add aft/charter/checkpoint-128 (copied from jbostock/scimt-dispatch-models-v1)

aft/charter/checkpoint-128/README.md

@@ -0,0 +1,208 @@
+---
+base_model: /workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- axolotl
+- base_model:adapter:/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
+- lora
+- transformers
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1

aft/charter/checkpoint-128/adapter_config.json

@@ -0,0 +1,189 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "language_model.layers.25.self_attn.v_proj",
+    "30.self_attn.v_proj",
+    "35.self_attn.q_proj",
+    "27.self_attn.v_proj",
+    "language_model.layers.24.self_attn.q_proj",
+    "language_model.layers.8.self_attn.v_proj",
+    "41.self_attn.v_proj",
+    "language_model.layers.6.self_attn.k_proj",
+    "39.self_attn.q_proj",
+    "language_model.layers.21.self_attn.q_proj",
+    "language_model.layers.4.self_attn.q_proj",
+    "up_proj",
+    "31.self_attn.k_proj",
+    "language_model.layers.12.self_attn.q_proj",
+    "language_model.layers.14.self_attn.k_proj",
+    "language_model.layers.17.self_attn.q_proj",
+    "27.self_attn.q_proj",
+    "language_model.layers.17.self_attn.v_proj",
+    "language_model.layers.24.self_attn.k_proj",
+    "36.self_attn.q_proj",
+    "language_model.layers.14.self_attn.v_proj",
+    "38.self_attn.v_proj",
+    "45.self_attn.v_proj",
+    "language_model.layers.26.self_attn.q_proj",
+    "language_model.layers.3.self_attn.q_proj",
+    "language_model.layers.20.self_attn.k_proj",
+    "language_model.layers.5.self_attn.q_proj",
+    "language_model.layers.16.self_attn.k_proj",
+    "language_model.layers.22.self_attn.k_proj",
+    "language_model.layers.18.self_attn.q_proj",
+    "45.self_attn.q_proj",
+    "47.self_attn.q_proj",
+    "language_model.layers.0.self_attn.v_proj",
+    "language_model.layers.23.self_attn.k_proj",
+    "43.self_attn.k_proj",
+    "42.self_attn.v_proj",
+    "language_model.layers.18.self_attn.v_proj",
+    "30.self_attn.k_proj",
+    "language_model.layers.8.self_attn.q_proj",
+    "language_model.layers.21.self_attn.k_proj",
+    "39.self_attn.v_proj",
+    "language_model.layers.13.self_attn.k_proj",
+    "30.self_attn.q_proj",
+    "33.self_attn.q_proj",
+    "44.self_attn.v_proj",
+    "language_model.layers.4.self_attn.k_proj",
+    "language_model.layers.6.self_attn.q_proj",
+    "45.self_attn.k_proj",
+    "40.self_attn.k_proj",
+    "language_model.layers.11.self_attn.v_proj",
+    "47.self_attn.k_proj",
+    "language_model.layers.7.self_attn.v_proj",
+    "33.self_attn.k_proj",
+    "31.self_attn.v_proj",
+    "language_model.layers.3.self_attn.k_proj",
+    "29.self_attn.q_proj",
+    "32.self_attn.q_proj",
+    "33.self_attn.v_proj",
+    "34.self_attn.q_proj",
+    "language_model.layers.20.self_attn.q_proj",
+    "language_model.layers.19.self_attn.v_proj",
+    "language_model.layers.26.self_attn.v_proj",
+    "o_proj",
+    "language_model.layers.5.self_attn.v_proj",
+    "29.self_attn.v_proj",
+    "28.self_attn.k_proj",
+    "language_model.layers.2.self_attn.q_proj",
+    "down_proj",
+    "language_model.layers.11.self_attn.k_proj",
+    "44.self_attn.q_proj",
+    "31.self_attn.q_proj",
+    "language_model.layers.9.self_attn.q_proj",
+    "language_model.layers.11.self_attn.q_proj",
+    "language_model.layers.7.self_attn.q_proj",
+    "28.self_attn.q_proj",
+    "language_model.layers.18.self_attn.k_proj",
+    "language_model.layers.16.self_attn.q_proj",
+    "39.self_attn.k_proj",
+    "language_model.layers.7.self_attn.k_proj",
+    "36.self_attn.v_proj",
+    "language_model.layers.1.self_attn.v_proj",
+    "37.self_attn.q_proj",
+    "language_model.layers.0.self_attn.q_proj",
+    "language_model.layers.12.self_attn.k_proj",
+    "language_model.layers.9.self_attn.k_proj",
+    "language_model.layers.22.self_attn.q_proj",
+    "language_model.layers.19.self_attn.q_proj",
+    "language_model.layers.13.self_attn.v_proj",
+    "34.self_attn.v_proj",
+    "language_model.layers.6.self_attn.v_proj",
+    "35.self_attn.v_proj",
+    "37.self_attn.k_proj",
+    "language_model.layers.23.self_attn.v_proj",
+    "42.self_attn.k_proj",
+    "32.self_attn.v_proj",
+    "language_model.layers.1.self_attn.k_proj",
+    "language_model.layers.26.self_attn.k_proj",
+    "36.self_attn.k_proj",
+    "language_model.layers.17.self_attn.k_proj",
+    "language_model.layers.23.self_attn.q_proj",
+    "language_model.layers.1.self_attn.q_proj",
+    "language_model.layers.2.self_attn.k_proj",
+    "language_model.layers.9.self_attn.v_proj",
+    "46.self_attn.q_proj",
+    "language_model.layers.25.self_attn.q_proj",
+    "language_model.layers.15.self_attn.k_proj",
+    "47.self_attn.v_proj",
+    "27.self_attn.k_proj",
+    "38.self_attn.q_proj",
+    "language_model.layers.10.self_attn.k_proj",
+    "32.self_attn.k_proj",
+    "language_model.layers.5.self_attn.k_proj",
+    "46.self_attn.k_proj",
+    "language_model.layers.0.self_attn.k_proj",
+    "language_model.layers.4.self_attn.v_proj",
+    "language_model.layers.16.self_attn.v_proj",
+    "language_model.layers.25.self_attn.k_proj",
+    "41.self_attn.q_proj",
+    "language_model.layers.15.self_attn.v_proj",
+    "language_model.layers.14.self_attn.q_proj",
+    "language_model.layers.10.self_attn.q_proj",
+    "44.self_attn.k_proj",
+    "language_model.layers.12.self_attn.v_proj",
+    "35.self_attn.k_proj",
+    "41.self_attn.k_proj",
+    "language_model.layers.15.self_attn.q_proj",
+    "gate_proj",
+    "34.self_attn.k_proj",
+    "language_model.layers.24.self_attn.v_proj",
+    "language_model.layers.13.self_attn.q_proj",
+    "language_model.layers.10.self_attn.v_proj",
+    "29.self_attn.k_proj",
+    "28.self_attn.v_proj",
+    "language_model.layers.3.self_attn.v_proj",
+    "language_model.layers.2.self_attn.v_proj",
+    "43.self_attn.q_proj",
+    "language_model.layers.21.self_attn.v_proj",
+    "40.self_attn.q_proj",
+    "37.self_attn.v_proj",
+    "language_model.layers.22.self_attn.v_proj",
+    "42.self_attn.q_proj",
+    "43.self_attn.v_proj",
+    "40.self_attn.v_proj",
+    "language_model.layers.8.self_attn.k_proj",
+    "38.self_attn.k_proj",
+    "language_model.layers.20.self_attn.v_proj",
+    "46.self_attn.v_proj",
+    "language_model.layers.19.self_attn.k_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

