Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,202 @@
+---
+base_model: THUDM/GLM-4-32B-0414
+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 0.15.1

adapter_config.json

@@ -0,0 +1,38 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "THUDM/GLM-4-32B-0414",
+  "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": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.25,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "v_proj",
+    "down_proj",
+    "k_proj",
+    "o_proj",
+    "gate_up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d75fe4bb3fa25b393887cd2c587586e6f3c749e6a8403e1c32e6d0d0b7c39d4
+size 231966248

global_step468/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:ed38079166248d0310efc2c1a242644621df5ff89de15cd54cb5540fbc804623
+size 88915560

global_step468/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92857ea017068615b662b0f1f98c3ff128dd764149b926cb585f567404f304e5
+size 88915560

global_step468/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9844097153b81beef73a256f8e6b32c77b728796bc8abfc8963a3e9a2232051c
+size 88915560

global_step468/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4dc8afd9eb16322de91bd3e69a66517bd08e02944f32650c84a1b9467da9f610
+size 88915560

global_step468/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d5f88f76f03cd9861ddaf9daac7b7e9fa0a383e862f33f41c48018264b3c20a
+size 88915560

global_step468/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:749c01fdd0f9352fe201ff9a06d3c45f16427e33f0e2eedf7a196793085130dd
+size 88915560

global_step468/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbae5b7d26140ece9ae562f5fbe79fc79d8bacaf844b7a8503bf699b87cc3912
+size 88915560

global_step468/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d042a9d2157a83b74de0f988a7593fbc06241a48aa4da0cda860d933ac17999f
+size 88915560

global_step468/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf3ec33b522be25973920c68249a1b975596ff4756489a0079bab6e79db6ab5f
+size 489169474

global_step468/zero_pp_rank_1_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04557168c909936295a6f6b331fca0b13bcea252dccef18d4303f25e39b31b10
+size 489169474

global_step468/zero_pp_rank_2_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5a8f95e320a1eed01df24f7a6146141169a0b6cd4e15d27560553d90230bdb9
+size 489169474

global_step468/zero_pp_rank_3_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89989033171cf71d3436960f87d3de6a3c8348f2c76993e27816f4035e4cbb81
+size 489169474

global_step468/zero_pp_rank_4_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6df15ce92d4b29ac02f950bd12e13a0e2bf7cb7c0c5d8bf0a25eb51e34b3d1f7
+size 489169474

global_step468/zero_pp_rank_5_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d758c3278cbd673ee550903e63eb867d90f1077ba56088a2b65f1ced1cb1b703
+size 489169474

global_step468/zero_pp_rank_6_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f116a3d032822c3e862a1243274b41a9e0cf4384e5b4d354283758a8561d70e0
+size 489169474

global_step468/zero_pp_rank_7_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b12272789068eb2500c6dbd646320f42515f1e35ca8ff39a390dff274e065750
+size 489169474

latest

@@ -0,0 +1 @@
+global_step468

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc64b408da4f8c559c35c409923e394f4a7977cbc762651abf99d46b7084b982
+size 15984

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2e314bda318014a74205de3842e71c4a10a9d3b84793883a67a220113b7f9f7
+size 15984

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:943124b2528590a47654d52819621dcb463433cb3d3c8933a0c176fb126c6a43
+size 15984

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9175feb38e77a12fe62809bfea253ca91f837eecd61bd00037cc5ed4140c3af6
+size 15984

rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1f36a31701184818cb920a0a0e24f13bebed551683d3d399296dd997f500762
+size 15984

rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d1c2d845078dbdfebc7922410718ed5993ac72778bd91ffc823852d269691c9
+size 15984

rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bda2fad69036d5a69510770ddfc0a5951a90286d6cf8495af3d8756be8b6106a
+size 15984

rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d2a50fc34b49cd1d88e811b1a1a7266bd79872ce6f7cf29d95ac978d7539a1f
+size 15984

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8a11b2b11819ea1201e976a4e9e83a257be0b6930063c975e94c18d9889b70d
+size 1064

special_tokens_map.json

@@ -0,0 +1,32 @@
+{
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "[MASK]",
+    "[gMASK]",
+    "[sMASK]",
+    "<sop>",
+    "<eop>",
+    "<|system|>",
+    "<|user|>",
+    "<|assistant|>",
+    "<|observation|>",
+    "<|begin_of_image|>",
+    "<|end_of_image|>",
+    "<|begin_of_video|>",
+    "<|end_of_video|>"
+  ],
+  "eos_token": {
+    "content": "<|user|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76ebeac0d8bd7879ead7b43c16b44981f277e47225de2bd7de9ae1a6cc664a8c
+size 19966496

tokenizer_config.json

@@ -0,0 +1,146 @@
+{
+  "added_tokens_decoder": {
+    "151329": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151330": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151331": {
+      "content": "[gMASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151332": {
+      "content": "[sMASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151333": {
+      "content": "<sop>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151334": {
+      "content": "<eop>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151335": {
+      "content": "<|system|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151336": {
+      "content": "<|user|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151337": {
+      "content": "<|assistant|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151338": {
+      "content": "<|observation|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151339": {
+      "content": "<|begin_of_image|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151340": {
+      "content": "<|end_of_image|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151341": {
+      "content": "<|begin_of_video|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151342": {
+      "content": "<|end_of_video|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "[MASK]",
+    "[gMASK]",
+    "[sMASK]",
+    "<sop>",
+    "<eop>",
+    "<|system|>",
+    "<|user|>",
+    "<|assistant|>",
+    "<|observation|>",
+    "<|begin_of_image|>",
+    "<|end_of_image|>",
+    "<|begin_of_video|>",
+    "<|end_of_video|>"
+  ],
+  "chat_template": "[gMASK]<sop>{%- for msg in messages %}{%- if msg.role == 'system' %}<|system|>\n{{ msg.content }}{%- elif msg.role == 'user' %}<|user|>\n{{ msg.content }}{%- elif msg.role == 'assistant' %}<|assistant|>\n{{ msg.content }}{%- endif %}{%- endfor %}{% if add_generation_prompt %}<|assistant|>{% else %}<|user|>{% endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "do_lower_case": false,
+  "eos_token": "<|user|>",
+  "extra_special_tokens": {},
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 128000,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "left",
+  "remove_space": false,
+  "tokenizer_class": "PreTrainedTokenizer"
+}

