Training in progress, step 320, checkpoint

.gitattributes

@@ -35,3 +35,4 @@ saved_model/**/* 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-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text

checkpoint-320/README.md

@@ -0,0 +1,202 @@
+---
+base_model: Qwen/Qwen3-32B
+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-320/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-32B",
+  "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",
+    "q_proj",
+    "o_proj",
+    "k_proj",
+    "gate_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-320/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bb675c4729be3d3049706a7dace85ab59285e7d68764aaa0bdf6a3a11c48f0a
+size 1073864104

checkpoint-320/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-320/global_step320/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:807b98a35a0359d9f38b631ae5d55a8afd11035c94f424e1931a7cdd5a573e79
+size 3238110642

checkpoint-320/global_step320/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f9b8423678700968f54646b287f7db54e64f0cfc6c030b6a5a60b868dea14a2
+size 1571050088

checkpoint-320/latest

@@ -0,0 +1 @@
+global_step320

checkpoint-320/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1099ccf770f3fe86dbfa662c104e097f640d4bbd73dcf9a5be318f4b67c3b622
+size 14244

checkpoint-320/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d71e1b5d3ddab0a29620e2b9a137a888b2b80b3e9c9431339d971b5708ccaee
+size 1064

checkpoint-320/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-320/tokenizer.json

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

checkpoint-320/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-320/trainer_state.json