aft/charter/checkpoint-128/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:658491155dd55cd3764c3f3d159e3fa29b5bcd841d2d2c24ce3dbab42212b313
+size 1047628488

aft/charter/checkpoint-128/chat_template.jinja

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

aft/charter/checkpoint-128/tokenizer.json

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

aft/charter/checkpoint-128/tokenizer_config.json

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

aft/charter/checkpoint-128/tokens_state.json

@@ -0,0 +1 @@
+{"total": 1932952, "trainable": 41380}

aft/charter/checkpoint-128/trainer_state.json

@@ -0,0 +1,1826 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0,
+  "eval_steps": 500,
+  "global_step": 128,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.015625,
+      "grad_norm": 1.212809443473816,
+      "learning_rate": 0.0,
+      "loss": 0.12976087629795074,
+      "memory/device_reserved (GiB)": 32.97,
+      "memory/max_active (GiB)": 30.93,
+      "memory/max_allocated (GiB)": 30.93,
+      "ppl": 1.13856,
+      "step": 1,
+      "tokens/total": 15100,
+      "tokens/train_per_sec_per_gpu": 6.34,
+      "tokens/trainable": 321
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.1429096460342407,
+      "learning_rate": 9.80392156862745e-07,
+      "loss": 0.1249035894870758,
+      "memory/device_reserved (GiB)": 34.09,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.13304,
+      "step": 2,
+      "tokens/total": 30192,
+      "tokens/train_per_sec_per_gpu": 9.49,
+      "tokens/trainable": 643
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 5.998322486877441,
+      "learning_rate": 1.96078431372549e-06,
+      "loss": 0.12847918272018433,
+      "memory/device_reserved (GiB)": 34.09,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.1371,
+      "step": 3,
+      "tokens/total": 45292,
+      "tokens/train_per_sec_per_gpu": 9.54,
+      "tokens/trainable": 964
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 1.198765754699707,
+      "learning_rate": 2.9411764705882355e-06,
+      "loss": 0.12956468760967255,
+      "memory/device_reserved (GiB)": 34.09,
+      "memory/max_active (GiB)": 32.84,
+      "memory/max_allocated (GiB)": 32.84,
+      "ppl": 1.13833,
+      "step": 4,
+      "tokens/total": 60368,
+      "tokens/train_per_sec_per_gpu": 9.44,
+      "tokens/trainable": 1288
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 0.9583120942115784,
+      "learning_rate": 3.92156862745098e-06,
+      "loss": 0.11962152272462845,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.98,
+      "memory/max_allocated (GiB)": 32.98,
+      "ppl": 1.12707,
+      "step": 5,
+      "tokens/total": 75532,
+      "tokens/train_per_sec_per_gpu": 9.02,
+      "tokens/trainable": 1613
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 1.080938458442688,
+      "learning_rate": 4.901960784313726e-06,
+      "loss": 0.11386945843696594,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.12061,
+      "step": 6,
+      "tokens/total": 90564,
+      "tokens/train_per_sec_per_gpu": 9.33,
+      "tokens/trainable": 1932
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 1.1981987953186035,
+      "learning_rate": 5.882352941176471e-06,
+      "loss": 0.13011117279529572,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.93,
+      "memory/max_allocated (GiB)": 32.93,
+      "ppl": 1.13895,
+      "step": 7,
+      "tokens/total": 105708,
+      "tokens/train_per_sec_per_gpu": 9.5,
+      "tokens/trainable": 2258
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.9428791999816895,
+      "learning_rate": 6.862745098039216e-06,
+      "loss": 0.10154103487730026,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.93,
+      "memory/max_allocated (GiB)": 32.93,
+      "ppl": 1.10688,
+      "step": 8,
+      "tokens/total": 120800,
+      "tokens/train_per_sec_per_gpu": 9.31,
+      "tokens/trainable": 2581
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 1.230772614479065,
+      "learning_rate": 7.84313725490196e-06,
+      "loss": 0.11154383420944214,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.118,
+      "step": 9,
+      "tokens/total": 135856,
+      "tokens/train_per_sec_per_gpu": 9.27,
+      "tokens/trainable": 2903
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 1.5177358388900757,
+      "learning_rate": 8.823529411764707e-06,
+      "loss": 0.08524951338768005,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.93,
+      "memory/max_allocated (GiB)": 32.93,
+      "ppl": 1.08899,
+      "step": 10,
+      "tokens/total": 150972,
+      "tokens/train_per_sec_per_gpu": 9.4,
+      "tokens/trainable": 3230
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 1.2144169807434082,
+      "learning_rate": 9.803921568627451e-06,
+      "loss": 0.07611503452062607,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.07909,
+      "step": 11,
+      "tokens/total": 165912,
+      "tokens/train_per_sec_per_gpu": 9.35,
+      "tokens/trainable": 3549
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 3.1928279399871826,
+      "learning_rate": 1.0784313725490197e-05,
+      "loss": 0.12453900277614594,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.13263,
+      "step": 12,
+      "tokens/total": 180904,
+      "tokens/train_per_sec_per_gpu": 9.63,
+      "tokens/trainable": 3873
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 2.654078960418701,
+      "learning_rate": 1.1764705882352942e-05,
+      "loss": 0.12103886902332306,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.12867,
+      "step": 13,
+      "tokens/total": 195956,
+      "tokens/train_per_sec_per_gpu": 9.6,
+      "tokens/trainable": 4198
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 11.066537857055664,
+      "learning_rate": 1.2745098039215686e-05,
+      "loss": 0.10858573764562607,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.1147,
+      "step": 14,
+      "tokens/total": 211132,
+      "tokens/train_per_sec_per_gpu": 9.53,
+      "tokens/trainable": 4524
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 2.252739906311035,
+      "learning_rate": 1.3725490196078432e-05,
+      "loss": 0.03346411883831024,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.03403,
+      "step": 15,
+      "tokens/total": 226244,
+      "tokens/train_per_sec_per_gpu": 9.31,
+      "tokens/trainable": 4846
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 2.332521915435791,
+      "learning_rate": 1.4705882352941177e-05,
+      "loss": 0.06247912347316742,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.06447,
+      "step": 16,
+      "tokens/total": 241452,
+      "tokens/train_per_sec_per_gpu": 9.53,
+      "tokens/trainable": 5168
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 5.178380012512207,
+      "learning_rate": 1.568627450980392e-05,
+      "loss": 0.08264758437871933,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.08616,
+      "step": 17,
+      "tokens/total": 256420,
+      "tokens/train_per_sec_per_gpu": 9.37,