trainer_state.json

@@ -0,0 +1,3310 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.504823151125402,
+  "eval_steps": 500,
+  "global_step": 468,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003215434083601286,
+      "grad_norm": 1.8139427901504777,
+      "learning_rate": 3.2258064516129035e-07,
+      "loss": 1.8795,
+      "step": 1
+    },
+    {
+      "epoch": 0.006430868167202572,
+      "grad_norm": 1.4827057247111375,
+      "learning_rate": 6.451612903225807e-07,
+      "loss": 1.9012,
+      "step": 2
+    },
+    {
+      "epoch": 0.00964630225080386,
+      "grad_norm": 2.246892281108816,
+      "learning_rate": 9.67741935483871e-07,
+      "loss": 2.051,
+      "step": 3
+    },
+    {
+      "epoch": 0.012861736334405145,
+      "grad_norm": 2.1678014328927557,
+      "learning_rate": 1.2903225806451614e-06,
+      "loss": 1.9149,
+      "step": 4
+    },
+    {
+      "epoch": 0.01607717041800643,
+      "grad_norm": 1.5486796271797767,
+      "learning_rate": 1.6129032258064516e-06,
+      "loss": 1.9074,
+      "step": 5
+    },
+    {
+      "epoch": 0.01929260450160772,
+      "grad_norm": 2.070045324451304,
+      "learning_rate": 1.935483870967742e-06,
+      "loss": 1.8017,
+      "step": 6
+    },
+    {
+      "epoch": 0.022508038585209004,
+      "grad_norm": 1.0691565532221945,
+      "learning_rate": 2.2580645161290324e-06,
+      "loss": 1.8634,
+      "step": 7
+    },
+    {
+      "epoch": 0.02572347266881029,
+      "grad_norm": 1.2700085661670921,
+      "learning_rate": 2.580645161290323e-06,
+      "loss": 1.8937,
+      "step": 8
+    },
+    {
+      "epoch": 0.028938906752411574,
+      "grad_norm": 1.4648464524181335,
+      "learning_rate": 2.903225806451613e-06,
+      "loss": 1.8333,
+      "step": 9
+    },
+    {
+      "epoch": 0.03215434083601286,
+      "grad_norm": 1.9789558033563477,
+      "learning_rate": 3.225806451612903e-06,
+      "loss": 1.9324,
+      "step": 10
+    },
+    {
+      "epoch": 0.03536977491961415,
+      "grad_norm": 1.3084400687826616,
+      "learning_rate": 3.5483870967741936e-06,
+      "loss": 1.9304,
+      "step": 11
+    },
+    {
+      "epoch": 0.03858520900321544,
+      "grad_norm": 1.4314638080369735,
+      "learning_rate": 3.870967741935484e-06,
+      "loss": 1.8559,
+      "step": 12
+    },
+    {
+      "epoch": 0.04180064308681672,
+      "grad_norm": 1.8675934280306556,
+      "learning_rate": 4.193548387096774e-06,
+      "loss": 1.8142,
+      "step": 13
+    },
+    {
+      "epoch": 0.04501607717041801,
+      "grad_norm": 2.2721373779660885,
+      "learning_rate": 4.516129032258065e-06,
+      "loss": 1.9085,
+      "step": 14
+    },
+    {
+      "epoch": 0.04823151125401929,
+      "grad_norm": 2.1557786439995144,
+      "learning_rate": 4.838709677419355e-06,
+      "loss": 1.9455,
+      "step": 15
+    },
+    {
+      "epoch": 0.05144694533762058,
+      "grad_norm": 1.581014411505333,
+      "learning_rate": 5.161290322580646e-06,
+      "loss": 1.8905,
+      "step": 16
+    },
+    {
+      "epoch": 0.05466237942122187,
+      "grad_norm": 1.1620382151650777,
+      "learning_rate": 5.483870967741936e-06,
+      "loss": 1.9314,
+      "step": 17
+    },
+    {
+      "epoch": 0.05787781350482315,
+      "grad_norm": 1.54505591125172,
+      "learning_rate": 5.806451612903226e-06,
+      "loss": 1.9688,
+      "step": 18
+    },
+    {
+      "epoch": 0.06109324758842444,
+      "grad_norm": 1.4057721389474997,
+      "learning_rate": 6.129032258064517e-06,
+      "loss": 1.9826,
+      "step": 19
+    },
+    {
+      "epoch": 0.06430868167202572,
+      "grad_norm": 1.9706307878507736,
+      "learning_rate": 6.451612903225806e-06,
+      "loss": 2.0072,
+      "step": 20
+    },
+    {
+      "epoch": 0.06752411575562701,
+      "grad_norm": 1.7236343733533435,
+      "learning_rate": 6.774193548387097e-06,
+      "loss": 1.9905,
+      "step": 21
+    },
+    {
+      "epoch": 0.0707395498392283,
+      "grad_norm": 2.0533016397870476,
+      "learning_rate": 7.096774193548387e-06,
+      "loss": 1.9434,
+      "step": 22
+    },
+    {
+      "epoch": 0.07395498392282958,
+      "grad_norm": 1.481031731548318,
+      "learning_rate": 7.419354838709678e-06,
+      "loss": 1.8396,
+      "step": 23
+    },
+    {
+      "epoch": 0.07717041800643087,
+      "grad_norm": 1.170446737350014,
+      "learning_rate": 7.741935483870968e-06,
+      "loss": 1.8728,
+      "step": 24
+    },
+    {
+      "epoch": 0.08038585209003216,
+      "grad_norm": 1.311786758926783,
+      "learning_rate": 8.064516129032258e-06,
+      "loss": 1.8793,
+      "step": 25
+    },
+    {
+      "epoch": 0.08360128617363344,
+      "grad_norm": 1.164142868037603,
+      "learning_rate": 8.387096774193549e-06,
+      "loss": 1.8844,
+      "step": 26
+    },
+    {
+      "epoch": 0.08681672025723473,
+      "grad_norm": 1.4646144613768468,
+      "learning_rate": 8.70967741935484e-06,
+      "loss": 2.0018,
+      "step": 27
+    },
+    {
+      "epoch": 0.09003215434083602,
+      "grad_norm": 2.754370389076398,
+      "learning_rate": 9.03225806451613e-06,
+      "loss": 1.934,
+      "step": 28
+    },
+    {
+      "epoch": 0.0932475884244373,
+      "grad_norm": 0.9737813594884901,
+      "learning_rate": 9.35483870967742e-06,
+      "loss": 1.7937,
+      "step": 29
+    },
+    {
+      "epoch": 0.09646302250803858,
+      "grad_norm": 0.8062538970359915,
+      "learning_rate": 9.67741935483871e-06,
+      "loss": 1.948,
+      "step": 30
+    },
+    {
+      "epoch": 0.09967845659163987,
+      "grad_norm": 0.9350863955119307,
+      "learning_rate": 1e-05,
+      "loss": 1.9635,
+      "step": 31
+    },
+    {
+      "epoch": 0.10289389067524116,
+      "grad_norm": 0.8234637750282952,
+      "learning_rate": 9.998305371970854e-06,
+      "loss": 1.7344,
+      "step": 32
+    },
+    {
+      "epoch": 0.10610932475884244,
+      "grad_norm": 1.0156697352772721,
+      "learning_rate": 9.996605566870334e-06,
+      "loss": 1.8349,
+      "step": 33
+    },
+    {
+      "epoch": 0.10932475884244373,
+      "grad_norm": 0.8584344086710765,
+      "learning_rate": 9.994900560938297e-06,
+      "loss": 1.8419,
+      "step": 34
+    },
+    {
+      "epoch": 0.11254019292604502,
+      "grad_norm": 0.7548443656059388,
+      "learning_rate": 9.993190330268982e-06,
+      "loss": 1.79,
+      "step": 35
+    },
+    {
+      "epoch": 0.1157556270096463,
+      "grad_norm": 0.9117562153771265,
+      "learning_rate": 9.99147485080989e-06,
+      "loss": 1.7731,
+      "step": 36
+    },
+    {
+      "epoch": 0.1189710610932476,
+      "grad_norm": 0.7857198552918699,
+      "learning_rate": 9.989754098360657e-06,
+      "loss": 1.8349,
+      "step": 37
+    },
+    {
+      "epoch": 0.12218649517684887,
+      "grad_norm": 0.4842461872351954,
+      "learning_rate": 9.988028048571917e-06,
+      "loss": 1.8621,
+      "step": 38
+    },
+    {
+      "epoch": 0.12540192926045016,
+      "grad_norm": 0.5546706406169316,
+      "learning_rate": 9.98629667694416e-06,
+      "loss": 1.872,
+      "step": 39
+    },
+    {
+      "epoch": 0.12861736334405144,
+      "grad_norm": 0.4743331575347531,
+      "learning_rate": 9.984559958826557e-06,
+      "loss": 1.8474,
+      "step": 40
+    },
+    {
+      "epoch": 0.13183279742765272,
+      "grad_norm": 0.6191091199067691,
+      "learning_rate": 9.982817869415808e-06,
+      "loss": 1.9006,
+      "step": 41
+    },
+    {
+      "epoch": 0.13504823151125403,
+      "grad_norm": 0.7350082816564365,
+      "learning_rate": 9.98107038375495e-06,
+      "loss": 1.9712,
+      "step": 42
+    },
+    {
+      "epoch": 0.1382636655948553,
+      "grad_norm": 1.84774696273094,
+      "learning_rate": 9.979317476732161e-06,
+      "loss": 1.7948,
+      "step": 43
+    },
+    {
+      "epoch": 0.1414790996784566,
+      "grad_norm": 1.6542501458240009,
+      "learning_rate": 9.97755912307958e-06,
+      "loss": 1.8911,
+      "step": 44
+    },
+    {
+      "epoch": 0.14469453376205788,
+      "grad_norm": 0.42138602432759076,
+      "learning_rate": 9.975795297372061e-06,
+      "loss": 1.7762,
+      "step": 45
+    },
+    {
+      "epoch": 0.14790996784565916,
+      "grad_norm": 0.6398366473724499,
+      "learning_rate": 9.974025974025974e-06,
+      "loss": 1.7533,
+      "step": 46
+    },
+    {
+      "epoch": 0.15112540192926044,
+      "grad_norm": 0.4090060859208197,
+      "learning_rate": 9.972251127297954e-06,
+      "loss": 1.7385,
+      "step": 47
+    },
+    {
+      "epoch": 0.15434083601286175,
+      "grad_norm": 0.5850115421043386,
+      "learning_rate": 9.970470731283656e-06,
+      "loss": 1.8072,
+      "step": 48
+    },
+    {
+      "epoch": 0.15755627009646303,
+      "grad_norm": 0.7238316463106957,
