Training in progress, step 348, checkpoint

.gitattributes

@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoint-348/tokenizer.json filter=lfs diff=lfs merge=lfs -text

checkpoint-348/README.md

@@ -0,0 +1,202 @@
+---
+base_model: Qwen/Qwen3-30B-A3B
+library_name: peft
+---
+
+# 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 0.15.2

checkpoint-348/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-30B-A3B",
+  "bias": "none",
+  "corda_config": null,
+  "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": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "k_proj",
+    "q_proj",
+    "gate_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-348/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fefb9ae46d523e5e0a0ad7ee9f2804bc4da7c0a59d1df2881f7fc33d2fc261ca
+size 6756184008

checkpoint-348/added_tokens.json

@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoint-348/global_step348/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c26224c2a2504f326ff8e6bc3674e2b4fda6aec49d24d90bc5aac7bfc5fcff9
+size 20362125410

checkpoint-348/global_step348/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d54b7da4cb4de6f59643cd609a1475186df00d7591f1a946567a82748f3e4ebe
+size 7316660630

checkpoint-348/latest

@@ -0,0 +1 @@
+global_step348

checkpoint-348/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3b848ab7ef0264d0508b072d69f710255cd0e7c030dac907e05d4a279a23d4e
+size 14244

checkpoint-348/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c26c90e5036214cbccdc225848656b5d5b067324cc261862b5f76b29c318bf58
+size 1064

checkpoint-348/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-348/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4
+size 11422654

checkpoint-348/tokenizer_config.json

@@ -0,0 +1,240 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% for message in messages %}{{'<|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|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-348/trainer_state.json