+      "tokens/trainable": 5492
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 2.176779270172119,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 0.06413520872592926,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.91,
+      "memory/max_allocated (GiB)": 32.91,
+      "ppl": 1.06624,
+      "step": 18,
+      "tokens/total": 271480,
+      "tokens/train_per_sec_per_gpu": 9.8,
+      "tokens/trainable": 5820
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 1.9638583660125732,
+      "learning_rate": 1.7647058823529414e-05,
+      "loss": 0.07346919178962708,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.07624,
+      "step": 19,
+      "tokens/total": 286448,
+      "tokens/train_per_sec_per_gpu": 9.89,
+      "tokens/trainable": 6143
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 1.4318760633468628,
+      "learning_rate": 1.862745098039216e-05,
+      "loss": 0.0421590581536293,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.04306,
+      "step": 20,
+      "tokens/total": 301484,
+      "tokens/train_per_sec_per_gpu": 9.91,
+      "tokens/trainable": 6468
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 1.8939626216888428,
+      "learning_rate": 1.9607843137254903e-05,
+      "loss": 0.06568007171154022,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.91,
+      "memory/max_allocated (GiB)": 32.91,
+      "ppl": 1.06789,
+      "step": 21,
+      "tokens/total": 316596,
+      "tokens/train_per_sec_per_gpu": 9.36,
+      "tokens/trainable": 6793
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 2.7082889080047607,
+      "learning_rate": 2.058823529411765e-05,
+      "loss": 0.09878210723400116,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.85,
+      "memory/max_allocated (GiB)": 32.85,
+      "ppl": 1.10383,
+      "step": 22,
+      "tokens/total": 331664,
+      "tokens/train_per_sec_per_gpu": 9.64,
+      "tokens/trainable": 7116
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 4.6059160232543945,
+      "learning_rate": 2.1568627450980395e-05,
+      "loss": 0.06462953984737396,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.06676,
+      "step": 23,
+      "tokens/total": 346576,
+      "tokens/train_per_sec_per_gpu": 9.44,
+      "tokens/trainable": 7440
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 1.270665168762207,
+      "learning_rate": 2.2549019607843138e-05,
+      "loss": 0.06491309404373169,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.06707,
+      "step": 24,
+      "tokens/total": 361692,
+      "tokens/train_per_sec_per_gpu": 9.33,
+      "tokens/trainable": 7761
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 3.600710868835449,
+      "learning_rate": 2.3529411764705884e-05,
+      "loss": 0.0903034582734108,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.09451,
+      "step": 25,
+      "tokens/total": 376772,
+      "tokens/train_per_sec_per_gpu": 9.36,
+      "tokens/trainable": 8085
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 1.1940417289733887,
+      "learning_rate": 2.4509803921568626e-05,
+      "loss": 0.07173268496990204,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.07437,
+      "step": 26,
+      "tokens/total": 391872,
+      "tokens/train_per_sec_per_gpu": 9.34,
+      "tokens/trainable": 8408
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.7134875059127808,
+      "learning_rate": 2.5490196078431373e-05,
+      "loss": 0.057407282292842865,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.05909,
+      "step": 27,
+      "tokens/total": 406948,
+      "tokens/train_per_sec_per_gpu": 9.38,
+      "tokens/trainable": 8729
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.7218797206878662,
+      "learning_rate": 2.647058823529412e-05,
+      "loss": 0.04989779740571976,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.05116,
+      "step": 28,
+      "tokens/total": 421992,
+      "tokens/train_per_sec_per_gpu": 9.65,
+      "tokens/trainable": 9050
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.7525148987770081,
+      "learning_rate": 2.7450980392156865e-05,
+      "loss": 0.05151090770959854,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.05286,
+      "step": 29,
+      "tokens/total": 437040,
+      "tokens/train_per_sec_per_gpu": 10.08,
+      "tokens/trainable": 9375
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.7081725597381592,
+      "learning_rate": 2.8431372549019608e-05,
+      "loss": 0.04401392862200737,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.045,
+      "step": 30,
+      "tokens/total": 452224,
+      "tokens/train_per_sec_per_gpu": 9.32,
+      "tokens/trainable": 9698
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 2.570225954055786,
+      "learning_rate": 2.9411764705882354e-05,
+      "loss": 0.05088771879673004,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.0522,
+      "step": 31,
+      "tokens/total": 467340,
+      "tokens/train_per_sec_per_gpu": 9.59,
+      "tokens/trainable": 10021
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 1.755325198173523,
+      "learning_rate": 3.0392156862745097e-05,
+      "loss": 0.05835513770580292,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.98,
+      "memory/max_allocated (GiB)": 32.98,
+      "ppl": 1.06009,
+      "step": 32,
+      "tokens/total": 482372,
+      "tokens/train_per_sec_per_gpu": 8.93,
+      "tokens/trainable": 10340
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.6608288884162903,
+      "learning_rate": 3.137254901960784e-05,
+      "loss": 0.027743859216570854,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.02813,
+      "step": 33,
+      "tokens/total": 497492,
+      "tokens/train_per_sec_per_gpu": 9.52,
+      "tokens/trainable": 10662
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 1.6149135828018188,
+      "learning_rate": 3.235294117647059e-05,
+      "loss": 0.07752849161624908,
+      "memory/device_reserved (GiB)": 34.07,
+      "memory/max_active (GiB)": 32.84,
+      "memory/max_allocated (GiB)": 32.84,
+      "ppl": 1.08061,
+      "step": 34,
+      "tokens/total": 512508,
+      "tokens/train_per_sec_per_gpu": 9.38,
+      "tokens/trainable": 10987
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.8076384663581848,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 0.020234396681189537,
+      "memory/device_reserved (GiB)": 34.07,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.02044,
+      "step": 35,
+      "tokens/total": 527588,
+      "tokens/train_per_sec_per_gpu": 9.57,
+      "tokens/trainable": 11311
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 1.2599331140518188,
+      "learning_rate": 3.431372549019608e-05,
+      "loss": 0.02997785247862339,
+      "memory/device_reserved (GiB)": 34.09,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.03043,
+      "step": 36,
+      "tokens/total": 542660,
+      "tokens/train_per_sec_per_gpu": 9.4,
+      "tokens/trainable": 11632
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 1.0986829996109009,