+      "learning_rate": 9.968684759916494e-06,
+      "loss": 1.8023,
+      "step": 49
+    },
+    {
+      "epoch": 0.1607717041800643,
+      "grad_norm": 0.46218123288707824,
+      "learning_rate": 9.966893186966372e-06,
+      "loss": 1.8054,
+      "step": 50
+    },
+    {
+      "epoch": 0.1639871382636656,
+      "grad_norm": 0.616105183333845,
+      "learning_rate": 9.965095986038394e-06,
+      "loss": 1.6704,
+      "step": 51
+    },
+    {
+      "epoch": 0.16720257234726688,
+      "grad_norm": 0.4516417429014012,
+      "learning_rate": 9.963293130571578e-06,
+      "loss": 1.8424,
+      "step": 52
+    },
+    {
+      "epoch": 0.17041800643086816,
+      "grad_norm": 0.77318360969182,
+      "learning_rate": 9.961484593837536e-06,
+      "loss": 1.757,
+      "step": 53
+    },
+    {
+      "epoch": 0.17363344051446947,
+      "grad_norm": 0.42668808439045164,
+      "learning_rate": 9.959670348939155e-06,
+      "loss": 1.7232,
+      "step": 54
+    },
+    {
+      "epoch": 0.17684887459807075,
+      "grad_norm": 0.461318262919585,
+      "learning_rate": 9.957850368809273e-06,
+      "loss": 1.7933,
+      "step": 55
+    },
+    {
+      "epoch": 0.18006430868167203,
+      "grad_norm": 0.7315165303506296,
+      "learning_rate": 9.956024626209323e-06,
+      "loss": 1.7174,
+      "step": 56
+    },
+    {
+      "epoch": 0.1832797427652733,
+      "grad_norm": 0.5547095768027991,
+      "learning_rate": 9.954193093727978e-06,
+      "loss": 1.6926,
+      "step": 57
+    },
+    {
+      "epoch": 0.1864951768488746,
+      "grad_norm": 0.5584801951510784,
+      "learning_rate": 9.952355743779778e-06,
+      "loss": 1.8614,
+      "step": 58
+    },
+    {
+      "epoch": 0.18971061093247588,
+      "grad_norm": 0.4283225621008119,
+      "learning_rate": 9.950512548603748e-06,
+      "loss": 1.7414,
+      "step": 59
+    },
+    {
+      "epoch": 0.19292604501607716,
+      "grad_norm": 0.6611404688540005,
+      "learning_rate": 9.948663480261995e-06,
+      "loss": 1.7605,
+      "step": 60
+    },
+    {
+      "epoch": 0.19614147909967847,
+      "grad_norm": 0.48587548617769466,
+      "learning_rate": 9.946808510638298e-06,
+      "loss": 1.724,
+      "step": 61
+    },
+    {
+      "epoch": 0.19935691318327975,
+      "grad_norm": 0.39342564370484284,
+      "learning_rate": 9.94494761143669e-06,
+      "loss": 1.8336,
+      "step": 62
+    },
+    {
+      "epoch": 0.20257234726688103,
+      "grad_norm": 0.4173189236198402,
+      "learning_rate": 9.943080754180008e-06,
+      "loss": 1.7816,
+      "step": 63
+    },
+    {
+      "epoch": 0.2057877813504823,
+      "grad_norm": 0.4658148025943206,
+      "learning_rate": 9.941207910208445e-06,
+      "loss": 1.6283,
+      "step": 64
+    },
+    {
+      "epoch": 0.2090032154340836,
+      "grad_norm": 0.517548892373079,
+      "learning_rate": 9.939329050678089e-06,
+      "loss": 1.8234,
+      "step": 65
+    },
+    {
+      "epoch": 0.21221864951768488,
+      "grad_norm": 0.46275765989156126,
+      "learning_rate": 9.937444146559429e-06,
+      "loss": 1.7447,
+      "step": 66
+    },
+    {
+      "epoch": 0.21543408360128619,
+      "grad_norm": 1.1081478758742653,
+      "learning_rate": 9.935553168635876e-06,
+      "loss": 1.7269,
+      "step": 67
+    },
+    {
+      "epoch": 0.21864951768488747,
+      "grad_norm": 0.3685207965444784,
+      "learning_rate": 9.933656087502243e-06,
+      "loss": 1.6526,
+      "step": 68
+    },
+    {
+      "epoch": 0.22186495176848875,
+      "grad_norm": 0.4915041140865339,
+      "learning_rate": 9.93175287356322e-06,
+      "loss": 1.7168,
+      "step": 69
+    },
+    {
+      "epoch": 0.22508038585209003,
+      "grad_norm": 0.5376826838504105,
+      "learning_rate": 9.92984349703184e-06,
+      "loss": 1.717,
+      "step": 70
+    },
+    {
+      "epoch": 0.2282958199356913,
+      "grad_norm": 0.5138524626951391,
+      "learning_rate": 9.927927927927928e-06,
+      "loss": 1.8743,
+      "step": 71
+    },
+    {
+      "epoch": 0.2315112540192926,
+      "grad_norm": 0.45983175614227173,
+      "learning_rate": 9.926006136076522e-06,
+      "loss": 1.5821,
+      "step": 72
+    },
+    {
+      "epoch": 0.2347266881028939,
+      "grad_norm": 0.43682547338676814,
+      "learning_rate": 9.924078091106291e-06,
+      "loss": 1.6565,
+      "step": 73
+    },
+    {
+      "epoch": 0.2379421221864952,
+      "grad_norm": 0.48445488077665566,
+      "learning_rate": 9.922143762447946e-06,
+      "loss": 1.8203,
+      "step": 74
+    },
+    {
+      "epoch": 0.24115755627009647,
+      "grad_norm": 0.43776824297139105,
+      "learning_rate": 9.920203119332609e-06,
+      "loss": 1.6981,
+      "step": 75
+    },
+    {
+      "epoch": 0.24437299035369775,
+      "grad_norm": 0.8119898531586365,
+      "learning_rate": 9.918256130790192e-06,
+      "loss": 1.7794,
+      "step": 76
+    },
+    {
+      "epoch": 0.24758842443729903,
+      "grad_norm": 0.3901084260561504,
+      "learning_rate": 9.916302765647746e-06,
+      "loss": 1.8199,
+      "step": 77
+    },
+    {
+      "epoch": 0.2508038585209003,
+      "grad_norm": 0.4594634220443229,
+      "learning_rate": 9.914342992527794e-06,
+      "loss": 1.7681,
+      "step": 78
+    },
+    {
+      "epoch": 0.2540192926045016,
+      "grad_norm": 0.4588533669296292,
+      "learning_rate": 9.91237677984666e-06,
+      "loss": 1.6955,
+      "step": 79
+    },
+    {
+      "epoch": 0.2572347266881029,
+      "grad_norm": 0.4616453710837963,
+      "learning_rate": 9.910404095812764e-06,
+      "loss": 1.5667,
+      "step": 80
+    },
+    {
+      "epoch": 0.2604501607717042,
+      "grad_norm": 0.6020248390199695,
+      "learning_rate": 9.90842490842491e-06,
+      "loss": 1.9007,
+      "step": 81
+    },
+    {
+      "epoch": 0.26366559485530544,
+      "grad_norm": 0.4110162307135801,
+      "learning_rate": 9.906439185470556e-06,
+      "loss": 1.8028,
+      "step": 82
+    },
+    {
+      "epoch": 0.26688102893890675,
+      "grad_norm": 0.5124129858836964,
+      "learning_rate": 9.904446894524072e-06,
+      "loss": 1.7442,
+      "step": 83
+    },
+    {
+      "epoch": 0.27009646302250806,
+      "grad_norm": 0.5964949481206979,
+      "learning_rate": 9.902448002944966e-06,
+      "loss": 1.5914,
+      "step": 84
+    },
+    {
+      "epoch": 0.2733118971061093,
+      "grad_norm": 0.42395305200907873,
+      "learning_rate": 9.900442477876108e-06,
+      "loss": 1.7459,
+      "step": 85
+    },
+    {
+      "epoch": 0.2765273311897106,
+      "grad_norm": 0.5072002009865632,
+      "learning_rate": 9.898430286241922e-06,
+      "loss": 1.8099,
+      "step": 86
+    },
+    {
+      "epoch": 0.2797427652733119,
+      "grad_norm": 0.45165823004316974,
+      "learning_rate": 9.896411394746579e-06,
+      "loss": 1.6525,
+      "step": 87
+    },
+    {
+      "epoch": 0.2829581993569132,
+      "grad_norm": 0.6306765126972796,
+      "learning_rate": 9.894385769872153e-06,
+      "loss": 1.732,
+      "step": 88
+    },
+    {
+      "epoch": 0.2861736334405145,
+      "grad_norm": 0.5739555807551489,
+      "learning_rate": 9.892353377876765e-06,
+      "loss": 1.6367,
+      "step": 89
+    },
+    {
+      "epoch": 0.28938906752411575,
+      "grad_norm": 0.42492816633261454,
+      "learning_rate": 9.890314184792715e-06,
+      "loss": 1.9578,
+      "step": 90
+    },
+    {
+      "epoch": 0.29260450160771706,
+      "grad_norm": 0.3618619515123821,
+      "learning_rate": 9.888268156424582e-06,
+      "loss": 1.7637,
+      "step": 91
+    },
+    {
+      "epoch": 0.2958199356913183,
+      "grad_norm": 0.39652453599536036,
+      "learning_rate": 9.886215258347324e-06,
+      "loss": 1.7515,
+      "step": 92
+    },
+    {
+      "epoch": 0.2990353697749196,
+      "grad_norm": 0.6725210809375401,
+      "learning_rate": 9.884155455904336e-06,
+      "loss": 1.8649,
+      "step": 93
+    },
+    {
+      "epoch": 0.3022508038585209,
+      "grad_norm": 0.36922904709086996,
+      "learning_rate": 9.882088714205503e-06,
+      "loss": 1.8891,
+      "step": 94
+    },
+    {
+      "epoch": 0.3054662379421222,
+      "grad_norm": 0.546943119320176,
+      "learning_rate": 9.880014998125235e-06,
+      "loss": 1.7453,
+      "step": 95
+    },
+    {
+      "epoch": 0.3086816720257235,
+      "grad_norm": 0.33061581393958167,
+      "learning_rate": 9.87793427230047e-06,
+      "loss": 1.7088,
+      "step": 96
+    },
+    {
+      "epoch": 0.31189710610932475,
+      "grad_norm": 0.532292966712963,
+      "learning_rate": 9.875846501128668e-06,
+      "loss": 1.7597,
+      "step": 97
+    },
+    {
+      "epoch": 0.31511254019292606,
+      "grad_norm": 0.3627002665021866,
+      "learning_rate": 9.873751648765783e-06,
+      "loss": 1.8064,