@@ -0,0 +1,2478 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.2003454231433506,
+  "eval_steps": 869,
+  "global_step": 348,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0005757052389176742,
+      "grad_norm": 2.557003974914551,
+      "learning_rate": 0.0,
+      "loss": 5.4277,
+      "step": 1
+    },
+    {
+      "epoch": 0.0005757052389176742,
+      "eval_loss": 5.319709300994873,
+      "eval_runtime": 1026.7022,
+      "eval_samples_per_second": 2.496,
+      "eval_steps_per_second": 2.496,
+      "step": 1
+    },
+    {
+      "epoch": 0.0011514104778353484,
+      "grad_norm": 2.985229969024658,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 5.7019,
+      "step": 2
+    },
+    {
+      "epoch": 0.0017271157167530224,
+      "grad_norm": 3.0353081226348877,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 6.1934,
+      "step": 3
+    },
+    {
+      "epoch": 0.002302820955670697,
+      "grad_norm": 3.724905490875244,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 5.4617,
+      "step": 4
+    },
+    {
+      "epoch": 0.0028785261945883708,
+      "grad_norm": 2.6505627632141113,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 5.4285,
+      "step": 5
+    },
+    {
+      "epoch": 0.0034542314335060447,
+      "grad_norm": 2.7363409996032715,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 5.8634,
+      "step": 6
+    },
+    {
+      "epoch": 0.004029936672423719,
+      "grad_norm": 3.082538366317749,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 4.7461,
+      "step": 7
+    },
+    {
+      "epoch": 0.004605641911341394,
+      "grad_norm": 9.095250129699707,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 7.5703,
+      "step": 8
+    },
+    {
+      "epoch": 0.0051813471502590676,
+      "grad_norm": 2.2597923278808594,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 5.3631,
+      "step": 9
+    },
+    {
+      "epoch": 0.0057570523891767415,
+      "grad_norm": 5.053525924682617,
+      "learning_rate": 3.6e-06,
+      "loss": 6.0132,
+      "step": 10
+    },
+    {
+      "epoch": 0.0063327576280944155,
+      "grad_norm": 2.7407820224761963,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 5.9776,
+      "step": 11
+    },
+    {
+      "epoch": 0.0069084628670120895,
+      "grad_norm": 2.4892263412475586,
+      "learning_rate": 4.4e-06,
+      "loss": 5.524,
+      "step": 12
+    },
+    {
+      "epoch": 0.007484168105929764,
+      "grad_norm": 2.5302274227142334,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 5.8044,
+      "step": 13
+    },
+    {
+      "epoch": 0.008059873344847437,
+      "grad_norm": 2.992504358291626,
+      "learning_rate": 5.2e-06,
+      "loss": 6.0307,
+      "step": 14
+    },
+    {
+      "epoch": 0.008635578583765112,
+      "grad_norm": 4.081608295440674,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 4.6732,
+      "step": 15
+    },
+    {
+      "epoch": 0.009211283822682787,
+      "grad_norm": 2.33296799659729,
+      "learning_rate": 6e-06,
+      "loss": 4.6356,
+      "step": 16
+    },
+    {
+      "epoch": 0.00978698906160046,
+      "grad_norm": 2.798452854156494,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 5.2941,
+      "step": 17
+    },
+    {
+      "epoch": 0.010362694300518135,
+      "grad_norm": 2.290029525756836,
+      "learning_rate": 6.800000000000001e-06,
+      "loss": 4.9405,
+      "step": 18
+    },
+    {
+      "epoch": 0.010938399539435808,
+      "grad_norm": 3.2164740562438965,
+      "learning_rate": 7.2e-06,
+      "loss": 5.6711,
+      "step": 19
+    },
+    {
+      "epoch": 0.011514104778353483,
+      "grad_norm": 2.4481987953186035,
+      "learning_rate": 7.6e-06,
+      "loss": 5.0366,
+      "step": 20
+    },
+    {
+      "epoch": 0.012089810017271158,
+      "grad_norm": 3.398063898086548,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 5.9377,
+      "step": 21
+    },
+    {
+      "epoch": 0.012665515256188831,
+      "grad_norm": 2.3936686515808105,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 5.4237,
+      "step": 22
+    },
+    {
+      "epoch": 0.013241220495106506,
+      "grad_norm": 2.7233810424804688,
+      "learning_rate": 8.8e-06,
+      "loss": 5.6551,
+      "step": 23
+    },
+    {
+      "epoch": 0.013816925734024179,
+      "grad_norm": 2.9957566261291504,
+      "learning_rate": 9.2e-06,
+      "loss": 4.7701,
+      "step": 24
+    },
+    {
+      "epoch": 0.014392630972941854,
+      "grad_norm": 6.397132396697998,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 6.4459,
+      "step": 25
+    },
+    {
+      "epoch": 0.014968336211859529,
+      "grad_norm": 3.0593409538269043,
+      "learning_rate": 1e-05,
+      "loss": 5.2758,
+      "step": 26
+    },
+    {
+      "epoch": 0.015544041450777202,
+      "grad_norm": 2.9723803997039795,
+      "learning_rate": 1.04e-05,
+      "loss": 5.6136,
+      "step": 27
+    },
+    {
+      "epoch": 0.016119746689694875,
+      "grad_norm": 2.03314471244812,
+      "learning_rate": 1.08e-05,
+      "loss": 5.3556,
+      "step": 28
+    },
+    {
+      "epoch": 0.01669545192861255,
+      "grad_norm": 1.777107834815979,
+      "learning_rate": 1.1200000000000001e-05,
+      "loss": 5.1061,
+      "step": 29
+    },
+    {
+      "epoch": 0.017271157167530225,
+      "grad_norm": 3.2192044258117676,
+      "learning_rate": 1.16e-05,
+      "loss": 5.2414,
+      "step": 30
+    },
+    {
+      "epoch": 0.017846862406447898,
+      "grad_norm": 3.924452066421509,
+      "learning_rate": 1.2e-05,
+      "loss": 5.2754,
+      "step": 31
+    },
+    {
+      "epoch": 0.018422567645365574,
+      "grad_norm": 3.5611093044281006,
+      "learning_rate": 1.24e-05,
+      "loss": 5.2817,
+      "step": 32
+    },
+    {
+      "epoch": 0.018998272884283247,
+      "grad_norm": 2.5194263458251953,
+      "learning_rate": 1.2800000000000001e-05,
+      "loss": 5.9063,
+      "step": 33
+    },
+    {
+      "epoch": 0.01957397812320092,
+      "grad_norm": 2.403895854949951,
+      "learning_rate": 1.32e-05,
+      "loss": 5.1161,
+      "step": 34
+    },
+    {
+      "epoch": 0.020149683362118594,
+      "grad_norm": 2.496400833129883,
+      "learning_rate": 1.3600000000000002e-05,
+      "loss": 5.3049,
+      "step": 35
+    },
+    {
+      "epoch": 0.02072538860103627,
+      "grad_norm": 3.0970828533172607,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 5.5807,
+      "step": 36
+    },
+    {
+      "epoch": 0.021301093839953943,
+      "grad_norm": 3.941403388977051,
+      "learning_rate": 1.44e-05,
+      "loss": 6.0418,
+      "step": 37
+    },
+    {
+      "epoch": 0.021876799078871616,
+      "grad_norm": 2.291431188583374,
+      "learning_rate": 1.48e-05,
+      "loss": 4.3686,
+      "step": 38
+    },
+    {
+      "epoch": 0.022452504317789293,
+      "grad_norm": 2.783054828643799,
+      "learning_rate": 1.52e-05,
+      "loss": 5.15,
+      "step": 39
+    },
+    {
+      "epoch": 0.023028209556706966,
+      "grad_norm": 3.579267978668213,
+      "learning_rate": 1.56e-05,
+      "loss": 5.7507,
+      "step": 40
+    },
+    {
+      "epoch": 0.02360391479562464,
+      "grad_norm": 3.5277323722839355,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 6.112,
+      "step": 41
+    },
+    {
+      "epoch": 0.024179620034542316,
+      "grad_norm": 2.5100817680358887,
+      "learning_rate": 1.6400000000000002e-05,
+      "loss": 5.2133,
+      "step": 42
+    },
+    {
+      "epoch": 0.02475532527345999,
+      "grad_norm": 2.3821561336517334,
+      "learning_rate": 1.6800000000000002e-05,
+      "loss": 6.0345,
+      "step": 43
+    },
+    {
+      "epoch": 0.025331030512377662,
+      "grad_norm": 3.0675108432769775,
+      "learning_rate": 1.7199999999999998e-05,
+      "loss": 5.2294,
+      "step": 44
+    },
+    {
+      "epoch": 0.025906735751295335,
+      "grad_norm": 2.8790383338928223,
+      "learning_rate": 1.76e-05,
+      "loss": 5.6393,
+      "step": 45
+    },
+    {
+      "epoch": 0.02648244099021301,
+      "grad_norm": 3.3649141788482666,
+      "learning_rate": 1.8e-05,
+      "loss": 6.014,
+      "step": 46
+    },
+    {
+      "epoch": 0.027058146229130685,
+      "grad_norm": 3.4695286750793457,
+      "learning_rate": 1.84e-05,
+      "loss": 5.3457,
+      "step": 47
+    },
+    {
+      "epoch": 0.027633851468048358,
+      "grad_norm": 3.303622245788574,
+      "learning_rate": 1.88e-05,
+      "loss": 5.593,
+      "step": 48
+    },
+    {
+      "epoch": 0.028209556706966035,
+      "grad_norm": 2.481895923614502,
+      "learning_rate": 1.9200000000000003e-05,
+      "loss": 5.1439,