@@ -0,0 +1,2298 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.2011313639220616,
+  "eval_steps": 160,
+  "global_step": 320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0006285355122564425,
+      "grad_norm": 2.96633243560791,
+      "learning_rate": 0.0,
+      "loss": 7.5468,
+      "step": 1
+    },
+    {
+      "epoch": 0.0006285355122564425,
+      "eval_loss": 7.076071739196777,
+      "eval_runtime": 1556.6109,
+      "eval_samples_per_second": 1.656,
+      "eval_steps_per_second": 1.656,
+      "step": 1
+    },
+    {
+      "epoch": 0.001257071024512885,
+      "grad_norm": 2.7123501300811768,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 6.3309,
+      "step": 2
+    },
+    {
+      "epoch": 0.0018856065367693275,
+      "grad_norm": 2.4243223667144775,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 6.7763,
+      "step": 3
+    },
+    {
+      "epoch": 0.00251414204902577,
+      "grad_norm": 2.319216728210449,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 6.8187,
+      "step": 4
+    },
+    {
+      "epoch": 0.0031426775612822125,
+      "grad_norm": 2.4935989379882812,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 7.0526,
+      "step": 5
+    },
+    {
+      "epoch": 0.003771213073538655,
+      "grad_norm": 2.1718926429748535,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 6.7646,
+      "step": 6
+    },
+    {
+      "epoch": 0.0043997485857950975,
+      "grad_norm": 2.177217960357666,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 6.6131,
+      "step": 7
+    },
+    {
+      "epoch": 0.00502828409805154,
+      "grad_norm": 2.9915874004364014,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 7.6544,
+      "step": 8
+    },
+    {
+      "epoch": 0.0056568196103079825,
+      "grad_norm": 2.6981074810028076,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 6.799,
+      "step": 9
+    },
+    {
+      "epoch": 0.006285355122564425,
+      "grad_norm": 4.512426376342773,
+      "learning_rate": 3.6e-06,
+      "loss": 8.3895,
+      "step": 10
+    },
+    {
+      "epoch": 0.0069138906348208675,
+      "grad_norm": 2.3968913555145264,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 6.3478,
+      "step": 11
+    },
+    {
+      "epoch": 0.00754242614707731,
+      "grad_norm": 1.9744027853012085,
+      "learning_rate": 4.4e-06,
+      "loss": 5.5536,
+      "step": 12
+    },
+    {
+      "epoch": 0.008170961659333752,
+      "grad_norm": 3.050507068634033,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 6.5491,
+      "step": 13
+    },
+    {
+      "epoch": 0.008799497171590195,
+      "grad_norm": 3.0034799575805664,
+      "learning_rate": 5.2e-06,
+      "loss": 7.2764,
+      "step": 14
+    },
+    {
+      "epoch": 0.009428032683846637,
+      "grad_norm": 2.323613405227661,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 6.8805,
+      "step": 15
+    },
+    {
+      "epoch": 0.01005656819610308,
+      "grad_norm": 3.129593849182129,
+      "learning_rate": 6e-06,
+      "loss": 7.3171,
+      "step": 16
+    },
+    {
+      "epoch": 0.010685103708359522,
+      "grad_norm": 2.296137571334839,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 6.4251,
+      "step": 17
+    },
+    {
+      "epoch": 0.011313639220615965,
+      "grad_norm": 2.637282133102417,
+      "learning_rate": 6.800000000000001e-06,
+      "loss": 6.6142,
+      "step": 18
+    },
+    {
+      "epoch": 0.011942174732872407,
+      "grad_norm": 2.1313271522521973,
+      "learning_rate": 7.2e-06,
+      "loss": 6.3841,
+      "step": 19
+    },
+    {
+      "epoch": 0.01257071024512885,
+      "grad_norm": 2.7492284774780273,
+      "learning_rate": 7.6e-06,
+      "loss": 6.8148,
+      "step": 20
+    },
+    {
+      "epoch": 0.013199245757385292,
+      "grad_norm": 2.945878267288208,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 6.9732,
+      "step": 21
+    },
+    {
+      "epoch": 0.013827781269641735,
+      "grad_norm": 3.709951162338257,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 7.7273,
+      "step": 22
+    },
+    {
+      "epoch": 0.014456316781898177,
+      "grad_norm": 3.023289203643799,
+      "learning_rate": 8.8e-06,
+      "loss": 7.0649,
+      "step": 23
+    },
+    {
+      "epoch": 0.01508485229415462,
+      "grad_norm": 3.163715124130249,
+      "learning_rate": 9.2e-06,
+      "loss": 6.9342,
+      "step": 24
+    },
+    {
+      "epoch": 0.01571338780641106,
+      "grad_norm": 4.114445686340332,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 8.119,
+      "step": 25
+    },
+    {
+      "epoch": 0.016341923318667503,
+      "grad_norm": 3.021068572998047,
+      "learning_rate": 1e-05,
+      "loss": 7.4067,
+      "step": 26
+    },
+    {
+      "epoch": 0.01697045883092395,
+      "grad_norm": 3.724407911300659,
+      "learning_rate": 1.04e-05,
+      "loss": 7.3869,
+      "step": 27
+    },
+    {
+      "epoch": 0.01759899434318039,
+      "grad_norm": 2.656257390975952,
+      "learning_rate": 1.08e-05,
+      "loss": 5.9961,
+      "step": 28
+    },
+    {
+      "epoch": 0.018227529855436832,
+      "grad_norm": 2.7785143852233887,
+      "learning_rate": 1.1200000000000001e-05,
+      "loss": 5.6596,
+      "step": 29
+    },
+    {
+      "epoch": 0.018856065367693273,
+      "grad_norm": 3.130934715270996,
+      "learning_rate": 1.16e-05,
+      "loss": 6.6092,
+      "step": 30
+    },
+    {
+      "epoch": 0.01948460087994972,
+      "grad_norm": 3.42301869392395,
+      "learning_rate": 1.2e-05,
+      "loss": 6.3373,
+      "step": 31
+    },
+    {
+      "epoch": 0.02011313639220616,
+      "grad_norm": 2.8691611289978027,
+      "learning_rate": 1.24e-05,
+      "loss": 6.5923,
+      "step": 32
+    },
+    {
+      "epoch": 0.020741671904462602,
+      "grad_norm": 2.917086601257324,
+      "learning_rate": 1.2800000000000001e-05,
+      "loss": 5.9773,
+      "step": 33
+    },
+    {
+      "epoch": 0.021370207416719043,
+      "grad_norm": 4.07196044921875,
+      "learning_rate": 1.32e-05,
+      "loss": 7.4423,
+      "step": 34
+    },
+    {
+      "epoch": 0.02199874292897549,
+      "grad_norm": 4.738312244415283,
+      "learning_rate": 1.3600000000000002e-05,
+      "loss": 7.5113,
+      "step": 35
+    },
+    {
+      "epoch": 0.02262727844123193,
+      "grad_norm": 3.898664712905884,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 6.8738,
+      "step": 36
+    },
+    {
+      "epoch": 0.023255813953488372,
+      "grad_norm": 3.7448792457580566,
+      "learning_rate": 1.44e-05,
+      "loss": 6.9615,
+      "step": 37
+    },
+    {
+      "epoch": 0.023884349465744813,
+      "grad_norm": 3.5938379764556885,
+      "learning_rate": 1.48e-05,
+      "loss": 6.0651,
+      "step": 38
+    },
+    {
+      "epoch": 0.02451288497800126,
+      "grad_norm": 4.253636360168457,
+      "learning_rate": 1.52e-05,
+      "loss": 6.9986,
+      "step": 39
+    },
+    {
+      "epoch": 0.0251414204902577,
+      "grad_norm": 4.985451698303223,
+      "learning_rate": 1.56e-05,
+      "loss": 7.252,
+      "step": 40
+    },
+    {
+      "epoch": 0.025769956002514142,
+      "grad_norm": 4.376275062561035,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 6.8091,
+      "step": 41
+    },
+    {
+      "epoch": 0.026398491514770583,
+      "grad_norm": 4.697645664215088,
+      "learning_rate": 1.6400000000000002e-05,
+      "loss": 6.4319,
+      "step": 42
+    },
+    {
+      "epoch": 0.02702702702702703,
+      "grad_norm": 5.258227348327637,
+      "learning_rate": 1.6800000000000002e-05,
+      "loss": 6.7672,
+      "step": 43
+    },
+    {
+      "epoch": 0.02765556253928347,
+      "grad_norm": 5.063000679016113,
+      "learning_rate": 1.7199999999999998e-05,
+      "loss": 6.3354,
+      "step": 44
+    },
+    {
+      "epoch": 0.028284098051539912,
+      "grad_norm": 4.573636531829834,
+      "learning_rate": 1.76e-05,
+      "loss": 6.3374,
+      "step": 45