+      "step": 98
+    },
+    {
+      "epoch": 0.3183279742765273,
+      "grad_norm": 0.3579021058819948,
+      "learning_rate": 9.871649679124198e-06,
+      "loss": 1.7573,
+      "step": 99
+    },
+    {
+      "epoch": 0.3215434083601286,
+      "grad_norm": 0.3611562992520892,
+      "learning_rate": 9.869540555870676e-06,
+      "loss": 1.814,
+      "step": 100
+    },
+    {
+      "epoch": 0.3247588424437299,
+      "grad_norm": 0.37816481014727843,
+      "learning_rate": 9.867424242424243e-06,
+      "loss": 1.8183,
+      "step": 101
+    },
+    {
+      "epoch": 0.3279742765273312,
+      "grad_norm": 0.38027583942322074,
+      "learning_rate": 9.86530070195409e-06,
+      "loss": 1.6795,
+      "step": 102
+    },
+    {
+      "epoch": 0.3311897106109325,
+      "grad_norm": 0.4653778897269012,
+      "learning_rate": 9.863169897377425e-06,
+      "loss": 1.6914,
+      "step": 103
+    },
+    {
+      "epoch": 0.33440514469453375,
+      "grad_norm": 0.3581142302428615,
+      "learning_rate": 9.86103179135732e-06,
+      "loss": 1.7623,
+      "step": 104
+    },
+    {
+      "epoch": 0.33762057877813506,
+      "grad_norm": 0.4345920198663122,
+      "learning_rate": 9.858886346300535e-06,
+      "loss": 1.6939,
+      "step": 105
+    },
+    {
+      "epoch": 0.3408360128617363,
+      "grad_norm": 0.42325366386457053,
+      "learning_rate": 9.856733524355302e-06,
+      "loss": 1.7802,
+      "step": 106
+    },
+    {
+      "epoch": 0.3440514469453376,
+      "grad_norm": 0.3886379815457078,
+      "learning_rate": 9.854573287409109e-06,
+      "loss": 1.6567,
+      "step": 107
+    },
+    {
+      "epoch": 0.34726688102893893,
+      "grad_norm": 0.30485293356067333,
+      "learning_rate": 9.852405597086448e-06,
+      "loss": 1.636,
+      "step": 108
+    },
+    {
+      "epoch": 0.3504823151125402,
+      "grad_norm": 0.3748853682253275,
+      "learning_rate": 9.850230414746545e-06,
+      "loss": 1.588,
+      "step": 109
+    },
+    {
+      "epoch": 0.3536977491961415,
+      "grad_norm": 0.41688250438538,
+      "learning_rate": 9.848047701481055e-06,
+      "loss": 1.6259,
+      "step": 110
+    },
+    {
+      "epoch": 0.35691318327974275,
+      "grad_norm": 0.5047715300641117,
+      "learning_rate": 9.845857418111755e-06,
+      "loss": 1.7629,
+      "step": 111
+    },
+    {
+      "epoch": 0.36012861736334406,
+      "grad_norm": 1.3947369892065562,
+      "learning_rate": 9.84365952518819e-06,
+      "loss": 1.755,
+      "step": 112
+    },
+    {
+      "epoch": 0.3633440514469453,
+      "grad_norm": 0.37395913731060204,
+      "learning_rate": 9.841453982985307e-06,
+      "loss": 1.7715,
+      "step": 113
+    },
+    {
+      "epoch": 0.3665594855305466,
+      "grad_norm": 0.6498894393940923,
+      "learning_rate": 9.839240751501066e-06,
+      "loss": 1.76,
+      "step": 114
+    },
+    {
+      "epoch": 0.36977491961414793,
+      "grad_norm": 0.32729381393912355,
+      "learning_rate": 9.837019790454017e-06,
+      "loss": 1.7168,
+      "step": 115
+    },
+    {
+      "epoch": 0.3729903536977492,
+      "grad_norm": 0.2880709453917275,
+      "learning_rate": 9.834791059280858e-06,
+      "loss": 1.8346,
+      "step": 116
+    },
+    {
+      "epoch": 0.3762057877813505,
+      "grad_norm": 0.3753012936624105,
+      "learning_rate": 9.832554517133957e-06,
+      "loss": 1.8204,
+      "step": 117
+    },
+    {
+      "epoch": 0.37942122186495175,
+      "grad_norm": 0.6853756846541161,
+      "learning_rate": 9.830310122878877e-06,
+      "loss": 1.7079,
+      "step": 118
+    },
+    {
+      "epoch": 0.38263665594855306,
+      "grad_norm": 0.37176959545385335,
+      "learning_rate": 9.828057835091833e-06,
+      "loss": 1.7688,
+      "step": 119
+    },
+    {
+      "epoch": 0.3858520900321543,
+      "grad_norm": 0.5291903456941895,
+      "learning_rate": 9.825797612057155e-06,
+      "loss": 1.7607,
+      "step": 120
+    },
+    {
+      "epoch": 0.3890675241157556,
+      "grad_norm": 0.6727781531506605,
+      "learning_rate": 9.823529411764706e-06,
+      "loss": 1.7175,
+      "step": 121
+    },
+    {
+      "epoch": 0.39228295819935693,
+      "grad_norm": 0.5075163995225351,
+      "learning_rate": 9.821253191907288e-06,
+      "loss": 1.7723,
+      "step": 122
+    },
+    {
+      "epoch": 0.3954983922829582,
+      "grad_norm": 0.4112992781424648,
+      "learning_rate": 9.818968909878002e-06,
+      "loss": 1.7529,
+      "step": 123
+    },
+    {
+      "epoch": 0.3987138263665595,
+      "grad_norm": 0.4775602103181259,
+      "learning_rate": 9.816676522767595e-06,
+      "loss": 1.6441,
+      "step": 124
+    },
+    {
+      "epoch": 0.40192926045016075,
+      "grad_norm": 0.7408080969119749,
+      "learning_rate": 9.81437598736177e-06,
+      "loss": 1.7677,
+      "step": 125
+    },
+    {
+      "epoch": 0.40514469453376206,
+      "grad_norm": 0.39439516285136644,
+      "learning_rate": 9.812067260138477e-06,
+      "loss": 1.6723,
+      "step": 126
+    },
+    {
+      "epoch": 0.40836012861736337,
+      "grad_norm": 0.35591432736328665,
+      "learning_rate": 9.809750297265161e-06,
+      "loss": 1.7665,
+      "step": 127
+    },
+    {
+      "epoch": 0.4115755627009646,
+      "grad_norm": 0.7717654585178413,
+      "learning_rate": 9.80742505459599e-06,
+      "loss": 1.7471,
+      "step": 128
+    },
+    {
+      "epoch": 0.41479099678456594,
+      "grad_norm": 0.38084935961821687,
+      "learning_rate": 9.805091487669056e-06,
+      "loss": 1.7191,
+      "step": 129
+    },
+    {
+      "epoch": 0.4180064308681672,
+      "grad_norm": 0.38431094171283037,
+      "learning_rate": 9.802749551703527e-06,
+      "loss": 1.753,
+      "step": 130
+    },
+    {
+      "epoch": 0.4212218649517685,
+      "grad_norm": 0.6945337606296567,
+      "learning_rate": 9.800399201596807e-06,
+      "loss": 1.8477,
+      "step": 131
+    },
+    {
+      "epoch": 0.42443729903536975,
+      "grad_norm": 0.28845445836156125,
+      "learning_rate": 9.798040391921617e-06,
+      "loss": 1.6586,
+      "step": 132
+    },
+    {
+      "epoch": 0.42765273311897106,
+      "grad_norm": 0.4385562306030188,
+      "learning_rate": 9.795673076923078e-06,
+      "loss": 1.7007,
+      "step": 133
+    },
+    {
+      "epoch": 0.43086816720257237,
+      "grad_norm": 0.35319986130974196,
+      "learning_rate": 9.793297210515755e-06,
+      "loss": 1.7712,
+      "step": 134
+    },
+    {
+      "epoch": 0.4340836012861736,
+      "grad_norm": 0.3341550532314632,
+      "learning_rate": 9.79091274628066e-06,
+      "loss": 1.6172,
+      "step": 135
+    },
+    {
+      "epoch": 0.43729903536977494,
+      "grad_norm": 0.44462279316358466,
+      "learning_rate": 9.788519637462236e-06,
+      "loss": 1.807,
+      "step": 136
+    },
+    {
+      "epoch": 0.4405144694533762,
+      "grad_norm": 0.3697826824798809,
+      "learning_rate": 9.786117836965296e-06,
+      "loss": 1.5715,
+      "step": 137
+    },
+    {
+      "epoch": 0.4437299035369775,
+      "grad_norm": 0.5000401238827548,
+      "learning_rate": 9.783707297351933e-06,
+      "loss": 1.6492,
+      "step": 138
+    },
+    {
+      "epoch": 0.44694533762057875,
+      "grad_norm": 0.32757243363816857,
+      "learning_rate": 9.781287970838397e-06,
+      "loss": 1.648,
+      "step": 139
+    },
+    {
+      "epoch": 0.45016077170418006,
+      "grad_norm": 1.3370622416951148,
+      "learning_rate": 9.778859809291946e-06,
+      "loss": 1.6093,
+      "step": 140
+    },
+    {
+      "epoch": 0.4533762057877814,
+      "grad_norm": 0.37062914068055847,
+      "learning_rate": 9.776422764227644e-06,
+      "loss": 1.7923,
+      "step": 141
+    },
+    {
+      "epoch": 0.4565916398713826,
+      "grad_norm": 0.47471068802805677,
+      "learning_rate": 9.773976786805133e-06,
+      "loss": 1.7107,
+      "step": 142
+    },
+    {
+      "epoch": 0.45980707395498394,
+      "grad_norm": 0.34897821399457957,
+      "learning_rate": 9.771521827825378e-06,
+      "loss": 1.7544,
+      "step": 143
+    },
+    {
+      "epoch": 0.4630225080385852,
+      "grad_norm": 0.5391221592013379,
+      "learning_rate": 9.769057837727365e-06,
+      "loss": 1.8124,
+      "step": 144
+    },
+    {
+      "epoch": 0.4662379421221865,
+      "grad_norm": 0.5422480089938636,
+      "learning_rate": 9.766584766584768e-06,
+      "loss": 1.6946,
+      "step": 145
+    },
+    {
+      "epoch": 0.4694533762057878,
+      "grad_norm": 0.3916809216288247,
+      "learning_rate": 9.764102564102566e-06,
+      "loss": 1.7273,
+      "step": 146
+    },
+    {
+      "epoch": 0.47266881028938906,
+      "grad_norm": 0.41161799735171617,
+      "learning_rate": 9.761611179613646e-06,
+      "loss": 1.6448,
+      "step": 147
+    },
+    {
