate": 1.0857142857142858e-05,
+      "loss": 0.3093,
+      "step": 322
+    },
+    {
+      "epoch": 8.972222222222221,
+      "grad_norm": 0.6883450150489807,
+      "learning_rate": 1.0571428571428572e-05,
+      "loss": 0.3187,
+      "step": 323
+    },
+    {
+      "epoch": 9.0,
+      "grad_norm": 1.3395229578018188,
+      "learning_rate": 1.0285714285714286e-05,
+      "loss": 0.3205,
+      "step": 324
+    },
+    {
+      "epoch": 9.027777777777779,
+      "grad_norm": 0.8341308832168579,
+      "learning_rate": 1e-05,
+      "loss": 0.3136,
+      "step": 325
+    },
+    {
+      "epoch": 9.055555555555555,
+      "grad_norm": 0.8046018481254578,
+      "learning_rate": 9.714285714285715e-06,
+      "loss": 0.3135,
+      "step": 326
+    },
+    {
+      "epoch": 9.083333333333334,
+      "grad_norm": 0.6076436638832092,
+      "learning_rate": 9.42857142857143e-06,
+      "loss": 0.3118,
+      "step": 327
+    },
+    {
+      "epoch": 9.11111111111111,
+      "grad_norm": 0.8679487705230713,
+      "learning_rate": 9.142857142857144e-06,
+      "loss": 0.3108,
+      "step": 328
+    },
+    {
+      "epoch": 9.13888888888889,
+      "grad_norm": 0.7275362014770508,
+      "learning_rate": 8.857142857142857e-06,
+      "loss": 0.3076,
+      "step": 329
+    },
+    {
+      "epoch": 9.166666666666666,
+      "grad_norm": 1.152782678604126,
+      "learning_rate": 8.571428571428573e-06,
+      "loss": 0.3153,
+      "step": 330
+    },
+    {
+      "epoch": 9.166666666666666,
+      "eval_loss": 0.34156906604766846,
+      "eval_runtime": 0.1043,
+      "eval_samples_per_second": 440.833,
+      "eval_steps_per_second": 9.583,
+      "step": 330
+    },
+    {
+      "epoch": 9.194444444444445,
+      "grad_norm": 1.1763490438461304,
+      "learning_rate": 8.285714285714285e-06,
+      "loss": 0.3114,
+      "step": 331
+    },
+    {
+      "epoch": 9.222222222222221,
+      "grad_norm": 0.7035480737686157,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.3091,
+      "step": 332
+    },
+    {
+      "epoch": 9.25,
+      "grad_norm": 0.7956453561782837,
+      "learning_rate": 7.714285714285714e-06,
+      "loss": 0.3173,
+      "step": 333
+    },
+    {
+      "epoch": 9.277777777777779,
+      "grad_norm": 0.7433366179466248,
+      "learning_rate": 7.428571428571429e-06,
+      "loss": 0.3069,
+      "step": 334
+    },
+    {
+      "epoch": 9.305555555555555,
+      "grad_norm": 1.6120795011520386,
+      "learning_rate": 7.142857142857143e-06,
+      "loss": 0.3142,
+      "step": 335
+    },
+    {
+      "epoch": 9.333333333333334,
+      "grad_norm": 0.7718474268913269,
+      "learning_rate": 6.857142857142858e-06,
+      "loss": 0.3066,
+      "step": 336
+    },
+    {
+      "epoch": 9.36111111111111,
+      "grad_norm": 0.7094411253929138,
+      "learning_rate": 6.5714285714285714e-06,
+      "loss": 0.3159,
+      "step": 337
+    },
+    {
+      "epoch": 9.38888888888889,
+      "grad_norm": 0.8119326829910278,
+      "learning_rate": 6.285714285714287e-06,
+      "loss": 0.3106,
+      "step": 338
+    },
+    {
+      "epoch": 9.416666666666666,
+      "grad_norm": 1.7396577596664429,
+      "learning_rate": 6e-06,
+      "loss": 0.3084,
+      "step": 339
+    },
+    {
+      "epoch": 9.444444444444445,
+      "grad_norm": 1.0242513418197632,
+      "learning_rate": 5.7142857142857145e-06,
+      "loss": 0.3101,
+      "step": 340
+    },
+    {
+      "epoch": 9.444444444444445,
+      "eval_loss": 0.3424239754676819,
+      "eval_runtime": 0.1047,
+      "eval_samples_per_second": 439.488,
+      "eval_steps_per_second": 9.554,
+      "step": 340
+    },
+    {
+      "epoch": 9.472222222222221,
+      "grad_norm": 1.1069552898406982,
+      "learning_rate": 5.428571428571429e-06,
+      "loss": 0.3151,
+      "step": 341
+    },
+    {
+      "epoch": 9.5,
+      "grad_norm": 0.7284579277038574,
+      "learning_rate": 5.142857142857143e-06,
+      "loss": 0.3063,
+      "step": 342
+    },
+    {
+      "epoch": 9.527777777777779,
+      "grad_norm": 0.6138132214546204,
+      "learning_rate": 4.857142857142858e-06,
+      "loss": 0.3096,
+      "step": 343
+    },
+    {
+      "epoch": 9.555555555555555,
+      "grad_norm": 0.7576848864555359,
+      "learning_rate": 4.571428571428572e-06,
+      "loss": 0.3142,
+      "step": 344
+    },
+    {
+      "epoch": 9.583333333333334,
+      "grad_norm": 0.5917118191719055,
+      "learning_rate": 4.285714285714286e-06,
+      "loss": 0.3111,
+      "step": 345
+    },
+    {
+      "epoch": 9.61111111111111,
+      "grad_norm": 1.052357792854309,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.3134,
+      "step": 346
+    },
+    {
+      "epoch": 9.63888888888889,
+      "grad_norm": 1.0170456171035767,
+      "learning_rate": 3.7142857142857146e-06,
+      "loss": 0.3057,
+      "step": 347
+    },
+    {
+      "epoch": 9.666666666666666,
+      "grad_norm": 0.9443879723548889,
+      "learning_rate": 3.428571428571429e-06,
+      "loss": 0.3078,
+      "step": 348
+    },
+    {
+      "epoch": 9.694444444444445,
+      "grad_norm": 0.589635968208313,
+      "learning_rate": 3.1428571428571433e-06,
+      "loss": 0.309,
+      "step": 349
+    },
+    {
+      "epoch": 9.722222222222221,
+      "grad_norm": 0.7533226609230042,
+      "learning_rate": 2.8571428571428573e-06,
+      "loss": 0.3151,
+      "step": 350
+    },
+    {
+      "epoch": 9.722222222222221,
+      "eval_loss": 0.34002798795700073,
+      "eval_runtime": 0.1051,
+      "eval_samples_per_second": 437.81,
+      "eval_steps_per_second": 9.518,
+      "step": 350
+    },
+    {
+      "epoch": 9.75,
+      "grad_norm": 0.997863233089447,
+      "learning_rate": 2.5714285714285716e-06,
+      "loss": 0.3066,
+      "step": 351
+    },
+    {
+      "epoch": 9.777777777777779,
+      "grad_norm": 0.7841188907623291,
+      "learning_rate": 2.285714285714286e-06,
+      "loss": 0.3133,
+      "step": 352
+    },
+    {
+      "epoch": 9.805555555555555,
+      "grad_norm": 0.8386639356613159,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.3081,
+      "step": 353
+    },
+    {
+      "epoch": 9.833333333333334,
+      "grad_norm": 0.9777269959449768,
+      "learning_rate": 1.7142857142857145e-06,
+      "loss": 0.3093,
+      "step": 354
+    },
+    {
+      "epoch": 9.86111111111111,
+      "grad_norm": 0.5559796094894409,
+      "learning_rate": 1.4285714285714286e-06,
+      "loss": 0.3086,
+      "step": 355
+    },
+    {
+      "epoch": 9.88888888888889,
+      "grad_norm": 0.8345446586608887,
+      "learning_rate": 1.142857142857143e-06,
+      "loss": 0.3134,
+      "step": 356
+    },
+    {
+      "epoch": 9.916666666666666,
+      "grad_norm": 0.6907045245170593,
+      "learning_rate": 8.571428571428572e-07,
+      "loss": 0.3085,
+      "step": 357
+    },
+    {
+      "epoch": 9.944444444444445,
+      "grad_norm": 0.9042662978172302,
+      "learning_rate": 5.714285714285715e-07,
+      "loss": 0.3112,
+      "step": 358
+    },
+    {
+      "epoch": 9.972222222222221,
+      "grad_norm": 0.790155291557312,
+      "learning_rate": 2.8571428571428575e-07,
+      "loss": 0.3134,
+      "step": 359
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 0.983366847038269,
+      "learning_rate": 0.0,
+      "loss": 0.3109,
+      "step": 360
+    },
+    {
+      "epoch": 10.0,
+      "eval_loss": 0.3411920666694641,
+      "eval_runtime": 0.1051,
+      "eval_samples_per_second": 437.492,
+      "eval_steps_per_second": 9.511,
+      "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
+}

2026.03.27/14.32.19_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:ac0c304e82c9922a614b94da74c2443ccb0f60cb553e6b706d97d2e18196d836
+size 6008

2026.03.27/14.32.19_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
+}

2026.03.27/14.32.19_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"
+}

2026.03.27/14.32.19_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:f55c1948674d165534e4d6baf48e4473a272ef058b49958b1718d72870e43e61
+size 539588496

2026.03.27/14.32.19_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 sha256:7863dd8b1cec54ea35c3840df161a34ee700911175e25f412f15c526ba1f4db9
+size 940

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/train_llm_workspace.log

@@ -0,0 +1,6 @@
+[2026-03-27 14:32:23,647][datasets.arrow_dataset][WARNING] - Setting TOKENIZERS_PARALLELISM=false for forked processes.
+[2026-03-27 14:32:24,779][datasets.arrow_dataset][WARNING] - Setting TOKENIZERS_PARALLELISM=false for forked processes.
+[2026-03-27 14:32:25,633][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/tmpwvwuxc5d/test.c -o /tmp/tmpwvwuxc5d/test.o
+[2026-03-27 14:32:25,686][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat /tmp/tmpwvwuxc5d/test.o -laio -o /tmp/tmpwvwuxc5d/a.out
+[2026-03-27 14:32:26,629][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/tmp56po9x84/test.c -o /tmp/tmp56po9x84/test.o
+[2026-03-27 14:32:26,682][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat /tmp/tmp56po9x84/test.o -L/usr/local/cuda -L/usr/local/cuda/lib64 -lcufile -o /tmp/tmp56po9x84/a.out