+    },
+    {
+      "epoch": 0.028912633563796353,
+      "grad_norm": 4.72340202331543,
+      "learning_rate": 1.8e-05,
+      "loss": 6.6553,
+      "step": 46
+    },
+    {
+      "epoch": 0.0295411690760528,
+      "grad_norm": 6.681248664855957,
+      "learning_rate": 1.84e-05,
+      "loss": 7.7157,
+      "step": 47
+    },
+    {
+      "epoch": 0.03016970458830924,
+      "grad_norm": 5.952408313751221,
+      "learning_rate": 1.88e-05,
+      "loss": 5.8215,
+      "step": 48
+    },
+    {
+      "epoch": 0.030798240100565682,
+      "grad_norm": 6.599308967590332,
+      "learning_rate": 1.9200000000000003e-05,
+      "loss": 6.921,
+      "step": 49
+    },
+    {
+      "epoch": 0.03142677561282212,
+      "grad_norm": 6.538867473602295,
+      "learning_rate": 1.9600000000000002e-05,
+      "loss": 7.1274,
+      "step": 50
+    },
+    {
+      "epoch": 0.03205531112507857,
+      "grad_norm": 5.91294527053833,
+      "learning_rate": 2e-05,
+      "loss": 6.7263,
+      "step": 51
+    },
+    {
+      "epoch": 0.03268384663733501,
+      "grad_norm": 7.943373203277588,
+      "learning_rate": 2.04e-05,
+      "loss": 7.4335,
+      "step": 52
+    },
+    {
+      "epoch": 0.03331238214959145,
+      "grad_norm": 7.023540496826172,
+      "learning_rate": 2.08e-05,
+      "loss": 6.3428,
+      "step": 53
+    },
+    {
+      "epoch": 0.0339409176618479,
+      "grad_norm": 7.031850814819336,
+      "learning_rate": 2.12e-05,
+      "loss": 6.2423,
+      "step": 54
+    },
+    {
+      "epoch": 0.034569453174104335,
+      "grad_norm": 6.891653537750244,
+      "learning_rate": 2.16e-05,
+      "loss": 6.4081,
+      "step": 55
+    },
+    {
+      "epoch": 0.03519798868636078,
+      "grad_norm": 8.165786743164062,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 7.1081,
+      "step": 56
+    },
+    {
+      "epoch": 0.035826524198617225,
+      "grad_norm": 9.146330833435059,
+      "learning_rate": 2.2400000000000002e-05,
+      "loss": 7.4309,
+      "step": 57
+    },
+    {
+      "epoch": 0.036455059710873663,
+      "grad_norm": 8.637526512145996,
+      "learning_rate": 2.2800000000000002e-05,
+      "loss": 6.9889,
+      "step": 58
+    },
+    {
+      "epoch": 0.03708359522313011,
+      "grad_norm": 8.397353172302246,
+      "learning_rate": 2.32e-05,
+      "loss": 5.8222,
+      "step": 59
+    },
+    {
+      "epoch": 0.03771213073538655,
+      "grad_norm": 9.219857215881348,
+      "learning_rate": 2.36e-05,
+      "loss": 7.0839,
+      "step": 60
+    },
+    {
+      "epoch": 0.03834066624764299,
+      "grad_norm": 8.762686729431152,
+      "learning_rate": 2.4e-05,
+      "loss": 6.192,
+      "step": 61
+    },
+    {
+      "epoch": 0.03896920175989944,
+      "grad_norm": 9.55370044708252,
+      "learning_rate": 2.44e-05,
+      "loss": 6.2674,
+      "step": 62
+    },
+    {
+      "epoch": 0.039597737272155875,
+      "grad_norm": 10.248711585998535,
+      "learning_rate": 2.48e-05,
+      "loss": 7.4737,
+      "step": 63
+    },
+    {
+      "epoch": 0.04022627278441232,
+      "grad_norm": 9.455551147460938,
+      "learning_rate": 2.5200000000000003e-05,
+      "loss": 6.5796,
+      "step": 64
+    },
+    {
+      "epoch": 0.04085480829666876,
+      "grad_norm": 10.217570304870605,
+      "learning_rate": 2.5600000000000002e-05,
+      "loss": 6.8329,
+      "step": 65
+    },
+    {
+      "epoch": 0.041483343808925204,
+      "grad_norm": 12.418697357177734,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 7.0127,
+      "step": 66
+    },
+    {
+      "epoch": 0.04211187932118165,
+      "grad_norm": 13.362143516540527,
+      "learning_rate": 2.64e-05,
+      "loss": 6.3027,
+      "step": 67
+    },
+    {
+      "epoch": 0.04274041483343809,
+      "grad_norm": 10.577826499938965,
+      "learning_rate": 2.6800000000000004e-05,
+      "loss": 6.2406,
+      "step": 68
+    },
+    {
+      "epoch": 0.04336895034569453,
+      "grad_norm": 13.15530776977539,
+      "learning_rate": 2.7200000000000004e-05,
+      "loss": 7.3093,
+      "step": 69
+    },
+    {
+      "epoch": 0.04399748585795098,
+      "grad_norm": 10.775976181030273,
+      "learning_rate": 2.7600000000000003e-05,
+      "loss": 5.9398,
+      "step": 70
+    },
+    {
+      "epoch": 0.044626021370207415,
+      "grad_norm": 12.925591468811035,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 6.7144,
+      "step": 71
+    },
+    {
+      "epoch": 0.04525455688246386,
+      "grad_norm": 12.626113891601562,
+      "learning_rate": 2.84e-05,
+      "loss": 6.5783,
+      "step": 72
+    },
+    {
+      "epoch": 0.0458830923947203,
+      "grad_norm": 14.04005241394043,
+      "learning_rate": 2.88e-05,
+      "loss": 6.1111,
+      "step": 73
+    },
+    {
+      "epoch": 0.046511627906976744,
+      "grad_norm": 14.279847145080566,
+      "learning_rate": 2.9199999999999998e-05,
+      "loss": 5.7717,
+      "step": 74
+    },
+    {
+      "epoch": 0.04714016341923319,
+      "grad_norm": 13.597288131713867,
+      "learning_rate": 2.96e-05,
+      "loss": 6.7714,
+      "step": 75
+    },
+    {
+      "epoch": 0.04776869893148963,
+      "grad_norm": 12.93455696105957,
+      "learning_rate": 3e-05,
+      "loss": 6.5082,
+      "step": 76
+    },
+    {
+      "epoch": 0.04839723444374607,
+      "grad_norm": 12.823902130126953,
+      "learning_rate": 3.04e-05,
+      "loss": 6.7108,
+      "step": 77
+    },
+    {
+      "epoch": 0.04902576995600252,
+      "grad_norm": 16.93589973449707,
+      "learning_rate": 3.08e-05,
+      "loss": 6.2773,
+      "step": 78
+    },
+    {
+      "epoch": 0.049654305468258955,
+      "grad_norm": 14.550610542297363,
+      "learning_rate": 3.12e-05,
+      "loss": 6.4329,
+      "step": 79
+    },
+    {
+      "epoch": 0.0502828409805154,
+      "grad_norm": 11.992222785949707,
+      "learning_rate": 3.16e-05,
+      "loss": 5.3369,
+      "step": 80
+    },
+    {
+      "epoch": 0.05091137649277184,
+      "grad_norm": 12.921990394592285,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 5.8783,
+      "step": 81
+    },
+    {
+      "epoch": 0.051539912005028284,
+      "grad_norm": 14.483002662658691,
+      "learning_rate": 3.24e-05,
+      "loss": 7.4091,
+      "step": 82
+    },
+    {
+      "epoch": 0.05216844751728473,
+      "grad_norm": 15.877086639404297,
+      "learning_rate": 3.2800000000000004e-05,
+      "loss": 6.9743,
+      "step": 83
+    },
+    {
+      "epoch": 0.05279698302954117,
+      "grad_norm": 15.85240650177002,
+      "learning_rate": 3.32e-05,
+      "loss": 7.0001,
+      "step": 84
+    },
+    {
+      "epoch": 0.05342551854179761,
+      "grad_norm": 17.0369815826416,
+      "learning_rate": 3.3600000000000004e-05,
+      "loss": 5.8022,
+      "step": 85
+    },
+    {
+      "epoch": 0.05405405405405406,
+      "grad_norm": 15.62733268737793,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 6.8172,
+      "step": 86
+    },
+    {
+      "epoch": 0.054682589566310495,
+      "grad_norm": 18.79668426513672,
+      "learning_rate": 3.4399999999999996e-05,
+      "loss": 6.6659,
+      "step": 87
+    },
+    {
+      "epoch": 0.05531112507856694,
+      "grad_norm": 15.78856086730957,
+      "learning_rate": 3.48e-05,
+      "loss": 5.7255,
+      "step": 88
+    },
+    {
+      "epoch": 0.05593966059082338,
+      "grad_norm": 13.196313858032227,
+      "learning_rate": 3.52e-05,
+      "loss": 5.8423,
+      "step": 89
+    },
+    {
+      "epoch": 0.056568196103079824,
+      "grad_norm": 16.4498233795166,
+      "learning_rate": 3.56e-05,
+      "loss": 5.4969,
+      "step": 90
+    },
+    {
+      "epoch": 0.05719673161533627,
+      "grad_norm": 15.243169784545898,
+      "learning_rate": 3.6e-05,
+      "loss": 6.0186,
+      "step": 91
+    },
+    {
+      "epoch": 0.05782526712759271,
+      "grad_norm": 17.294044494628906,
+      "learning_rate": 3.6400000000000004e-05,
+      "loss": 6.8408,