+      "epoch": 0.4758842443729904,
+      "grad_norm": 0.36292020284907417,
+      "learning_rate": 9.759110562075355e-06,
+      "loss": 1.895,
+      "step": 148
+    },
+    {
+      "epoch": 0.4790996784565916,
+      "grad_norm": 0.301095836637533,
+      "learning_rate": 9.756600660066009e-06,
+      "loss": 1.6524,
+      "step": 149
+    },
+    {
+      "epoch": 0.48231511254019294,
+      "grad_norm": 0.39210820948008196,
+      "learning_rate": 9.75408142178136e-06,
+      "loss": 1.7231,
+      "step": 150
+    },
+    {
+      "epoch": 0.4855305466237942,
+      "grad_norm": 0.3866944738105665,
+      "learning_rate": 9.751552795031057e-06,
+      "loss": 1.6906,
+      "step": 151
+    },
+    {
+      "epoch": 0.4887459807073955,
+      "grad_norm": 0.49765003846519,
+      "learning_rate": 9.749014727235013e-06,
+      "loss": 1.5538,
+      "step": 152
+    },
+    {
+      "epoch": 0.4919614147909968,
+      "grad_norm": 0.6646611871216477,
+      "learning_rate": 9.746467165419785e-06,
+      "loss": 1.7296,
+      "step": 153
+    },
+    {
+      "epoch": 0.49517684887459806,
+      "grad_norm": 0.5269792497712498,
+      "learning_rate": 9.743910056214867e-06,
+      "loss": 1.7508,
+      "step": 154
+    },
+    {
+      "epoch": 0.4983922829581994,
+      "grad_norm": 0.6358378390742749,
+      "learning_rate": 9.741343345848979e-06,
+      "loss": 1.788,
+      "step": 155
+    },
+    {
+      "epoch": 0.5016077170418006,
+      "grad_norm": 0.8033471060172549,
+      "learning_rate": 9.73876698014629e-06,
+      "loss": 1.6622,
+      "step": 156
+    },
+    {
+      "epoch": 0.5048231511254019,
+      "grad_norm": 0.4300110838337297,
+      "learning_rate": 9.736180904522614e-06,
+      "loss": 1.6958,
+      "step": 157
+    },
+    {
+      "epoch": 0.5080385852090032,
+      "grad_norm": 0.44075312332289185,
+      "learning_rate": 9.733585063981542e-06,
+      "loss": 1.7791,
+      "step": 158
+    },
+    {
+      "epoch": 0.5112540192926045,
+      "grad_norm": 0.5552324541921604,
+      "learning_rate": 9.730979403110552e-06,
+      "loss": 1.7309,
+      "step": 159
+    },
+    {
+      "epoch": 0.5144694533762058,
+      "grad_norm": 0.5760086447839781,
+      "learning_rate": 9.72836386607707e-06,
+      "loss": 1.7043,
+      "step": 160
+    },
+    {
+      "epoch": 0.5176848874598071,
+      "grad_norm": 0.6120385835619392,
+      "learning_rate": 9.725738396624473e-06,
+      "loss": 1.7058,
+      "step": 161
+    },
+    {
+      "epoch": 0.5209003215434084,
+      "grad_norm": 0.3247721035171554,
+      "learning_rate": 9.723102938068061e-06,
+      "loss": 1.7192,
+      "step": 162
+    },
+    {
+      "epoch": 0.5241157556270096,
+      "grad_norm": 0.5504225309819054,
+      "learning_rate": 9.72045743329098e-06,
+      "loss": 1.6469,
+      "step": 163
+    },
+    {
+      "epoch": 0.5273311897106109,
+      "grad_norm": 0.6909293644017759,
+      "learning_rate": 9.717801824740082e-06,
+      "loss": 1.7844,
+      "step": 164
+    },
+    {
+      "epoch": 0.5305466237942122,
+      "grad_norm": 0.41329021494056956,
+      "learning_rate": 9.715136054421768e-06,
+      "loss": 1.7064,
+      "step": 165
+    },
+    {
+      "epoch": 0.5337620578778135,
+      "grad_norm": 0.4124736887649105,
+      "learning_rate": 9.712460063897765e-06,
+      "loss": 1.701,
+      "step": 166
+    },
+    {
+      "epoch": 0.5369774919614148,
+      "grad_norm": 0.44751507511817246,
+      "learning_rate": 9.709773794280837e-06,
+      "loss": 1.727,
+      "step": 167
+    },
+    {
+      "epoch": 0.5401929260450161,
+      "grad_norm": 0.4570947974943733,
+      "learning_rate": 9.707077186230491e-06,
+      "loss": 1.7531,
+      "step": 168
+    },
+    {
+      "epoch": 0.5434083601286174,
+      "grad_norm": 0.456791576642582,
+      "learning_rate": 9.704370179948586e-06,
+      "loss": 1.7289,
+      "step": 169
+    },
+    {
+      "epoch": 0.5466237942122186,
+      "grad_norm": 0.4897609545957888,
+      "learning_rate": 9.701652715174931e-06,
+      "loss": 1.7474,
+      "step": 170
+    },
+    {
+      "epoch": 0.5498392282958199,
+      "grad_norm": 0.3329368691084365,
+      "learning_rate": 9.698924731182796e-06,
+      "loss": 1.736,
+      "step": 171
+    },
+    {
+      "epoch": 0.5530546623794212,
+      "grad_norm": 0.3585759438897442,
+      "learning_rate": 9.696186166774403e-06,
+      "loss": 1.6931,
+      "step": 172
+    },
+    {
+      "epoch": 0.5562700964630225,
+      "grad_norm": 0.3628593026322015,
+      "learning_rate": 9.69343696027634e-06,
+      "loss": 1.6662,
+      "step": 173
+    },
+    {
+      "epoch": 0.5594855305466238,
+      "grad_norm": 0.455211826907985,
+      "learning_rate": 9.690677049534935e-06,
+      "loss": 1.736,
+      "step": 174
+    },
+    {
+      "epoch": 0.5627009646302251,
+      "grad_norm": 0.45923586331559213,
+      "learning_rate": 9.687906371911575e-06,
+      "loss": 1.7457,
+      "step": 175
+    },
+    {
+      "epoch": 0.5659163987138264,
+      "grad_norm": 0.521091953112185,
+      "learning_rate": 9.68512486427796e-06,
+      "loss": 1.6473,
+      "step": 176
+    },
+    {
+      "epoch": 0.5691318327974276,
+      "grad_norm": 0.37370844400034475,
+      "learning_rate": 9.682332463011316e-06,
+      "loss": 1.7925,
+      "step": 177
+    },
+    {
+      "epoch": 0.572347266881029,
+      "grad_norm": 0.33252276561485655,
+      "learning_rate": 9.679529103989536e-06,
+      "loss": 1.6674,
+      "step": 178
+    },
+    {
+      "epoch": 0.5755627009646302,
+      "grad_norm": 0.4194621576600804,
+      "learning_rate": 9.676714722586282e-06,
+      "loss": 1.6637,
+      "step": 179
+    },
+    {
+      "epoch": 0.5787781350482315,
+      "grad_norm": 0.36101868848046603,
+      "learning_rate": 9.67388925366601e-06,
+      "loss": 1.7907,
+      "step": 180
+    },
+    {
+      "epoch": 0.5819935691318328,
+      "grad_norm": 0.5995838884168876,
+      "learning_rate": 9.671052631578948e-06,
+      "loss": 1.6472,
+      "step": 181
+    },
+    {
+      "epoch": 0.5852090032154341,
+      "grad_norm": 0.31639066572509666,
+      "learning_rate": 9.668204790156012e-06,
+      "loss": 1.6797,
+      "step": 182
+    },
+    {
+      "epoch": 0.5884244372990354,
+      "grad_norm": 0.3286228611937227,
+      "learning_rate": 9.665345662703655e-06,
+      "loss": 1.729,
+      "step": 183
+    },
+    {
+      "epoch": 0.5916398713826366,
+      "grad_norm": 0.4145896260724765,
+      "learning_rate": 9.662475181998677e-06,
+      "loss": 1.6494,
+      "step": 184
+    },
+    {
+      "epoch": 0.594855305466238,
+      "grad_norm": 0.3841298222071092,
+      "learning_rate": 9.659593280282937e-06,
+      "loss": 1.8569,
+      "step": 185
+    },
+    {
+      "epoch": 0.5980707395498392,
+      "grad_norm": 0.3600521829909006,
+      "learning_rate": 9.65669988925803e-06,
+      "loss": 1.6805,
+      "step": 186
+    },
+    {
+      "epoch": 0.6012861736334405,
+      "grad_norm": 0.4282854580776211,
+      "learning_rate": 9.653794940079893e-06,
+      "loss": 1.8267,
+      "step": 187
+    },
+    {
+      "epoch": 0.6045016077170418,
+      "grad_norm": 0.6247529464498389,
+      "learning_rate": 9.650878363353347e-06,
+      "loss": 1.7073,
+      "step": 188
+    },
+    {
+      "epoch": 0.6077170418006431,
+      "grad_norm": 0.9115210250050622,
+      "learning_rate": 9.647950089126561e-06,
+      "loss": 1.7859,
+      "step": 189
+    },
+    {
+      "epoch": 0.6109324758842444,
+      "grad_norm": 0.4755718564603684,
+      "learning_rate": 9.645010046885466e-06,
+      "loss": 1.7883,
+      "step": 190
+    },
+    {
+      "epoch": 0.6141479099678456,
+      "grad_norm": 0.4523639938975236,
+      "learning_rate": 9.642058165548099e-06,
+      "loss": 1.6162,
+      "step": 191
+    },
+    {
+      "epoch": 0.617363344051447,
+      "grad_norm": 0.467081135964763,
+      "learning_rate": 9.639094373458865e-06,
+      "loss": 1.5402,
+      "step": 192
+    },
+    {
+      "epoch": 0.6205787781350482,
+      "grad_norm": 0.3670253791248765,
+      "learning_rate": 9.63611859838275e-06,
+      "loss": 1.7617,
+      "step": 193
+    },
+    {
+      "epoch": 0.6237942122186495,
+      "grad_norm": 1.1166686525871807,
+      "learning_rate": 9.633130767499438e-06,
+      "loss": 1.5991,
+      "step": 194
+    },
+    {
+      "epoch": 0.6270096463022508,
+      "grad_norm": 0.34857847452000734,
+      "learning_rate": 9.630130807397385e-06,
+      "loss": 1.7124,
+      "step": 195
+    },
+    {
+      "epoch": 0.6302250803858521,
+      "grad_norm": 0.3433159356058392,
+      "learning_rate": 9.627118644067797e-06,
+      "loss": 1.6253,
+      "step": 196
+    },
+    {
+      "epoch": 0.6334405144694534,