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/debug-internal.log

@@ -0,0 +1,17 @@
+{"time":"2026-03-27T14:32:20.620246342+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-03-27T14:32:20.620253938+08:00","level":"INFO","msg":"created symlink","path":"/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/logs/debug-core.log"}
+{"time":"2026-03-27T14:32:20.724793938+08:00","level":"INFO","msg":"created new stream","id":"0k6mearn"}
+{"time":"2026-03-27T14:32:20.724817259+08:00","level":"INFO","msg":"stream: started","id":"0k6mearn"}
+{"time":"2026-03-27T14:32:20.725168562+08:00","level":"INFO","msg":"sender: started","stream_id":"0k6mearn"}
+{"time":"2026-03-27T14:32:20.725226533+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"0k6mearn"}}
+{"time":"2026-03-27T14:32:20.725446167+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"0k6mearn"}}
+{"time":"2026-03-27T14:32:21.694533142+08:00","level":"INFO","msg":"Starting system monitor"}
+{"time":"2026-03-27T14:37:07.025330756+08:00","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2026-03-27T14:37:07.027139427+08:00","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2026-03-27T14:37:08.025582467+08:00","level":"INFO","msg":"handler: operation stats","stats":{"operations":[{"desc":"uploading wandb-summary.json","runtime_seconds":0.172485726,"progress":"494B/494B"},{"desc":"saving job artifact","runtime_seconds":0.001363647}],"total_operations":2}}
+{"time":"2026-03-27T14:37:12.457040628+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-03-27T14:37:14.154706692+08:00","level":"INFO","msg":"stream: closing","id":"0k6mearn"}
+{"time":"2026-03-27T14:37:14.154740921+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"0k6mearn"}}
+{"time":"2026-03-27T14:37:14.154772565+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"0k6mearn"}}
+{"time":"2026-03-27T14:37:14.154798203+08:00","level":"INFO","msg":"sender: closed","stream_id":"0k6mearn"}
+{"time":"2026-03-27T14:37:14.154919487+08:00","level":"INFO","msg":"stream: closed","id":"0k6mearn"}

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/debug.log

@@ -0,0 +1,35 @@
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Configure stats pid to 2671052
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Loading settings from /home/chyang/.config/wandb/settings
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Loading settings from /tmp2/chyang/workspace/LLM-BC/wandb/settings
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'llmbc/workspace/train_llm_workspace.py', 'program_abspath': '/tmp2/chyang/workspace/LLM-BC/llmbc/workspace/train_llm_workspace.py', 'program': '/tmp2/chyang/workspace/LLM-BC/./llmbc/workspace/train_llm_workspace.py'}
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_init.py:_log_setup():533] Logging user logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/logs/debug.log
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_init.py:_log_setup():534] Logging internal logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/logs/debug-internal.log
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_init.py:init():619] calling init triggers
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [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': '2026.03.27-14.32.19_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/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1', 'wandb_name_base': '2026.03.27-14.32.19_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': True}, '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/2026.03.27/02.42.34_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/2026.03.27/14.32.19_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_init.py:init():669] starting backend
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_init.py:init():673] sending inform_init request
+2026-03-27 14:32:20,616 INFO    MainThread:2671052 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-03-27 14:32:20,617 INFO    MainThread:2671052 [wandb_init.py:init():686] backend started and connected
+2026-03-27 14:32:20,623 INFO    MainThread:2671052 [wandb_init.py:init():781] updated telemetry
+2026-03-27 14:32:20,645 INFO    MainThread:2671052 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-03-27 14:32:21,691 INFO    MainThread:2671052 [wandb_init.py:init():867] starting run threads in backend
+2026-03-27 14:32:21,800 INFO    MainThread:2671052 [wandb_run.py:_console_start():2451] atexit reg
+2026-03-27 14:32:21,800 INFO    MainThread:2671052 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-03-27 14:32:21,800 INFO    MainThread:2671052 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-03-27 14:32:21,800 INFO    MainThread:2671052 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-03-27 14:32:21,801 INFO    MainThread:2671052 [wandb_init.py:init():911] run started, returning control to user process
+2026-03-27 14:32:28,013 INFO    MainThread:2671052 [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/2026.03.27/14.32.19_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/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/runs/Mar27_14-32-25_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/2026.03.27/14.32.19_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 14:32:28,015 INFO    MainThread:2671052 [wandb_config.py:__setitem__():154] config set model/num_parameters = 134889408 - <bound method Run._config_callback of <wandb.sdk.wandb_run.Run object at 0x76464270ed60>>
+2026-03-27 14:32:28,015 INFO    MainThread:2671052 [wandb_run.py:_config_callback():1389] config_cb model/num_parameters 134889408 None
+2026-03-27 14:37:07,024 INFO    MainThread:2671052 [wandb_run.py:_finish():2146] finishing run chyang25-national-taiwan-university/llm_module_finetuning/0k6mearn
+2026-03-27 14:37:07,024 INFO    MainThread:2671052 [wandb_run.py:_atexit_cleanup():2414] got exitcode: 0
+2026-03-27 14:37:07,024 INFO    MainThread:2671052 [wandb_run.py:_restore():2396] restore
+2026-03-27 14:37:07,024 INFO    MainThread:2671052 [wandb_run.py:_restore():2402] restore done
+2026-03-27 14:37:14,148 INFO    MainThread:2671052 [wandb_run.py:_footer_history_summary_info():3963] rendering history
+2026-03-27 14:37:14,150 INFO    MainThread:2671052 [wandb_run.py:_footer_history_summary_info():3995] rendering summary
+2026-03-27 14:37:14,154 INFO    MainThread:2671052 [wandb_run.py:_footer_sync_info():3922] logging synced files

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/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": eval/steps_per_second
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": eval/runtime
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": eval/loss
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": train/grad_norm
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": train/epoch
+              "5": 2
+              "6":
+                - 1
+                - 3
+              "7": []
+            - "1": eval/samples_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/2026.03.27/02.42.34_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/2026.03.27/14.32.19_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: 2026.03.27-14.32.19_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/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/runs/Mar27_14-32-25_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/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1
+        wandb_name_base: 2026.03.27-14.32.19_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/2026.03.27/14.32.19_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/2026.03.27/14.32.19_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: true
+            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