+      "step": 92
+    },
+    {
+      "epoch": 0.05845380263984915,
+      "grad_norm": 19.32530975341797,
+      "learning_rate": 3.68e-05,
+      "loss": 6.5562,
+      "step": 93
+    },
+    {
+      "epoch": 0.0590823381521056,
+      "grad_norm": 20.01514434814453,
+      "learning_rate": 3.72e-05,
+      "loss": 6.4273,
+      "step": 94
+    },
+    {
+      "epoch": 0.059710873664362035,
+      "grad_norm": 19.90467643737793,
+      "learning_rate": 3.76e-05,
+      "loss": 7.2277,
+      "step": 95
+    },
+    {
+      "epoch": 0.06033940917661848,
+      "grad_norm": 17.582975387573242,
+      "learning_rate": 3.8e-05,
+      "loss": 7.0173,
+      "step": 96
+    },
+    {
+      "epoch": 0.06096794468887492,
+      "grad_norm": 17.471969604492188,
+      "learning_rate": 3.8400000000000005e-05,
+      "loss": 6.6919,
+      "step": 97
+    },
+    {
+      "epoch": 0.061596480201131364,
+      "grad_norm": 16.06781005859375,
+      "learning_rate": 3.88e-05,
+      "loss": 6.54,
+      "step": 98
+    },
+    {
+      "epoch": 0.06222501571338781,
+      "grad_norm": 17.99164390563965,
+      "learning_rate": 3.9200000000000004e-05,
+      "loss": 5.9168,
+      "step": 99
+    },
+    {
+      "epoch": 0.06285355122564425,
+      "grad_norm": 16.01885986328125,
+      "learning_rate": 3.960000000000001e-05,
+      "loss": 5.9236,
+      "step": 100
+    },
+    {
+      "epoch": 0.06348208673790069,
+      "grad_norm": 9.705466270446777,
+      "learning_rate": 4e-05,
+      "loss": 4.4168,
+      "step": 101
+    },
+    {
+      "epoch": 0.06411062225015714,
+      "grad_norm": 21.02065086364746,
+      "learning_rate": 4.0400000000000006e-05,
+      "loss": 6.7312,
+      "step": 102
+    },
+    {
+      "epoch": 0.06473915776241358,
+      "grad_norm": 16.56360626220703,
+      "learning_rate": 4.08e-05,
+      "loss": 6.5865,
+      "step": 103
+    },
+    {
+      "epoch": 0.06536769327467001,
+      "grad_norm": 17.319807052612305,
+      "learning_rate": 4.12e-05,
+      "loss": 6.4122,
+      "step": 104
+    },
+    {
+      "epoch": 0.06599622878692646,
+      "grad_norm": 21.3114070892334,
+      "learning_rate": 4.16e-05,
+      "loss": 6.9886,
+      "step": 105
+    },
+    {
+      "epoch": 0.0666247642991829,
+      "grad_norm": 21.147327423095703,
+      "learning_rate": 4.2e-05,
+      "loss": 6.7595,
+      "step": 106
+    },
+    {
+      "epoch": 0.06725329981143935,
+      "grad_norm": 16.264938354492188,
+      "learning_rate": 4.24e-05,
+      "loss": 6.114,
+      "step": 107
+    },
+    {
+      "epoch": 0.0678818353236958,
+      "grad_norm": 20.424713134765625,
+      "learning_rate": 4.2800000000000004e-05,
+      "loss": 6.9508,
+      "step": 108
+    },
+    {
+      "epoch": 0.06851037083595223,
+      "grad_norm": 19.48565673828125,
+      "learning_rate": 4.32e-05,
+      "loss": 6.7488,
+      "step": 109
+    },
+    {
+      "epoch": 0.06913890634820867,
+      "grad_norm": 17.111894607543945,
+      "learning_rate": 4.36e-05,
+      "loss": 6.5613,
+      "step": 110
+    },
+    {
+      "epoch": 0.06976744186046512,
+      "grad_norm": 14.604440689086914,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 6.0278,
+      "step": 111
+    },
+    {
+      "epoch": 0.07039597737272156,
+      "grad_norm": 19.691083908081055,
+      "learning_rate": 4.44e-05,
+      "loss": 6.6524,
+      "step": 112
+    },
+    {
+      "epoch": 0.071024512884978,
+      "grad_norm": 20.216869354248047,
+      "learning_rate": 4.4800000000000005e-05,
+      "loss": 7.1919,
+      "step": 113
+    },
+    {
+      "epoch": 0.07165304839723445,
+      "grad_norm": 12.34801197052002,
+      "learning_rate": 4.52e-05,
+      "loss": 5.8613,
+      "step": 114
+    },
+    {
+      "epoch": 0.07228158390949088,
+      "grad_norm": 10.295191764831543,
+      "learning_rate": 4.5600000000000004e-05,
+      "loss": 5.2475,
+      "step": 115
+    },
+    {
+      "epoch": 0.07291011942174733,
+      "grad_norm": 15.049263954162598,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 6.1295,
+      "step": 116
+    },
+    {
+      "epoch": 0.07353865493400377,
+      "grad_norm": 12.287694931030273,
+      "learning_rate": 4.64e-05,
+      "loss": 5.7615,
+      "step": 117
+    },
+    {
+      "epoch": 0.07416719044626022,
+      "grad_norm": 12.11177921295166,
+      "learning_rate": 4.6800000000000006e-05,
+      "loss": 5.8763,
+      "step": 118
+    },
+    {
+      "epoch": 0.07479572595851666,
+      "grad_norm": 11.103803634643555,
+      "learning_rate": 4.72e-05,
+      "loss": 5.5072,
+      "step": 119
+    },
+    {
+      "epoch": 0.0754242614707731,
+      "grad_norm": 10.065237998962402,
+      "learning_rate": 4.76e-05,
+      "loss": 5.4734,
+      "step": 120
+    },
+    {
+      "epoch": 0.07605279698302954,
+      "grad_norm": 15.039546012878418,
+      "learning_rate": 4.8e-05,
+      "loss": 6.2005,
+      "step": 121
+    },
+    {
+      "epoch": 0.07668133249528598,
+      "grad_norm": 10.25068473815918,
+      "learning_rate": 4.8400000000000004e-05,
+      "loss": 5.4703,
+      "step": 122
+    },
+    {
+      "epoch": 0.07730986800754243,
+      "grad_norm": 13.106711387634277,
+      "learning_rate": 4.88e-05,
+      "loss": 6.6926,
+      "step": 123
+    },
+    {
+      "epoch": 0.07793840351979887,
+      "grad_norm": 7.927108287811279,
+      "learning_rate": 4.92e-05,
+      "loss": 5.3666,
+      "step": 124
+    },
+    {
+      "epoch": 0.0785669390320553,
+      "grad_norm": 10.937745094299316,
+      "learning_rate": 4.96e-05,
+      "loss": 6.1474,
+      "step": 125
+    },
+    {
+      "epoch": 0.07919547454431175,
+      "grad_norm": 10.88867473602295,
+      "learning_rate": 5e-05,
+      "loss": 6.1572,
+      "step": 126
+    },
+    {
+      "epoch": 0.0798240100565682,
+      "grad_norm": 11.629264831542969,
+      "learning_rate": 5.0400000000000005e-05,
+      "loss": 6.391,
+      "step": 127
+    },
+    {
+      "epoch": 0.08045254556882464,
+      "grad_norm": 10.296239852905273,
+      "learning_rate": 5.08e-05,
+      "loss": 5.9652,
+      "step": 128
+    },
+    {
+      "epoch": 0.08108108108108109,
+      "grad_norm": 11.038286209106445,
+      "learning_rate": 5.1200000000000004e-05,
+      "loss": 6.1295,
+      "step": 129
+    },
+    {
+      "epoch": 0.08170961659333752,
+      "grad_norm": 9.234803199768066,
+      "learning_rate": 5.16e-05,
+      "loss": 5.796,
+      "step": 130
+    },
+    {
+      "epoch": 0.08233815210559396,
+      "grad_norm": 8.689288139343262,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 5.3704,
+      "step": 131
+    },
+    {
+      "epoch": 0.08296668761785041,
+      "grad_norm": 15.41921615600586,
+      "learning_rate": 5.2400000000000007e-05,
+      "loss": 6.0926,
+      "step": 132
+    },
+    {
+      "epoch": 0.08359522313010685,
+      "grad_norm": 8.419554710388184,
+      "learning_rate": 5.28e-05,
+      "loss": 5.8071,
+      "step": 133
+    },
+    {
+      "epoch": 0.0842237586423633,
+      "grad_norm": 8.644979476928711,
+      "learning_rate": 5.3200000000000006e-05,
+      "loss": 5.9138,
+      "step": 134
+    },
+    {
+      "epoch": 0.08485229415461974,
+      "grad_norm": 11.236026763916016,
+      "learning_rate": 5.360000000000001e-05,
+      "loss": 5.9092,
+      "step": 135
+    },
+    {
+      "epoch": 0.08548082966687617,
+      "grad_norm": 9.669437408447266,
+      "learning_rate": 5.4000000000000005e-05,
+      "loss": 5.3778,
+      "step": 136
+    },
+    {
+      "epoch": 0.08610936517913262,
+      "grad_norm": 10.624286651611328,
+      "learning_rate": 5.440000000000001e-05,
+      "loss": 5.4098,
+      "step": 137
+    },
+    {
+      "epoch": 0.08673790069138906,
+      "grad_norm": 8.831917762756348,
+      "learning_rate": 5.4800000000000004e-05,
+      "loss": 5.3806,
+      "step": 138
+    },
+    {
+      "epoch": 0.08736643620364551,
+      "grad_norm": 8.536581993103027,