+      "grad_norm": 0.3810201491490459,
+      "learning_rate": 9.624094202898552e-06,
+      "loss": 1.7483,
+      "step": 197
+    },
+    {
+      "epoch": 0.6366559485530546,
+      "grad_norm": 0.3968780260221282,
+      "learning_rate": 9.62105740866803e-06,
+      "loss": 1.6065,
+      "step": 198
+    },
+    {
+      "epoch": 0.639871382636656,
+      "grad_norm": 0.5280487842852907,
+      "learning_rate": 9.618008185538883e-06,
+      "loss": 1.7588,
+      "step": 199
+    },
+    {
+      "epoch": 0.6430868167202572,
+      "grad_norm": 0.33732179856044275,
+      "learning_rate": 9.61494645705172e-06,
+      "loss": 1.6858,
+      "step": 200
+    },
+    {
+      "epoch": 0.6463022508038585,
+      "grad_norm": 0.4502107673996015,
+      "learning_rate": 9.611872146118723e-06,
+      "loss": 1.6584,
+      "step": 201
+    },
+    {
+      "epoch": 0.6495176848874598,
+      "grad_norm": 0.34373667243464945,
+      "learning_rate": 9.608785175017159e-06,
+      "loss": 1.7167,
+      "step": 202
+    },
+    {
+      "epoch": 0.6527331189710611,
+      "grad_norm": 0.46241506664930326,
+      "learning_rate": 9.605685465382853e-06,
+      "loss": 1.5874,
+      "step": 203
+    },
+    {
+      "epoch": 0.6559485530546624,
+      "grad_norm": 0.4457300415234676,
+      "learning_rate": 9.602572938203539e-06,
+      "loss": 1.5274,
+      "step": 204
+    },
+    {
+      "epoch": 0.6591639871382636,
+      "grad_norm": 0.3277148970485814,
+      "learning_rate": 9.599447513812156e-06,
+      "loss": 1.6461,
+      "step": 205
+    },
+    {
+      "epoch": 0.662379421221865,
+      "grad_norm": 0.4340323200722544,
+      "learning_rate": 9.596309111880047e-06,
+      "loss": 1.6812,
+      "step": 206
+    },
+    {
+      "epoch": 0.6655948553054662,
+      "grad_norm": 0.7144218287537519,
+      "learning_rate": 9.59315765141008e-06,
+      "loss": 1.7375,
+      "step": 207
+    },
+    {
+      "epoch": 0.6688102893890675,
+      "grad_norm": 0.6873856761013734,
+      "learning_rate": 9.589993050729674e-06,
+      "loss": 1.5441,
+      "step": 208
+    },
+    {
+      "epoch": 0.6720257234726688,
+      "grad_norm": 0.34753669260660897,
+      "learning_rate": 9.586815227483751e-06,
+      "loss": 1.7679,
+      "step": 209
+    },
+    {
+      "epoch": 0.6752411575562701,
+      "grad_norm": 0.46276953757117595,
+      "learning_rate": 9.583624098627588e-06,
+      "loss": 1.7122,
+      "step": 210
+    },
+    {
+      "epoch": 0.6784565916398714,
+      "grad_norm": 0.5296020589067518,
+      "learning_rate": 9.58041958041958e-06,
+      "loss": 1.7018,
+      "step": 211
+    },
+    {
+      "epoch": 0.6816720257234726,
+      "grad_norm": 0.4589413825352081,
+      "learning_rate": 9.577201588413924e-06,
+      "loss": 1.5787,
+      "step": 212
+    },
+    {
+      "epoch": 0.684887459807074,
+      "grad_norm": 0.6458129947828045,
+      "learning_rate": 9.573970037453184e-06,
+      "loss": 1.7081,
+      "step": 213
+    },
+    {
+      "epoch": 0.6881028938906752,
+      "grad_norm": 0.3837501853542465,
+      "learning_rate": 9.570724841660802e-06,
+      "loss": 1.7253,
+      "step": 214
+    },
+    {
+      "epoch": 0.6913183279742765,
+      "grad_norm": 2.684251622776636,
+      "learning_rate": 9.567465914433475e-06,
+      "loss": 1.6053,
+      "step": 215
+    },
+    {
+      "epoch": 0.6945337620578779,
+      "grad_norm": 0.3424217529372007,
+      "learning_rate": 9.564193168433452e-06,
+      "loss": 1.6773,
+      "step": 216
+    },
+    {
+      "epoch": 0.6977491961414791,
+      "grad_norm": 0.4204093374494449,
+      "learning_rate": 9.560906515580737e-06,
+      "loss": 1.6485,
+      "step": 217
+    },
+    {
+      "epoch": 0.7009646302250804,
+      "grad_norm": 0.9274816452954984,
+      "learning_rate": 9.557605867045187e-06,
+      "loss": 1.7436,
+      "step": 218
+    },
+    {
+      "epoch": 0.7041800643086816,
+      "grad_norm": 0.48902637197532123,
+      "learning_rate": 9.554291133238503e-06,
+      "loss": 1.7407,
+      "step": 219
+    },
+    {
+      "epoch": 0.707395498392283,
+      "grad_norm": 0.42302486731986544,
+      "learning_rate": 9.55096222380613e-06,
+      "loss": 1.7201,
+      "step": 220
+    },
+    {
+      "epoch": 0.7106109324758842,
+      "grad_norm": 0.6816522904850881,
+      "learning_rate": 9.547619047619049e-06,
+      "loss": 1.7592,
+      "step": 221
+    },
+    {
+      "epoch": 0.7138263665594855,
+      "grad_norm": 0.8034229829352186,
+      "learning_rate": 9.54426151276545e-06,
+      "loss": 1.6816,
+      "step": 222
+    },
+    {
+      "epoch": 0.7170418006430869,
+      "grad_norm": 0.5603448935112179,
+      "learning_rate": 9.540889526542325e-06,
+      "loss": 1.6417,
+      "step": 223
+    },
+    {
+      "epoch": 0.7202572347266881,
+      "grad_norm": 0.4187234758855297,
+      "learning_rate": 9.537502995446921e-06,
+      "loss": 1.7746,
+      "step": 224
+    },
+    {
+      "epoch": 0.7234726688102894,
+      "grad_norm": 0.36227230998153437,
+      "learning_rate": 9.53410182516811e-06,
+      "loss": 1.6367,
+      "step": 225
+    },
+    {
+      "epoch": 0.7266881028938906,
+      "grad_norm": 0.5106622344593317,
+      "learning_rate": 9.530685920577617e-06,
+      "loss": 1.7035,
+      "step": 226
+    },
+    {
+      "epoch": 0.729903536977492,
+      "grad_norm": 0.4027558733746282,
+      "learning_rate": 9.527255185721179e-06,
+      "loss": 1.6249,
+      "step": 227
+    },
+    {
+      "epoch": 0.7331189710610932,
+      "grad_norm": 0.5110606055911895,
+      "learning_rate": 9.523809523809525e-06,
+      "loss": 1.5942,
+      "step": 228
+    },
+    {
+      "epoch": 0.7363344051446945,
+      "grad_norm": 0.6758839413871981,
+      "learning_rate": 9.520348837209304e-06,
+      "loss": 1.7247,
+      "step": 229
+    },
+    {
+      "epoch": 0.7395498392282959,
+      "grad_norm": 1.1906021750644638,
+      "learning_rate": 9.516873027433845e-06,
+      "loss": 1.6526,
+      "step": 230
+    },
+    {
+      "epoch": 0.7427652733118971,
+      "grad_norm": 0.38273734006160703,
+      "learning_rate": 9.51338199513382e-06,
+      "loss": 1.6519,
+      "step": 231
+    },
+    {
+      "epoch": 0.7459807073954984,
+      "grad_norm": 0.40341824248617164,
+      "learning_rate": 9.509875640087784e-06,
+      "loss": 1.6969,
+      "step": 232
+    },
+    {
+      "epoch": 0.7491961414790996,
+      "grad_norm": 0.815911593005578,
+      "learning_rate": 9.506353861192572e-06,
+      "loss": 1.7442,
+      "step": 233
+    },
+    {
+      "epoch": 0.752411575562701,
+      "grad_norm": 2.2094840582917437,
+      "learning_rate": 9.502816556453589e-06,
+      "loss": 1.7439,
+      "step": 234
+    },
+    {
+      "epoch": 0.7556270096463023,
+      "grad_norm": 0.7519003593570723,
+      "learning_rate": 9.499263622974963e-06,
+      "loss": 1.7495,
+      "step": 235
+    },
+    {
+      "epoch": 0.7588424437299035,
+      "grad_norm": 0.37514672367476304,
+      "learning_rate": 9.49569495694957e-06,
+      "loss": 1.5284,
+      "step": 236
+    },
+    {
+      "epoch": 0.7620578778135049,
+      "grad_norm": 0.6226506367026134,
+      "learning_rate": 9.492110453648916e-06,
+      "loss": 1.6313,
+      "step": 237
+    },
+    {
+      "epoch": 0.7652733118971061,
+      "grad_norm": 0.447336191003546,
+      "learning_rate": 9.4885100074129e-06,
+      "loss": 1.6154,
+      "step": 238
+    },
+    {
+      "epoch": 0.7684887459807074,
+      "grad_norm": 0.36714620010490284,
+      "learning_rate": 9.484893511639425e-06,
+      "loss": 1.6554,
+      "step": 239
+    },
+    {
+      "epoch": 0.7717041800643086,
+      "grad_norm": 0.36304191332375657,
+      "learning_rate": 9.48126085877389e-06,
+      "loss": 1.5677,
+      "step": 240
+    },
+    {
+      "epoch": 0.77491961414791,
+      "grad_norm": 0.3145137130101893,
+      "learning_rate": 9.477611940298507e-06,
+      "loss": 1.7632,
+      "step": 241
+    },
+    {
+      "epoch": 0.7781350482315113,
+      "grad_norm": 0.4324951978035683,
+      "learning_rate": 9.473946646721517e-06,
+      "loss": 1.8002,
+      "step": 242
+    },
+    {
+      "epoch": 0.7813504823151125,
+      "grad_norm": 0.4831266498517749,
+      "learning_rate": 9.470264867566218e-06,
+      "loss": 1.7129,
+      "step": 243
+    },
+    {
+      "epoch": 0.7845659163987139,
+      "grad_norm": 0.6026850921759767,
+      "learning_rate": 9.46656649135988e-06,
+      "loss": 1.7952,
+      "step": 244
+    },
+    {
+      "epoch": 0.7877813504823151,
+      "grad_norm": 0.34281770743037326,
+      "learning_rate": 9.46285140562249e-06,
+      "loss": 1.7296,
+      "step": 245
+    },
+    {
+      "epoch": 0.7909967845659164,
+      "grad_norm": 0.4317358788821195,
+      "learning_rate": 9.459119496855347e-06,