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/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/2026.03.27/02.42.34_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, 22636.87it/s]
+Setting TOKENIZERS_PARALLELISM=false for forked processes.
+[2026-03-27 14:32:23,647][datasets.arrow_dataset][WARNING] - Setting TOKENIZERS_PARALLELISM=false for forked processes.
+Map (num_proc=4): 100%|███████████████████████████████████████████████████████████████████████| 4585/4585 [00:00<00:00, 4851.45 examples/s]
+Setting TOKENIZERS_PARALLELISM=false for forked processes.
+[2026-03-27 14:32:24,779][datasets.arrow_dataset][WARNING] - Setting TOKENIZERS_PARALLELISM=false for forked processes.
+Map (num_proc=4): 100%|███████████████████████████████████████████████████████████████████████| 4585/4585 [00:00<00:00, 9674.89 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 14:32:25,500] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cuda (auto detect)
+[2026-03-27 14:32:25,633][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/tmpwvwuxc5d/test.c -o /tmp/tmpwvwuxc5d/test.o
+[2026-03-27 14:32:25,686][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat /tmp/tmpwvwuxc5d/test.o -laio -o /tmp/tmpwvwuxc5d/a.out
+[2026-03-27 14:32:26,629][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/tmp56po9x84/test.c -o /tmp/tmp56po9x84/test.o
+[2026-03-27 14:32:26,682][root][INFO] - gcc -pthread -B /home/chyang/miniconda3/envs/llm-bc/compiler_compat /tmp/tmp56po9x84/test.o -L/usr/local/cuda -L/usr/local/cuda/lib64 -lcufile -o /tmp/tmp56po9x84/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': 1.3459, 'grad_norm': 41.524925231933594, 'learning_rate': 1e-05, 'epoch': 0.03}
+{'loss': 1.3676, 'grad_norm': 36.50713348388672, 'learning_rate': 2e-05, 'epoch': 0.06}
+{'loss': 1.3473, 'grad_norm': 101.92433166503906, 'learning_rate': 3e-05, 'epoch': 0.08}
+{'loss': 1.334, 'grad_norm': 107.9959945678711, 'learning_rate': 4e-05, 'epoch': 0.11}
+{'loss': 1.3542, 'grad_norm': 51.27205276489258, 'learning_rate': 5e-05, 'epoch': 0.14}
+{'loss': 1.3031, 'grad_norm': 46.98747253417969, 'learning_rate': 6e-05, 'epoch': 0.17}
+{'loss': 1.2489, 'grad_norm': 163.22494506835938, 'learning_rate': 7e-05, 'epoch': 0.19}
+{'loss': 1.0557, 'grad_norm': 45.806182861328125, 'learning_rate': 8e-05, 'epoch': 0.22}
+{'loss': 0.8285, 'grad_norm': 27.931575775146484, 'learning_rate': 9e-05, 'epoch': 0.25}
+{'loss': 0.8731, 'grad_norm': 55.34366226196289, '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': 0.8334420323371887, 'eval_runtime': 0.1041, 'eval_samples_per_second': 442.063, 'eval_steps_per_second': 9.61, 'epoch': 0.28}
+{'loss': 0.7469, 'grad_norm': 27.75310516357422, 'learning_rate': 9.971428571428571e-05, 'epoch': 0.31}
+{'loss': 0.6513, 'grad_norm': 24.505470275878906, 'learning_rate': 9.942857142857144e-05, 'epoch': 0.33}
+{'loss': 0.5855, 'grad_norm': 8.471323013305664, 'learning_rate': 9.914285714285715e-05, 'epoch': 0.36}
+{'loss': 0.5901, 'grad_norm': 12.284788131713867, 'learning_rate': 9.885714285714286e-05, 'epoch': 0.39}
+{'loss': 0.558, 'grad_norm': 6.142014026641846, 'learning_rate': 9.857142857142858e-05, 'epoch': 0.42}
+{'loss': 0.5863, 'grad_norm': 13.09057903289795, 'learning_rate': 9.828571428571429e-05, 'epoch': 0.44}
+{'loss': 0.5416, 'grad_norm': 8.136122703552246, 'learning_rate': 9.8e-05, 'epoch': 0.47}
+{'loss': 0.5398, 'grad_norm': 7.257413387298584, 'learning_rate': 9.771428571428572e-05, 'epoch': 0.5}
+{'loss': 0.4912, 'grad_norm': 2.9581573009490967, 'learning_rate': 9.742857142857143e-05, 'epoch': 0.53}
+{'loss': 0.4961, 'grad_norm': 3.979379892349243, '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': 0.4880354404449463, 'eval_runtime': 0.1036, 'eval_samples_per_second': 443.898, 'eval_steps_per_second': 9.65, 'epoch': 0.56}
+{'loss': 0.4898, 'grad_norm': 2.800405502319336, 'learning_rate': 9.685714285714286e-05, 'epoch': 0.58}
+{'loss': 0.4617, 'grad_norm': 2.456360340118408, 'learning_rate': 9.657142857142858e-05, 'epoch': 0.61}
+{'loss': 0.4847, 'grad_norm': 3.3507285118103027, 'learning_rate': 9.628571428571429e-05, 'epoch': 0.64}
+{'loss': 0.4609, 'grad_norm': 2.5141634941101074, 'learning_rate': 9.6e-05, 'epoch': 0.67}
+{'loss': 0.4369, 'grad_norm': 2.4233858585357666, 'learning_rate': 9.571428571428573e-05, 'epoch': 0.69}
+{'loss': 0.4452, 'grad_norm': 2.399322986602783, 'learning_rate': 9.542857142857143e-05, 'epoch': 0.72}
+{'loss': 0.4326, 'grad_norm': 1.9068751335144043, 'learning_rate': 9.514285714285714e-05, 'epoch': 0.75}
+{'loss': 0.4309, 'grad_norm': 1.8973027467727661, 'learning_rate': 9.485714285714287e-05, 'epoch': 0.78}
+{'loss': 0.4216, 'grad_norm': 1.655167579650879, 'learning_rate': 9.457142857142858e-05, 'epoch': 0.81}
+{'loss': 0.4191, 'grad_norm': 1.3332226276397705, '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': 0.43893250823020935, 'eval_runtime': 0.104, 'eval_samples_per_second': 442.414, 'eval_steps_per_second': 9.618, 'epoch': 0.83}
+{'loss': 0.4206, 'grad_norm': 1.7973644733428955, 'learning_rate': 9.4e-05, 'epoch': 0.86}
+{'loss': 0.424, 'grad_norm': 1.4797940254211426, 'learning_rate': 9.371428571428572e-05, 'epoch': 0.89}
+{'loss': 0.4135, 'grad_norm': 1.4304866790771484, 'learning_rate': 9.342857142857143e-05, 'epoch': 0.92}
+{'loss': 0.4225, 'grad_norm': 1.3090794086456299, 'learning_rate': 9.314285714285715e-05, 'epoch': 0.94}
+{'loss': 0.4083, 'grad_norm': 1.271675944328308, 'learning_rate': 9.285714285714286e-05, 'epoch': 0.97}
+{'loss': 0.4146, 'grad_norm': 1.5309550762176514, 'learning_rate': 9.257142857142858e-05, 'epoch': 1.0}
+{'loss': 0.3898, 'grad_norm': 1.4727030992507935, 'learning_rate': 9.228571428571429e-05, 'epoch': 1.03}
+{'loss': 0.4118, 'grad_norm': 1.752487301826477, 'learning_rate': 9.200000000000001e-05, 'epoch': 1.06}
+{'loss': 0.405, 'grad_norm': 1.634020447731018, 'learning_rate': 9.171428571428572e-05, 'epoch': 1.08}
+{'loss': 0.3834, 'grad_norm': 1.0251728296279907, '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': 0.41799336671829224, 'eval_runtime': 0.104, 'eval_samples_per_second': 442.159, 'eval_steps_per_second': 9.612, 'epoch': 1.11}
+{'loss': 0.3938, 'grad_norm': 1.767491340637207, 'learning_rate': 9.114285714285716e-05, 'epoch': 1.14}
+{'loss': 0.3988, 'grad_norm': 1.4250415563583374, 'learning_rate': 9.085714285714286e-05, 'epoch': 1.17}
+{'loss': 0.4027, 'grad_norm': 2.269926071166992, 'learning_rate': 9.057142857142857e-05, 'epoch': 1.19}
+{'loss': 0.4105, 'grad_norm': 1.5418059825897217, 'learning_rate': 9.028571428571428e-05, 'epoch': 1.22}
+{'loss': 0.3888, 'grad_norm': 1.3003933429718018, 'learning_rate': 9e-05, 'epoch': 1.25}
+{'loss': 0.4034, 'grad_norm': 1.33571195602417, 'learning_rate': 8.971428571428571e-05, 'epoch': 1.28}
+{'loss': 0.3938, 'grad_norm': 1.3438071012496948, 'learning_rate': 8.942857142857142e-05, 'epoch': 1.31}
+{'loss': 0.3917, 'grad_norm': 1.3643842935562134, 'learning_rate': 8.914285714285715e-05, 'epoch': 1.33}
+{'loss': 0.3786, 'grad_norm': 1.1976509094238281, 'learning_rate': 8.885714285714286e-05, 'epoch': 1.36}
+{'loss': 0.3858, 'grad_norm': 1.463449239730835, 'learning_rate': 8.857142857142857e-05, 'epoch': 1.39}
+ 17%|████████████████▊                                                                                    | 60/360 [00:46<03:48,  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': 0.39549750089645386, 'eval_runtime': 0.1046, 'eval_samples_per_second': 439.872, 'eval_steps_per_second': 9.562, 'epoch': 1.39}
+{'loss': 0.3784, 'grad_norm': 1.3630038499832153, 'learning_rate': 8.828571428571429e-05, 'epoch': 1.42}
+{'loss': 0.3696, 'grad_norm': 1.2265862226486206, 'learning_rate': 8.800000000000001e-05, 'epoch': 1.44}
+{'loss': 0.3762, 'grad_norm': 1.042543649673462, 'learning_rate': 8.771428571428572e-05, 'epoch': 1.47}
+{'loss': 0.3813, 'grad_norm': 1.2349590063095093, 'learning_rate': 8.742857142857144e-05, 'epoch': 1.5}
+{'loss': 0.3983, 'grad_norm': 1.9418014287948608, 'learning_rate': 8.714285714285715e-05, 'epoch': 1.53}
+{'loss': 0.3726, 'grad_norm': 0.9780258536338806, 'learning_rate': 8.685714285714286e-05, 'epoch': 1.56}
+{'loss': 0.3801, 'grad_norm': 1.3302329778671265, 'learning_rate': 8.657142857142858e-05, 'epoch': 1.58}
+{'loss': 0.3835, 'grad_norm': 1.0229287147521973, 'learning_rate': 8.62857142857143e-05, 'epoch': 1.61}
+{'loss': 0.3698, 'grad_norm': 1.0397497415542603, 'learning_rate': 8.6e-05, 'epoch': 1.64}
+{'loss': 0.3752, 'grad_norm': 0.9683515429496765, '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': 0.3742421269416809, 'eval_runtime': 0.1043, 'eval_samples_per_second': 441.127, 'eval_steps_per_second': 9.59, 'epoch': 1.67}
+{'loss': 0.3621, 'grad_norm': 0.7956013679504395, 'learning_rate': 8.542857142857144e-05, 'epoch': 1.69}
+{'loss': 0.3685, 'grad_norm': 1.0049488544464111, 'learning_rate': 8.514285714285714e-05, 'epoch': 1.72}
+{'loss': 0.3692, 'grad_norm': 0.9522997140884399, 'learning_rate': 8.485714285714285e-05, 'epoch': 1.75}
+{'loss': 0.3677, 'grad_norm': 1.6391457319259644, 'learning_rate': 8.457142857142858e-05, 'epoch': 1.78}
+{'loss': 0.3587, 'grad_norm': 0.7660079002380371, 'learning_rate': 8.428571428571429e-05, 'epoch': 1.81}
+{'loss': 0.3561, 'grad_norm': 0.8130621314048767, 'learning_rate': 8.4e-05, 'epoch': 1.83}
+{'loss': 0.3703, 'grad_norm': 2.256351947784424, 'learning_rate': 8.371428571428572e-05, 'epoch': 1.86}
+{'loss': 0.3733, 'grad_norm': 2.5743725299835205, 'learning_rate': 8.342857142857143e-05, 'epoch': 1.89}
+{'loss': 0.3691, 'grad_norm': 1.5146273374557495, 'learning_rate': 8.314285714285715e-05, 'epoch': 1.92}