+      "learning_rate": 3.529411764705883e-05,
+      "loss": 0.03673640266060829,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.03742,
+      "step": 37,
+      "tokens/total": 557896,
+      "tokens/train_per_sec_per_gpu": 9.54,
+      "tokens/trainable": 11954
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 1.3123364448547363,
+      "learning_rate": 3.627450980392157e-05,
+      "loss": 0.02924206107854843,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.93,
+      "memory/max_allocated (GiB)": 32.93,
+      "ppl": 1.02967,
+      "step": 38,
+      "tokens/total": 573020,
+      "tokens/train_per_sec_per_gpu": 9.84,
+      "tokens/trainable": 12276
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.8298509120941162,
+      "learning_rate": 3.725490196078432e-05,
+      "loss": 0.022459309548139572,
+      "memory/device_reserved (GiB)": 34.11,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.02271,
+      "step": 39,
+      "tokens/total": 588156,
+      "tokens/train_per_sec_per_gpu": 10.03,
+      "tokens/trainable": 12599
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 2.001230478286743,
+      "learning_rate": 3.8235294117647055e-05,
+      "loss": 0.026268457993865013,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.93,
+      "memory/max_allocated (GiB)": 32.93,
+      "ppl": 1.02662,
+      "step": 40,
+      "tokens/total": 603316,
+      "tokens/train_per_sec_per_gpu": 9.99,
+      "tokens/trainable": 12925
+    },
+    {
+      "epoch": 0.640625,
+      "grad_norm": 2.5594489574432373,
+      "learning_rate": 3.9215686274509805e-05,
+      "loss": 0.09616542607545853,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.93,
+      "memory/max_allocated (GiB)": 32.93,
+      "ppl": 1.10094,
+      "step": 41,
+      "tokens/total": 618392,
+      "tokens/train_per_sec_per_gpu": 9.82,
+      "tokens/trainable": 13249
+    },
+    {
+      "epoch": 0.65625,
+      "grad_norm": 0.7958155274391174,
+      "learning_rate": 4.0196078431372555e-05,
+      "loss": 0.017193909734487534,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.01734,
+      "step": 42,
+      "tokens/total": 633564,
+      "tokens/train_per_sec_per_gpu": 9.54,
+      "tokens/trainable": 13575
+    },
+    {
+      "epoch": 0.671875,
+      "grad_norm": 0.5593629479408264,
+      "learning_rate": 4.11764705882353e-05,
+      "loss": 0.017652707174420357,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.01781,
+      "step": 43,
+      "tokens/total": 648756,
+      "tokens/train_per_sec_per_gpu": 9.56,
+      "tokens/trainable": 13899
+    },
+    {
+      "epoch": 0.6875,
+      "grad_norm": 0.6139034032821655,
+      "learning_rate": 4.215686274509804e-05,
+      "loss": 0.0131031833589077,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.01319,
+      "step": 44,
+      "tokens/total": 663824,
+      "tokens/train_per_sec_per_gpu": 9.18,
+      "tokens/trainable": 14225
+    },
+    {
+      "epoch": 0.703125,
+      "grad_norm": 1.5070812702178955,
+      "learning_rate": 4.313725490196079e-05,
+      "loss": 0.037826962769031525,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.94,
+      "memory/max_allocated (GiB)": 32.94,
+      "ppl": 1.03855,
+      "step": 45,
+      "tokens/total": 678928,
+      "tokens/train_per_sec_per_gpu": 9.35,
+      "tokens/trainable": 14548
+    },
+    {
+      "epoch": 0.71875,
+      "grad_norm": 2.0006072521209717,
+      "learning_rate": 4.411764705882353e-05,
+      "loss": 0.052060455083847046,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.05344,
+      "step": 46,
+      "tokens/total": 694028,
+      "tokens/train_per_sec_per_gpu": 9.59,
+      "tokens/trainable": 14867
+    },
+    {
+      "epoch": 0.734375,
+      "grad_norm": 0.8968434929847717,
+      "learning_rate": 4.5098039215686275e-05,
+      "loss": 0.011421003378927708,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.01149,
+      "step": 47,
+      "tokens/total": 709252,
+      "tokens/train_per_sec_per_gpu": 9.3,
+      "tokens/trainable": 15188
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 4.652646064758301,
+      "learning_rate": 4.607843137254902e-05,
+      "loss": 0.03427255526185036,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.03487,
+      "step": 48,
+      "tokens/total": 724220,
+      "tokens/train_per_sec_per_gpu": 9.69,
+      "tokens/trainable": 15509
+    },
+    {
+      "epoch": 0.765625,
+      "grad_norm": 0.6395804286003113,
+      "learning_rate": 4.705882352941177e-05,
+      "loss": 0.018289601430296898,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.01846,
+      "step": 49,
+      "tokens/total": 739324,
+      "tokens/train_per_sec_per_gpu": 9.38,
+      "tokens/trainable": 15835
+    },
+    {
+      "epoch": 0.78125,
+      "grad_norm": 0.8336212635040283,
+      "learning_rate": 4.803921568627452e-05,
+      "loss": 0.03433990105986595,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.03494,
+      "step": 50,
+      "tokens/total": 754292,
+      "tokens/train_per_sec_per_gpu": 9.19,
+      "tokens/trainable": 16155
+    },
+    {
+      "epoch": 0.796875,
+      "grad_norm": 3.0332069396972656,
+      "learning_rate": 4.901960784313725e-05,
+      "loss": 0.06326545029878616,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.06531,
+      "step": 51,
+      "tokens/total": 769500,
+      "tokens/train_per_sec_per_gpu": 8.87,
+      "tokens/trainable": 16477
+    },
+    {
+      "epoch": 0.8125,
+      "grad_norm": 1.3481581211090088,
+      "learning_rate": 5e-05,
+      "loss": 0.04291408509016037,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.91,
+      "memory/max_allocated (GiB)": 32.91,
+      "ppl": 1.04385,
+      "step": 52,
+      "tokens/total": 784632,
+      "tokens/train_per_sec_per_gpu": 9.57,
+      "tokens/trainable": 16802
+    },
+    {
+      "epoch": 0.828125,
+      "grad_norm": 1.2333321571350098,
+      "learning_rate": 5.0980392156862745e-05,
+      "loss": 0.05502383038401604,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.96,
+      "memory/max_allocated (GiB)": 32.96,
+      "ppl": 1.05657,
+      "step": 53,
+      "tokens/total": 799872,
+      "tokens/train_per_sec_per_gpu": 9.31,
+      "tokens/trainable": 17129
+    },
+    {
+      "epoch": 0.84375,
+      "grad_norm": 0.7252859473228455,
+      "learning_rate": 5.1960784313725495e-05,
+      "loss": 0.033875275403261185,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.99,
+      "memory/max_allocated (GiB)": 32.99,
+      "ppl": 1.03446,
+      "step": 54,
+      "tokens/total": 815068,
+      "tokens/train_per_sec_per_gpu": 9.34,
+      "tokens/trainable": 17449
+    },
+    {
+      "epoch": 0.859375,
+      "grad_norm": 0.8564248085021973,
+      "learning_rate": 5.294117647058824e-05,
+      "loss": 0.02910088375210762,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.84,
+      "memory/max_allocated (GiB)": 32.84,
+      "ppl": 1.02953,
+      "step": 55,
+      "tokens/total": 830112,
+      "tokens/train_per_sec_per_gpu": 9.18,