+      "loss": 1.6606,
+      "step": 246
+    },
+    {
+      "epoch": 0.7942122186495176,
+      "grad_norm": 0.5439289773575391,
+      "learning_rate": 9.455370650529502e-06,
+      "loss": 1.7122,
+      "step": 247
+    },
+    {
+      "epoch": 0.797427652733119,
+      "grad_norm": 0.6700615541858532,
+      "learning_rate": 9.451604751074045e-06,
+      "loss": 1.6289,
+      "step": 248
+    },
+    {
+      "epoch": 0.8006430868167203,
+      "grad_norm": 0.31265590919433695,
+      "learning_rate": 9.447821681864235e-06,
+      "loss": 1.6228,
+      "step": 249
+    },
+    {
+      "epoch": 0.8038585209003215,
+      "grad_norm": 0.5328314957245512,
+      "learning_rate": 9.444021325209444e-06,
+      "loss": 1.8026,
+      "step": 250
+    },
+    {
+      "epoch": 0.8070739549839229,
+      "grad_norm": 0.43702581760266385,
+      "learning_rate": 9.440203562340969e-06,
+      "loss": 1.6136,
+      "step": 251
+    },
+    {
+      "epoch": 0.8102893890675241,
+      "grad_norm": 1.0070017597669874,
+      "learning_rate": 9.436368273399642e-06,
+      "loss": 1.6565,
+      "step": 252
+    },
+    {
+      "epoch": 0.8135048231511254,
+      "grad_norm": 0.5960613625871756,
+      "learning_rate": 9.432515337423314e-06,
+      "loss": 1.6371,
+      "step": 253
+    },
+    {
+      "epoch": 0.8167202572347267,
+      "grad_norm": 0.7616016258050006,
+      "learning_rate": 9.428644632334104e-06,
+      "loss": 1.6788,
+      "step": 254
+    },
+    {
+      "epoch": 0.819935691318328,
+      "grad_norm": 0.4218802074142177,
+      "learning_rate": 9.424756034925528e-06,
+      "loss": 1.81,
+      "step": 255
+    },
+    {
+      "epoch": 0.8231511254019293,
+      "grad_norm": 0.414457316699709,
+      "learning_rate": 9.420849420849423e-06,
+      "loss": 1.5819,
+      "step": 256
+    },
+    {
+      "epoch": 0.8263665594855305,
+      "grad_norm": 0.6347364275878894,
+      "learning_rate": 9.416924664602683e-06,
+      "loss": 1.7309,
+      "step": 257
+    },
+    {
+      "epoch": 0.8295819935691319,
+      "grad_norm": 0.4134627156259147,
+      "learning_rate": 9.412981639513837e-06,
+      "loss": 1.7553,
+      "step": 258
+    },
+    {
+      "epoch": 0.8327974276527331,
+      "grad_norm": 0.45271039800234564,
+      "learning_rate": 9.409020217729394e-06,
+      "loss": 1.606,
+      "step": 259
+    },
+    {
+      "epoch": 0.8360128617363344,
+      "grad_norm": 0.5653328484002546,
+      "learning_rate": 9.405040270200053e-06,
+      "loss": 1.6987,
+      "step": 260
+    },
+    {
+      "epoch": 0.8392282958199357,
+      "grad_norm": 0.4474839885592963,
+      "learning_rate": 9.401041666666667e-06,
+      "loss": 1.6809,
+      "step": 261
+    },
+    {
+      "epoch": 0.842443729903537,
+      "grad_norm": 0.46052107266954306,
+      "learning_rate": 9.397024275646046e-06,
+      "loss": 1.7181,
+      "step": 262
+    },
+    {
+      "epoch": 0.8456591639871383,
+      "grad_norm": 0.34499847597431416,
+      "learning_rate": 9.392987964416537e-06,
+      "loss": 1.7356,
+      "step": 263
+    },
+    {
+      "epoch": 0.8488745980707395,
+      "grad_norm": 0.5932683349504281,
+      "learning_rate": 9.388932599003411e-06,
+      "loss": 1.6987,
+      "step": 264
+    },
+    {
+      "epoch": 0.8520900321543409,
+      "grad_norm": 0.3989077412640957,
+      "learning_rate": 9.38485804416404e-06,
+      "loss": 1.6705,
+      "step": 265
+    },
+    {
+      "epoch": 0.8553054662379421,
+      "grad_norm": 0.4019284799432194,
+      "learning_rate": 9.38076416337286e-06,
+      "loss": 1.6986,
+      "step": 266
+    },
+    {
+      "epoch": 0.8585209003215434,
+      "grad_norm": 0.4142993445208996,
+      "learning_rate": 9.37665081880613e-06,
+      "loss": 1.6694,
+      "step": 267
+    },
+    {
+      "epoch": 0.8617363344051447,
+      "grad_norm": 0.6988031516188687,
+      "learning_rate": 9.37251787132645e-06,
+      "loss": 1.6771,
+      "step": 268
+    },
+    {
+      "epoch": 0.864951768488746,
+      "grad_norm": 0.40592122849069995,
+      "learning_rate": 9.368365180467092e-06,
+      "loss": 1.6458,
+      "step": 269
+    },
+    {
+      "epoch": 0.8681672025723473,
+      "grad_norm": 0.4902266091891843,
+      "learning_rate": 9.364192604416068e-06,
+      "loss": 1.716,
+      "step": 270
+    },
+    {
+      "epoch": 0.8713826366559485,
+      "grad_norm": 0.6296846798064156,
+      "learning_rate": 9.36e-06,
+      "loss": 1.6063,
+      "step": 271
+    },
+    {
+      "epoch": 0.8745980707395499,
+      "grad_norm": 0.5174782113353117,
+      "learning_rate": 9.355787222667737e-06,
+      "loss": 1.7948,
+      "step": 272
+    },
+    {
+      "epoch": 0.8778135048231511,
+      "grad_norm": 0.3730293067600126,
+      "learning_rate": 9.351554126473742e-06,
+      "loss": 1.7841,
+      "step": 273
+    },
+    {
+      "epoch": 0.8810289389067524,
+      "grad_norm": 0.647732251841464,
+      "learning_rate": 9.347300564061242e-06,
+      "loss": 1.5601,
+      "step": 274
+    },
+    {
+      "epoch": 0.8842443729903537,
+      "grad_norm": 0.7875287738906409,
+      "learning_rate": 9.343026386645127e-06,
+      "loss": 1.56,
+      "step": 275
+    },
+    {
+      "epoch": 0.887459807073955,
+      "grad_norm": 0.3720534077091662,
+      "learning_rate": 9.338731443994604e-06,
+      "loss": 1.5762,
+      "step": 276
+    },
+    {
+      "epoch": 0.8906752411575563,
+      "grad_norm": 0.801553893175062,
+      "learning_rate": 9.334415584415585e-06,
+      "loss": 1.5802,
+      "step": 277
+    },
+    {
+      "epoch": 0.8938906752411575,
+      "grad_norm": 0.49810159757057776,
+      "learning_rate": 9.330078654732847e-06,
+      "loss": 1.691,
+      "step": 278
+    },
+    {
+      "epoch": 0.8971061093247589,
+      "grad_norm": 0.33154338296209374,
+      "learning_rate": 9.325720500271888e-06,
+      "loss": 1.5735,
+      "step": 279
+    },
+    {
+      "epoch": 0.9003215434083601,
+      "grad_norm": 0.3793613901994727,
+      "learning_rate": 9.321340964840556e-06,
+      "loss": 1.7236,
+      "step": 280
+    },
+    {
+      "epoch": 0.9035369774919614,
+      "grad_norm": 0.4801354458812132,
+      "learning_rate": 9.316939890710383e-06,
+      "loss": 1.5809,
+      "step": 281
+    },
+    {
+      "epoch": 0.9067524115755627,
+      "grad_norm": 0.44437107250799746,
+      "learning_rate": 9.312517118597646e-06,
+      "loss": 1.7521,
+      "step": 282
+    },
+    {
+      "epoch": 0.909967845659164,
+      "grad_norm": 0.45242065176072754,
+      "learning_rate": 9.308072487644152e-06,
+      "loss": 1.6321,
+      "step": 283
+    },
+    {
+      "epoch": 0.9131832797427653,
+      "grad_norm": 0.4109067424412524,
+      "learning_rate": 9.303605835397742e-06,
+      "loss": 1.6303,
+      "step": 284
+    },
+    {
+      "epoch": 0.9163987138263665,
+      "grad_norm": 0.41773466853125474,
+      "learning_rate": 9.299116997792495e-06,
+      "loss": 1.7004,
+      "step": 285
+    },
+    {
+      "epoch": 0.9196141479099679,
+      "grad_norm": 0.39787310083891125,
+      "learning_rate": 9.294605809128632e-06,
+      "loss": 1.6362,
+      "step": 286
+    },
+    {
+      "epoch": 0.9228295819935691,
+      "grad_norm": 0.9588254824242197,
+      "learning_rate": 9.290072102052135e-06,
+      "loss": 1.586,
+      "step": 287
+    },
+    {
+      "epoch": 0.9260450160771704,
+      "grad_norm": 0.565145734756142,
+      "learning_rate": 9.285515707534057e-06,
+      "loss": 1.7777,
+      "step": 288
+    },
+    {
+      "epoch": 0.9292604501607717,
+      "grad_norm": 0.4506680876469492,
+      "learning_rate": 9.2809364548495e-06,
+      "loss": 1.6013,
+      "step": 289
+    },
+    {
+      "epoch": 0.932475884244373,
+      "grad_norm": 0.4501165211988009,
+      "learning_rate": 9.2763341715563e-06,
+      "loss": 1.6312,
+      "step": 290
+    },
+    {
+      "epoch": 0.9356913183279743,
+      "grad_norm": 0.3543433812557552,
+      "learning_rate": 9.27170868347339e-06,
+      "loss": 1.6267,
+      "step": 291
+    },
+    {
+      "epoch": 0.9389067524115756,
+      "grad_norm": 0.377909128143483,
+      "learning_rate": 9.267059814658803e-06,
+      "loss": 1.688,
+      "step": 292
+    },
+    {
+      "epoch": 0.9421221864951769,
+      "grad_norm": 0.392041296511637,
+      "learning_rate": 9.262387387387389e-06,
+      "loss": 1.6751,
+      "step": 293
+    },
+    {
+      "epoch": 0.9453376205787781,
+      "grad_norm": 0.4629012501293565,
+      "learning_rate": 9.257691222128141e-06,
+      "loss": 1.7538,
+      "step": 294
+    },
+    {
+      "epoch": 0.9485530546623794,
+      "grad_norm": 0.3520264198695876,
+      "learning_rate": 9.252971137521222e-06,
+      "loss": 1.5865,
+      "step": 295
+    },
+    {