+{'loss': 0.3685, 'grad_norm': 2.370126962661743, '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': 0.37778788805007935, 'eval_runtime': 0.1038, 'eval_samples_per_second': 443.059, 'eval_steps_per_second': 9.632, 'epoch': 1.94}
+{'loss': 0.36, 'grad_norm': 2.041372537612915, 'learning_rate': 8.257142857142858e-05, 'epoch': 1.97}
+{'loss': 0.3819, 'grad_norm': 1.340862512588501, 'learning_rate': 8.228571428571429e-05, 'epoch': 2.0}
+{'loss': 0.3614, 'grad_norm': 1.614141821861267, 'learning_rate': 8.2e-05, 'epoch': 2.03}
+{'loss': 0.3558, 'grad_norm': 1.172598958015442, 'learning_rate': 8.171428571428572e-05, 'epoch': 2.06}
+{'loss': 0.3646, 'grad_norm': 1.5925283432006836, 'learning_rate': 8.142857142857143e-05, 'epoch': 2.08}
+{'loss': 0.3647, 'grad_norm': 1.7977277040481567, 'learning_rate': 8.114285714285714e-05, 'epoch': 2.11}
+{'loss': 0.3599, 'grad_norm': 1.2800759077072144, 'learning_rate': 8.085714285714287e-05, 'epoch': 2.14}
+{'loss': 0.3661, 'grad_norm': 1.8917945623397827, 'learning_rate': 8.057142857142857e-05, 'epoch': 2.17}
+{'loss': 0.3715, 'grad_norm': 2.058119058609009, 'learning_rate': 8.028571428571428e-05, 'epoch': 2.19}
+{'loss': 0.3531, 'grad_norm': 0.8664758801460266, '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': 0.3683662414550781, 'eval_runtime': 0.1045, 'eval_samples_per_second': 440.203, 'eval_steps_per_second': 9.57, 'epoch': 2.22}
+{'loss': 0.3591, 'grad_norm': 1.1171331405639648, 'learning_rate': 7.971428571428572e-05, 'epoch': 2.25}
+{'loss': 0.3554, 'grad_norm': 1.2088688611984253, 'learning_rate': 7.942857142857143e-05, 'epoch': 2.28}
+{'loss': 0.3488, 'grad_norm': 0.7355909943580627, 'learning_rate': 7.914285714285715e-05, 'epoch': 2.31}
+{'loss': 0.3516, 'grad_norm': 1.5655473470687866, 'learning_rate': 7.885714285714286e-05, 'epoch': 2.33}
+{'loss': 0.3531, 'grad_norm': 1.1028951406478882, 'learning_rate': 7.857142857142858e-05, 'epoch': 2.36}
+{'loss': 0.3566, 'grad_norm': 1.0214101076126099, 'learning_rate': 7.828571428571429e-05, 'epoch': 2.39}
+{'loss': 0.3525, 'grad_norm': 1.0004141330718994, 'learning_rate': 7.800000000000001e-05, 'epoch': 2.42}
+{'loss': 0.364, 'grad_norm': 1.1501681804656982, 'learning_rate': 7.771428571428572e-05, 'epoch': 2.44}
+{'loss': 0.3452, 'grad_norm': 0.8693153858184814, 'learning_rate': 7.742857142857143e-05, 'epoch': 2.47}
+{'loss': 0.348, 'grad_norm': 1.0027663707733154, 'learning_rate': 7.714285714285715e-05, 'epoch': 2.5}
+ 28%|███████████████████████████▊                                                                        | 100/360 [01:17<03:19,  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': 0.36503931879997253, 'eval_runtime': 0.1049, 'eval_samples_per_second': 438.538, 'eval_steps_per_second': 9.533, 'epoch': 2.5}
+{'loss': 0.3605, 'grad_norm': 1.3100874423980713, 'learning_rate': 7.685714285714286e-05, 'epoch': 2.53}
+{'loss': 0.3473, 'grad_norm': 1.0220280885696411, 'learning_rate': 7.657142857142857e-05, 'epoch': 2.56}
+{'loss': 0.3631, 'grad_norm': 1.4616146087646484, 'learning_rate': 7.62857142857143e-05, 'epoch': 2.58}
+{'loss': 0.3426, 'grad_norm': 1.0069721937179565, 'learning_rate': 7.6e-05, 'epoch': 2.61}
+{'loss': 0.3367, 'grad_norm': 0.8820877075195312, 'learning_rate': 7.571428571428571e-05, 'epoch': 2.64}
+{'loss': 0.3553, 'grad_norm': 1.2850284576416016, 'learning_rate': 7.542857142857144e-05, 'epoch': 2.67}
+{'loss': 0.3436, 'grad_norm': 0.8124020099639893, 'learning_rate': 7.514285714285715e-05, 'epoch': 2.69}
+{'loss': 0.3504, 'grad_norm': 0.9645909667015076, 'learning_rate': 7.485714285714285e-05, 'epoch': 2.72}
+{'loss': 0.3586, 'grad_norm': 1.253502368927002, 'learning_rate': 7.457142857142856e-05, 'epoch': 2.75}
+{'loss': 0.3403, 'grad_norm': 0.7083765864372253, 'learning_rate': 7.428571428571429e-05, 'epoch': 2.78}
+ 31%|██████████████████████████████▌                                                                     | 110/360 [01:24<02:56,  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': 0.3600272834300995, 'eval_runtime': 0.1045, 'eval_samples_per_second': 440.303, 'eval_steps_per_second': 9.572, 'epoch': 2.78}
+{'loss': 0.3489, 'grad_norm': 0.8545306921005249, 'learning_rate': 7.4e-05, 'epoch': 2.81}
+{'loss': 0.3383, 'grad_norm': 1.097291350364685, 'learning_rate': 7.371428571428572e-05, 'epoch': 2.83}
+{'loss': 0.3395, 'grad_norm': 0.7652900218963623, 'learning_rate': 7.342857142857144e-05, 'epoch': 2.86}
+{'loss': 0.3513, 'grad_norm': 0.9357858896255493, 'learning_rate': 7.314285714285715e-05, 'epoch': 2.89}
+{'loss': 0.3434, 'grad_norm': 1.1487563848495483, 'learning_rate': 7.285714285714286e-05, 'epoch': 2.92}
+{'loss': 0.3464, 'grad_norm': 1.1131272315979004, 'learning_rate': 7.257142857142858e-05, 'epoch': 2.94}
+{'loss': 0.3484, 'grad_norm': 1.0483183860778809, 'learning_rate': 7.228571428571429e-05, 'epoch': 2.97}
+{'loss': 0.3533, 'grad_norm': 1.248903751373291, 'learning_rate': 7.2e-05, 'epoch': 3.0}
+{'loss': 0.3389, 'grad_norm': 0.8581050634384155, 'learning_rate': 7.171428571428572e-05, 'epoch': 3.03}
+{'loss': 0.347, 'grad_norm': 1.1233102083206177, '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': 0.3557908535003662, 'eval_runtime': 0.1045, 'eval_samples_per_second': 440.002, 'eval_steps_per_second': 9.565, 'epoch': 3.06}
+{'loss': 0.3425, 'grad_norm': 1.056329369544983, 'learning_rate': 7.114285714285714e-05, 'epoch': 3.08}
+{'loss': 0.3398, 'grad_norm': 0.6349842548370361, 'learning_rate': 7.085714285714285e-05, 'epoch': 3.11}
+{'loss': 0.3413, 'grad_norm': 1.2684061527252197, 'learning_rate': 7.057142857142858e-05, 'epoch': 3.14}
+{'loss': 0.3504, 'grad_norm': 0.9766147136688232, 'learning_rate': 7.028571428571428e-05, 'epoch': 3.17}
+{'loss': 0.3461, 'grad_norm': 1.1366608142852783, 'learning_rate': 7e-05, 'epoch': 3.19}
+{'loss': 0.3441, 'grad_norm': 1.1576147079467773, 'learning_rate': 6.971428571428572e-05, 'epoch': 3.22}
+{'loss': 0.3382, 'grad_norm': 0.896751880645752, 'learning_rate': 6.942857142857143e-05, 'epoch': 3.25}
+{'loss': 0.3406, 'grad_norm': 0.7345131039619446, 'learning_rate': 6.914285714285715e-05, 'epoch': 3.28}
+{'loss': 0.34, 'grad_norm': 0.8541216254234314, 'learning_rate': 6.885714285714286e-05, 'epoch': 3.31}
+{'loss': 0.3389, 'grad_norm': 0.6585732102394104, 'learning_rate': 6.857142857142858e-05, 'epoch': 3.33}
+ 36%|████████████████████████████████████                                                                | 130/360 [01:40<02:56,  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': 0.35226884484291077, 'eval_runtime': 0.1048, 'eval_samples_per_second': 438.732, 'eval_steps_per_second': 9.538, 'epoch': 3.33}
+{'loss': 0.3462, 'grad_norm': 1.3044718503952026, 'learning_rate': 6.828571428571429e-05, 'epoch': 3.36}
+{'loss': 0.3374, 'grad_norm': 0.6839655041694641, 'learning_rate': 6.800000000000001e-05, 'epoch': 3.39}
+{'loss': 0.3461, 'grad_norm': 1.3587740659713745, 'learning_rate': 6.771428571428572e-05, 'epoch': 3.42}
+{'loss': 0.3409, 'grad_norm': 1.2411038875579834, 'learning_rate': 6.742857142857143e-05, 'epoch': 3.44}
+{'loss': 0.3337, 'grad_norm': 1.1246317625045776, 'learning_rate': 6.714285714285714e-05, 'epoch': 3.47}
+{'loss': 0.3363, 'grad_norm': 0.8248189687728882, 'learning_rate': 6.685714285714286e-05, 'epoch': 3.5}
+{'loss': 0.3421, 'grad_norm': 0.9192841649055481, 'learning_rate': 6.657142857142857e-05, 'epoch': 3.53}
+{'loss': 0.3418, 'grad_norm': 1.1318869590759277, 'learning_rate': 6.628571428571428e-05, 'epoch': 3.56}
+{'loss': 0.342, 'grad_norm': 1.0143375396728516, 'learning_rate': 6.6e-05, 'epoch': 3.58}
+{'loss': 0.3321, 'grad_norm': 0.726323127746582, 'learning_rate': 6.571428571428571e-05, 'epoch': 3.61}
+ 39%|██████████████████████████████████████▉                                                             | 140/360 [01:48<02:51,  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': 0.3512151837348938, 'eval_runtime': 0.1051, 'eval_samples_per_second': 437.793, 'eval_steps_per_second': 9.517, 'epoch': 3.61}
+{'loss': 0.3379, 'grad_norm': 0.9356390833854675, 'learning_rate': 6.542857142857142e-05, 'epoch': 3.64}
+{'loss': 0.334, 'grad_norm': 1.051357388496399, 'learning_rate': 6.514285714285715e-05, 'epoch': 3.67}
+{'loss': 0.3433, 'grad_norm': 0.7245939373970032, 'learning_rate': 6.485714285714286e-05, 'epoch': 3.69}
+{'loss': 0.333, 'grad_norm': 1.5335993766784668, 'learning_rate': 6.457142857142856e-05, 'epoch': 3.72}
+{'loss': 0.3345, 'grad_norm': 0.6289878487586975, 'learning_rate': 6.428571428571429e-05, 'epoch': 3.75}
+{'loss': 0.3416, 'grad_norm': 0.7878074645996094, 'learning_rate': 6.400000000000001e-05, 'epoch': 3.78}
+{'loss': 0.3436, 'grad_norm': 0.8280585408210754, 'learning_rate': 6.371428571428572e-05, 'epoch': 3.81}
+{'loss': 0.3365, 'grad_norm': 0.7266003489494324, 'learning_rate': 6.342857142857143e-05, 'epoch': 3.83}
+{'loss': 0.3426, 'grad_norm': 1.3992286920547485, 'learning_rate': 6.314285714285715e-05, 'epoch': 3.86}
+{'loss': 0.3332, 'grad_norm': 0.7563316822052002, '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': 0.3488697409629822, 'eval_runtime': 0.105, 'eval_samples_per_second': 437.99, 'eval_steps_per_second': 9.522, 'epoch': 3.89}
+{'loss': 0.3313, 'grad_norm': 0.896052896976471, 'learning_rate': 6.257142857142857e-05, 'epoch': 3.92}
+{'loss': 0.3365, 'grad_norm': 1.2708873748779297, 'learning_rate': 6.22857142857143e-05, 'epoch': 3.94}
+{'loss': 0.3357, 'grad_norm': 1.7695562839508057, 'learning_rate': 6.2e-05, 'epoch': 3.97}
+{'loss': 0.3424, 'grad_norm': 1.4549791812896729, 'learning_rate': 6.171428571428571e-05, 'epoch': 4.0}
+{'loss': 0.3375, 'grad_norm': 0.9521704912185669, 'learning_rate': 6.142857142857143e-05, 'epoch': 4.03}
+{'loss': 0.3361, 'grad_norm': 1.7584255933761597, 'learning_rate': 6.114285714285714e-05, 'epoch': 4.06}