+      "step": 49
+    },
+    {
+      "epoch": 0.028785261945883708,
+      "grad_norm": 2.888579845428467,
+      "learning_rate": 1.9600000000000002e-05,
+      "loss": 4.6318,
+      "step": 50
+    },
+    {
+      "epoch": 0.02936096718480138,
+      "grad_norm": 3.4528300762176514,
+      "learning_rate": 2e-05,
+      "loss": 5.0376,
+      "step": 51
+    },
+    {
+      "epoch": 0.029936672423719057,
+      "grad_norm": 3.6751370429992676,
+      "learning_rate": 2.04e-05,
+      "loss": 4.9183,
+      "step": 52
+    },
+    {
+      "epoch": 0.03051237766263673,
+      "grad_norm": 3.382035970687866,
+      "learning_rate": 2.08e-05,
+      "loss": 5.499,
+      "step": 53
+    },
+    {
+      "epoch": 0.031088082901554404,
+      "grad_norm": 2.8802406787872314,
+      "learning_rate": 2.12e-05,
+      "loss": 5.3177,
+      "step": 54
+    },
+    {
+      "epoch": 0.03166378814047208,
+      "grad_norm": 6.158539772033691,
+      "learning_rate": 2.16e-05,
+      "loss": 6.2133,
+      "step": 55
+    },
+    {
+      "epoch": 0.03223949337938975,
+      "grad_norm": 2.599864959716797,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 5.3691,
+      "step": 56
+    },
+    {
+      "epoch": 0.03281519861830743,
+      "grad_norm": 3.4526188373565674,
+      "learning_rate": 2.2400000000000002e-05,
+      "loss": 5.3801,
+      "step": 57
+    },
+    {
+      "epoch": 0.0333909038572251,
+      "grad_norm": 9.494807243347168,
+      "learning_rate": 2.2800000000000002e-05,
+      "loss": 7.3116,
+      "step": 58
+    },
+    {
+      "epoch": 0.033966609096142776,
+      "grad_norm": 4.3456130027771,
+      "learning_rate": 2.32e-05,
+      "loss": 4.7467,
+      "step": 59
+    },
+    {
+      "epoch": 0.03454231433506045,
+      "grad_norm": 3.8471431732177734,
+      "learning_rate": 2.36e-05,
+      "loss": 5.2742,
+      "step": 60
+    },
+    {
+      "epoch": 0.03511801957397812,
+      "grad_norm": 3.985994815826416,
+      "learning_rate": 2.4e-05,
+      "loss": 5.4615,
+      "step": 61
+    },
+    {
+      "epoch": 0.035693724812895795,
+      "grad_norm": 9.588626861572266,
+      "learning_rate": 2.44e-05,
+      "loss": 6.8261,
+      "step": 62
+    },
+    {
+      "epoch": 0.03626943005181347,
+      "grad_norm": 5.3343915939331055,
+      "learning_rate": 2.48e-05,
+      "loss": 6.0899,
+      "step": 63
+    },
+    {
+      "epoch": 0.03684513529073115,
+      "grad_norm": 5.611617088317871,
+      "learning_rate": 2.5200000000000003e-05,
+      "loss": 6.4523,
+      "step": 64
+    },
+    {
+      "epoch": 0.03742084052964882,
+      "grad_norm": 4.497012615203857,
+      "learning_rate": 2.5600000000000002e-05,
+      "loss": 4.787,
+      "step": 65
+    },
+    {
+      "epoch": 0.037996545768566495,
+      "grad_norm": 5.032821178436279,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 5.6337,
+      "step": 66
+    },
+    {
+      "epoch": 0.03857225100748417,
+      "grad_norm": 3.732733726501465,
+      "learning_rate": 2.64e-05,
+      "loss": 5.5212,
+      "step": 67
+    },
+    {
+      "epoch": 0.03914795624640184,
+      "grad_norm": 4.3597517013549805,
+      "learning_rate": 2.6800000000000004e-05,
+      "loss": 4.647,
+      "step": 68
+    },
+    {
+      "epoch": 0.039723661485319514,
+      "grad_norm": 5.359225273132324,
+      "learning_rate": 2.7200000000000004e-05,
+      "loss": 5.7052,
+      "step": 69
+    },
+    {
+      "epoch": 0.04029936672423719,
+      "grad_norm": 4.9161601066589355,
+      "learning_rate": 2.7600000000000003e-05,
+      "loss": 5.3191,
+      "step": 70
+    },
+    {
+      "epoch": 0.04087507196315487,
+      "grad_norm": 4.137385368347168,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 5.1797,
+      "step": 71
+    },
+    {
+      "epoch": 0.04145077720207254,
+      "grad_norm": 4.728359699249268,
+      "learning_rate": 2.84e-05,
+      "loss": 5.1125,
+      "step": 72
+    },
+    {
+      "epoch": 0.042026482440990214,
+      "grad_norm": 4.568793773651123,
+      "learning_rate": 2.88e-05,
+      "loss": 5.7705,
+      "step": 73
+    },
+    {
+      "epoch": 0.04260218767990789,
+      "grad_norm": 4.931026935577393,
+      "learning_rate": 2.9199999999999998e-05,
+      "loss": 5.1052,
+      "step": 74
+    },
+    {
+      "epoch": 0.04317789291882556,
+      "grad_norm": 4.697461128234863,
+      "learning_rate": 2.96e-05,
+      "loss": 5.1404,
+      "step": 75
+    },
+    {
+      "epoch": 0.04375359815774323,
+      "grad_norm": 6.393320083618164,
+      "learning_rate": 3e-05,
+      "loss": 6.2212,
+      "step": 76
+    },
+    {
+      "epoch": 0.04432930339666091,
+      "grad_norm": 5.876922607421875,
+      "learning_rate": 3.04e-05,
+      "loss": 5.7775,
+      "step": 77
+    },
+    {
+      "epoch": 0.044905008635578586,
+      "grad_norm": 4.749701499938965,
+      "learning_rate": 3.08e-05,
+      "loss": 4.7321,
+      "step": 78
+    },
+    {
+      "epoch": 0.04548071387449626,
+      "grad_norm": 4.894115447998047,
+      "learning_rate": 3.12e-05,
+      "loss": 5.2017,
+      "step": 79
+    },
+    {
+      "epoch": 0.04605641911341393,
+      "grad_norm": 5.125804424285889,
+      "learning_rate": 3.16e-05,
+      "loss": 5.1661,
+      "step": 80
+    },
+    {
+      "epoch": 0.046632124352331605,
+      "grad_norm": 7.571075439453125,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 6.1439,
+      "step": 81
+    },
+    {
+      "epoch": 0.04720782959124928,
+      "grad_norm": 4.469061374664307,
+      "learning_rate": 3.24e-05,
+      "loss": 5.1732,
+      "step": 82
+    },
+    {
+      "epoch": 0.04778353483016695,
+      "grad_norm": 4.565371513366699,
+      "learning_rate": 3.2800000000000004e-05,
+      "loss": 5.4892,
+      "step": 83
+    },
+    {
+      "epoch": 0.04835924006908463,
+      "grad_norm": 5.844489097595215,
+      "learning_rate": 3.32e-05,
+      "loss": 5.875,
+      "step": 84
+    },
+    {
+      "epoch": 0.048934945308002305,
+      "grad_norm": 10.564720153808594,
+      "learning_rate": 3.3600000000000004e-05,
+      "loss": 5.9008,
+      "step": 85
+    },
+    {
+      "epoch": 0.04951065054691998,
+      "grad_norm": 6.923472881317139,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 5.4949,
+      "step": 86
+    },
+    {
+      "epoch": 0.05008635578583765,
+      "grad_norm": 6.902386665344238,
+      "learning_rate": 3.4399999999999996e-05,
+      "loss": 4.9801,
+      "step": 87
+    },
+    {
+      "epoch": 0.050662061024755324,
+      "grad_norm": 8.239148139953613,
+      "learning_rate": 3.48e-05,
+      "loss": 5.6578,
+      "step": 88
+    },
+    {
+      "epoch": 0.051237766263673,
+      "grad_norm": 6.162630081176758,
+      "learning_rate": 3.52e-05,
+      "loss": 4.9911,
+      "step": 89
+    },
+    {
+      "epoch": 0.05181347150259067,
+      "grad_norm": 7.2612433433532715,
+      "learning_rate": 3.56e-05,
+      "loss": 5.7976,
+      "step": 90
+    },
+    {
+      "epoch": 0.05238917674150835,
+      "grad_norm": 6.149419784545898,
+      "learning_rate": 3.6e-05,
+      "loss": 4.9756,
+      "step": 91
+    },
+    {
+      "epoch": 0.05296488198042602,
+      "grad_norm": 7.4116106033325195,
+      "learning_rate": 3.6400000000000004e-05,
+      "loss": 5.5805,
+      "step": 92
+    },
+    {
+      "epoch": 0.0535405872193437,
+      "grad_norm": 5.512300491333008,
+      "learning_rate": 3.68e-05,
+      "loss": 4.5575,
+      "step": 93
+    },
+    {
+      "epoch": 0.05411629245826137,
+      "grad_norm": 14.799551963806152,
+      "learning_rate": 3.72e-05,
+      "loss": 5.2244,
+      "step": 94
+    },
+    {
+      "epoch": 0.05469199769717904,
+      "grad_norm": 9.756938934326172,
+      "learning_rate": 3.76e-05,
+      "loss": 4.8444,
+      "step": 95
+    },
+    {
+      "epoch": 0.055267702936096716,
+      "grad_norm": 6.400147914886475,
+      "learning_rate": 3.8e-05,
+      "loss": 5.5091,
+      "step": 96
+    },
+    {
+      "epoch": 0.055843408175014396,
+      "grad_norm": 8.406181335449219,
+      "learning_rate": 3.8400000000000005e-05,
+      "loss": 5.2641,
+      "step": 97
+    },
+    {
+      "epoch": 0.05641911341393207,
+      "grad_norm": 6.860042572021484,
+      "learning_rate": 3.88e-05,
+      "loss": 5.2917,
+      "step": 98
+    },
+    {
+      "epoch": 0.05699481865284974,
+      "grad_norm": 7.542653560638428,
+      "learning_rate": 3.9200000000000004e-05,
+      "loss": 5.1584,
+      "step": 99
+    },
+    {
+      "epoch": 0.057570523891767415,
+      "grad_norm": 8.149137496948242,
+      "learning_rate": 3.960000000000001e-05,