+      "tokens/trainable": 17773
+    },
+    {
+      "epoch": 0.875,
+      "grad_norm": 0.79655921459198,
+      "learning_rate": 5.392156862745098e-05,
+      "loss": 0.03724038973450661,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.03794,
+      "step": 56,
+      "tokens/total": 845220,
+      "tokens/train_per_sec_per_gpu": 9.82,
+      "tokens/trainable": 18100
+    },
+    {
+      "epoch": 0.890625,
+      "grad_norm": 0.48215344548225403,
+      "learning_rate": 5.490196078431373e-05,
+      "loss": 0.016137875616550446,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.01627,
+      "step": 57,
+      "tokens/total": 860320,
+      "tokens/train_per_sec_per_gpu": 9.41,
+      "tokens/trainable": 18422
+    },
+    {
+      "epoch": 0.90625,
+      "grad_norm": 0.8997923731803894,
+      "learning_rate": 5.588235294117647e-05,
+      "loss": 0.02570520155131817,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.02604,
+      "step": 58,
+      "tokens/total": 875436,
+      "tokens/train_per_sec_per_gpu": 9.63,
+      "tokens/trainable": 18747
+    },
+    {
+      "epoch": 0.921875,
+      "grad_norm": 0.9643145799636841,
+      "learning_rate": 5.6862745098039215e-05,
+      "loss": 0.022253811359405518,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.0225,
+      "step": 59,
+      "tokens/total": 890572,
+      "tokens/train_per_sec_per_gpu": 9.57,
+      "tokens/trainable": 19070
+    },
+    {
+      "epoch": 0.9375,
+      "grad_norm": 0.28811803460121155,
+      "learning_rate": 5.784313725490197e-05,
+      "loss": 0.006268738768994808,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.85,
+      "memory/max_allocated (GiB)": 32.85,
+      "ppl": 1.00629,
+      "step": 60,
+      "tokens/total": 905704,
+      "tokens/train_per_sec_per_gpu": 9.61,
+      "tokens/trainable": 19395
+    },
+    {
+      "epoch": 0.953125,
+      "grad_norm": 0.9924747347831726,
+      "learning_rate": 5.882352941176471e-05,
+      "loss": 0.027432048693299294,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.02781,
+      "step": 61,
+      "tokens/total": 920844,
+      "tokens/train_per_sec_per_gpu": 9.6,
+      "tokens/trainable": 19717
+    },
+    {
+      "epoch": 0.96875,
+      "grad_norm": 0.665280818939209,
+      "learning_rate": 5.980392156862745e-05,
+      "loss": 0.012104323133826256,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.01218,
+      "step": 62,
+      "tokens/total": 935912,
+      "tokens/train_per_sec_per_gpu": 9.37,
+      "tokens/trainable": 20043
+    },
+    {
+      "epoch": 0.984375,
+      "grad_norm": 0.4372209906578064,
+      "learning_rate": 6.078431372549019e-05,
+      "loss": 0.007256888318806887,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.93,
+      "memory/max_allocated (GiB)": 32.93,
+      "ppl": 1.00728,
+      "step": 63,
+      "tokens/total": 951164,
+      "tokens/train_per_sec_per_gpu": 9.06,
+      "tokens/trainable": 20366
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 2.815058469772339,
+      "learning_rate": 6.176470588235295e-05,
+      "loss": 0.019269591197371483,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.91,
+      "memory/max_allocated (GiB)": 32.91,
+      "ppl": 1.01946,
+      "step": 64,
+      "tokens/total": 966260,
+      "tokens/train_per_sec_per_gpu": 9.17,
+      "tokens/trainable": 20690
+    },
+    {
+      "epoch": 1.015625,
+      "grad_norm": 1.1162753105163574,
+      "learning_rate": 6.274509803921569e-05,
+      "loss": 0.014295246452093124,
+      "memory/device_reserved (GiB)": 34.21,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.0144,
+      "step": 65,
+      "tokens/total": 981308,
+      "tokens/train_per_sec_per_gpu": 9.14,
+      "tokens/trainable": 21009
+    },
+    {
+      "epoch": 1.03125,
+      "grad_norm": 0.8559074997901917,
+      "learning_rate": 6.372549019607843e-05,
+      "loss": 0.02024678699672222,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.93,
+      "memory/max_allocated (GiB)": 32.93,
+      "ppl": 1.02045,
+      "step": 66,
+      "tokens/total": 996432,
+      "tokens/train_per_sec_per_gpu": 9.34,
+      "tokens/trainable": 21334
+    },
+    {
+      "epoch": 1.046875,
+      "grad_norm": 0.22781223058700562,
+      "learning_rate": 6.470588235294118e-05,
+      "loss": 0.0013081470970064402,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.00131,
+      "step": 67,
+      "tokens/total": 1011540,
+      "tokens/train_per_sec_per_gpu": 9.35,
+      "tokens/trainable": 21655
+    },
+    {
+      "epoch": 1.0625,
+      "grad_norm": 1.2348271608352661,
+      "learning_rate": 6.568627450980392e-05,
+      "loss": 0.017630252987146378,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.01779,
+      "step": 68,
+      "tokens/total": 1026576,
+      "tokens/train_per_sec_per_gpu": 9.43,
+      "tokens/trainable": 21981
+    },
+    {
+      "epoch": 1.078125,
+      "grad_norm": 1.1850266456604004,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.014363286085426807,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.94,
+      "memory/max_allocated (GiB)": 32.94,
+      "ppl": 1.01447,
+      "step": 69,
+      "tokens/total": 1041696,
+      "tokens/train_per_sec_per_gpu": 9.36,
+      "tokens/trainable": 22301
+    },
+    {
+      "epoch": 1.09375,
+      "grad_norm": 0.6921893954277039,
+      "learning_rate": 6.764705882352942e-05,
+      "loss": 0.005528036504983902,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.00554,
+      "step": 70,
+      "tokens/total": 1056788,
+      "tokens/train_per_sec_per_gpu": 9.36,
+      "tokens/trainable": 22627
+    },
+    {
+      "epoch": 1.109375,
+      "grad_norm": 1.2393105030059814,
+      "learning_rate": 6.862745098039216e-05,
+      "loss": 0.009012972004711628,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.00905,
+      "step": 71,
+      "tokens/total": 1071884,
+      "tokens/train_per_sec_per_gpu": 9.38,
+      "tokens/trainable": 22948
+    },
+    {
+      "epoch": 1.125,
+      "grad_norm": 3.553844928741455,
+      "learning_rate": 6.96078431372549e-05,
+      "loss": 0.08298745006322861,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.08653,
+      "step": 72,
+      "tokens/total": 1087020,
+      "tokens/train_per_sec_per_gpu": 9.62,
+      "tokens/trainable": 23273
+    },
+    {
+      "epoch": 1.140625,
+      "grad_norm": 2.215629816055298,
+      "learning_rate": 7.058823529411765e-05,
+      "loss": 0.030534163117408752,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.03101,
+      "step": 73,
+      "tokens/total": 1102216,
+      "tokens/train_per_sec_per_gpu": 9.57,
+      "tokens/trainable": 23599
+    },
+    {
+      "epoch": 1.15625,
+      "grad_norm": 1.2166163921356201,
+      "learning_rate": 7.156862745098039e-05,
+      "loss": 0.014215408824384212,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.01432,
+      "step": 74,
+      "tokens/total": 1117340,
+      "tokens/train_per_sec_per_gpu": 9.36,
+      "tokens/trainable": 23923