+{'loss': 0.3346, 'grad_norm': 0.7483154535293579, 'learning_rate': 6.085714285714286e-05, 'epoch': 4.08}
+{'loss': 0.3274, 'grad_norm': 1.067897081375122, 'learning_rate': 6.0571428571428576e-05, 'epoch': 4.11}
+{'loss': 0.3257, 'grad_norm': 0.9337431788444519, 'learning_rate': 6.028571428571429e-05, 'epoch': 4.14}
+{'loss': 0.3356, 'grad_norm': 0.9089264273643494, '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': 0.35016533732414246, 'eval_runtime': 0.1051, 'eval_samples_per_second': 437.834, 'eval_steps_per_second': 9.518, 'epoch': 4.17}
+{'loss': 0.3379, 'grad_norm': 1.1468346118927002, 'learning_rate': 5.9714285714285724e-05, 'epoch': 4.19}
+{'loss': 0.3323, 'grad_norm': 1.1817346811294556, 'learning_rate': 5.9428571428571434e-05, 'epoch': 4.22}
+{'loss': 0.336, 'grad_norm': 1.4314862489700317, 'learning_rate': 5.914285714285714e-05, 'epoch': 4.25}
+{'loss': 0.3362, 'grad_norm': 1.1927298307418823, 'learning_rate': 5.885714285714285e-05, 'epoch': 4.28}
+{'loss': 0.3304, 'grad_norm': 0.8057120442390442, 'learning_rate': 5.8571428571428575e-05, 'epoch': 4.31}
+{'loss': 0.3403, 'grad_norm': 1.638046145439148, 'learning_rate': 5.828571428571429e-05, 'epoch': 4.33}
+{'loss': 0.332, 'grad_norm': 0.874083399772644, 'learning_rate': 5.8e-05, 'epoch': 4.36}
+{'loss': 0.3204, 'grad_norm': 0.7660898566246033, 'learning_rate': 5.771428571428572e-05, 'epoch': 4.39}
+{'loss': 0.3413, 'grad_norm': 1.310951828956604, 'learning_rate': 5.742857142857143e-05, 'epoch': 4.42}
+{'loss': 0.3356, 'grad_norm': 0.7157622575759888, 'learning_rate': 5.714285714285714e-05, 'epoch': 4.44}
+ 47%|███████████████████████████████████████████████▏                                                    | 170/360 [02:11<02:26,  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': 0.35388267040252686, 'eval_runtime': 0.1051, 'eval_samples_per_second': 437.501, 'eval_steps_per_second': 9.511, 'epoch': 4.44}
+{'loss': 0.3373, 'grad_norm': 0.996565043926239, 'learning_rate': 5.6857142857142865e-05, 'epoch': 4.47}
+{'loss': 0.3314, 'grad_norm': 0.6834873557090759, 'learning_rate': 5.6571428571428574e-05, 'epoch': 4.5}
+{'loss': 0.3298, 'grad_norm': 1.1225472688674927, 'learning_rate': 5.628571428571428e-05, 'epoch': 4.53}
+{'loss': 0.3355, 'grad_norm': 1.5352915525436401, 'learning_rate': 5.6000000000000006e-05, 'epoch': 4.56}
+{'loss': 0.3359, 'grad_norm': 0.8244190216064453, 'learning_rate': 5.571428571428572e-05, 'epoch': 4.58}
+{'loss': 0.3344, 'grad_norm': 1.3828486204147339, 'learning_rate': 5.542857142857143e-05, 'epoch': 4.61}
+{'loss': 0.3278, 'grad_norm': 0.8235058784484863, 'learning_rate': 5.514285714285714e-05, 'epoch': 4.64}
+{'loss': 0.3268, 'grad_norm': 0.7643269300460815, 'learning_rate': 5.485714285714286e-05, 'epoch': 4.67}
+{'loss': 0.3318, 'grad_norm': 0.8052845001220703, 'learning_rate': 5.457142857142857e-05, 'epoch': 4.69}
+{'loss': 0.3372, 'grad_norm': 1.0252516269683838, 'learning_rate': 5.428571428571428e-05, 'epoch': 4.72}
+ 50%|██████████████████████████████████████████████████                                                  | 180/360 [02:18<01:57,  1.53it/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': 0.3490443825721741, 'eval_runtime': 0.1055, 'eval_samples_per_second': 436.194, 'eval_steps_per_second': 9.482, 'epoch': 4.72}
+{'loss': 0.3373, 'grad_norm': 1.2792561054229736, 'learning_rate': 5.4000000000000005e-05, 'epoch': 4.75}
+{'loss': 0.3346, 'grad_norm': 1.143295407295227, 'learning_rate': 5.3714285714285714e-05, 'epoch': 4.78}
+{'loss': 0.3354, 'grad_norm': 1.3516252040863037, 'learning_rate': 5.342857142857143e-05, 'epoch': 4.81}
+{'loss': 0.3379, 'grad_norm': 1.4615265130996704, 'learning_rate': 5.314285714285715e-05, 'epoch': 4.83}
+{'loss': 0.3291, 'grad_norm': 0.9049354195594788, 'learning_rate': 5.285714285714286e-05, 'epoch': 4.86}
+{'loss': 0.3309, 'grad_norm': 1.4232890605926514, 'learning_rate': 5.257142857142857e-05, 'epoch': 4.89}
+{'loss': 0.3297, 'grad_norm': 0.7629730701446533, 'learning_rate': 5.2285714285714294e-05, 'epoch': 4.92}
+{'loss': 0.3424, 'grad_norm': 1.2539515495300293, 'learning_rate': 5.2000000000000004e-05, 'epoch': 4.94}
+{'loss': 0.3331, 'grad_norm': 1.3451244831085205, 'learning_rate': 5.171428571428571e-05, 'epoch': 4.97}
+{'loss': 0.3315, 'grad_norm': 1.4241085052490234, 'learning_rate': 5.142857142857143e-05, 'epoch': 5.0}
+ 53%|████████████████████████████████████████████████████▊                                               | 190/360 [02:26<02:11,  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': 0.35601839423179626, 'eval_runtime': 0.1052, 'eval_samples_per_second': 437.465, 'eval_steps_per_second': 9.51, 'epoch': 5.0}
+{'loss': 0.3266, 'grad_norm': 1.3330923318862915, 'learning_rate': 5.1142857142857145e-05, 'epoch': 5.03}
+{'loss': 0.3247, 'grad_norm': 0.8518350124359131, 'learning_rate': 5.085714285714286e-05, 'epoch': 5.06}
+{'loss': 0.3322, 'grad_norm': 1.5773379802703857, 'learning_rate': 5.057142857142857e-05, 'epoch': 5.08}
+{'loss': 0.3361, 'grad_norm': 1.9808964729309082, 'learning_rate': 5.028571428571429e-05, 'epoch': 5.11}
+{'loss': 0.3326, 'grad_norm': 0.6624701023101807, 'learning_rate': 5e-05, 'epoch': 5.14}
+{'loss': 0.3294, 'grad_norm': 1.3144952058792114, 'learning_rate': 4.971428571428572e-05, 'epoch': 5.17}
+{'loss': 0.3249, 'grad_norm': 1.4999899864196777, 'learning_rate': 4.942857142857143e-05, 'epoch': 5.19}
+{'loss': 0.3295, 'grad_norm': 0.6328815817832947, 'learning_rate': 4.9142857142857144e-05, 'epoch': 5.22}
+{'loss': 0.3325, 'grad_norm': 1.7996755838394165, 'learning_rate': 4.885714285714286e-05, 'epoch': 5.25}
+{'loss': 0.3294, 'grad_norm': 1.4809287786483765, 'learning_rate': 4.8571428571428576e-05, 'epoch': 5.28}
+ 56%|███████████████████████████████████████████████████████▌                                            | 200/360 [02:34<02:03,  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': 0.34694546461105347, 'eval_runtime': 0.1055, 'eval_samples_per_second': 435.923, 'eval_steps_per_second': 9.477, 'epoch': 5.28}
+{'loss': 0.3256, 'grad_norm': 1.0571750402450562, 'learning_rate': 4.828571428571429e-05, 'epoch': 5.31}
+{'loss': 0.3302, 'grad_norm': 1.0487523078918457, 'learning_rate': 4.8e-05, 'epoch': 5.33}
+{'loss': 0.3261, 'grad_norm': 0.762173593044281, 'learning_rate': 4.771428571428572e-05, 'epoch': 5.36}
+{'loss': 0.3267, 'grad_norm': 0.8178277611732483, 'learning_rate': 4.742857142857143e-05, 'epoch': 5.39}
+{'loss': 0.326, 'grad_norm': 1.104577898979187, 'learning_rate': 4.714285714285714e-05, 'epoch': 5.42}
+{'loss': 0.3308, 'grad_norm': 0.7541172504425049, 'learning_rate': 4.685714285714286e-05, 'epoch': 5.44}
+{'loss': 0.3323, 'grad_norm': 1.1193814277648926, 'learning_rate': 4.6571428571428575e-05, 'epoch': 5.47}
+{'loss': 0.3251, 'grad_norm': 0.6032533049583435, 'learning_rate': 4.628571428571429e-05, 'epoch': 5.5}
+{'loss': 0.3287, 'grad_norm': 0.7051272988319397, 'learning_rate': 4.600000000000001e-05, 'epoch': 5.53}
+{'loss': 0.3274, 'grad_norm': 0.8189981579780579, 'learning_rate': 4.5714285714285716e-05, 'epoch': 5.56}
+ 58%|██████████████████████████████████████████████████████████▎                                         | 210/360 [02:41<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': 0.34395119547843933, 'eval_runtime': 0.1046, 'eval_samples_per_second': 439.963, 'eval_steps_per_second': 9.564, 'epoch': 5.56}
+{'loss': 0.3243, 'grad_norm': 0.7370961904525757, 'learning_rate': 4.542857142857143e-05, 'epoch': 5.58}
+{'loss': 0.3323, 'grad_norm': 0.9540398716926575, 'learning_rate': 4.514285714285714e-05, 'epoch': 5.61}
+{'loss': 0.3282, 'grad_norm': 0.6874322891235352, 'learning_rate': 4.485714285714286e-05, 'epoch': 5.64}
+{'loss': 0.3328, 'grad_norm': 0.8128199577331543, 'learning_rate': 4.4571428571428574e-05, 'epoch': 5.67}
+{'loss': 0.3257, 'grad_norm': 0.9183529615402222, 'learning_rate': 4.428571428571428e-05, 'epoch': 5.69}
+{'loss': 0.3284, 'grad_norm': 0.9220953583717346, 'learning_rate': 4.4000000000000006e-05, 'epoch': 5.72}
+{'loss': 0.3304, 'grad_norm': 1.4376723766326904, 'learning_rate': 4.371428571428572e-05, 'epoch': 5.75}
+{'loss': 0.33, 'grad_norm': 1.0230571031570435, 'learning_rate': 4.342857142857143e-05, 'epoch': 5.78}
+{'loss': 0.335, 'grad_norm': 1.4990969896316528, 'learning_rate': 4.314285714285715e-05, 'epoch': 5.81}
+{'loss': 0.3375, 'grad_norm': 0.9291542172431946, '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': 0.35131019353866577, 'eval_runtime': 0.1053, 'eval_samples_per_second': 436.997, 'eval_steps_per_second': 9.5, 'epoch': 5.83}
+{'loss': 0.3372, 'grad_norm': 2.625933885574341, 'learning_rate': 4.257142857142857e-05, 'epoch': 5.86}
+{'loss': 0.3284, 'grad_norm': 1.0259548425674438, 'learning_rate': 4.228571428571429e-05, 'epoch': 5.89}
+{'loss': 0.329, 'grad_norm': 1.4596794843673706, 'learning_rate': 4.2e-05, 'epoch': 5.92}
+{'loss': 0.3268, 'grad_norm': 1.1902427673339844, 'learning_rate': 4.1714285714285714e-05, 'epoch': 5.94}
+{'loss': 0.3336, 'grad_norm': 1.7438991069793701, 'learning_rate': 4.1428571428571437e-05, 'epoch': 5.97}
+{'loss': 0.3338, 'grad_norm': 2.127594470977783, 'learning_rate': 4.1142857142857146e-05, 'epoch': 6.0}
+{'loss': 0.3351, 'grad_norm': 2.1449036598205566, 'learning_rate': 4.085714285714286e-05, 'epoch': 6.03}
+{'loss': 0.3351, 'grad_norm': 2.090463161468506, 'learning_rate': 4.057142857142857e-05, 'epoch': 6.06}
+{'loss': 0.3289, 'grad_norm': 1.173915147781372, 'learning_rate': 4.028571428571429e-05, 'epoch': 6.08}
+{'loss': 0.3286, 'grad_norm': 1.5039347410202026, 'learning_rate': 4e-05, 'epoch': 6.11}