+      "learning_rate": 5.520000000000001e-05,
+      "loss": 5.9237,
+      "step": 139
+    },
+    {
+      "epoch": 0.08799497171590195,
+      "grad_norm": 7.873721599578857,
+      "learning_rate": 5.560000000000001e-05,
+      "loss": 5.423,
+      "step": 140
+    },
+    {
+      "epoch": 0.08862350722815839,
+      "grad_norm": 8.635188102722168,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 5.6118,
+      "step": 141
+    },
+    {
+      "epoch": 0.08925204274041483,
+      "grad_norm": 13.39542293548584,
+      "learning_rate": 5.6399999999999995e-05,
+      "loss": 5.2728,
+      "step": 142
+    },
+    {
+      "epoch": 0.08988057825267128,
+      "grad_norm": 12.998862266540527,
+      "learning_rate": 5.68e-05,
+      "loss": 6.9434,
+      "step": 143
+    },
+    {
+      "epoch": 0.09050911376492772,
+      "grad_norm": 8.149964332580566,
+      "learning_rate": 5.72e-05,
+      "loss": 5.3559,
+      "step": 144
+    },
+    {
+      "epoch": 0.09113764927718417,
+      "grad_norm": 7.66946268081665,
+      "learning_rate": 5.76e-05,
+      "loss": 5.5603,
+      "step": 145
+    },
+    {
+      "epoch": 0.0917661847894406,
+      "grad_norm": 9.277702331542969,
+      "learning_rate": 5.8e-05,
+      "loss": 4.9751,
+      "step": 146
+    },
+    {
+      "epoch": 0.09239472030169704,
+      "grad_norm": 7.442458152770996,
+      "learning_rate": 5.8399999999999997e-05,
+      "loss": 5.1653,
+      "step": 147
+    },
+    {
+      "epoch": 0.09302325581395349,
+      "grad_norm": 8.340120315551758,
+      "learning_rate": 5.88e-05,
+      "loss": 5.5422,
+      "step": 148
+    },
+    {
+      "epoch": 0.09365179132620993,
+      "grad_norm": 6.774919033050537,
+      "learning_rate": 5.92e-05,
+      "loss": 5.0187,
+      "step": 149
+    },
+    {
+      "epoch": 0.09428032683846638,
+      "grad_norm": 8.363201141357422,
+      "learning_rate": 5.96e-05,
+      "loss": 4.9803,
+      "step": 150
+    },
+    {
+      "epoch": 0.09490886235072282,
+      "grad_norm": 7.182234764099121,
+      "learning_rate": 6e-05,
+      "loss": 5.2913,
+      "step": 151
+    },
+    {
+      "epoch": 0.09553739786297925,
+      "grad_norm": 9.065616607666016,
+      "learning_rate": 6.04e-05,
+      "loss": 5.8604,
+      "step": 152
+    },
+    {
+      "epoch": 0.0961659333752357,
+      "grad_norm": 7.823053359985352,
+      "learning_rate": 6.08e-05,
+      "loss": 5.7326,
+      "step": 153
+    },
+    {
+      "epoch": 0.09679446888749214,
+      "grad_norm": 8.177785873413086,
+      "learning_rate": 6.12e-05,
+      "loss": 5.5663,
+      "step": 154
+    },
+    {
+      "epoch": 0.09742300439974859,
+      "grad_norm": 8.24718952178955,
+      "learning_rate": 6.16e-05,
+      "loss": 5.5606,
+      "step": 155
+    },
+    {
+      "epoch": 0.09805153991200503,
+      "grad_norm": 10.260727882385254,
+      "learning_rate": 6.2e-05,
+      "loss": 5.7043,
+      "step": 156
+    },
+    {
+      "epoch": 0.09868007542426147,
+      "grad_norm": 11.80507755279541,
+      "learning_rate": 6.24e-05,
+      "loss": 6.0268,
+      "step": 157
+    },
+    {
+      "epoch": 0.09930861093651791,
+      "grad_norm": 9.024588584899902,
+      "learning_rate": 6.280000000000001e-05,
+      "loss": 6.2249,
+      "step": 158
+    },
+    {
+      "epoch": 0.09993714644877436,
+      "grad_norm": 6.814016342163086,
+      "learning_rate": 6.32e-05,
+      "loss": 5.0123,
+      "step": 159
+    },
+    {
+      "epoch": 0.1005656819610308,
+      "grad_norm": 6.701868534088135,
+      "learning_rate": 6.36e-05,
+      "loss": 4.9993,
+      "step": 160
+    },
+    {
+      "epoch": 0.1005656819610308,
+      "eval_loss": 5.605074882507324,
+      "eval_runtime": 1565.1736,
+      "eval_samples_per_second": 1.647,
+      "eval_steps_per_second": 1.647,
+      "step": 160
+    },
+    {
+      "epoch": 0.10119421747328725,
+      "grad_norm": 14.352143287658691,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 5.8635,
+      "step": 161
+    },
+    {
+      "epoch": 0.10182275298554368,
+      "grad_norm": 7.681995868682861,
+      "learning_rate": 6.440000000000001e-05,
+      "loss": 5.2773,
+      "step": 162
+    },
+    {
+      "epoch": 0.10245128849780012,
+      "grad_norm": 10.032212257385254,
+      "learning_rate": 6.48e-05,
+      "loss": 5.4718,
+      "step": 163
+    },
+    {
+      "epoch": 0.10307982401005657,
+      "grad_norm": 6.617798805236816,
+      "learning_rate": 6.52e-05,
+      "loss": 4.7259,
+      "step": 164
+    },
+    {
+      "epoch": 0.10370835952231301,
+      "grad_norm": 7.318296432495117,
+      "learning_rate": 6.560000000000001e-05,
+      "loss": 4.9979,
+      "step": 165
+    },
+    {
+      "epoch": 0.10433689503456946,
+      "grad_norm": 9.494800567626953,
+      "learning_rate": 6.6e-05,
+      "loss": 5.4222,
+      "step": 166
+    },
+    {
+      "epoch": 0.1049654305468259,
+      "grad_norm": 6.2663798332214355,
+      "learning_rate": 6.64e-05,
+      "loss": 4.806,
+      "step": 167
+    },
+    {
+      "epoch": 0.10559396605908233,
+      "grad_norm": 7.666187286376953,
+      "learning_rate": 6.680000000000001e-05,
+      "loss": 5.583,
+      "step": 168
+    },
+    {
+      "epoch": 0.10622250157133878,
+      "grad_norm": 6.545588493347168,
+      "learning_rate": 6.720000000000001e-05,
+      "loss": 5.4927,
+      "step": 169
+    },
+    {
+      "epoch": 0.10685103708359522,
+      "grad_norm": 7.650016784667969,
+      "learning_rate": 6.76e-05,
+      "loss": 5.6347,
+      "step": 170
+    },
+    {
+      "epoch": 0.10747957259585167,
+      "grad_norm": 6.05757474899292,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 4.8323,
+      "step": 171
+    },
+    {
+      "epoch": 0.10810810810810811,
+      "grad_norm": 8.37306022644043,
+      "learning_rate": 6.840000000000001e-05,
+      "loss": 5.5312,
+      "step": 172
+    },
+    {
+      "epoch": 0.10873664362036455,
+      "grad_norm": 6.950132369995117,
+      "learning_rate": 6.879999999999999e-05,
+      "loss": 5.216,
+      "step": 173
+    },
+    {
+      "epoch": 0.10936517913262099,
+      "grad_norm": 7.643744945526123,
+      "learning_rate": 6.92e-05,
+      "loss": 5.374,
+      "step": 174
+    },
+    {
+      "epoch": 0.10999371464487744,
+      "grad_norm": 12.621402740478516,
+      "learning_rate": 6.96e-05,
+      "loss": 6.3553,
+      "step": 175
+    },
+    {
+      "epoch": 0.11062225015713388,
+      "grad_norm": 6.861866474151611,
+      "learning_rate": 7e-05,
+      "loss": 5.0186,
+      "step": 176
+    },
+    {
+      "epoch": 0.11125078566939033,
+      "grad_norm": 7.023342609405518,
+      "learning_rate": 7.04e-05,
+      "loss": 4.7693,
+      "step": 177
+    },
+    {
+      "epoch": 0.11187932118164676,
+      "grad_norm": 7.627071380615234,
+      "learning_rate": 7.08e-05,
+      "loss": 5.5354,
+      "step": 178
+    },
+    {
+      "epoch": 0.1125078566939032,
+      "grad_norm": 7.475245952606201,
+      "learning_rate": 7.12e-05,
+      "loss": 5.5916,
+      "step": 179
+    },
+    {
+      "epoch": 0.11313639220615965,
+      "grad_norm": 6.348559379577637,
+      "learning_rate": 7.16e-05,
+      "loss": 4.7914,
+      "step": 180
+    },
+    {
+      "epoch": 0.11376492771841609,
+      "grad_norm": 6.895409107208252,
+      "learning_rate": 7.2e-05,
+      "loss": 5.4136,
+      "step": 181
+    },
+    {
+      "epoch": 0.11439346323067254,
+      "grad_norm": 5.746918201446533,
+      "learning_rate": 7.24e-05,
+      "loss": 4.8071,
+      "step": 182
+    },
+    {
+      "epoch": 0.11502199874292897,
+      "grad_norm": 7.965058326721191,
+      "learning_rate": 7.280000000000001e-05,
+      "loss": 4.6486,
+      "step": 183
+    },
+    {
+      "epoch": 0.11565053425518541,
+      "grad_norm": 8.059706687927246,
+      "learning_rate": 7.32e-05,
+      "loss": 5.3917,
+      "step": 184
+    },
+    {
+      "epoch": 0.11627906976744186,