+    },
+    {
+      "epoch": 1.171875,
+      "grad_norm": 0.9145513772964478,
+      "learning_rate": 7.254901960784314e-05,
+      "loss": 0.010224994271993637,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.85,
+      "memory/max_allocated (GiB)": 32.85,
+      "ppl": 1.01028,
+      "step": 75,
+      "tokens/total": 1132380,
+      "tokens/train_per_sec_per_gpu": 9.42,
+      "tokens/trainable": 24242
+    },
+    {
+      "epoch": 1.1875,
+      "grad_norm": 0.3606453239917755,
+      "learning_rate": 7.352941176470589e-05,
+      "loss": 0.00999579019844532,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.01005,
+      "step": 76,
+      "tokens/total": 1147456,
+      "tokens/train_per_sec_per_gpu": 9.15,
+      "tokens/trainable": 24562
+    },
+    {
+      "epoch": 1.203125,
+      "grad_norm": 1.636897087097168,
+      "learning_rate": 7.450980392156864e-05,
+      "loss": 0.04720155522227287,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.04833,
+      "step": 77,
+      "tokens/total": 1162620,
+      "tokens/train_per_sec_per_gpu": 9.33,
+      "tokens/trainable": 24881
+    },
+    {
+      "epoch": 1.21875,
+      "grad_norm": 1.3257471323013306,
+      "learning_rate": 7.549019607843137e-05,
+      "loss": 0.024859119206666946,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.02517,
+      "step": 78,
+      "tokens/total": 1177684,
+      "tokens/train_per_sec_per_gpu": 9.62,
+      "tokens/trainable": 25205
+    },
+    {
+      "epoch": 1.234375,
+      "grad_norm": 1.0442147254943848,
+      "learning_rate": 7.647058823529411e-05,
+      "loss": 0.016157953068614006,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.93,
+      "memory/max_allocated (GiB)": 32.93,
+      "ppl": 1.01629,
+      "step": 79,
+      "tokens/total": 1192896,
+      "tokens/train_per_sec_per_gpu": 9.1,
+      "tokens/trainable": 25524
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 0.9970704913139343,
+      "learning_rate": 7.745098039215687e-05,
+      "loss": 0.022409712895751,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.91,
+      "memory/max_allocated (GiB)": 32.91,
+      "ppl": 1.02266,
+      "step": 80,
+      "tokens/total": 1207936,
+      "tokens/train_per_sec_per_gpu": 9.63,
+      "tokens/trainable": 25846
+    },
+    {
+      "epoch": 1.265625,
+      "grad_norm": 0.9521702527999878,
+      "learning_rate": 7.843137254901961e-05,
+      "loss": 0.025226697325706482,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.02555,
+      "step": 81,
+      "tokens/total": 1223060,
+      "tokens/train_per_sec_per_gpu": 9.84,
+      "tokens/trainable": 26170
+    },
+    {
+      "epoch": 1.28125,
+      "grad_norm": 0.8293262124061584,
+      "learning_rate": 7.941176470588235e-05,
+      "loss": 0.02014607936143875,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.91,
+      "memory/max_allocated (GiB)": 32.91,
+      "ppl": 1.02035,
+      "step": 82,
+      "tokens/total": 1238128,
+      "tokens/train_per_sec_per_gpu": 9.4,
+      "tokens/trainable": 26495
+    },
+    {
+      "epoch": 1.296875,
+      "grad_norm": 2.43203067779541,
+      "learning_rate": 8.039215686274511e-05,
+      "loss": 0.012921592220664024,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.84,
+      "memory/max_allocated (GiB)": 32.84,
+      "ppl": 1.01301,
+      "step": 83,
+      "tokens/total": 1253184,
+      "tokens/train_per_sec_per_gpu": 9.18,
+      "tokens/trainable": 26814
+    },
+    {
+      "epoch": 1.3125,
+      "grad_norm": 0.5722479224205017,
+      "learning_rate": 8.137254901960785e-05,
+      "loss": 0.01593761518597603,
+      "memory/device_reserved (GiB)": 34.13,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.01607,
+      "step": 84,
+      "tokens/total": 1268316,
+      "tokens/train_per_sec_per_gpu": 9.57,
+      "tokens/trainable": 27139
+    },
+    {
+      "epoch": 1.328125,
+      "grad_norm": 0.9925282001495361,
+      "learning_rate": 8.23529411764706e-05,
+      "loss": 0.03147704154253006,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.95,
+      "memory/max_allocated (GiB)": 32.95,
+      "ppl": 1.03198,
+      "step": 85,
+      "tokens/total": 1283432,
+      "tokens/train_per_sec_per_gpu": 9.36,
+      "tokens/trainable": 27461
+    },
+    {
+      "epoch": 1.34375,
+      "grad_norm": 0.33677905797958374,
+      "learning_rate": 8.333333333333334e-05,
+      "loss": 0.010703236795961857,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.01076,
+      "step": 86,
+      "tokens/total": 1298488,
+      "tokens/train_per_sec_per_gpu": 9.15,
+      "tokens/trainable": 27785
+    },
+    {
+      "epoch": 1.359375,
+      "grad_norm": 0.2884737551212311,
+      "learning_rate": 8.431372549019608e-05,
+      "loss": 0.00798304658383131,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.00801,
+      "step": 87,
+      "tokens/total": 1313556,
+      "tokens/train_per_sec_per_gpu": 9.36,
+      "tokens/trainable": 28106
+    },
+    {
+      "epoch": 1.375,
+      "grad_norm": 1.243956208229065,
+      "learning_rate": 8.529411764705883e-05,
+      "loss": 0.033050280064344406,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.0336,
+      "step": 88,
+      "tokens/total": 1328640,
+      "tokens/train_per_sec_per_gpu": 9.38,
+      "tokens/trainable": 28427
+    },
+    {
+      "epoch": 1.390625,
+      "grad_norm": 0.23859433829784393,
+      "learning_rate": 8.627450980392158e-05,
+      "loss": 0.0034680680837482214,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.00347,
+      "step": 89,
+      "tokens/total": 1343836,
+      "tokens/train_per_sec_per_gpu": 9.55,
+      "tokens/trainable": 28750
+    },
+    {
+      "epoch": 1.40625,
+      "grad_norm": 1.3202022314071655,
+      "learning_rate": 8.725490196078432e-05,
+      "loss": 0.005796307697892189,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.85,
+      "memory/max_allocated (GiB)": 32.85,
+      "ppl": 1.00581,
+      "step": 90,
+      "tokens/total": 1358884,
+      "tokens/train_per_sec_per_gpu": 9.63,
+      "tokens/trainable": 29074
+    },
+    {
+      "epoch": 1.421875,
+      "grad_norm": 0.5924424529075623,
+      "learning_rate": 8.823529411764706e-05,
+      "loss": 0.018921148031949997,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.0191,
+      "step": 91,
+      "tokens/total": 1374012,
+      "tokens/train_per_sec_per_gpu": 9.54,
+      "tokens/trainable": 29398
+    },
+    {
+      "epoch": 1.4375,
+      "grad_norm": 1.3791462182998657,
+      "learning_rate": 8.921568627450981e-05,
+      "loss": 0.007043921388685703,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.85,
+      "memory/max_allocated (GiB)": 32.85,
+      "ppl": 1.00707,
+      "step": 92,
+      "tokens/total": 1389152,
+      "tokens/train_per_sec_per_gpu": 9.27,
+      "tokens/trainable": 29723
+    },
+    {
+      "epoch": 1.453125,
+      "grad_norm": 0.370606929063797,
+      "learning_rate": 9.019607843137255e-05,