+ 64%|███████████████████████████████████████████████████████████████▉                                    | 230/360 [02:57<01:39,  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': 0.3502294719219208, 'eval_runtime': 0.105, 'eval_samples_per_second': 438.259, 'eval_steps_per_second': 9.527, 'epoch': 6.11}
+{'loss': 0.327, 'grad_norm': 1.5905152559280396, 'learning_rate': 3.971428571428571e-05, 'epoch': 6.14}
+{'loss': 0.3266, 'grad_norm': 1.1944066286087036, 'learning_rate': 3.942857142857143e-05, 'epoch': 6.17}
+{'loss': 0.3309, 'grad_norm': 1.469502329826355, 'learning_rate': 3.9142857142857145e-05, 'epoch': 6.19}
+{'loss': 0.3343, 'grad_norm': 2.0030901432037354, 'learning_rate': 3.885714285714286e-05, 'epoch': 6.22}
+{'loss': 0.3275, 'grad_norm': 1.3164979219436646, 'learning_rate': 3.857142857142858e-05, 'epoch': 6.25}
+{'loss': 0.3277, 'grad_norm': 0.7995334267616272, 'learning_rate': 3.8285714285714286e-05, 'epoch': 6.28}
+{'loss': 0.3242, 'grad_norm': 1.2250066995620728, 'learning_rate': 3.8e-05, 'epoch': 6.31}
+{'loss': 0.3313, 'grad_norm': 1.2443605661392212, 'learning_rate': 3.771428571428572e-05, 'epoch': 6.33}
+{'loss': 0.3221, 'grad_norm': 0.8322944641113281, 'learning_rate': 3.742857142857143e-05, 'epoch': 6.36}
+{'loss': 0.3331, 'grad_norm': 0.8514747023582458, 'learning_rate': 3.7142857142857143e-05, 'epoch': 6.39}
+ 67%|██████████████████████████████████████████████████████████████████▋                                 | 240/360 [03:05<01:32,  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': 0.34429529309272766, 'eval_runtime': 0.1052, 'eval_samples_per_second': 437.304, 'eval_steps_per_second': 9.507, 'epoch': 6.39}
+{'loss': 0.3266, 'grad_norm': 0.9474199414253235, 'learning_rate': 3.685714285714286e-05, 'epoch': 6.42}
+{'loss': 0.3285, 'grad_norm': 1.4075018167495728, 'learning_rate': 3.6571428571428576e-05, 'epoch': 6.44}
+{'loss': 0.3302, 'grad_norm': 1.2337582111358643, 'learning_rate': 3.628571428571429e-05, 'epoch': 6.47}
+{'loss': 0.317, 'grad_norm': 0.7781947255134583, 'learning_rate': 3.6e-05, 'epoch': 6.5}
+{'loss': 0.3211, 'grad_norm': 1.1268919706344604, 'learning_rate': 3.571428571428572e-05, 'epoch': 6.53}
+{'loss': 0.3253, 'grad_norm': 1.742037057876587, 'learning_rate': 3.5428571428571426e-05, 'epoch': 6.56}
+{'loss': 0.3173, 'grad_norm': 0.7068340182304382, 'learning_rate': 3.514285714285714e-05, 'epoch': 6.58}
+{'loss': 0.3218, 'grad_norm': 1.5968836545944214, 'learning_rate': 3.485714285714286e-05, 'epoch': 6.61}
+{'loss': 0.3298, 'grad_norm': 2.1051909923553467, 'learning_rate': 3.4571428571428574e-05, 'epoch': 6.64}
+{'loss': 0.3196, 'grad_norm': 0.9735121130943298, 'learning_rate': 3.428571428571429e-05, 'epoch': 6.67}
+ 69%|████████████████████████████████████████████████████████████████████████████████████▋                                     | 250/360 [03:12<01:24,  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': 0.35019850730895996, 'eval_runtime': 0.1049, 'eval_samples_per_second': 438.63, 'eval_steps_per_second': 9.535, 'epoch': 6.67}
+{'loss': 0.3303, 'grad_norm': 1.281566858291626, 'learning_rate': 3.4000000000000007e-05, 'epoch': 6.69}
+{'loss': 0.3309, 'grad_norm': 1.067278265953064, 'learning_rate': 3.3714285714285716e-05, 'epoch': 6.72}
+{'loss': 0.3246, 'grad_norm': 1.42350435256958, 'learning_rate': 3.342857142857143e-05, 'epoch': 6.75}
+{'loss': 0.3212, 'grad_norm': 1.0641676187515259, 'learning_rate': 3.314285714285714e-05, 'epoch': 6.78}
+{'loss': 0.3231, 'grad_norm': 1.0593169927597046, 'learning_rate': 3.285714285714286e-05, 'epoch': 6.81}
+{'loss': 0.3186, 'grad_norm': 1.6785848140716553, 'learning_rate': 3.257142857142857e-05, 'epoch': 6.83}
+{'loss': 0.3337, 'grad_norm': 1.6702518463134766, 'learning_rate': 3.228571428571428e-05, 'epoch': 6.86}
+{'loss': 0.3285, 'grad_norm': 1.1052055358886719, 'learning_rate': 3.2000000000000005e-05, 'epoch': 6.89}
+{'loss': 0.3258, 'grad_norm': 1.2286932468414307, 'learning_rate': 3.1714285714285715e-05, 'epoch': 6.92}
+{'loss': 0.3264, 'grad_norm': 1.6575883626937866, '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': 0.343667209148407, 'eval_runtime': 0.105, 'eval_samples_per_second': 438.131, 'eval_steps_per_second': 9.525, 'epoch': 6.94}
+{'loss': 0.319, 'grad_norm': 1.004503607749939, 'learning_rate': 3.114285714285715e-05, 'epoch': 6.97}
+{'loss': 0.3211, 'grad_norm': 1.1171271800994873, 'learning_rate': 3.0857142857142856e-05, 'epoch': 7.0}
+{'loss': 0.3184, 'grad_norm': 0.989422619342804, 'learning_rate': 3.057142857142857e-05, 'epoch': 7.03}
+{'loss': 0.315, 'grad_norm': 0.727877676486969, 'learning_rate': 3.0285714285714288e-05, 'epoch': 7.06}
+{'loss': 0.3207, 'grad_norm': 0.7329328060150146, 'learning_rate': 3e-05, 'epoch': 7.08}
+{'loss': 0.3259, 'grad_norm': 1.305487871170044, 'learning_rate': 2.9714285714285717e-05, 'epoch': 7.11}
+{'loss': 0.3247, 'grad_norm': 0.8338795900344849, 'learning_rate': 2.9428571428571426e-05, 'epoch': 7.14}
+{'loss': 0.3192, 'grad_norm': 0.5776103138923645, 'learning_rate': 2.9142857142857146e-05, 'epoch': 7.17}
+{'loss': 0.3221, 'grad_norm': 1.2487804889678955, 'learning_rate': 2.885714285714286e-05, 'epoch': 7.19}
+{'loss': 0.3204, 'grad_norm': 0.954796314239502, 'learning_rate': 2.857142857142857e-05, 'epoch': 7.22}
+ 75%|███████████████████████████████████████████████████████████████████████████████████████████▌                              | 270/360 [03:28<01:09,  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': 0.3448273539543152, 'eval_runtime': 0.1048, 'eval_samples_per_second': 439.048, 'eval_steps_per_second': 9.545, 'epoch': 7.22}
+{'loss': 0.3159, 'grad_norm': 1.0453810691833496, 'learning_rate': 2.8285714285714287e-05, 'epoch': 7.25}
+{'loss': 0.32, 'grad_norm': 0.6628478169441223, 'learning_rate': 2.8000000000000003e-05, 'epoch': 7.28}
+{'loss': 0.3227, 'grad_norm': 2.0091917514801025, 'learning_rate': 2.7714285714285716e-05, 'epoch': 7.31}
+{'loss': 0.3227, 'grad_norm': 1.5404771566390991, 'learning_rate': 2.742857142857143e-05, 'epoch': 7.33}
+{'loss': 0.3221, 'grad_norm': 0.8551548719406128, 'learning_rate': 2.714285714285714e-05, 'epoch': 7.36}
+{'loss': 0.3176, 'grad_norm': 0.816592812538147, 'learning_rate': 2.6857142857142857e-05, 'epoch': 7.39}
+{'loss': 0.328, 'grad_norm': 1.5894920825958252, 'learning_rate': 2.6571428571428576e-05, 'epoch': 7.42}
+{'loss': 0.3227, 'grad_norm': 1.1515345573425293, 'learning_rate': 2.6285714285714286e-05, 'epoch': 7.44}
+{'loss': 0.311, 'grad_norm': 0.6154195070266724, 'learning_rate': 2.6000000000000002e-05, 'epoch': 7.47}
+{'loss': 0.3195, 'grad_norm': 0.7660549879074097, 'learning_rate': 2.5714285714285714e-05, 'epoch': 7.5}
+ 78%|██████████████████████████████████████████████████████████████████████████████████████████████▉                           | 280/360 [03:35<01:00,  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': 0.3412049412727356, 'eval_runtime': 0.1048, 'eval_samples_per_second': 438.732, 'eval_steps_per_second': 9.538, 'epoch': 7.5}
+{'loss': 0.3192, 'grad_norm': 1.6722830533981323, 'learning_rate': 2.542857142857143e-05, 'epoch': 7.53}
+{'loss': 0.3262, 'grad_norm': 1.0368938446044922, 'learning_rate': 2.5142857142857147e-05, 'epoch': 7.56}
+{'loss': 0.3247, 'grad_norm': 1.0102812051773071, 'learning_rate': 2.485714285714286e-05, 'epoch': 7.58}
+{'loss': 0.3253, 'grad_norm': 0.9800832271575928, 'learning_rate': 2.4571428571428572e-05, 'epoch': 7.61}
+{'loss': 0.3116, 'grad_norm': 0.741663932800293, 'learning_rate': 2.4285714285714288e-05, 'epoch': 7.64}
+{'loss': 0.3195, 'grad_norm': 1.0429283380508423, 'learning_rate': 2.4e-05, 'epoch': 7.67}
+{'loss': 0.3178, 'grad_norm': 0.8469560146331787, 'learning_rate': 2.3714285714285717e-05, 'epoch': 7.69}
+{'loss': 0.3154, 'grad_norm': 0.8006932139396667, 'learning_rate': 2.342857142857143e-05, 'epoch': 7.72}
+{'loss': 0.3203, 'grad_norm': 0.8269030451774597, 'learning_rate': 2.3142857142857145e-05, 'epoch': 7.75}
+{'loss': 0.314, 'grad_norm': 0.7541152834892273, 'learning_rate': 2.2857142857142858e-05, 'epoch': 7.78}
+ 81%|██████████████████████████████████████████████████████████████████████████████████████████████████▎                       | 290/360 [03:43<00:49,  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': 0.3413321375846863, 'eval_runtime': 0.1045, 'eval_samples_per_second': 440.216, 'eval_steps_per_second': 9.57, 'epoch': 7.78}
+{'loss': 0.3201, 'grad_norm': 1.1791499853134155, 'learning_rate': 2.257142857142857e-05, 'epoch': 7.81}
+{'loss': 0.3164, 'grad_norm': 0.8804281949996948, 'learning_rate': 2.2285714285714287e-05, 'epoch': 7.83}
+{'loss': 0.3194, 'grad_norm': 0.7627995014190674, 'learning_rate': 2.2000000000000003e-05, 'epoch': 7.86}
+{'loss': 0.3272, 'grad_norm': 1.013617753982544, 'learning_rate': 2.1714285714285715e-05, 'epoch': 7.89}
+{'loss': 0.3184, 'grad_norm': 0.9311711192131042, 'learning_rate': 2.1428571428571428e-05, 'epoch': 7.92}
+{'loss': 0.3218, 'grad_norm': 0.9507532119750977, 'learning_rate': 2.1142857142857144e-05, 'epoch': 7.94}
+{'loss': 0.3245, 'grad_norm': 0.8252774477005005, 'learning_rate': 2.0857142857142857e-05, 'epoch': 7.97}
+{'loss': 0.316, 'grad_norm': 0.9303626418113708, 'learning_rate': 2.0571428571428573e-05, 'epoch': 8.0}
+{'loss': 0.3141, 'grad_norm': 1.1884357929229736, 'learning_rate': 2.0285714285714286e-05, 'epoch': 8.03}
+{'loss': 0.3211, 'grad_norm': 0.9842406511306763, 'learning_rate': 2e-05, 'epoch': 8.06}
+ 83%|█████████████████████████████████████████████████████████████████████████████████████████████████████▋                    | 300/360 [03:51<00:46,  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': 0.34589651226997375, 'eval_runtime': 0.1044, 'eval_samples_per_second': 440.747, 'eval_steps_per_second': 9.581, 'epoch': 8.06}