+      "loss": 5.5326,
+      "step": 100
+    },
+    {
+      "epoch": 0.05814622913068509,
+      "grad_norm": 5.590121269226074,
+      "learning_rate": 4e-05,
+      "loss": 5.2789,
+      "step": 101
+    },
+    {
+      "epoch": 0.05872193436960276,
+      "grad_norm": 7.877676010131836,
+      "learning_rate": 4.0400000000000006e-05,
+      "loss": 4.8526,
+      "step": 102
+    },
+    {
+      "epoch": 0.059297639608520435,
+      "grad_norm": 5.773808479309082,
+      "learning_rate": 4.08e-05,
+      "loss": 5.033,
+      "step": 103
+    },
+    {
+      "epoch": 0.059873344847438115,
+      "grad_norm": 6.092824935913086,
+      "learning_rate": 4.12e-05,
+      "loss": 4.8936,
+      "step": 104
+    },
+    {
+      "epoch": 0.06044905008635579,
+      "grad_norm": 5.934675693511963,
+      "learning_rate": 4.16e-05,
+      "loss": 4.4764,
+      "step": 105
+    },
+    {
+      "epoch": 0.06102475532527346,
+      "grad_norm": 5.622652530670166,
+      "learning_rate": 4.2e-05,
+      "loss": 5.1344,
+      "step": 106
+    },
+    {
+      "epoch": 0.061600460564191134,
+      "grad_norm": 7.697418212890625,
+      "learning_rate": 4.24e-05,
+      "loss": 5.2087,
+      "step": 107
+    },
+    {
+      "epoch": 0.06217616580310881,
+      "grad_norm": 5.204082489013672,
+      "learning_rate": 4.2800000000000004e-05,
+      "loss": 4.6294,
+      "step": 108
+    },
+    {
+      "epoch": 0.06275187104202648,
+      "grad_norm": 6.288537979125977,
+      "learning_rate": 4.32e-05,
+      "loss": 5.3009,
+      "step": 109
+    },
+    {
+      "epoch": 0.06332757628094415,
+      "grad_norm": 6.717288017272949,
+      "learning_rate": 4.36e-05,
+      "loss": 5.5392,
+      "step": 110
+    },
+    {
+      "epoch": 0.06390328151986183,
+      "grad_norm": 5.432399272918701,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 4.3602,
+      "step": 111
+    },
+    {
+      "epoch": 0.0644789867587795,
+      "grad_norm": 6.823062896728516,
+      "learning_rate": 4.44e-05,
+      "loss": 5.7343,
+      "step": 112
+    },
+    {
+      "epoch": 0.06505469199769717,
+      "grad_norm": 6.532074928283691,
+      "learning_rate": 4.4800000000000005e-05,
+      "loss": 5.0605,
+      "step": 113
+    },
+    {
+      "epoch": 0.06563039723661486,
+      "grad_norm": 5.982126712799072,
+      "learning_rate": 4.52e-05,
+      "loss": 5.2182,
+      "step": 114
+    },
+    {
+      "epoch": 0.06620610247553253,
+      "grad_norm": 5.759943962097168,
+      "learning_rate": 4.5600000000000004e-05,
+      "loss": 4.9098,
+      "step": 115
+    },
+    {
+      "epoch": 0.0667818077144502,
+      "grad_norm": 5.147834300994873,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 4.8671,
+      "step": 116
+    },
+    {
+      "epoch": 0.06735751295336788,
+      "grad_norm": 8.015042304992676,
+      "learning_rate": 4.64e-05,
+      "loss": 5.7445,
+      "step": 117
+    },
+    {
+      "epoch": 0.06793321819228555,
+      "grad_norm": 7.161843299865723,
+      "learning_rate": 4.6800000000000006e-05,
+      "loss": 5.9092,
+      "step": 118
+    },
+    {
+      "epoch": 0.06850892343120323,
+      "grad_norm": 9.394163131713867,
+      "learning_rate": 4.72e-05,
+      "loss": 4.7243,
+      "step": 119
+    },
+    {
+      "epoch": 0.0690846286701209,
+      "grad_norm": 4.96219539642334,
+      "learning_rate": 4.76e-05,
+      "loss": 4.7233,
+      "step": 120
+    },
+    {
+      "epoch": 0.06966033390903857,
+      "grad_norm": 6.473387241363525,
+      "learning_rate": 4.8e-05,
+      "loss": 5.1295,
+      "step": 121
+    },
+    {
+      "epoch": 0.07023603914795624,
+      "grad_norm": 6.797422885894775,
+      "learning_rate": 4.8400000000000004e-05,
+      "loss": 4.7697,
+      "step": 122
+    },
+    {
+      "epoch": 0.07081174438687392,
+      "grad_norm": 6.656020641326904,
+      "learning_rate": 4.88e-05,
+      "loss": 5.2377,
+      "step": 123
+    },
+    {
+      "epoch": 0.07138744962579159,
+      "grad_norm": 5.552718639373779,
+      "learning_rate": 4.92e-05,
+      "loss": 4.4741,
+      "step": 124
+    },
+    {
+      "epoch": 0.07196315486470926,
+      "grad_norm": 6.101820468902588,
+      "learning_rate": 4.96e-05,
+      "loss": 4.4192,
+      "step": 125
+    },
+    {
+      "epoch": 0.07253886010362694,
+      "grad_norm": 7.695935249328613,
+      "learning_rate": 5e-05,
+      "loss": 5.4128,
+      "step": 126
+    },
+    {
+      "epoch": 0.07311456534254462,
+      "grad_norm": 6.9946208000183105,
+      "learning_rate": 5.0400000000000005e-05,
+      "loss": 5.4829,
+      "step": 127
+    },
+    {
+      "epoch": 0.0736902705814623,
+      "grad_norm": 16.10480308532715,
+      "learning_rate": 5.08e-05,
+      "loss": 4.6945,
+      "step": 128
+    },
+    {
+      "epoch": 0.07426597582037997,
+      "grad_norm": 5.313148021697998,
+      "learning_rate": 5.1200000000000004e-05,
+      "loss": 4.2429,
+      "step": 129
+    },
+    {
+      "epoch": 0.07484168105929764,
+      "grad_norm": 5.506260871887207,
+      "learning_rate": 5.16e-05,
+      "loss": 4.7241,
+      "step": 130
+    },
+    {
+      "epoch": 0.07541738629821532,
+      "grad_norm": 5.655925273895264,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 5.4156,
+      "step": 131
+    },
+    {
+      "epoch": 0.07599309153713299,
+      "grad_norm": 6.528857231140137,
+      "learning_rate": 5.2400000000000007e-05,
+      "loss": 5.3606,
+      "step": 132
+    },
+    {
+      "epoch": 0.07656879677605066,
+      "grad_norm": 5.360299110412598,
+      "learning_rate": 5.28e-05,
+      "loss": 5.0686,
+      "step": 133
+    },
+    {
+      "epoch": 0.07714450201496834,
+      "grad_norm": 5.301785945892334,
+      "learning_rate": 5.3200000000000006e-05,
+      "loss": 4.845,
+      "step": 134
+    },
+    {
+      "epoch": 0.07772020725388601,
+      "grad_norm": 4.986385345458984,
+      "learning_rate": 5.360000000000001e-05,
+      "loss": 5.1493,
+      "step": 135
+    },
+    {
+      "epoch": 0.07829591249280368,
+      "grad_norm": 5.200460433959961,
+      "learning_rate": 5.4000000000000005e-05,
+      "loss": 4.781,
+      "step": 136
+    },
+    {
+      "epoch": 0.07887161773172136,
+      "grad_norm": 7.154032230377197,
+      "learning_rate": 5.440000000000001e-05,
+      "loss": 5.8801,
+      "step": 137
+    },
+    {
+      "epoch": 0.07944732297063903,
+      "grad_norm": 4.641168117523193,
+      "learning_rate": 5.4800000000000004e-05,
+      "loss": 5.1929,
+      "step": 138
+    },
+    {
+      "epoch": 0.0800230282095567,
+      "grad_norm": 4.8809123039245605,
+      "learning_rate": 5.520000000000001e-05,
+      "loss": 5.0221,
+      "step": 139
+    },
+    {
+      "epoch": 0.08059873344847437,
+      "grad_norm": 5.0507402420043945,
+      "learning_rate": 5.560000000000001e-05,
+      "loss": 4.8543,
+      "step": 140
+    },
+    {
+      "epoch": 0.08117443868739206,
+      "grad_norm": 6.459733963012695,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 5.051,
+      "step": 141
+    },
+    {
+      "epoch": 0.08175014392630973,
+      "grad_norm": 6.107847690582275,
+      "learning_rate": 5.6399999999999995e-05,
+      "loss": 4.8338,
+      "step": 142
+    },
+    {
+      "epoch": 0.08232584916522741,
+      "grad_norm": 6.28361701965332,
+      "learning_rate": 5.68e-05,
+      "loss": 5.1373,
+      "step": 143
+    },
+    {
+      "epoch": 0.08290155440414508,
+      "grad_norm": 4.957414627075195,
+      "learning_rate": 5.72e-05,
+      "loss": 4.8154,
+      "step": 144
+    },
+    {
+      "epoch": 0.08347725964306275,
+      "grad_norm": 4.774332046508789,
+      "learning_rate": 5.76e-05,
+      "loss": 4.7262,
+      "step": 145
+    },
+    {
+      "epoch": 0.08405296488198043,
+      "grad_norm": 7.41762113571167,
+      "learning_rate": 5.8e-05,
+      "loss": 5.5137,
+      "step": 146
+    },
+    {
+      "epoch": 0.0846286701208981,
+      "grad_norm": 7.484424591064453,
+      "learning_rate": 5.8399999999999997e-05,
+      "loss": 5.766,
+      "step": 147
+    },
+    {
+      "epoch": 0.08520437535981577,
+      "grad_norm": 4.917182922363281,
+      "learning_rate": 5.88e-05,
+      "loss": 5.0193,
+      "step": 148
+    },
+    {
+      "epoch": 0.08578008059873345,
+      "grad_norm": 4.608645915985107,
+      "learning_rate": 5.92e-05,
+      "loss": 5.0873,
+      "step": 149
+    },
+    {
+      "epoch": 0.08635578583765112,
+      "grad_norm": 6.5947794914245605,
+      "learning_rate": 5.96e-05,
+      "loss": 4.9855,
+      "step": 150
+    },
+    {
+      "epoch": 0.08693149107656879,