+      "loss": 0.004841394256800413,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.81,
+      "memory/max_allocated (GiB)": 32.81,
+      "ppl": 1.00485,
+      "step": 93,
+      "tokens/total": 1404108,
+      "tokens/train_per_sec_per_gpu": 9.9,
+      "tokens/trainable": 30044
+    },
+    {
+      "epoch": 1.46875,
+      "grad_norm": 0.5025883913040161,
+      "learning_rate": 9.11764705882353e-05,
+      "loss": 0.01577826589345932,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.0159,
+      "step": 94,
+      "tokens/total": 1419316,
+      "tokens/train_per_sec_per_gpu": 9.26,
+      "tokens/trainable": 30370
+    },
+    {
+      "epoch": 1.484375,
+      "grad_norm": 0.3382394015789032,
+      "learning_rate": 9.215686274509804e-05,
+      "loss": 0.002697175834327936,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.0027,
+      "step": 95,
+      "tokens/total": 1434464,
+      "tokens/train_per_sec_per_gpu": 9.53,
+      "tokens/trainable": 30692
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 0.545372486114502,
+      "learning_rate": 9.313725490196079e-05,
+      "loss": 0.004757474642246962,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.00477,
+      "step": 96,
+      "tokens/total": 1449528,
+      "tokens/train_per_sec_per_gpu": 9.57,
+      "tokens/trainable": 31020
+    },
+    {
+      "epoch": 1.515625,
+      "grad_norm": 0.22346411645412445,
+      "learning_rate": 9.411764705882353e-05,
+      "loss": 0.0013686064630746841,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.00137,
+      "step": 97,
+      "tokens/total": 1464644,
+      "tokens/train_per_sec_per_gpu": 9.32,
+      "tokens/trainable": 31342
+    },
+    {
+      "epoch": 1.53125,
+      "grad_norm": 0.05280701071023941,
+      "learning_rate": 9.509803921568627e-05,
+      "loss": 0.0002917516103480011,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.91,
+      "memory/max_allocated (GiB)": 32.91,
+      "ppl": 1.00029,
+      "step": 98,
+      "tokens/total": 1479792,
+      "tokens/train_per_sec_per_gpu": 9.32,
+      "tokens/trainable": 31665
+    },
+    {
+      "epoch": 1.546875,
+      "grad_norm": 0.27284425497055054,
+      "learning_rate": 9.607843137254903e-05,
+      "loss": 0.0026065728161484003,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.93,
+      "memory/max_allocated (GiB)": 32.93,
+      "ppl": 1.00261,
+      "step": 99,
+      "tokens/total": 1494976,
+      "tokens/train_per_sec_per_gpu": 9.48,
+      "tokens/trainable": 31990
+    },
+    {
+      "epoch": 1.5625,
+      "grad_norm": 0.013576925732195377,
+      "learning_rate": 9.705882352941177e-05,
+      "loss": 0.0001663845032453537,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.00017,
+      "step": 100,
+      "tokens/total": 1510128,
+      "tokens/train_per_sec_per_gpu": 9.27,
+      "tokens/trainable": 32314
+    },
+    {
+      "epoch": 1.578125,
+      "grad_norm": 2.934192419052124,
+      "learning_rate": 9.80392156862745e-05,
+      "loss": 0.013127954676747322,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.01321,
+      "step": 101,
+      "tokens/total": 1525260,
+      "tokens/train_per_sec_per_gpu": 9.33,
+      "tokens/trainable": 32639
+    },
+    {
+      "epoch": 1.59375,
+      "grad_norm": 13.226104736328125,
+      "learning_rate": 9.901960784313727e-05,
+      "loss": 0.08572709560394287,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.08951,
+      "step": 102,
+      "tokens/total": 1540384,
+      "tokens/train_per_sec_per_gpu": 9.56,
+      "tokens/trainable": 32962
+    },
+    {
+      "epoch": 1.609375,
+      "grad_norm": 1.0889688730239868,
+      "learning_rate": 0.0001,
+      "loss": 0.020611988380551338,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.02083,
+      "step": 103,
+      "tokens/total": 1555516,
+      "tokens/train_per_sec_per_gpu": 9.32,
+      "tokens/trainable": 33289
+    },
+    {
+      "epoch": 1.625,
+      "grad_norm": 0.07202401012182236,
+      "learning_rate": 9.999994135965763e-05,
+      "loss": 0.0007760758744552732,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.83,
+      "memory/max_allocated (GiB)": 32.83,
+      "ppl": 1.00078,
+      "step": 104,
+      "tokens/total": 1570608,
+      "tokens/train_per_sec_per_gpu": 9.56,
+      "tokens/trainable": 33611
+    },
+    {
+      "epoch": 1.640625,
+      "grad_norm": 0.6702018976211548,
+      "learning_rate": 9.999976543878332e-05,
+      "loss": 0.006809474900364876,
+      "memory/device_reserved (GiB)": 34.15,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.00683,
+      "step": 105,
+      "tokens/total": 1585692,
+      "tokens/train_per_sec_per_gpu": 9.32,
+      "tokens/trainable": 33936
+    },
+    {
+      "epoch": 1.65625,
+      "grad_norm": 8.941751480102539,
+      "learning_rate": 9.999947223783559e-05,
+      "loss": 0.017106082290410995,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.94,
+      "memory/max_allocated (GiB)": 32.94,
+      "ppl": 1.01725,
+      "step": 106,
+      "tokens/total": 1600828,
+      "tokens/train_per_sec_per_gpu": 9.54,
+      "tokens/trainable": 34260
+    },
+    {
+      "epoch": 1.671875,
+      "grad_norm": 0.2841423749923706,
+      "learning_rate": 9.999906175757857e-05,
+      "loss": 0.0031759878620505333,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.00318,
+      "step": 107,
+      "tokens/total": 1615928,
+      "tokens/train_per_sec_per_gpu": 9.51,
+      "tokens/trainable": 34586
+    },
+    {
+      "epoch": 1.6875,
+      "grad_norm": 1.337603211402893,
+      "learning_rate": 9.99985339990821e-05,
+      "loss": 0.02129254676401615,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.02152,
+      "step": 108,
+      "tokens/total": 1631080,
+      "tokens/train_per_sec_per_gpu": 9.56,
+      "tokens/trainable": 34909
+    },
+    {
+      "epoch": 1.703125,
+      "grad_norm": 0.18884946405887604,
+      "learning_rate": 9.999788896372159e-05,
+      "loss": 0.0026449498254805803,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.00265,
+      "step": 109,
+      "tokens/total": 1646176,
+      "tokens/train_per_sec_per_gpu": 9.28,
+      "tokens/trainable": 35231
+    },
+    {
+      "epoch": 1.71875,
+      "grad_norm": 1.0585309267044067,
+      "learning_rate": 9.999712665317819e-05,
+      "loss": 0.0119435153901577,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.01202,
+      "step": 110,
+      "tokens/total": 1661304,
+      "tokens/train_per_sec_per_gpu": 9.58,
+      "tokens/trainable": 35558
+    },
+    {
+      "epoch": 1.734375,
+      "grad_norm": 0.3974311947822571,
+      "learning_rate": 9.999624706943868e-05,
+      "loss": 0.0038674657698720694,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.00387,
+      "step": 111,
+      "tokens/total": 1676396,
+      "tokens/train_per_sec_per_gpu": 9.77,
+      "tokens/trainable": 35884
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 0.7487826943397522,
+      "learning_rate": 9.999525021479544e-05,
+      "loss": 0.010628800839185715,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.01069,