+{'loss': 0.3111, 'grad_norm': 1.2224154472351074, 'learning_rate': 1.9714285714285714e-05, 'epoch': 8.08}
+{'loss': 0.3175, 'grad_norm': 0.9548574090003967, 'learning_rate': 1.942857142857143e-05, 'epoch': 8.11}
+{'loss': 0.3178, 'grad_norm': 0.8031886219978333, 'learning_rate': 1.9142857142857143e-05, 'epoch': 8.14}
+{'loss': 0.3189, 'grad_norm': 0.8493176698684692, 'learning_rate': 1.885714285714286e-05, 'epoch': 8.17}
+{'loss': 0.3119, 'grad_norm': 0.8125531077384949, 'learning_rate': 1.8571428571428572e-05, 'epoch': 8.19}
+{'loss': 0.3113, 'grad_norm': 0.6368350982666016, 'learning_rate': 1.8285714285714288e-05, 'epoch': 8.22}
+{'loss': 0.318, 'grad_norm': 1.0392847061157227, 'learning_rate': 1.8e-05, 'epoch': 8.25}
+{'loss': 0.3234, 'grad_norm': 0.7722540497779846, 'learning_rate': 1.7714285714285713e-05, 'epoch': 8.28}
+{'loss': 0.3172, 'grad_norm': 0.740352988243103, 'learning_rate': 1.742857142857143e-05, 'epoch': 8.31}
+{'loss': 0.3161, 'grad_norm': 1.4656850099563599, 'learning_rate': 1.7142857142857145e-05, 'epoch': 8.33}
+ 86%|█████████████████████████████████████████████████████████████████████████████████████████████████████████                 | 310/360 [03:58<00:38,  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': 0.3438817262649536, 'eval_runtime': 0.1048, 'eval_samples_per_second': 439.01, 'eval_steps_per_second': 9.544, 'epoch': 8.33}
+{'loss': 0.3154, 'grad_norm': 0.9987584948539734, 'learning_rate': 1.6857142857142858e-05, 'epoch': 8.36}
+{'loss': 0.3081, 'grad_norm': 0.7025516629219055, 'learning_rate': 1.657142857142857e-05, 'epoch': 8.39}
+{'loss': 0.3114, 'grad_norm': 0.6634803414344788, 'learning_rate': 1.6285714285714287e-05, 'epoch': 8.42}
+{'loss': 0.3184, 'grad_norm': 1.4024748802185059, 'learning_rate': 1.6000000000000003e-05, 'epoch': 8.44}
+{'loss': 0.3221, 'grad_norm': 1.4321742057800293, 'learning_rate': 1.5714285714285715e-05, 'epoch': 8.47}
+{'loss': 0.3182, 'grad_norm': 1.052692174911499, 'learning_rate': 1.5428571428571428e-05, 'epoch': 8.5}
+{'loss': 0.3207, 'grad_norm': 1.5870689153671265, 'learning_rate': 1.5142857142857144e-05, 'epoch': 8.53}
+{'loss': 0.3172, 'grad_norm': 1.1841577291488647, 'learning_rate': 1.4857142857142858e-05, 'epoch': 8.56}
+{'loss': 0.3123, 'grad_norm': 0.6852739453315735, 'learning_rate': 1.4571428571428573e-05, 'epoch': 8.58}
+{'loss': 0.3192, 'grad_norm': 0.6914618015289307, 'learning_rate': 1.4285714285714285e-05, 'epoch': 8.61}
+ 89%|████████████████████████████��███████████████████████████████████████████████████████████████████████████████▍             | 320/360 [04:06<00:30,  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': 0.34119004011154175, 'eval_runtime': 0.105, 'eval_samples_per_second': 438.008, 'eval_steps_per_second': 9.522, 'epoch': 8.61}
+{'loss': 0.3165, 'grad_norm': 0.8360849618911743, 'learning_rate': 1.4000000000000001e-05, 'epoch': 8.64}
+{'loss': 0.3147, 'grad_norm': 0.7968160510063171, 'learning_rate': 1.3714285714285716e-05, 'epoch': 8.67}
+{'loss': 0.3146, 'grad_norm': 0.9014184474945068, 'learning_rate': 1.3428571428571429e-05, 'epoch': 8.69}
+{'loss': 0.3124, 'grad_norm': 0.7425902485847473, 'learning_rate': 1.3142857142857143e-05, 'epoch': 8.72}
+{'loss': 0.313, 'grad_norm': 1.4130454063415527, 'learning_rate': 1.2857142857142857e-05, 'epoch': 8.75}
+{'loss': 0.3221, 'grad_norm': 1.8442883491516113, 'learning_rate': 1.2571428571428573e-05, 'epoch': 8.78}
+{'loss': 0.3186, 'grad_norm': 0.8699185252189636, 'learning_rate': 1.2285714285714286e-05, 'epoch': 8.81}
+{'loss': 0.3139, 'grad_norm': 0.7799740433692932, 'learning_rate': 1.2e-05, 'epoch': 8.83}
+{'loss': 0.3121, 'grad_norm': 0.7385247349739075, 'learning_rate': 1.1714285714285715e-05, 'epoch': 8.86}
+{'loss': 0.3119, 'grad_norm': 1.3332445621490479, 'learning_rate': 1.1428571428571429e-05, 'epoch': 8.89}
+ 92%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████▊          | 330/360 [04:14<00:22,  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': 0.3395558297634125, 'eval_runtime': 0.1048, 'eval_samples_per_second': 439.113, 'eval_steps_per_second': 9.546, 'epoch': 8.89}
+{'loss': 0.3145, 'grad_norm': 0.7556406855583191, 'learning_rate': 1.1142857142857143e-05, 'epoch': 8.92}
+{'loss': 0.3093, 'grad_norm': 1.078497290611267, 'learning_rate': 1.0857142857142858e-05, 'epoch': 8.94}
+{'loss': 0.3187, 'grad_norm': 0.6883450150489807, 'learning_rate': 1.0571428571428572e-05, 'epoch': 8.97}
+{'loss': 0.3205, 'grad_norm': 1.3395229578018188, 'learning_rate': 1.0285714285714286e-05, 'epoch': 9.0}
+{'loss': 0.3136, 'grad_norm': 0.8341308832168579, 'learning_rate': 1e-05, 'epoch': 9.03}
+{'loss': 0.3135, 'grad_norm': 0.8046018481254578, 'learning_rate': 9.714285714285715e-06, 'epoch': 9.06}
+{'loss': 0.3118, 'grad_norm': 0.6076436638832092, 'learning_rate': 9.42857142857143e-06, 'epoch': 9.08}
+{'loss': 0.3108, 'grad_norm': 0.8679487705230713, 'learning_rate': 9.142857142857144e-06, 'epoch': 9.11}
+{'loss': 0.3076, 'grad_norm': 0.7275362014770508, 'learning_rate': 8.857142857142857e-06, 'epoch': 9.14}
+{'loss': 0.3153, 'grad_norm': 1.152782678604126, 'learning_rate': 8.571428571428573e-06, 'epoch': 9.17}
+ 94%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████▏      | 340/360 [04:21<00:15,  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': 0.34156906604766846, 'eval_runtime': 0.1043, 'eval_samples_per_second': 440.833, 'eval_steps_per_second': 9.583, 'epoch': 9.17}
+{'loss': 0.3114, 'grad_norm': 1.1763490438461304, 'learning_rate': 8.285714285714285e-06, 'epoch': 9.19}
+{'loss': 0.3091, 'grad_norm': 0.7035480737686157, 'learning_rate': 8.000000000000001e-06, 'epoch': 9.22}
+{'loss': 0.3173, 'grad_norm': 0.7956453561782837, 'learning_rate': 7.714285714285714e-06, 'epoch': 9.25}
+{'loss': 0.3069, 'grad_norm': 0.7433366179466248, 'learning_rate': 7.428571428571429e-06, 'epoch': 9.28}
+{'loss': 0.3142, 'grad_norm': 1.6120795011520386, 'learning_rate': 7.142857142857143e-06, 'epoch': 9.31}
+{'loss': 0.3066, 'grad_norm': 0.7718474268913269, 'learning_rate': 6.857142857142858e-06, 'epoch': 9.33}
+{'loss': 0.3159, 'grad_norm': 0.7094411253929138, 'learning_rate': 6.5714285714285714e-06, 'epoch': 9.36}
+{'loss': 0.3106, 'grad_norm': 0.8119326829910278, 'learning_rate': 6.285714285714287e-06, 'epoch': 9.39}
+{'loss': 0.3084, 'grad_norm': 1.7396577596664429, 'learning_rate': 6e-06, 'epoch': 9.42}
+{'loss': 0.3101, 'grad_norm': 1.0242513418197632, 'learning_rate': 5.7142857142857145e-06, 'epoch': 9.44}
+ 97%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▌   | 350/360 [04:29<00:07,  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': 0.3424239754676819, 'eval_runtime': 0.1047, 'eval_samples_per_second': 439.488, 'eval_steps_per_second': 9.554, 'epoch': 9.44}
+{'loss': 0.3151, 'grad_norm': 1.1069552898406982, 'learning_rate': 5.428571428571429e-06, 'epoch': 9.47}
+{'loss': 0.3063, 'grad_norm': 0.7284579277038574, 'learning_rate': 5.142857142857143e-06, 'epoch': 9.5}
+{'loss': 0.3096, 'grad_norm': 0.6138132214546204, 'learning_rate': 4.857142857142858e-06, 'epoch': 9.53}
+{'loss': 0.3142, 'grad_norm': 0.7576848864555359, 'learning_rate': 4.571428571428572e-06, 'epoch': 9.56}
+{'loss': 0.3111, 'grad_norm': 0.5917118191719055, 'learning_rate': 4.285714285714286e-06, 'epoch': 9.58}
+{'loss': 0.3134, 'grad_norm': 1.052357792854309, 'learning_rate': 4.000000000000001e-06, 'epoch': 9.61}
+{'loss': 0.3057, 'grad_norm': 1.0170456171035767, 'learning_rate': 3.7142857142857146e-06, 'epoch': 9.64}
+{'loss': 0.3078, 'grad_norm': 0.9443879723548889, 'learning_rate': 3.428571428571429e-06, 'epoch': 9.67}
+{'loss': 0.309, 'grad_norm': 0.589635968208313, 'learning_rate': 3.1428571428571433e-06, 'epoch': 9.69}
+{'loss': 0.3151, 'grad_norm': 0.7533226609230042, 'learning_rate': 2.8571428571428573e-06, 'epoch': 9.72}
+100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 360/360 [04:37<00:00,  1.55it/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': 0.34002798795700073, 'eval_runtime': 0.1051, 'eval_samples_per_second': 437.81, 'eval_steps_per_second': 9.518, 'epoch': 9.72}
+{'loss': 0.3066, 'grad_norm': 0.997863233089447, 'learning_rate': 2.5714285714285716e-06, 'epoch': 9.75}
+{'loss': 0.3133, 'grad_norm': 0.7841188907623291, 'learning_rate': 2.285714285714286e-06, 'epoch': 9.78}
+{'loss': 0.3081, 'grad_norm': 0.8386639356613159, 'learning_rate': 2.0000000000000003e-06, 'epoch': 9.81}
+{'loss': 0.3093, 'grad_norm': 0.9777269959449768, 'learning_rate': 1.7142857142857145e-06, 'epoch': 9.83}
+{'loss': 0.3086, 'grad_norm': 0.5559796094894409, 'learning_rate': 1.4285714285714286e-06, 'epoch': 9.86}
+{'loss': 0.3134, 'grad_norm': 0.8345446586608887, 'learning_rate': 1.142857142857143e-06, 'epoch': 9.89}
+{'loss': 0.3085, 'grad_norm': 0.6907045245170593, 'learning_rate': 8.571428571428572e-07, 'epoch': 9.92}
+{'loss': 0.3112, 'grad_norm': 0.9042662978172302, 'learning_rate': 5.714285714285715e-07, 'epoch': 9.94}
+{'loss': 0.3134, 'grad_norm': 0.790155291557312, 'learning_rate': 2.8571428571428575e-07, 'epoch': 9.97}
+{'loss': 0.3109, 'grad_norm': 0.983366847038269, 'learning_rate': 0.0, 'epoch': 10.0}
+100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 360/360 [04:39<00:00,  1.29it/s]
+                                                                                                                                                                 
+{'eval_loss': 0.3411920666694641, 'eval_runtime': 0.1051, 'eval_samples_per_second': 437.492, 'eval_steps_per_second': 9.511, 'epoch': 10.0}
+{'train_runtime': 279.0126, 'train_samples_per_second': 162.681, 'train_steps_per_second': 1.29, 'train_loss': 0.37004996778236493, 'epoch': 10.0}