+      "grad_norm": 3.8302507400512695,
+      "learning_rate": 6e-05,
+      "loss": 3.7953,
+      "step": 151
+    },
+    {
+      "epoch": 0.08750719631548647,
+      "grad_norm": 3.6352171897888184,
+      "learning_rate": 6.04e-05,
+      "loss": 4.1647,
+      "step": 152
+    },
+    {
+      "epoch": 0.08808290155440414,
+      "grad_norm": 4.818563461303711,
+      "learning_rate": 6.08e-05,
+      "loss": 4.2128,
+      "step": 153
+    },
+    {
+      "epoch": 0.08865860679332183,
+      "grad_norm": 7.7323503494262695,
+      "learning_rate": 6.12e-05,
+      "loss": 5.4562,
+      "step": 154
+    },
+    {
+      "epoch": 0.0892343120322395,
+      "grad_norm": 5.785284996032715,
+      "learning_rate": 6.16e-05,
+      "loss": 4.8956,
+      "step": 155
+    },
+    {
+      "epoch": 0.08981001727115717,
+      "grad_norm": 6.181385040283203,
+      "learning_rate": 6.2e-05,
+      "loss": 5.2373,
+      "step": 156
+    },
+    {
+      "epoch": 0.09038572251007485,
+      "grad_norm": 6.015028476715088,
+      "learning_rate": 6.24e-05,
+      "loss": 4.3663,
+      "step": 157
+    },
+    {
+      "epoch": 0.09096142774899252,
+      "grad_norm": 4.41657829284668,
+      "learning_rate": 6.280000000000001e-05,
+      "loss": 4.5991,
+      "step": 158
+    },
+    {
+      "epoch": 0.09153713298791019,
+      "grad_norm": 6.5107622146606445,
+      "learning_rate": 6.32e-05,
+      "loss": 4.8784,
+      "step": 159
+    },
+    {
+      "epoch": 0.09211283822682786,
+      "grad_norm": 4.11070442199707,
+      "learning_rate": 6.36e-05,
+      "loss": 4.6766,
+      "step": 160
+    },
+    {
+      "epoch": 0.09268854346574554,
+      "grad_norm": 8.204343795776367,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 5.5088,
+      "step": 161
+    },
+    {
+      "epoch": 0.09326424870466321,
+      "grad_norm": 3.9389288425445557,
+      "learning_rate": 6.440000000000001e-05,
+      "loss": 4.3476,
+      "step": 162
+    },
+    {
+      "epoch": 0.09383995394358088,
+      "grad_norm": 5.597643852233887,
+      "learning_rate": 6.48e-05,
+      "loss": 4.9976,
+      "step": 163
+    },
+    {
+      "epoch": 0.09441565918249856,
+      "grad_norm": 8.994287490844727,
+      "learning_rate": 6.52e-05,
+      "loss": 5.5959,
+      "step": 164
+    },
+    {
+      "epoch": 0.09499136442141623,
+      "grad_norm": 5.60779333114624,
+      "learning_rate": 6.560000000000001e-05,
+      "loss": 4.6283,
+      "step": 165
+    },
+    {
+      "epoch": 0.0955670696603339,
+      "grad_norm": 4.319982528686523,
+      "learning_rate": 6.6e-05,
+      "loss": 4.041,
+      "step": 166
+    },
+    {
+      "epoch": 0.09614277489925158,
+      "grad_norm": 5.684337615966797,
+      "learning_rate": 6.64e-05,
+      "loss": 4.8941,
+      "step": 167
+    },
+    {
+      "epoch": 0.09671848013816926,
+      "grad_norm": 3.872518539428711,
+      "learning_rate": 6.680000000000001e-05,
+      "loss": 4.2242,
+      "step": 168
+    },
+    {
+      "epoch": 0.09729418537708694,
+      "grad_norm": 4.826557636260986,
+      "learning_rate": 6.720000000000001e-05,
+      "loss": 4.8546,
+      "step": 169
+    },
+    {
+      "epoch": 0.09786989061600461,
+      "grad_norm": 4.660156726837158,
+      "learning_rate": 6.76e-05,
+      "loss": 4.3797,
+      "step": 170
+    },
+    {
+      "epoch": 0.09844559585492228,
+      "grad_norm": 4.616059303283691,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 4.7293,
+      "step": 171
+    },
+    {
+      "epoch": 0.09902130109383996,
+      "grad_norm": 7.685507774353027,
+      "learning_rate": 6.840000000000001e-05,
+      "loss": 5.6251,
+      "step": 172
+    },
+    {
+      "epoch": 0.09959700633275763,
+      "grad_norm": 7.424576282501221,
+      "learning_rate": 6.879999999999999e-05,
+      "loss": 4.8253,
+      "step": 173
+    },
+    {
+      "epoch": 0.1001727115716753,
+      "grad_norm": 4.379521369934082,
+      "learning_rate": 6.92e-05,
+      "loss": 4.5287,
+      "step": 174
+    },
+    {
+      "epoch": 0.10074841681059298,
+      "grad_norm": 4.753964424133301,
+      "learning_rate": 6.96e-05,
+      "loss": 4.5554,
+      "step": 175
+    },
+    {
+      "epoch": 0.10132412204951065,
+      "grad_norm": 4.559609413146973,
+      "learning_rate": 7e-05,
+      "loss": 4.5615,
+      "step": 176
+    },
+    {
+      "epoch": 0.10189982728842832,
+      "grad_norm": 5.178406238555908,
+      "learning_rate": 7.04e-05,
+      "loss": 4.6344,
+      "step": 177
+    },
+    {
+      "epoch": 0.102475532527346,
+      "grad_norm": 7.4183526039123535,
+      "learning_rate": 7.08e-05,
+      "loss": 4.5451,
+      "step": 178
+    },
+    {
+      "epoch": 0.10305123776626367,
+      "grad_norm": 5.832037448883057,
+      "learning_rate": 7.12e-05,
+      "loss": 4.7097,
+      "step": 179
+    },
+    {
+      "epoch": 0.10362694300518134,
+      "grad_norm": 4.9681925773620605,
+      "learning_rate": 7.16e-05,
+      "loss": 4.6288,
+      "step": 180
+    },
+    {
+      "epoch": 0.10420264824409903,
+      "grad_norm": 4.886664867401123,
+      "learning_rate": 7.2e-05,
+      "loss": 4.7019,
+      "step": 181
+    },
+    {
+      "epoch": 0.1047783534830167,
+      "grad_norm": 4.668741226196289,
+      "learning_rate": 7.24e-05,
+      "loss": 4.4534,
+      "step": 182
+    },
+    {
+      "epoch": 0.10535405872193437,
+      "grad_norm": 7.459389686584473,
+      "learning_rate": 7.280000000000001e-05,
+      "loss": 5.4758,
+      "step": 183
+    },
+    {
+      "epoch": 0.10592976396085205,
+      "grad_norm": 31.545869827270508,
+      "learning_rate": 7.32e-05,
+      "loss": 6.179,
+      "step": 184
+    },
+    {
+      "epoch": 0.10650546919976972,
+      "grad_norm": 9.739182472229004,
+      "learning_rate": 7.36e-05,
+      "loss": 4.9662,
+      "step": 185
+    },
+    {
+      "epoch": 0.1070811744386874,
+      "grad_norm": 4.12076997756958,
+      "learning_rate": 7.4e-05,
+      "loss": 3.88,
+      "step": 186
+    },
+    {
+      "epoch": 0.10765687967760507,
+      "grad_norm": 5.808717727661133,
+      "learning_rate": 7.44e-05,
+      "loss": 4.6157,
+      "step": 187
+    },
+    {
+      "epoch": 0.10823258491652274,
+      "grad_norm": 3.6208741664886475,
+      "learning_rate": 7.48e-05,
+      "loss": 3.9156,
+      "step": 188
+    },
+    {
+      "epoch": 0.10880829015544041,
+      "grad_norm": 4.674955368041992,
+      "learning_rate": 7.52e-05,
+      "loss": 4.4751,
+      "step": 189
+    },
+    {
+      "epoch": 0.10938399539435809,
+      "grad_norm": 5.331599235534668,
+      "learning_rate": 7.560000000000001e-05,
+      "loss": 4.3887,
+      "step": 190
+    },
+    {
+      "epoch": 0.10995970063327576,
+      "grad_norm": 5.1405534744262695,
+      "learning_rate": 7.6e-05,
+      "loss": 4.9114,
+      "step": 191
+    },
+    {
+      "epoch": 0.11053540587219343,
+      "grad_norm": 3.7066593170166016,
+      "learning_rate": 7.64e-05,
+      "loss": 3.8948,
+      "step": 192
+    },
+    {
+      "epoch": 0.1111111111111111,
+      "grad_norm": 5.185431003570557,
+      "learning_rate": 7.680000000000001e-05,
+      "loss": 4.232,
+      "step": 193
+    },
+    {
+      "epoch": 0.11168681635002879,
+      "grad_norm": 4.900607585906982,
+      "learning_rate": 7.72e-05,
+      "loss": 4.667,
+      "step": 194
+    },
+    {
+      "epoch": 0.11226252158894647,
+      "grad_norm": 5.091091632843018,
+      "learning_rate": 7.76e-05,
+      "loss": 4.3946,
+      "step": 195
+    },
+    {
+      "epoch": 0.11283822682786414,
+      "grad_norm": 4.859619617462158,
+      "learning_rate": 7.800000000000001e-05,
+      "loss": 4.6306,
+      "step": 196
+    },
+    {
+      "epoch": 0.11341393206678181,
+      "grad_norm": 3.544200897216797,
+      "learning_rate": 7.840000000000001e-05,
+      "loss": 4.2118,
+      "step": 197
+    },
+    {
+      "epoch": 0.11398963730569948,
+      "grad_norm": 8.28862190246582,
+      "learning_rate": 7.88e-05,
+      "loss": 4.4431,
+      "step": 198
+    },
+    {
+      "epoch": 0.11456534254461716,
+      "grad_norm": 6.373688220977783,
+      "learning_rate": 7.920000000000001e-05,
+      "loss": 4.7554,
+      "step": 199
+    },
+    {
+      "epoch": 0.11514104778353483,
+      "grad_norm": 6.8544392585754395,
+      "learning_rate": 7.960000000000001e-05,
+      "loss": 4.8723,
+      "step": 200
+    },
+    {
+      "epoch": 0.1157167530224525,
+      "grad_norm": 7.207869052886963,
+      "learning_rate": 8e-05,
+      "loss": 4.1096,
+      "step": 201
+    },
+    {
+      "epoch": 0.11629245826137018,
+      "grad_norm": 4.9073333740234375,