+      "step": 112,
+      "tokens/total": 1691460,
+      "tokens/train_per_sec_per_gpu": 9.34,
+      "tokens/trainable": 36210
+    },
+    {
+      "epoch": 1.765625,
+      "grad_norm": 0.35798656940460205,
+      "learning_rate": 9.99941360918465e-05,
+      "loss": 0.0034732844214886427,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.86,
+      "memory/max_allocated (GiB)": 32.86,
+      "ppl": 1.00348,
+      "step": 113,
+      "tokens/total": 1706496,
+      "tokens/train_per_sec_per_gpu": 9.58,
+      "tokens/trainable": 36533
+    },
+    {
+      "epoch": 1.78125,
+      "grad_norm": 0.09821026027202606,
+      "learning_rate": 9.999290470349554e-05,
+      "loss": 0.0006078437436372042,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.9,
+      "memory/max_allocated (GiB)": 32.9,
+      "ppl": 1.00061,
+      "step": 114,
+      "tokens/total": 1721644,
+      "tokens/train_per_sec_per_gpu": 9.27,
+      "tokens/trainable": 36854
+    },
+    {
+      "epoch": 1.796875,
+      "grad_norm": 0.5430122017860413,
+      "learning_rate": 9.999155605295188e-05,
+      "loss": 0.005998114589601755,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.84,
+      "memory/max_allocated (GiB)": 32.84,
+      "ppl": 1.00602,
+      "step": 115,
+      "tokens/total": 1736692,
+      "tokens/train_per_sec_per_gpu": 9.15,
+      "tokens/trainable": 37176
+    },
+    {
+      "epoch": 1.8125,
+      "grad_norm": 4.369235515594482,
+      "learning_rate": 9.999009014373037e-05,
+      "loss": 0.026876695454120636,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.91,
+      "memory/max_allocated (GiB)": 32.91,
+      "ppl": 1.02724,
+      "step": 116,
+      "tokens/total": 1751844,
+      "tokens/train_per_sec_per_gpu": 8.78,
+      "tokens/trainable": 37498
+    },
+    {
+      "epoch": 1.828125,
+      "grad_norm": 0.1437114179134369,
+      "learning_rate": 9.998850697965153e-05,
+      "loss": 0.0015325542772188783,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.94,
+      "memory/max_allocated (GiB)": 32.94,
+      "ppl": 1.00153,
+      "step": 117,
+      "tokens/total": 1766996,
+      "tokens/train_per_sec_per_gpu": 9.75,
+      "tokens/trainable": 37825
+    },
+    {
+      "epoch": 1.84375,
+      "grad_norm": 3.2782604694366455,
+      "learning_rate": 9.99868065648415e-05,
+      "loss": 0.012170067057013512,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.01224,
+      "step": 118,
+      "tokens/total": 1781996,
+      "tokens/train_per_sec_per_gpu": 9.61,
+      "tokens/trainable": 38145
+    },
+    {
+      "epoch": 1.859375,
+      "grad_norm": 0.21812424063682556,
+      "learning_rate": 9.998498890373194e-05,
+      "loss": 0.0009808371542021632,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.89,
+      "memory/max_allocated (GiB)": 32.89,
+      "ppl": 1.00098,
+      "step": 119,
+      "tokens/total": 1797092,
+      "tokens/train_per_sec_per_gpu": 9.56,
+      "tokens/trainable": 38469
+    },
+    {
+      "epoch": 1.875,
+      "grad_norm": 1.1771652698516846,
+      "learning_rate": 9.998305400106009e-05,
+      "loss": 0.023559479042887688,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.91,
+      "memory/max_allocated (GiB)": 32.91,
+      "ppl": 1.02384,
+      "step": 120,
+      "tokens/total": 1812120,
+      "tokens/train_per_sec_per_gpu": 9.14,
+      "tokens/trainable": 38793
+    },
+    {
+      "epoch": 1.890625,
+      "grad_norm": 0.26324164867401123,
+      "learning_rate": 9.998100186186879e-05,
+      "loss": 0.0008098048274405301,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.92,
+      "memory/max_allocated (GiB)": 32.92,
+      "ppl": 1.00081,
+      "step": 121,
+      "tokens/total": 1827244,
+      "tokens/train_per_sec_per_gpu": 9.55,
+      "tokens/trainable": 39119
+    },
+    {
+      "epoch": 1.90625,
+      "grad_norm": 1.4897570610046387,
+      "learning_rate": 9.997883249150642e-05,
+      "loss": 0.011786782182753086,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.94,
+      "memory/max_allocated (GiB)": 32.94,
+      "ppl": 1.01186,
+      "step": 122,
+      "tokens/total": 1842452,
+      "tokens/train_per_sec_per_gpu": 9.25,
+      "tokens/trainable": 39445
+    },
+    {
+      "epoch": 1.921875,
+      "grad_norm": 1.2045236825942993,
+      "learning_rate": 9.99765458956268e-05,
+      "loss": 0.009280101396143436,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.85,
+      "memory/max_allocated (GiB)": 32.85,
+      "ppl": 1.00932,
+      "step": 123,
+      "tokens/total": 1857472,
+      "tokens/train_per_sec_per_gpu": 9.14,
+      "tokens/trainable": 39765
+    },
+    {
+      "epoch": 1.9375,
+      "grad_norm": 0.04349978268146515,
+      "learning_rate": 9.997414208018943e-05,
+      "loss": 0.0004488311242312193,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.88,
+      "memory/max_allocated (GiB)": 32.88,
+      "ppl": 1.00045,
+      "step": 124,
+      "tokens/total": 1872560,
+      "tokens/train_per_sec_per_gpu": 9.59,
+      "tokens/trainable": 40088
+    },
+    {
+      "epoch": 1.953125,
+      "grad_norm": 0.14775650203227997,
+      "learning_rate": 9.997162105145916e-05,
+      "loss": 0.0016349845100194216,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.95,
+      "memory/max_allocated (GiB)": 32.95,
+      "ppl": 1.00164,
+      "step": 125,
+      "tokens/total": 1887756,
+      "tokens/train_per_sec_per_gpu": 9.26,
+      "tokens/trainable": 40416
+    },
+    {
+      "epoch": 1.96875,
+      "grad_norm": 0.582293689250946,
+      "learning_rate": 9.996898281600642e-05,
+      "loss": 0.008356460370123386,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.91,
+      "memory/max_allocated (GiB)": 32.91,
+      "ppl": 1.00839,
+      "step": 126,
+      "tokens/total": 1902868,
+      "tokens/train_per_sec_per_gpu": 9.32,
+      "tokens/trainable": 40738
+    },
+    {
+      "epoch": 1.984375,
+      "grad_norm": 0.16206955909729004,
+      "learning_rate": 9.996622738070706e-05,
+      "loss": 0.0015829011099413037,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.91,
+      "memory/max_allocated (GiB)": 32.91,
+      "ppl": 1.00158,
+      "step": 127,
+      "tokens/total": 1917972,
+      "tokens/train_per_sec_per_gpu": 9.57,
+      "tokens/trainable": 41057
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.0851789116859436,
+      "learning_rate": 9.996335475274243e-05,
+      "loss": 0.000537098734639585,
+      "memory/device_reserved (GiB)": 34.19,
+      "memory/max_active (GiB)": 32.87,
+      "memory/max_allocated (GiB)": 32.87,
+      "ppl": 1.00054,
+      "step": 128,
+      "tokens/total": 1932952,
+      "tokens/train_per_sec_per_gpu": 9.6,
+      "tokens/trainable": 41380
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 2048,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 32,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.326501019236718e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}