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/files/wandb-metadata.json

@@ -0,0 +1,49 @@
+{
+  "os": "Linux-6.8.0-101-generic-x86_64-with-glibc2.35",
+  "python": "3.9.20",
+  "startedAt": "2026-03-27T06:32:20.617442Z",
+  "program": "/tmp2/chyang/workspace/LLM-BC/./llmbc/workspace/train_llm_workspace.py",
+  "codePath": "llmbc/workspace/train_llm_workspace.py",
+  "git": {
+    "remote": "https://github.com/CHYang25/LLM-BC.git",
+    "commit": "2e811563d66038dc86e99909ecf9bbd3f9974991"
+  },
+  "email": "chris920325@gmail.com",
+  "root": "/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1",
+  "host": "A6000-2",
+  "username": "chyang",
+  "executable": "/home/chyang/miniconda3/envs/llm-bc/bin/python3",
+  "codePathLocal": "llmbc/workspace/train_llm_workspace.py",
+  "cpu_count": 12,
+  "cpu_count_logical": 24,
+  "gpu": "NVIDIA RTX A6000",
+  "gpu_count": 2,
+  "disk": {
+    "/": {
+      "total": "1967317549056",
+      "used": "785037066240"
+    }
+  },
+  "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"
+}

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/files/wandb-summary.json

@@ -0,0 +1 @@
+{"train_steps_per_second":1.29,"eval/steps_per_second":9.511,"train/loss":0.3109,"train/epoch":10,"_step":396,"eval/loss":0.3411920666694641,"_wandb":{"runtime":286},"eval/runtime":0.1051,"total_flos":3.83135455600128e+15,"train/grad_norm":0.983366847038269,"train_samples_per_second":162.681,"train/learning_rate":0,"_runtime":286.405640629,"train_loss":0.37004996778236493,"train_runtime":279.0126,"eval/samples_per_second":437.492,"_timestamp":1.7745934270229735e+09,"train/global_step":360}

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/logs/debug-core.log

@@ -0,0 +1,16 @@
+{"time":"2026-03-27T14:32:19.977209485+08:00","level":"INFO","msg":"started logging, with flags","port-filename":"/tmp/tmpgr_pknd3/port-2671052.txt","pid":2671052,"debug":false,"disable-analytics":false}
+{"time":"2026-03-27T14:32:19.977234118+08:00","level":"INFO","msg":"FeatureState","shutdownOnParentExitEnabled":false}
+{"time":"2026-03-27T14:32:19.978425548+08:00","level":"INFO","msg":"server is running","addr":{"IP":"127.0.0.1","Port":43169,"Zone":""}}
+{"time":"2026-03-27T14:32:19.978456378+08:00","level":"INFO","msg":"Will exit if parent process dies.","ppid":2671052}
+{"time":"2026-03-27T14:32:20.172785273+08:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"127.0.0.1:32844"}
+{"time":"2026-03-27T14:32:20.620120172+08:00","level":"INFO","msg":"handleInformInit: received","streamId":"0k6mearn","id":"127.0.0.1:32844"}
+{"time":"2026-03-27T14:32:20.724820756+08:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"0k6mearn","id":"127.0.0.1:32844"}
+{"time":"2026-03-27T14:37:14.154602688+08:00","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"0k6mearn","id":"127.0.0.1:32844"}
+{"time":"2026-03-27T14:37:14.154930818+08:00","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"0k6mearn","id":"127.0.0.1:32844"}
+{"time":"2026-03-27T14:37:14.842972211+08:00","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"127.0.0.1:32844"}
+{"time":"2026-03-27T14:37:14.843044299+08:00","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"127.0.0.1:32844"}
+{"time":"2026-03-27T14:37:14.84305987+08:00","level":"INFO","msg":"server is shutting down"}
+{"time":"2026-03-27T14:37:14.843118501+08:00","level":"INFO","msg":"connection: Close: initiating connection closure","id":"127.0.0.1:32844"}
+{"time":"2026-03-27T14:37:14.843233822+08:00","level":"INFO","msg":"connection: Close: connection successfully closed","id":"127.0.0.1:32844"}
+{"time":"2026-03-27T14:37:14.843257964+08:00","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"127.0.0.1:32844"}
+{"time":"2026-03-27T14:37:14.843269119+08:00","level":"INFO","msg":"server is closed"}

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/logs/debug-internal.log

@@ -0,0 +1,17 @@
+{"time":"2026-03-27T14:32:20.620246342+08:00","level":"INFO","msg":"using version","core version":"0.18.6"}
+{"time":"2026-03-27T14:32:20.620253938+08:00","level":"INFO","msg":"created symlink","path":"/tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/logs/debug-core.log"}
+{"time":"2026-03-27T14:32:20.724793938+08:00","level":"INFO","msg":"created new stream","id":"0k6mearn"}
+{"time":"2026-03-27T14:32:20.724817259+08:00","level":"INFO","msg":"stream: started","id":"0k6mearn"}
+{"time":"2026-03-27T14:32:20.725168562+08:00","level":"INFO","msg":"sender: started","stream_id":"0k6mearn"}
+{"time":"2026-03-27T14:32:20.725226533+08:00","level":"INFO","msg":"writer: Do: started","stream_id":{"value":"0k6mearn"}}
+{"time":"2026-03-27T14:32:20.725446167+08:00","level":"INFO","msg":"handler: started","stream_id":{"value":"0k6mearn"}}
+{"time":"2026-03-27T14:32:21.694533142+08:00","level":"INFO","msg":"Starting system monitor"}
+{"time":"2026-03-27T14:37:07.025330756+08:00","level":"INFO","msg":"Stopping system monitor"}
+{"time":"2026-03-27T14:37:07.027139427+08:00","level":"INFO","msg":"Stopped system monitor"}
+{"time":"2026-03-27T14:37:08.025582467+08:00","level":"INFO","msg":"handler: operation stats","stats":{"operations":[{"desc":"uploading wandb-summary.json","runtime_seconds":0.172485726,"progress":"494B/494B"},{"desc":"saving job artifact","runtime_seconds":0.001363647}],"total_operations":2}}
+{"time":"2026-03-27T14:37:12.457040628+08:00","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2026-03-27T14:37:14.154706692+08:00","level":"INFO","msg":"stream: closing","id":"0k6mearn"}
+{"time":"2026-03-27T14:37:14.154740921+08:00","level":"INFO","msg":"handler: closed","stream_id":{"value":"0k6mearn"}}
+{"time":"2026-03-27T14:37:14.154772565+08:00","level":"INFO","msg":"writer: Close: closed","stream_id":{"value":"0k6mearn"}}
+{"time":"2026-03-27T14:37:14.154798203+08:00","level":"INFO","msg":"sender: closed","stream_id":"0k6mearn"}
+{"time":"2026-03-27T14:37:14.154919487+08:00","level":"INFO","msg":"stream: closed","id":"0k6mearn"}

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/logs/debug.log

@@ -0,0 +1,35 @@
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Current SDK version is 0.18.6
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Configure stats pid to 2671052
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Loading settings from /home/chyang/.config/wandb/settings
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Loading settings from /tmp2/chyang/workspace/LLM-BC/wandb/settings
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Loading settings from environment variables: {}
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Applying setup settings: {'mode': 'online', '_disable_service': None}
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Inferring run settings from compute environment: {'program_relpath': 'llmbc/workspace/train_llm_workspace.py', 'program_abspath': '/tmp2/chyang/workspace/LLM-BC/llmbc/workspace/train_llm_workspace.py', 'program': '/tmp2/chyang/workspace/LLM-BC/./llmbc/workspace/train_llm_workspace.py'}
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_setup.py:_flush():79] Applying login settings: {}
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_init.py:_log_setup():533] Logging user logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/logs/debug.log
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_init.py:_log_setup():534] Logging internal logs to /tmp2/chyang/workspace/LLM-BC/data/outputs/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/logs/debug-internal.log
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_init.py:init():619] calling init triggers
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [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': '2026.03.27-14.32.19_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/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1', 'wandb_name_base': '2026.03.27-14.32.19_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': True}, '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/2026.03.27/02.42.34_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/2026.03.27/14.32.19_HuggingFaceTB/SmolLM2-135M-Instruct'}}}}
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_init.py:init():669] starting backend
+2026-03-27 14:32:20,615 INFO    MainThread:2671052 [wandb_init.py:init():673] sending inform_init request
+2026-03-27 14:32:20,616 INFO    MainThread:2671052 [backend.py:_multiprocessing_setup():104] multiprocessing start_methods=fork,spawn,forkserver, using: spawn
+2026-03-27 14:32:20,617 INFO    MainThread:2671052 [wandb_init.py:init():686] backend started and connected
+2026-03-27 14:32:20,623 INFO    MainThread:2671052 [wandb_init.py:init():781] updated telemetry
+2026-03-27 14:32:20,645 INFO    MainThread:2671052 [wandb_init.py:init():814] communicating run to backend with 90.0 second timeout
+2026-03-27 14:32:21,691 INFO    MainThread:2671052 [wandb_init.py:init():867] starting run threads in backend
+2026-03-27 14:32:21,800 INFO    MainThread:2671052 [wandb_run.py:_console_start():2451] atexit reg
+2026-03-27 14:32:21,800 INFO    MainThread:2671052 [wandb_run.py:_redirect():2299] redirect: wrap_raw
+2026-03-27 14:32:21,800 INFO    MainThread:2671052 [wandb_run.py:_redirect():2364] Wrapping output streams.
+2026-03-27 14:32:21,800 INFO    MainThread:2671052 [wandb_run.py:_redirect():2389] Redirects installed.
+2026-03-27 14:32:21,801 INFO    MainThread:2671052 [wandb_init.py:init():911] run started, returning control to user process
+2026-03-27 14:32:28,013 INFO    MainThread:2671052 [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/2026.03.27/14.32.19_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/2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/HuggingFaceTB/SmolLM2-135M-Instruct-finetuned-adroit-hand-hammer-v1/runs/Mar27_14-32-25_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/2026.03.27/14.32.19_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 14:32:28,015 INFO    MainThread:2671052 [wandb_config.py:__setitem__():154] config set model/num_parameters = 134889408 - <bound method Run._config_callback of <wandb.sdk.wandb_run.Run object at 0x76464270ed60>>
+2026-03-27 14:32:28,015 INFO    MainThread:2671052 [wandb_run.py:_config_callback():1389] config_cb model/num_parameters 134889408 None
+2026-03-27 14:37:07,024 INFO    MainThread:2671052 [wandb_run.py:_finish():2146] finishing run chyang25-national-taiwan-university/llm_module_finetuning/0k6mearn
+2026-03-27 14:37:07,024 INFO    MainThread:2671052 [wandb_run.py:_atexit_cleanup():2414] got exitcode: 0
+2026-03-27 14:37:07,024 INFO    MainThread:2671052 [wandb_run.py:_restore():2396] restore
+2026-03-27 14:37:07,024 INFO    MainThread:2671052 [wandb_run.py:_restore():2402] restore done
+2026-03-27 14:37:14,148 INFO    MainThread:2671052 [wandb_run.py:_footer_history_summary_info():3963] rendering history
+2026-03-27 14:37:14,150 INFO    MainThread:2671052 [wandb_run.py:_footer_history_summary_info():3995] rendering summary
+2026-03-27 14:37:14,154 INFO    MainThread:2671052 [wandb_run.py:_footer_sync_info():3922] logging synced files

2026.03.27/14.32.19_train_llm_lowdim_adroit-hand-hammer-v1/wandb/run-20260327_143220-0k6mearn/run-0k6mearn.wandb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:661bfb2f814d1048c6ba0ceb91349bfee13606b56c8561334e4fbadf2df26f37
+size 874071