+      "learning_rate": 8.04e-05,
+      "loss": 3.6834,
+      "step": 202
+    },
+    {
+      "epoch": 0.11686816350028785,
+      "grad_norm": 6.523554801940918,
+      "learning_rate": 8.080000000000001e-05,
+      "loss": 4.4934,
+      "step": 203
+    },
+    {
+      "epoch": 0.11744386873920552,
+      "grad_norm": 9.581537246704102,
+      "learning_rate": 8.120000000000001e-05,
+      "loss": 4.8199,
+      "step": 204
+    },
+    {
+      "epoch": 0.1180195739781232,
+      "grad_norm": 5.319664001464844,
+      "learning_rate": 8.16e-05,
+      "loss": 4.0881,
+      "step": 205
+    },
+    {
+      "epoch": 0.11859527921704087,
+      "grad_norm": 7.609442710876465,
+      "learning_rate": 8.2e-05,
+      "loss": 5.1011,
+      "step": 206
+    },
+    {
+      "epoch": 0.11917098445595854,
+      "grad_norm": 5.437283515930176,
+      "learning_rate": 8.24e-05,
+      "loss": 4.7683,
+      "step": 207
+    },
+    {
+      "epoch": 0.11974668969487623,
+      "grad_norm": 9.015962600708008,
+      "learning_rate": 8.28e-05,
+      "loss": 5.1197,
+      "step": 208
+    },
+    {
+      "epoch": 0.1203223949337939,
+      "grad_norm": 5.41486120223999,
+      "learning_rate": 8.32e-05,
+      "loss": 4.2228,
+      "step": 209
+    },
+    {
+      "epoch": 0.12089810017271158,
+      "grad_norm": 4.068630218505859,
+      "learning_rate": 8.36e-05,
+      "loss": 3.9683,
+      "step": 210
+    },
+    {
+      "epoch": 0.12147380541162925,
+      "grad_norm": 4.818974494934082,
+      "learning_rate": 8.4e-05,
+      "loss": 4.3969,
+      "step": 211
+    },
+    {
+      "epoch": 0.12204951065054692,
+      "grad_norm": 8.309637069702148,
+      "learning_rate": 8.44e-05,
+      "loss": 4.8983,
+      "step": 212
+    },
+    {
+      "epoch": 0.1226252158894646,
+      "grad_norm": 5.997379302978516,
+      "learning_rate": 8.48e-05,
+      "loss": 4.6983,
+      "step": 213
+    },
+    {
+      "epoch": 0.12320092112838227,
+      "grad_norm": 6.416568279266357,
+      "learning_rate": 8.52e-05,
+      "loss": 4.6,
+      "step": 214
+    },
+    {
+      "epoch": 0.12377662636729994,
+      "grad_norm": 5.038214206695557,
+      "learning_rate": 8.560000000000001e-05,
+      "loss": 4.1803,
+      "step": 215
+    },
+    {
+      "epoch": 0.12435233160621761,
+      "grad_norm": 5.035988807678223,
+      "learning_rate": 8.6e-05,
+      "loss": 4.1585,
+      "step": 216
+    },
+    {
+      "epoch": 0.12492803684513529,
+      "grad_norm": 6.7663726806640625,
+      "learning_rate": 8.64e-05,
+      "loss": 4.4256,
+      "step": 217
+    },
+    {
+      "epoch": 0.12550374208405296,
+      "grad_norm": 5.394269943237305,
+      "learning_rate": 8.680000000000001e-05,
+      "loss": 3.9008,
+      "step": 218
+    },
+    {
+      "epoch": 0.12607944732297063,
+      "grad_norm": 5.4501800537109375,
+      "learning_rate": 8.72e-05,
+      "loss": 3.9869,
+      "step": 219
+    },
+    {
+      "epoch": 0.1266551525618883,
+      "grad_norm": 4.7380170822143555,
+      "learning_rate": 8.76e-05,
+      "loss": 4.0876,
+      "step": 220
+    },
+    {
+      "epoch": 0.12723085780080598,
+      "grad_norm": 6.059116840362549,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 4.147,
+      "step": 221
+    },
+    {
+      "epoch": 0.12780656303972365,
+      "grad_norm": 5.5021586418151855,
+      "learning_rate": 8.840000000000001e-05,
+      "loss": 4.4547,
+      "step": 222
+    },
+    {
+      "epoch": 0.12838226827864133,
+      "grad_norm": 4.760106563568115,
+      "learning_rate": 8.88e-05,
+      "loss": 4.075,
+      "step": 223
+    },
+    {
+      "epoch": 0.128957973517559,
+      "grad_norm": 7.5847649574279785,
+      "learning_rate": 8.92e-05,
+      "loss": 4.6163,
+      "step": 224
+    },
+    {
+      "epoch": 0.12953367875647667,
+      "grad_norm": 6.257955074310303,
+      "learning_rate": 8.960000000000001e-05,
+      "loss": 4.6043,
+      "step": 225
+    },
+    {
+      "epoch": 0.13010938399539435,
+      "grad_norm": 7.368046283721924,
+      "learning_rate": 9e-05,
+      "loss": 4.7961,
+      "step": 226
+    },
+    {
+      "epoch": 0.13068508923431202,
+      "grad_norm": 4.385096549987793,
+      "learning_rate": 9.04e-05,
+      "loss": 4.1968,
+      "step": 227
+    },
+    {
+      "epoch": 0.13126079447322972,
+      "grad_norm": 6.34293794631958,
+      "learning_rate": 9.080000000000001e-05,
+      "loss": 4.3076,
+      "step": 228
+    },
+    {
+      "epoch": 0.1318364997121474,
+      "grad_norm": 6.403743267059326,
+      "learning_rate": 9.120000000000001e-05,
+      "loss": 3.8917,
+      "step": 229
+    },
+    {
+      "epoch": 0.13241220495106507,
+      "grad_norm": 6.792156219482422,
+      "learning_rate": 9.16e-05,
+      "loss": 3.9843,
+      "step": 230
+    },
+    {
+      "epoch": 0.13298791018998274,
+      "grad_norm": 8.062408447265625,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 4.2562,
+      "step": 231
+    },
+    {
+      "epoch": 0.1335636154289004,
+      "grad_norm": 8.513936042785645,
+      "learning_rate": 9.240000000000001e-05,
+      "loss": 4.6536,
+      "step": 232
+    },
+    {
+      "epoch": 0.13413932066781808,
+      "grad_norm": 5.92789363861084,
+      "learning_rate": 9.28e-05,
+      "loss": 4.104,
+      "step": 233
+    },
+    {
+      "epoch": 0.13471502590673576,
+      "grad_norm": 44.009300231933594,
+      "learning_rate": 9.320000000000002e-05,
+      "loss": 4.8297,
+      "step": 234
+    },
+    {
+      "epoch": 0.13529073114565343,
+      "grad_norm": 5.342921257019043,
+      "learning_rate": 9.360000000000001e-05,
+      "loss": 4.0662,
+      "step": 235
+    },
+    {
+      "epoch": 0.1358664363845711,
+      "grad_norm": 5.618771076202393,
+      "learning_rate": 9.4e-05,
+      "loss": 4.1692,
+      "step": 236
+    },
+    {
+      "epoch": 0.13644214162348878,
+      "grad_norm": 6.6655473709106445,
+      "learning_rate": 9.44e-05,
+      "loss": 4.2759,
+      "step": 237
+    },
+    {
+      "epoch": 0.13701784686240645,
+      "grad_norm": 6.415508270263672,
+      "learning_rate": 9.48e-05,
+      "loss": 4.025,
+      "step": 238
+    },
+    {
+      "epoch": 0.13759355210132412,
+      "grad_norm": 62.65280532836914,
+      "learning_rate": 9.52e-05,
+      "loss": 5.3187,
+      "step": 239
+    },
+    {
+      "epoch": 0.1381692573402418,
+      "grad_norm": 5.9870147705078125,
+      "learning_rate": 9.56e-05,
+      "loss": 4.3549,
+      "step": 240
+    },
+    {
+      "epoch": 0.13874496257915947,
+      "grad_norm": 6.323814868927002,
+      "learning_rate": 9.6e-05,
+      "loss": 4.0618,
+      "step": 241
+    },
+    {
+      "epoch": 0.13932066781807714,
+      "grad_norm": 7.25873327255249,
+      "learning_rate": 9.64e-05,
+      "loss": 4.6113,
+      "step": 242
+    },
+    {
+      "epoch": 0.13989637305699482,
+      "grad_norm": 6.708962440490723,
+      "learning_rate": 9.680000000000001e-05,
+      "loss": 4.2734,
+      "step": 243
+    },
+    {
+      "epoch": 0.1404720782959125,
+      "grad_norm": 6.766256332397461,
+      "learning_rate": 9.72e-05,
+      "loss": 3.8169,
+      "step": 244
+    },
+    {
+      "epoch": 0.14104778353483016,
+      "grad_norm": 9.25779914855957,
+      "learning_rate": 9.76e-05,
+      "loss": 4.0823,
+      "step": 245
+    },
+    {
+      "epoch": 0.14162348877374784,
+      "grad_norm": 6.24402379989624,
+      "learning_rate": 9.8e-05,
+      "loss": 3.9761,
+      "step": 246
+    },
+    {
+      "epoch": 0.1421991940126655,
+      "grad_norm": 4.627258777618408,
+      "learning_rate": 9.84e-05,
+      "loss": 3.3376,
+      "step": 247
+    },
+    {
+      "epoch": 0.14277489925158318,
+      "grad_norm": 6.5364766120910645,
+      "learning_rate": 9.88e-05,
+      "loss": 3.9101,
+      "step": 248
+    },
+    {
+      "epoch": 0.14335060449050085,
+      "grad_norm": 6.722381591796875,
+      "learning_rate": 9.92e-05,
+      "loss": 4.2916,
+      "step": 249
+    },
+    {
+      "epoch": 0.14392630972941853,
+      "grad_norm": 7.2800493240356445,
+      "learning_rate": 9.960000000000001e-05,
+      "loss": 4.1714,
+      "step": 250
+    },
+    {
+      "epoch": 0.1445020149683362,
+      "grad_norm": 9.137832641601562,
+      "learning_rate": 0.0001,
+      "loss": 3.9733,
+      "step": 251
+    },
+    {
+      "epoch": 0.14507772020725387,
+      "grad_norm": 5.290084362030029,
+      "learning_rate": 0.0001004,
+      "loss": 3.8465,
+      "step": 252
+    },
+    {
+      "epoch": 0.14565342544617155,
+      "grad_norm": 7.146475791931152,