+      "grad_norm": 7.2255353927612305,
+      "learning_rate": 7.36e-05,
+      "loss": 5.0876,
+      "step": 185
+    },
+    {
+      "epoch": 0.1169076052796983,
+      "grad_norm": 6.53950834274292,
+      "learning_rate": 7.4e-05,
+      "loss": 4.7656,
+      "step": 186
+    },
+    {
+      "epoch": 0.11753614079195475,
+      "grad_norm": 6.379930019378662,
+      "learning_rate": 7.44e-05,
+      "loss": 4.6405,
+      "step": 187
+    },
+    {
+      "epoch": 0.1181646763042112,
+      "grad_norm": 7.332822322845459,
+      "learning_rate": 7.48e-05,
+      "loss": 4.6839,
+      "step": 188
+    },
+    {
+      "epoch": 0.11879321181646763,
+      "grad_norm": 7.975095272064209,
+      "learning_rate": 7.52e-05,
+      "loss": 4.8748,
+      "step": 189
+    },
+    {
+      "epoch": 0.11942174732872407,
+      "grad_norm": 7.8162360191345215,
+      "learning_rate": 7.560000000000001e-05,
+      "loss": 4.9994,
+      "step": 190
+    },
+    {
+      "epoch": 0.12005028284098052,
+      "grad_norm": 7.749350070953369,
+      "learning_rate": 7.6e-05,
+      "loss": 4.7996,
+      "step": 191
+    },
+    {
+      "epoch": 0.12067881835323696,
+      "grad_norm": 12.814620018005371,
+      "learning_rate": 7.64e-05,
+      "loss": 5.4858,
+      "step": 192
+    },
+    {
+      "epoch": 0.1213073538654934,
+      "grad_norm": 7.985275745391846,
+      "learning_rate": 7.680000000000001e-05,
+      "loss": 4.9178,
+      "step": 193
+    },
+    {
+      "epoch": 0.12193588937774984,
+      "grad_norm": 7.037015438079834,
+      "learning_rate": 7.72e-05,
+      "loss": 4.4436,
+      "step": 194
+    },
+    {
+      "epoch": 0.12256442489000628,
+      "grad_norm": 12.540748596191406,
+      "learning_rate": 7.76e-05,
+      "loss": 5.3947,
+      "step": 195
+    },
+    {
+      "epoch": 0.12319296040226273,
+      "grad_norm": 8.698318481445312,
+      "learning_rate": 7.800000000000001e-05,
+      "loss": 4.6521,
+      "step": 196
+    },
+    {
+      "epoch": 0.12382149591451917,
+      "grad_norm": 6.837226867675781,
+      "learning_rate": 7.840000000000001e-05,
+      "loss": 4.4502,
+      "step": 197
+    },
+    {
+      "epoch": 0.12445003142677562,
+      "grad_norm": 6.902530670166016,
+      "learning_rate": 7.88e-05,
+      "loss": 5.2111,
+      "step": 198
+    },
+    {
+      "epoch": 0.12507856693903205,
+      "grad_norm": 7.714503288269043,
+      "learning_rate": 7.920000000000001e-05,
+      "loss": 4.5983,
+      "step": 199
+    },
+    {
+      "epoch": 0.1257071024512885,
+      "grad_norm": 9.488588333129883,
+      "learning_rate": 7.960000000000001e-05,
+      "loss": 5.496,
+      "step": 200
+    },
+    {
+      "epoch": 0.12633563796354494,
+      "grad_norm": 9.41787052154541,
+      "learning_rate": 8e-05,
+      "loss": 4.8966,
+      "step": 201
+    },
+    {
+      "epoch": 0.12696417347580138,
+      "grad_norm": 7.63663911819458,
+      "learning_rate": 8.04e-05,
+      "loss": 4.6692,
+      "step": 202
+    },
+    {
+      "epoch": 0.12759270898805783,
+      "grad_norm": 8.535087585449219,
+      "learning_rate": 8.080000000000001e-05,
+      "loss": 5.0543,
+      "step": 203
+    },
+    {
+      "epoch": 0.12822124450031427,
+      "grad_norm": 8.822395324707031,
+      "learning_rate": 8.120000000000001e-05,
+      "loss": 4.7249,
+      "step": 204
+    },
+    {
+      "epoch": 0.12884978001257072,
+      "grad_norm": 7.7394561767578125,
+      "learning_rate": 8.16e-05,
+      "loss": 4.5474,
+      "step": 205
+    },
+    {
+      "epoch": 0.12947831552482716,
+      "grad_norm": 9.459196090698242,
+      "learning_rate": 8.2e-05,
+      "loss": 5.3748,
+      "step": 206
+    },
+    {
+      "epoch": 0.13010685103708358,
+      "grad_norm": 6.900613784790039,
+      "learning_rate": 8.24e-05,
+      "loss": 4.5716,
+      "step": 207
+    },
+    {
+      "epoch": 0.13073538654934003,
+      "grad_norm": 7.961105823516846,
+      "learning_rate": 8.28e-05,
+      "loss": 4.5067,
+      "step": 208
+    },
+    {
+      "epoch": 0.13136392206159647,
+      "grad_norm": 10.71306037902832,
+      "learning_rate": 8.32e-05,
+      "loss": 4.8309,
+      "step": 209
+    },
+    {
+      "epoch": 0.13199245757385292,
+      "grad_norm": 8.900117874145508,
+      "learning_rate": 8.36e-05,
+      "loss": 5.0913,
+      "step": 210
+    },
+    {
+      "epoch": 0.13262099308610936,
+      "grad_norm": 9.863144874572754,
+      "learning_rate": 8.4e-05,
+      "loss": 4.9469,
+      "step": 211
+    },
+    {
+      "epoch": 0.1332495285983658,
+      "grad_norm": 7.620909690856934,
+      "learning_rate": 8.44e-05,
+      "loss": 4.484,
+      "step": 212
+    },
+    {
+      "epoch": 0.13387806411062225,
+      "grad_norm": 10.667424201965332,
+      "learning_rate": 8.48e-05,
+      "loss": 5.1474,
+      "step": 213
+    },
+    {
+      "epoch": 0.1345065996228787,
+      "grad_norm": 11.249619483947754,
+      "learning_rate": 8.52e-05,
+      "loss": 4.6222,
+      "step": 214
+    },
+    {
+      "epoch": 0.13513513513513514,
+      "grad_norm": 8.625935554504395,
+      "learning_rate": 8.560000000000001e-05,
+      "loss": 4.4776,
+      "step": 215
+    },
+    {
+      "epoch": 0.1357636706473916,
+      "grad_norm": 8.138617515563965,
+      "learning_rate": 8.6e-05,
+      "loss": 4.7134,
+      "step": 216
+    },
+    {
+      "epoch": 0.13639220615964803,
+      "grad_norm": 7.485641956329346,
+      "learning_rate": 8.64e-05,
+      "loss": 4.8201,
+      "step": 217
+    },
+    {
+      "epoch": 0.13702074167190445,
+      "grad_norm": 11.87485408782959,
+      "learning_rate": 8.680000000000001e-05,
+      "loss": 4.7288,
+      "step": 218
+    },
+    {
+      "epoch": 0.1376492771841609,
+      "grad_norm": 8.551025390625,
+      "learning_rate": 8.72e-05,
+      "loss": 4.3091,
+      "step": 219
+    },
+    {
+      "epoch": 0.13827781269641734,
+      "grad_norm": 10.597816467285156,
+      "learning_rate": 8.76e-05,
+      "loss": 4.9491,
+      "step": 220
+    },
+    {
+      "epoch": 0.13890634820867379,
+      "grad_norm": 84.34934997558594,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 7.8641,
+      "step": 221
+    },
+    {
+      "epoch": 0.13953488372093023,
+      "grad_norm": 14.08538818359375,
+      "learning_rate": 8.840000000000001e-05,
+      "loss": 5.1718,
+      "step": 222
+    },
+    {
+      "epoch": 0.14016341923318668,
+      "grad_norm": 8.5634183883667,
+      "learning_rate": 8.88e-05,
+      "loss": 4.5042,
+      "step": 223
+    },
+    {
+      "epoch": 0.14079195474544312,
+      "grad_norm": 9.070499420166016,
+      "learning_rate": 8.92e-05,
+      "loss": 4.684,
+      "step": 224
+    },
+    {
+      "epoch": 0.14142049025769957,
+      "grad_norm": 13.369037628173828,
+      "learning_rate": 8.960000000000001e-05,
+      "loss": 4.2649,
+      "step": 225
+    },
+    {
+      "epoch": 0.142049025769956,
+      "grad_norm": 8.797863960266113,
+      "learning_rate": 9e-05,
+      "loss": 4.0673,
+      "step": 226
+    },
+    {
+      "epoch": 0.14267756128221246,
+      "grad_norm": 11.846273422241211,
+      "learning_rate": 9.04e-05,
+      "loss": 4.3157,
+      "step": 227
+    },
+    {
+      "epoch": 0.1433060967944689,
+      "grad_norm": 9.263139724731445,
+      "learning_rate": 9.080000000000001e-05,
+      "loss": 4.2926,
+      "step": 228
+    },
+    {
+      "epoch": 0.14393463230672532,
+      "grad_norm": 6.980598449707031,
+      "learning_rate": 9.120000000000001e-05,
+      "loss": 4.5284,
+      "step": 229
+    },
+    {
+      "epoch": 0.14456316781898176,
+      "grad_norm": 7.539294719696045,
+      "learning_rate": 9.16e-05,
+      "loss": 4.0574,
+      "step": 230
+    },
+    {
+      "epoch": 0.1451917033312382,
+      "grad_norm": 10.551584243774414,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 4.8214,
+      "step": 231
+    },
+    {
+      "epoch": 0.14582023884349465,
+      "grad_norm": 8.377208709716797,
+      "learning_rate": 9.240000000000001e-05,
+      "loss": 4.1492,
+      "step": 232
+    },
+    {
+      "epoch": 0.1464487743557511,
+      "grad_norm": 8.809621810913086,
+      "learning_rate": 9.28e-05,
+      "loss": 3.9481,
+      "step": 233
+    },
+    {
+      "epoch": 0.14707730986800754,
+      "grad_norm": 15.098812103271484,
+      "learning_rate": 9.320000000000002e-05,
+      "loss": 4.5192,
+      "step": 234
+    },
+    {
+      "epoch": 0.147705845380264,
+      "grad_norm": 16.269895553588867,
+      "learning_rate": 9.360000000000001e-05,
+      "loss": 4.5962,
+      "step": 235
+    },
+    {
+      "epoch": 0.14833438089252043,
+      "grad_norm": 9.021276473999023,
+      "learning_rate": 9.4e-05,
+      "loss": 4.6729,
+      "step": 236
+    },
+    {
+      "epoch": 0.14896291640477688,
+      "grad_norm": 10.110734939575195,
+      "learning_rate": 9.44e-05,
+      "loss": 4.5957,
+      "step": 237
+    },
+    {
+      "epoch": 0.14959145191703332,
+      "grad_norm": 10.254793167114258,
+      "learning_rate": 9.48e-05,
+      "loss": 4.1534,
+      "step": 238
+    },
+    {
+      "epoch": 0.15021998742928974,
+      "grad_norm": 10.400528907775879,
+      "learning_rate": 9.52e-05,
+      "loss": 4.5035,
+      "step": 239
+    },
+    {
+      "epoch": 0.1508485229415462,
+      "grad_norm": 13.529309272766113,
+      "learning_rate": 9.56e-05,
+      "loss": 4.7124,
+      "step": 240
+    },
+    {
+      "epoch": 0.15147705845380263,
+      "grad_norm": 9.187389373779297,
+      "learning_rate": 9.6e-05,
+      "loss": 4.2564,
+      "step": 241
+    },
+    {
+      "epoch": 0.15210559396605908,
+      "grad_norm": 9.760189056396484,
+      "learning_rate": 9.64e-05,
+      "loss": 4.5863,
+      "step": 242
+    },
+    {
+      "epoch": 0.15273412947831552,
+      "grad_norm": 11.033064842224121,
+      "learning_rate": 9.680000000000001e-05,
+      "loss": 4.4069,
+      "step": 243
+    },
+    {
+      "epoch": 0.15336266499057197,
+      "grad_norm": 8.79299259185791,
+      "learning_rate": 9.72e-05,
+      "loss": 4.2232,
+      "step": 244
+    },
+    {
+      "epoch": 0.1539912005028284,
+      "grad_norm": 10.526970863342285,
+      "learning_rate": 9.76e-05,
+      "loss": 4.6675,
+      "step": 245
+    },
+    {
+      "epoch": 0.15461973601508486,
+      "grad_norm": 14.993115425109863,
+      "learning_rate": 9.8e-05,
+      "loss": 4.3974,
+      "step": 246
+    },
+    {
+      "epoch": 0.1552482715273413,
+      "grad_norm": 8.852916717529297,
+      "learning_rate": 9.84e-05,
+      "loss": 4.1023,
+      "step": 247
+    },
+    {
+      "epoch": 0.15587680703959775,
+      "grad_norm": 17.305910110473633,
+      "learning_rate": 9.88e-05,
+      "loss": 4.2657,
+      "step": 248
+    },
+    {
+      "epoch": 0.1565053425518542,
+      "grad_norm": 12.877272605895996,
+      "learning_rate": 9.92e-05,
+      "loss": 4.318,
+      "step": 249
+    },
+    {
+      "epoch": 0.1571338780641106,
+      "grad_norm": 12.16046142578125,
+      "learning_rate": 9.960000000000001e-05,
+      "loss": 4.0387,
+      "step": 250
+    },
+    {
+      "epoch": 0.15776241357636706,
+      "grad_norm": 13.384797096252441,
+      "learning_rate": 0.0001,
+      "loss": 4.3935,
+      "step": 251
+    },
+    {
+      "epoch": 0.1583909490886235,
+      "grad_norm": 12.947620391845703,
+      "learning_rate": 0.0001004,
+      "loss": 4.2752,
+      "step": 252
+    },
+    {
+      "epoch": 0.15901948460087995,
+      "grad_norm": 13.004393577575684,
+      "learning_rate": 0.00010080000000000001,
+      "loss": 4.4285,
+      "step": 253
+    },
+    {
+      "epoch": 0.1596480201131364,
+      "grad_norm": 10.64500617980957,
+      "learning_rate": 0.00010120000000000001,
+      "loss": 4.3775,
+      "step": 254
+    },
+    {
+      "epoch": 0.16027655562539284,
+      "grad_norm": 10.208504676818848,
+      "learning_rate": 0.0001016,
+      "loss": 4.3942,
+      "step": 255
+    },
+    {
+      "epoch": 0.16090509113764928,
+      "grad_norm": 10.392770767211914,
+      "learning_rate": 0.00010200000000000001,
+      "loss": 3.964,
+      "step": 256
+    },
+    {
+      "epoch": 0.16153362664990573,
+      "grad_norm": 11.704667091369629,
+      "learning_rate": 0.00010240000000000001,
+      "loss": 3.2764,
+      "step": 257
+    },
+    {
+      "epoch": 0.16216216216216217,
+      "grad_norm": 11.466705322265625,
+      "learning_rate": 0.0001028,
+      "loss": 4.3286,
+      "step": 258
+    },
+    {
+      "epoch": 0.16279069767441862,
+      "grad_norm": 11.332155227661133,
+      "learning_rate": 0.0001032,
+      "loss": 4.4598,
+      "step": 259
+    },
+    {
+      "epoch": 0.16341923318667503,
+      "grad_norm": 10.597373962402344,
+      "learning_rate": 0.00010360000000000001,
+      "loss": 4.3576,
+      "step": 260
+    },
+    {
+      "epoch": 0.16404776869893148,
+      "grad_norm": 13.826549530029297,
+      "learning_rate": 0.00010400000000000001,
+      "loss": 4.1209,
+      "step": 261
+    },
+    {
+      "epoch": 0.16467630421118792,
+      "grad_norm": 12.812018394470215,
+      "learning_rate": 0.0001044,
+      "loss": 4.3951,
+      "step": 262
+    },
+    {
+      "epoch": 0.16530483972344437,
+      "grad_norm": 11.573945045471191,
+      "learning_rate": 0.00010480000000000001,
+      "loss": 4.6093,
+      "step": 263
+    },
+    {
+      "epoch": 0.16593337523570081,
+      "grad_norm": 9.195262908935547,
+      "learning_rate": 0.00010520000000000001,
+      "loss": 4.0342,
+      "step": 264
+    },
+    {
+      "epoch": 0.16656191074795726,
+      "grad_norm": 10.87674331665039,
+      "learning_rate": 0.0001056,
+      "loss": 3.9368,
+      "step": 265
+    },
+    {
+      "epoch": 0.1671904462602137,
+      "grad_norm": 11.914413452148438,
+      "learning_rate": 0.00010600000000000002,
+      "loss": 4.0043,
+      "step": 266
+    },
+    {
+      "epoch": 0.16781898177247015,
+      "grad_norm": 11.688831329345703,
+      "learning_rate": 0.00010640000000000001,
+      "loss": 3.6894,
+      "step": 267
+    },
+    {
+      "epoch": 0.1684475172847266,
+      "grad_norm": 17.798152923583984,
+      "learning_rate": 0.00010680000000000001,
+      "loss": 4.4073,
+      "step": 268
+    },
+    {
+      "epoch": 0.16907605279698304,
+      "grad_norm": 11.202817916870117,
+      "learning_rate": 0.00010720000000000002,
+      "loss": 3.9713,
+      "step": 269
+    },
+    {
+      "epoch": 0.16970458830923948,
+      "grad_norm": 11.484283447265625,
+      "learning_rate": 0.00010760000000000001,
+      "loss": 3.7401,
+      "step": 270
+    },
+    {
+      "epoch": 0.1703331238214959,
+      "grad_norm": 8.625189781188965,
+      "learning_rate": 0.00010800000000000001,
+      "loss": 3.6847,
+      "step": 271
+    },
+    {
+      "epoch": 0.17096165933375235,
+      "grad_norm": 14.114002227783203,
+      "learning_rate": 0.00010840000000000002,
+      "loss": 4.1115,
+      "step": 272
+    },
+    {
+      "epoch": 0.1715901948460088,
+      "grad_norm": 12.885270118713379,
+      "learning_rate": 0.00010880000000000002,
+      "loss": 3.5002,
+      "step": 273
+    },
+    {
+      "epoch": 0.17221873035826524,
+      "grad_norm": 13.46296501159668,
+      "learning_rate": 0.00010920000000000001,
+      "loss": 4.0249,
+      "step": 274
+    },
+    {
+      "epoch": 0.17284726587052168,
+      "grad_norm": 17.22674560546875,
+      "learning_rate": 0.00010960000000000001,
+      "loss": 4.4484,
+      "step": 275
+    },
+    {
+      "epoch": 0.17347580138277813,
+      "grad_norm": 13.292183876037598,
+      "learning_rate": 0.00011000000000000002,
+      "loss": 3.9718,
+      "step": 276
+    },
+    {
+      "epoch": 0.17410433689503457,
+      "grad_norm": 35.4398307800293,
+      "learning_rate": 0.00011040000000000001,
+      "loss": 4.0052,
+      "step": 277
+    },
+    {
+      "epoch": 0.17473287240729102,