+      "learning_rate": 0.00010080000000000001,
+      "loss": 4.154,
+      "step": 253
+    },
+    {
+      "epoch": 0.14622913068508925,
+      "grad_norm": 5.462000370025635,
+      "learning_rate": 0.00010120000000000001,
+      "loss": 3.8403,
+      "step": 254
+    },
+    {
+      "epoch": 0.14680483592400692,
+      "grad_norm": 8.053996086120605,
+      "learning_rate": 0.0001016,
+      "loss": 4.224,
+      "step": 255
+    },
+    {
+      "epoch": 0.1473805411629246,
+      "grad_norm": 56.904518127441406,
+      "learning_rate": 0.00010200000000000001,
+      "loss": 5.3512,
+      "step": 256
+    },
+    {
+      "epoch": 0.14795624640184227,
+      "grad_norm": 67.7396469116211,
+      "learning_rate": 0.00010240000000000001,
+      "loss": 4.136,
+      "step": 257
+    },
+    {
+      "epoch": 0.14853195164075994,
+      "grad_norm": 5.19423770904541,
+      "learning_rate": 0.0001028,
+      "loss": 3.6272,
+      "step": 258
+    },
+    {
+      "epoch": 0.1491076568796776,
+      "grad_norm": 6.946446418762207,
+      "learning_rate": 0.0001032,
+      "loss": 3.7617,
+      "step": 259
+    },
+    {
+      "epoch": 0.1496833621185953,
+      "grad_norm": 6.839754104614258,
+      "learning_rate": 0.00010360000000000001,
+      "loss": 4.2895,
+      "step": 260
+    },
+    {
+      "epoch": 0.15025906735751296,
+      "grad_norm": 7.3253254890441895,
+      "learning_rate": 0.00010400000000000001,
+      "loss": 4.0997,
+      "step": 261
+    },
+    {
+      "epoch": 0.15083477259643063,
+      "grad_norm": 6.981521129608154,
+      "learning_rate": 0.0001044,
+      "loss": 3.4663,
+      "step": 262
+    },
+    {
+      "epoch": 0.1514104778353483,
+      "grad_norm": 6.424066543579102,
+      "learning_rate": 0.00010480000000000001,
+      "loss": 4.0914,
+      "step": 263
+    },
+    {
+      "epoch": 0.15198618307426598,
+      "grad_norm": 6.7790398597717285,
+      "learning_rate": 0.00010520000000000001,
+      "loss": 4.0818,
+      "step": 264
+    },
+    {
+      "epoch": 0.15256188831318365,
+      "grad_norm": 7.887113094329834,
+      "learning_rate": 0.0001056,
+      "loss": 4.3784,
+      "step": 265
+    },
+    {
+      "epoch": 0.15313759355210133,
+      "grad_norm": 8.3016939163208,
+      "learning_rate": 0.00010600000000000002,
+      "loss": 3.7843,
+      "step": 266
+    },
+    {
+      "epoch": 0.153713298791019,
+      "grad_norm": 10.073237419128418,
+      "learning_rate": 0.00010640000000000001,
+      "loss": 4.0118,
+      "step": 267
+    },
+    {
+      "epoch": 0.15428900402993667,
+      "grad_norm": 6.9664106369018555,
+      "learning_rate": 0.00010680000000000001,
+      "loss": 3.8644,
+      "step": 268
+    },
+    {
+      "epoch": 0.15486470926885434,
+      "grad_norm": 8.479534149169922,
+      "learning_rate": 0.00010720000000000002,
+      "loss": 3.7009,
+      "step": 269
+    },
+    {
+      "epoch": 0.15544041450777202,
+      "grad_norm": 8.317602157592773,
+      "learning_rate": 0.00010760000000000001,
+      "loss": 3.7018,
+      "step": 270
+    },
+    {
+      "epoch": 0.1560161197466897,
+      "grad_norm": 6.020889759063721,
+      "learning_rate": 0.00010800000000000001,
+      "loss": 3.656,
+      "step": 271
+    },
+    {
+      "epoch": 0.15659182498560736,
+      "grad_norm": 7.147673606872559,
+      "learning_rate": 0.00010840000000000002,
+      "loss": 3.9216,
+      "step": 272
+    },
+    {
+      "epoch": 0.15716753022452504,
+      "grad_norm": 5.485556125640869,
+      "learning_rate": 0.00010880000000000002,
+      "loss": 3.4732,
+      "step": 273
+    },
+    {
+      "epoch": 0.1577432354634427,
+      "grad_norm": 7.432086944580078,
+      "learning_rate": 0.00010920000000000001,
+      "loss": 3.423,
+      "step": 274
+    },
+    {
+      "epoch": 0.15831894070236038,
+      "grad_norm": 6.897833824157715,
+      "learning_rate": 0.00010960000000000001,
+      "loss": 3.6169,
+      "step": 275
+    },
+    {
+      "epoch": 0.15889464594127806,
+      "grad_norm": 7.707437992095947,
+      "learning_rate": 0.00011000000000000002,
+      "loss": 3.6883,
+      "step": 276
+    },
+    {
+      "epoch": 0.15947035118019573,
+      "grad_norm": 5.546234607696533,
+      "learning_rate": 0.00011040000000000001,
+      "loss": 3.8388,
+      "step": 277
+    },
+    {
+      "epoch": 0.1600460564191134,
+      "grad_norm": 10.001431465148926,
+      "learning_rate": 0.00011080000000000001,
+      "loss": 3.372,
+      "step": 278
+    },
+    {
+      "epoch": 0.16062176165803108,
+      "grad_norm": 8.793180465698242,
+      "learning_rate": 0.00011120000000000002,
+      "loss": 3.7929,
+      "step": 279
+    },
+    {
+      "epoch": 0.16119746689694875,
+      "grad_norm": 8.189177513122559,
+      "learning_rate": 0.00011160000000000002,
+      "loss": 4.0091,
+      "step": 280
+    },
+    {
+      "epoch": 0.16177317213586645,
+      "grad_norm": 6.998697280883789,
+      "learning_rate": 0.00011200000000000001,
+      "loss": 3.648,
+      "step": 281
+    },
+    {
+      "epoch": 0.16234887737478412,
+      "grad_norm": 8.115317344665527,
+      "learning_rate": 0.00011240000000000002,
+      "loss": 4.0327,
+      "step": 282
+    },
+    {
+      "epoch": 0.1629245826137018,
+      "grad_norm": 7.597106456756592,
+      "learning_rate": 0.00011279999999999999,
+      "loss": 3.7811,
+      "step": 283
+    },
+    {
+      "epoch": 0.16350028785261947,
+      "grad_norm": 6.518374443054199,
+      "learning_rate": 0.0001132,
+      "loss": 3.3359,
+      "step": 284
+    },
+    {
+      "epoch": 0.16407599309153714,
+      "grad_norm": 6.962795257568359,
+      "learning_rate": 0.0001136,
+      "loss": 3.3726,
+      "step": 285
+    },
+    {
+      "epoch": 0.16465169833045482,
+      "grad_norm": 8.1845703125,
+      "learning_rate": 0.00011399999999999999,
+      "loss": 4.0042,
+      "step": 286
+    },
+    {
+      "epoch": 0.1652274035693725,
+      "grad_norm": 6.869271755218506,
+      "learning_rate": 0.0001144,
+      "loss": 3.4989,
+      "step": 287
+    },
+    {
+      "epoch": 0.16580310880829016,
+      "grad_norm": 12.261098861694336,
+      "learning_rate": 0.0001148,
+      "loss": 4.1045,
+      "step": 288
+    },
+    {
+      "epoch": 0.16637881404720783,
+      "grad_norm": 6.912962913513184,
+      "learning_rate": 0.0001152,
+      "loss": 3.6853,
+      "step": 289
+    },
+    {
+      "epoch": 0.1669545192861255,
+      "grad_norm": 8.545379638671875,
+      "learning_rate": 0.00011559999999999999,
+      "loss": 3.8903,
+      "step": 290
+    },
+    {
+      "epoch": 0.16753022452504318,
+      "grad_norm": 15.040228843688965,
+      "learning_rate": 0.000116,
+      "loss": 3.4079,
+      "step": 291
+    },
+    {
+      "epoch": 0.16810592976396085,
+      "grad_norm": 7.038132667541504,
+      "learning_rate": 0.0001164,
+      "loss": 3.7119,
+      "step": 292
+    },
+    {
+      "epoch": 0.16868163500287853,
+      "grad_norm": 6.259817123413086,
+      "learning_rate": 0.00011679999999999999,
+      "loss": 3.4931,
+      "step": 293
+    },
+    {
+      "epoch": 0.1692573402417962,
+      "grad_norm": 6.947351455688477,
+      "learning_rate": 0.0001172,
+      "loss": 3.677,
+      "step": 294
+    },
+    {
+      "epoch": 0.16983304548071387,
+      "grad_norm": 14.260014533996582,
+      "learning_rate": 0.0001176,
+      "loss": 3.9591,
+      "step": 295
+    },
+    {
+      "epoch": 0.17040875071963155,
+      "grad_norm": 6.70070743560791,
+      "learning_rate": 0.000118,
+      "loss": 3.2433,
+      "step": 296
+    },
+    {
+      "epoch": 0.17098445595854922,
+      "grad_norm": 11.697699546813965,
+      "learning_rate": 0.0001184,
+      "loss": 4.0909,
+      "step": 297
+    },
+    {
+      "epoch": 0.1715601611974669,
+      "grad_norm": 10.029029846191406,
+      "learning_rate": 0.0001188,
+      "loss": 3.5743,
+      "step": 298
+    },
+    {
+      "epoch": 0.17213586643638457,
+      "grad_norm": 6.6930365562438965,
+      "learning_rate": 0.0001192,
+      "loss": 3.2007,
+      "step": 299
+    },
+    {
+      "epoch": 0.17271157167530224,
+      "grad_norm": 21.772619247436523,
+      "learning_rate": 0.00011960000000000001,
+      "loss": 3.8505,
+      "step": 300
+    },
+    {
+      "epoch": 0.1732872769142199,
+      "grad_norm": 9.126256942749023,
+      "learning_rate": 0.00012,
+      "loss": 3.5777,
+      "step": 301
+    },
+    {
+      "epoch": 0.17386298215313759,
+      "grad_norm": 7.574469566345215,
+      "learning_rate": 0.0001204,
+      "loss": 3.5329,
+      "step": 302
+    },
+    {
+      "epoch": 0.17443868739205526,