+      "grad_norm": 19.460407257080078,
+      "learning_rate": 0.00011080000000000001,
+      "loss": 4.1892,
+      "step": 278
+    },
+    {
+      "epoch": 0.17536140791954746,
+      "grad_norm": 15.537309646606445,
+      "learning_rate": 0.00011120000000000002,
+      "loss": 3.9878,
+      "step": 279
+    },
+    {
+      "epoch": 0.1759899434318039,
+      "grad_norm": 16.156564712524414,
+      "learning_rate": 0.00011160000000000002,
+      "loss": 3.8753,
+      "step": 280
+    },
+    {
+      "epoch": 0.17661847894406035,
+      "grad_norm": 17.241336822509766,
+      "learning_rate": 0.00011200000000000001,
+      "loss": 3.9479,
+      "step": 281
+    },
+    {
+      "epoch": 0.17724701445631677,
+      "grad_norm": 20.027856826782227,
+      "learning_rate": 0.00011240000000000002,
+      "loss": 3.9856,
+      "step": 282
+    },
+    {
+      "epoch": 0.17787554996857322,
+      "grad_norm": 11.107442855834961,
+      "learning_rate": 0.00011279999999999999,
+      "loss": 3.5669,
+      "step": 283
+    },
+    {
+      "epoch": 0.17850408548082966,
+      "grad_norm": 14.415189743041992,
+      "learning_rate": 0.0001132,
+      "loss": 4.1763,
+      "step": 284
+    },
+    {
+      "epoch": 0.1791326209930861,
+      "grad_norm": 16.054901123046875,
+      "learning_rate": 0.0001136,
+      "loss": 4.1089,
+      "step": 285
+    },
+    {
+      "epoch": 0.17976115650534255,
+      "grad_norm": 17.718029022216797,
+      "learning_rate": 0.00011399999999999999,
+      "loss": 3.7917,
+      "step": 286
+    },
+    {
+      "epoch": 0.180389692017599,
+      "grad_norm": 15.520782470703125,
+      "learning_rate": 0.0001144,
+      "loss": 3.9876,
+      "step": 287
+    },
+    {
+      "epoch": 0.18101822752985544,
+      "grad_norm": 19.573312759399414,
+      "learning_rate": 0.0001148,
+      "loss": 3.9456,
+      "step": 288
+    },
+    {
+      "epoch": 0.1816467630421119,
+      "grad_norm": 11.45984935760498,
+      "learning_rate": 0.0001152,
+      "loss": 3.5074,
+      "step": 289
+    },
+    {
+      "epoch": 0.18227529855436833,
+      "grad_norm": 21.8663387298584,
+      "learning_rate": 0.00011559999999999999,
+      "loss": 4.0081,
+      "step": 290
+    },
+    {
+      "epoch": 0.18290383406662478,
+      "grad_norm": 16.874753952026367,
+      "learning_rate": 0.000116,
+      "loss": 3.9281,
+      "step": 291
+    },
+    {
+      "epoch": 0.1835323695788812,
+      "grad_norm": 16.98707389831543,
+      "learning_rate": 0.0001164,
+      "loss": 3.9644,
+      "step": 292
+    },
+    {
+      "epoch": 0.18416090509113764,
+      "grad_norm": 12.668580055236816,
+      "learning_rate": 0.00011679999999999999,
+      "loss": 3.3014,
+      "step": 293
+    },
+    {
+      "epoch": 0.18478944060339408,
+      "grad_norm": 11.530134201049805,
+      "learning_rate": 0.0001172,
+      "loss": 3.3005,
+      "step": 294
+    },
+    {
+      "epoch": 0.18541797611565053,
+      "grad_norm": 53.85222244262695,
+      "learning_rate": 0.0001176,
+      "loss": 4.4674,
+      "step": 295
+    },
+    {
+      "epoch": 0.18604651162790697,
+      "grad_norm": 12.703718185424805,
+      "learning_rate": 0.000118,
+      "loss": 3.629,
+      "step": 296
+    },
+    {
+      "epoch": 0.18667504714016342,
+      "grad_norm": 18.4884090423584,
+      "learning_rate": 0.0001184,
+      "loss": 3.7944,
+      "step": 297
+    },
+    {
+      "epoch": 0.18730358265241986,
+      "grad_norm": 18.730792999267578,
+      "learning_rate": 0.0001188,
+      "loss": 3.6521,
+      "step": 298
+    },
+    {
+      "epoch": 0.1879321181646763,
+      "grad_norm": 10.74968433380127,
+      "learning_rate": 0.0001192,
+      "loss": 3.6895,
+      "step": 299
+    },
+    {
+      "epoch": 0.18856065367693275,
+      "grad_norm": 13.541333198547363,
+      "learning_rate": 0.00011960000000000001,
+      "loss": 3.4517,
+      "step": 300
+    },
+    {
+      "epoch": 0.1891891891891892,
+      "grad_norm": 20.38814926147461,
+      "learning_rate": 0.00012,
+      "loss": 3.3987,
+      "step": 301
+    },
+    {
+      "epoch": 0.18981772470144564,
+      "grad_norm": 21.178226470947266,
+      "learning_rate": 0.0001204,
+      "loss": 3.6073,
+      "step": 302
+    },
+    {
+      "epoch": 0.19044626021370206,
+      "grad_norm": 11.370046615600586,
+      "learning_rate": 0.0001208,
+      "loss": 3.3391,
+      "step": 303
+    },
+    {
+      "epoch": 0.1910747957259585,
+      "grad_norm": 10.467302322387695,
+      "learning_rate": 0.0001212,
+      "loss": 3.7676,
+      "step": 304
+    },
+    {
+      "epoch": 0.19170333123821495,
+      "grad_norm": 11.307904243469238,
+      "learning_rate": 0.0001216,
+      "loss": 3.4048,
+      "step": 305
+    },
+    {
+      "epoch": 0.1923318667504714,
+      "grad_norm": 20.741159439086914,
+      "learning_rate": 0.000122,
+      "loss": 3.1971,
+      "step": 306
+    },
+    {
+      "epoch": 0.19296040226272784,
+      "grad_norm": 22.384532928466797,
+      "learning_rate": 0.0001224,
+      "loss": 3.2315,
+      "step": 307
+    },
+    {
+      "epoch": 0.1935889377749843,
+      "grad_norm": 19.073701858520508,
+      "learning_rate": 0.0001228,
+      "loss": 3.3426,
+      "step": 308
+    },
+    {
+      "epoch": 0.19421747328724073,
+      "grad_norm": 20.864418029785156,
+      "learning_rate": 0.0001232,
+      "loss": 3.4605,
+      "step": 309
+    },
+    {
+      "epoch": 0.19484600879949718,
+      "grad_norm": 15.888052940368652,
+      "learning_rate": 0.0001236,
+      "loss": 3.5554,
+      "step": 310
+    },
+    {
+      "epoch": 0.19547454431175362,
+      "grad_norm": 13.650964736938477,
+      "learning_rate": 0.000124,
+      "loss": 3.3372,
+      "step": 311
+    },
+    {
+      "epoch": 0.19610307982401007,
+      "grad_norm": 20.19765853881836,
+      "learning_rate": 0.00012440000000000002,
+      "loss": 3.4044,
+      "step": 312
+    },
+    {
+      "epoch": 0.19673161533626649,
+      "grad_norm": 18.977705001831055,
+      "learning_rate": 0.0001248,
+      "loss": 3.5597,
+      "step": 313
+    },
+    {
+      "epoch": 0.19736015084852293,
+      "grad_norm": 13.097780227661133,
+      "learning_rate": 0.0001252,
+      "loss": 3.2755,
+      "step": 314
+    },
+    {
+      "epoch": 0.19798868636077938,
+      "grad_norm": 17.84178924560547,
+      "learning_rate": 0.00012560000000000002,
+      "loss": 3.2663,
+      "step": 315
+    },
+    {
+      "epoch": 0.19861722187303582,
+      "grad_norm": 14.159675598144531,
+      "learning_rate": 0.000126,
+      "loss": 3.4946,
+      "step": 316
+    },
+    {
+      "epoch": 0.19924575738529227,
+      "grad_norm": 11.891308784484863,
+      "learning_rate": 0.0001264,
+      "loss": 2.8523,
+      "step": 317
+    },
+    {
+      "epoch": 0.1998742928975487,
+      "grad_norm": 15.97802734375,
+      "learning_rate": 0.00012680000000000002,
+      "loss": 3.2556,
+      "step": 318
+    },
+    {
+      "epoch": 0.20050282840980516,
+      "grad_norm": 13.670428276062012,
+      "learning_rate": 0.0001272,
+      "loss": 3.4079,
+      "step": 319
+    },
+    {
+      "epoch": 0.2011313639220616,
+      "grad_norm": 15.742596626281738,
+      "learning_rate": 0.0001276,
+      "loss": 3.358,
+      "step": 320
+    },
+    {
+      "epoch": 0.2011313639220616,
+      "eval_loss": 2.595961093902588,
+      "eval_runtime": 1561.7692,
+      "eval_samples_per_second": 1.651,
+      "eval_steps_per_second": 1.651,
+      "step": 320
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 1591,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 160,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.0920991543381197e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-320/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bd0dbc47a9812516a7dc00a4bccf2e85152bb7b0be6caacda2e7ed6fd24ad7b
+size 8824

checkpoint-320/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)