+      "grad_norm": 6.436075687408447,
+      "learning_rate": 0.0001208,
+      "loss": 3.279,
+      "step": 303
+    },
+    {
+      "epoch": 0.17501439263097293,
+      "grad_norm": 5.945929527282715,
+      "learning_rate": 0.0001212,
+      "loss": 3.4338,
+      "step": 304
+    },
+    {
+      "epoch": 0.1755900978698906,
+      "grad_norm": 5.7057785987854,
+      "learning_rate": 0.0001216,
+      "loss": 3.2369,
+      "step": 305
+    },
+    {
+      "epoch": 0.17616580310880828,
+      "grad_norm": 9.411810874938965,
+      "learning_rate": 0.000122,
+      "loss": 3.5364,
+      "step": 306
+    },
+    {
+      "epoch": 0.17674150834772595,
+      "grad_norm": 8.872260093688965,
+      "learning_rate": 0.0001224,
+      "loss": 3.7803,
+      "step": 307
+    },
+    {
+      "epoch": 0.17731721358664365,
+      "grad_norm": 46.1115837097168,
+      "learning_rate": 0.0001228,
+      "loss": 3.7188,
+      "step": 308
+    },
+    {
+      "epoch": 0.17789291882556132,
+      "grad_norm": 48.33805465698242,
+      "learning_rate": 0.0001232,
+      "loss": 3.7491,
+      "step": 309
+    },
+    {
+      "epoch": 0.178468624064479,
+      "grad_norm": 7.272097587585449,
+      "learning_rate": 0.0001236,
+      "loss": 3.559,
+      "step": 310
+    },
+    {
+      "epoch": 0.17904432930339667,
+      "grad_norm": 7.471408367156982,
+      "learning_rate": 0.000124,
+      "loss": 3.6014,
+      "step": 311
+    },
+    {
+      "epoch": 0.17962003454231434,
+      "grad_norm": 11.095893859863281,
+      "learning_rate": 0.00012440000000000002,
+      "loss": 3.5741,
+      "step": 312
+    },
+    {
+      "epoch": 0.18019573978123202,
+      "grad_norm": 8.782601356506348,
+      "learning_rate": 0.0001248,
+      "loss": 3.2475,
+      "step": 313
+    },
+    {
+      "epoch": 0.1807714450201497,
+      "grad_norm": 7.485610485076904,
+      "learning_rate": 0.0001252,
+      "loss": 3.0304,
+      "step": 314
+    },
+    {
+      "epoch": 0.18134715025906736,
+      "grad_norm": 7.794425964355469,
+      "learning_rate": 0.00012560000000000002,
+      "loss": 2.9428,
+      "step": 315
+    },
+    {
+      "epoch": 0.18192285549798504,
+      "grad_norm": 6.470662593841553,
+      "learning_rate": 0.000126,
+      "loss": 3.4341,
+      "step": 316
+    },
+    {
+      "epoch": 0.1824985607369027,
+      "grad_norm": 10.054426193237305,
+      "learning_rate": 0.0001264,
+      "loss": 2.941,
+      "step": 317
+    },
+    {
+      "epoch": 0.18307426597582038,
+      "grad_norm": 93.38629150390625,
+      "learning_rate": 0.00012680000000000002,
+      "loss": 4.2291,
+      "step": 318
+    },
+    {
+      "epoch": 0.18364997121473806,
+      "grad_norm": 9.805968284606934,
+      "learning_rate": 0.0001272,
+      "loss": 3.0641,
+      "step": 319
+    },
+    {
+      "epoch": 0.18422567645365573,
+      "grad_norm": 6.104334831237793,
+      "learning_rate": 0.0001276,
+      "loss": 3.0856,
+      "step": 320
+    },
+    {
+      "epoch": 0.1848013816925734,
+      "grad_norm": 8.24195384979248,
+      "learning_rate": 0.00012800000000000002,
+      "loss": 3.0774,
+      "step": 321
+    },
+    {
+      "epoch": 0.18537708693149108,
+      "grad_norm": 6.327628135681152,
+      "learning_rate": 0.0001284,
+      "loss": 3.0826,
+      "step": 322
+    },
+    {
+      "epoch": 0.18595279217040875,
+      "grad_norm": 11.529990196228027,
+      "learning_rate": 0.00012880000000000001,
+      "loss": 3.7882,
+      "step": 323
+    },
+    {
+      "epoch": 0.18652849740932642,
+      "grad_norm": 9.700762748718262,
+      "learning_rate": 0.00012920000000000002,
+      "loss": 3.4958,
+      "step": 324
+    },
+    {
+      "epoch": 0.1871042026482441,
+      "grad_norm": 10.289152145385742,
+      "learning_rate": 0.0001296,
+      "loss": 3.3652,
+      "step": 325
+    },
+    {
+      "epoch": 0.18767990788716177,
+      "grad_norm": 6.888269901275635,
+      "learning_rate": 0.00013000000000000002,
+      "loss": 3.1086,
+      "step": 326
+    },
+    {
+      "epoch": 0.18825561312607944,
+      "grad_norm": 9.220719337463379,
+      "learning_rate": 0.0001304,
+      "loss": 3.5314,
+      "step": 327
+    },
+    {
+      "epoch": 0.1888313183649971,
+      "grad_norm": 9.044048309326172,
+      "learning_rate": 0.0001308,
+      "loss": 2.943,
+      "step": 328
+    },
+    {
+      "epoch": 0.1894070236039148,
+      "grad_norm": 11.338268280029297,
+      "learning_rate": 0.00013120000000000002,
+      "loss": 3.4617,
+      "step": 329
+    },
+    {
+      "epoch": 0.18998272884283246,
+      "grad_norm": 5.949525833129883,
+      "learning_rate": 0.0001316,
+      "loss": 2.8324,
+      "step": 330
+    },
+    {
+      "epoch": 0.19055843408175013,
+      "grad_norm": 9.158703804016113,
+      "learning_rate": 0.000132,
+      "loss": 3.1961,
+      "step": 331
+    },
+    {
+      "epoch": 0.1911341393206678,
+      "grad_norm": 8.708706855773926,
+      "learning_rate": 0.00013240000000000002,
+      "loss": 3.1941,
+      "step": 332
+    },
+    {
+      "epoch": 0.19170984455958548,
+      "grad_norm": 10.610583305358887,
+      "learning_rate": 0.0001328,
+      "loss": 3.3617,
+      "step": 333
+    },
+    {
+      "epoch": 0.19228554979850315,
+      "grad_norm": 8.023892402648926,
+      "learning_rate": 0.0001332,
+      "loss": 3.1775,
+      "step": 334
+    },
+    {
+      "epoch": 0.19286125503742085,
+      "grad_norm": 7.895623683929443,
+      "learning_rate": 0.00013360000000000002,
+      "loss": 3.1033,
+      "step": 335
+    },
+    {
+      "epoch": 0.19343696027633853,
+      "grad_norm": 6.376975059509277,
+      "learning_rate": 0.000134,
+      "loss": 2.808,
+      "step": 336
+    },
+    {
+      "epoch": 0.1940126655152562,
+      "grad_norm": 5.185142993927002,
+      "learning_rate": 0.00013440000000000001,
+      "loss": 2.8337,
+      "step": 337
+    },
+    {
+      "epoch": 0.19458837075417387,
+      "grad_norm": 6.408693790435791,
+      "learning_rate": 0.00013480000000000002,
+      "loss": 3.0604,
+      "step": 338
+    },
+    {
+      "epoch": 0.19516407599309155,
+      "grad_norm": 21.610239028930664,
+      "learning_rate": 0.0001352,
+      "loss": 3.431,
+      "step": 339
+    },
+    {
+      "epoch": 0.19573978123200922,
+      "grad_norm": 9.485398292541504,
+      "learning_rate": 0.00013560000000000002,
+      "loss": 3.2208,
+      "step": 340
+    },
+    {
+      "epoch": 0.1963154864709269,
+      "grad_norm": 6.460340976715088,
+      "learning_rate": 0.00013600000000000003,
+      "loss": 2.793,
+      "step": 341
+    },
+    {
+      "epoch": 0.19689119170984457,
+      "grad_norm": 5.64215612411499,
+      "learning_rate": 0.0001364,
+      "loss": 2.8589,
+      "step": 342
+    },
+    {
+      "epoch": 0.19746689694876224,
+      "grad_norm": 6.9033427238464355,
+      "learning_rate": 0.00013680000000000002,
+      "loss": 3.1031,
+      "step": 343
+    },
+    {
+      "epoch": 0.1980426021876799,
+      "grad_norm": 5.724493980407715,
+      "learning_rate": 0.00013720000000000003,
+      "loss": 2.8605,
+      "step": 344
+    },
+    {
+      "epoch": 0.19861830742659758,
+      "grad_norm": 15.779448509216309,
+      "learning_rate": 0.00013759999999999998,
+      "loss": 3.2151,
+      "step": 345
+    },
+    {
+      "epoch": 0.19919401266551526,
+      "grad_norm": 6.960752964019775,
+      "learning_rate": 0.000138,
+      "loss": 2.8537,
+      "step": 346
+    },
+    {
+      "epoch": 0.19976971790443293,
+      "grad_norm": 8.871850967407227,
+      "learning_rate": 0.0001384,
+      "loss": 2.7536,
+      "step": 347
+    },
+    {
+      "epoch": 0.2003454231433506,
+      "grad_norm": 6.670348644256592,
+      "learning_rate": 0.00013879999999999999,
+      "loss": 2.9525,
+      "step": 348
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 1737,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 348,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.022631528658895e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-348/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34e6c29f7b8116a03a0ae69dc83decfe763ea99a6fb5eb5ddd5b64f81364fadc
+size 8824

checkpoint-348/zero_to_fp32.py

@@ -0,0 +1,674 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import json
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering Sharded Weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        state_dict_split = split_torch_state_dict_into_shards(state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard, output_path)
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)