+      "epoch": 0.9517684887459807,
+      "grad_norm": 0.8994489642586122,
+      "learning_rate": 9.248226950354609e-06,
+      "loss": 1.598,
+      "step": 296
+    },
+    {
+      "epoch": 0.954983922829582,
+      "grad_norm": 0.5378656826194321,
+      "learning_rate": 9.243458475540387e-06,
+      "loss": 1.663,
+      "step": 297
+    },
+    {
+      "epoch": 0.9581993569131833,
+      "grad_norm": 0.6360011252456891,
+      "learning_rate": 9.238665526090677e-06,
+      "loss": 1.6476,
+      "step": 298
+    },
+    {
+      "epoch": 0.9614147909967846,
+      "grad_norm": 0.4475516604413354,
+      "learning_rate": 9.233847913093197e-06,
+      "loss": 1.7376,
+      "step": 299
+    },
+    {
+      "epoch": 0.9646302250803859,
+      "grad_norm": 0.36689815703760215,
+      "learning_rate": 9.229005445686443e-06,
+      "loss": 1.7903,
+      "step": 300
+    },
+    {
+      "epoch": 0.9678456591639871,
+      "grad_norm": 0.6497688600569227,
+      "learning_rate": 9.224137931034482e-06,
+      "loss": 1.7326,
+      "step": 301
+    },
+    {
+      "epoch": 0.9710610932475884,
+      "grad_norm": 0.5278755924846331,
+      "learning_rate": 9.219245174301355e-06,
+      "loss": 1.5999,
+      "step": 302
+    },
+    {
+      "epoch": 0.9742765273311897,
+      "grad_norm": 0.7408394306218721,
+      "learning_rate": 9.214326978625072e-06,
+      "loss": 1.716,
+      "step": 303
+    },
+    {
+      "epoch": 0.977491961414791,
+      "grad_norm": 0.49833419873915763,
+      "learning_rate": 9.209383145091225e-06,
+      "loss": 1.645,
+      "step": 304
+    },
+    {
+      "epoch": 0.9807073954983923,
+      "grad_norm": 0.48566788490986423,
+      "learning_rate": 9.204413472706157e-06,
+      "loss": 1.7228,
+      "step": 305
+    },
+    {
+      "epoch": 0.9839228295819936,
+      "grad_norm": 0.33255489425334983,
+      "learning_rate": 9.199417758369725e-06,
+      "loss": 1.6768,
+      "step": 306
+    },
+    {
+      "epoch": 0.9871382636655949,
+      "grad_norm": 0.37814223485208215,
+      "learning_rate": 9.194395796847637e-06,
+      "loss": 1.7724,
+      "step": 307
+    },
+    {
+      "epoch": 0.9903536977491961,
+      "grad_norm": 0.477011074691142,
+      "learning_rate": 9.189347380743341e-06,
+      "loss": 1.4795,
+      "step": 308
+    },
+    {
+      "epoch": 0.9935691318327974,
+      "grad_norm": 0.4895940006411262,
+      "learning_rate": 9.184272300469483e-06,
+      "loss": 1.6117,
+      "step": 309
+    },
+    {
+      "epoch": 0.9967845659163987,
+      "grad_norm": 1.4139407357066591,
+      "learning_rate": 9.179170344218889e-06,
+      "loss": 1.6346,
+      "step": 310
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.6013595587806265,
+      "learning_rate": 9.174041297935104e-06,
+      "loss": 1.7778,
+      "step": 311
+    },
+    {
+      "epoch": 1.0032154340836013,
+      "grad_norm": 0.3247241569415427,
+      "learning_rate": 9.168884945282463e-06,
+      "loss": 1.6117,
+      "step": 312
+    },
+    {
+      "epoch": 1.0064308681672025,
+      "grad_norm": 0.4134569918286294,
+      "learning_rate": 9.163701067615658e-06,
+      "loss": 1.6031,
+      "step": 313
+    },
+    {
+      "epoch": 1.0096463022508038,
+      "grad_norm": 0.35906074261193893,
+      "learning_rate": 9.158489443948855e-06,
+      "loss": 1.6898,
+      "step": 314
+    },
+    {
+      "epoch": 1.0128617363344052,
+      "grad_norm": 0.3200371761779238,
+      "learning_rate": 9.15324985092427e-06,
+      "loss": 1.5302,
+      "step": 315
+    },
+    {
+      "epoch": 1.0160771704180065,
+      "grad_norm": 2.687902950840855,
+      "learning_rate": 9.147982062780271e-06,
+      "loss": 1.6216,
+      "step": 316
+    },
+    {
+      "epoch": 1.0192926045016077,
+      "grad_norm": 0.3884366277890246,
+      "learning_rate": 9.142685851318945e-06,
+      "loss": 1.5919,
+      "step": 317
+    },
+    {
+      "epoch": 1.022508038585209,
+      "grad_norm": 0.5992984151297521,
+      "learning_rate": 9.137360985873159e-06,
+      "loss": 1.7046,
+      "step": 318
+    },
+    {
+      "epoch": 1.0257234726688103,
+      "grad_norm": 0.9231033246198187,
+      "learning_rate": 9.132007233273057e-06,
+      "loss": 1.5955,
+      "step": 319
+    },
+    {
+      "epoch": 1.0289389067524115,
+      "grad_norm": 0.6011589166427741,
+      "learning_rate": 9.126624357812029e-06,
+      "loss": 1.6585,
+      "step": 320
+    },
+    {
+      "epoch": 1.0321543408360128,
+      "grad_norm": 0.4449266184028303,
+      "learning_rate": 9.121212121212124e-06,
+      "loss": 1.6662,
+      "step": 321
+    },
+    {
+      "epoch": 1.0353697749196142,
+      "grad_norm": 0.3792587689355056,
+      "learning_rate": 9.115770282588878e-06,
+      "loss": 1.725,
+      "step": 322
+    },
+    {
+      "epoch": 1.0385852090032155,
+      "grad_norm": 0.6352367505908573,
+      "learning_rate": 9.110298598415602e-06,
+      "loss": 1.719,
+      "step": 323
+    },
+    {
+      "epoch": 1.0418006430868167,
+      "grad_norm": 0.4941310259709336,
+      "learning_rate": 9.104796822487015e-06,
+      "loss": 1.7285,
+      "step": 324
+    },
+    {
+      "epoch": 1.045016077170418,
+      "grad_norm": 0.3988174469434485,
+      "learning_rate": 9.099264705882353e-06,
+      "loss": 1.6026,
+      "step": 325
+    },
+    {
+      "epoch": 1.0482315112540193,
+      "grad_norm": 0.37465191631335226,
+      "learning_rate": 9.093701996927803e-06,
+      "loss": 1.7684,
+      "step": 326
+    },
+    {
+      "epoch": 1.0514469453376205,
+      "grad_norm": 0.35360546929948017,
+      "learning_rate": 9.08810844115835e-06,
+      "loss": 1.861,
+      "step": 327
+    },
+    {
+      "epoch": 1.0546623794212218,
+      "grad_norm": 0.3687758841597626,
+      "learning_rate": 9.082483781278962e-06,
+      "loss": 1.6025,
+      "step": 328
+    },
+    {
+      "epoch": 1.0578778135048232,
+      "grad_norm": 0.4030524533633393,
+      "learning_rate": 9.076827757125155e-06,
+      "loss": 1.6729,
+      "step": 329
+    },
+    {
+      "epoch": 1.0610932475884245,
+      "grad_norm": 0.40031179933253463,
+      "learning_rate": 9.071140105622864e-06,
+      "loss": 1.7872,
+      "step": 330
+    },
+    {
+      "epoch": 1.0643086816720257,
+      "grad_norm": 0.3869462069975008,
+      "learning_rate": 9.065420560747664e-06,
+      "loss": 1.5474,
+      "step": 331
+    },
+    {
+      "epoch": 1.067524115755627,
+      "grad_norm": 0.37677780233171493,
+      "learning_rate": 9.059668853483288e-06,
+      "loss": 1.6805,
+      "step": 332
+    },
+    {
+      "epoch": 1.0707395498392283,
+      "grad_norm": 0.6628520504842822,
+      "learning_rate": 9.053884711779449e-06,
+      "loss": 1.7545,
+      "step": 333
+    },
+    {
+      "epoch": 1.0739549839228295,
+      "grad_norm": 0.33892446270593685,
+      "learning_rate": 9.048067860508956e-06,
+      "loss": 1.4833,
+      "step": 334
+    },
+    {
+      "epoch": 1.077170418006431,
+      "grad_norm": 0.38980433283323407,
+      "learning_rate": 9.042218021424073e-06,
+      "loss": 1.7001,
+      "step": 335
+    },
+    {
+      "epoch": 1.0803858520900322,
+      "grad_norm": 0.3460248454198814,
+      "learning_rate": 9.036334913112166e-06,
+      "loss": 1.7802,
+      "step": 336
+    },
+    {
+      "epoch": 1.0836012861736335,
+      "grad_norm": 0.8595770845976156,
+      "learning_rate": 9.030418250950572e-06,
+      "loss": 1.59,
+      "step": 337
+    },
+    {
+      "epoch": 1.0868167202572347,
+      "grad_norm": 0.3993362492113389,
+      "learning_rate": 9.024467747060694e-06,
+      "loss": 1.5678,
+      "step": 338
+    },
+    {
+      "epoch": 1.090032154340836,
+      "grad_norm": 0.4055047973666988,
+      "learning_rate": 9.018483110261313e-06,
+      "loss": 1.6281,
+      "step": 339
+    },
+    {
+      "epoch": 1.0932475884244373,
+      "grad_norm": 0.3431967651327368,
+      "learning_rate": 9.012464046021093e-06,
+      "loss": 1.7847,
+      "step": 340
+    },
+    {
+      "epoch": 1.0964630225080385,
+      "grad_norm": 0.6667661624149999,
+      "learning_rate": 9.006410256410258e-06,
+      "loss": 1.616,
+      "step": 341
+    },
+    {
+      "epoch": 1.09967845659164,
+      "grad_norm": 0.44360815958355554,
+      "learning_rate": 9.000321440051431e-06,
+      "loss": 1.6944,
+      "step": 342
+    },
+    {
+      "epoch": 1.1028938906752412,
+      "grad_norm": 0.5789231882394865,
+      "learning_rate": 8.994197292069633e-06,
+      "loss": 1.6589,
+      "step": 343
+    },
+    {
+      "epoch": 1.1061093247588425,
+      "grad_norm": 0.3696163005381853,
+      "learning_rate": 8.988037504041384e-06,
+      "loss": 1.6103,
+      "step": 344
+    },
+    {
+      "epoch": 1.1093247588424437,
+      "grad_norm": 0.48191850907372585,
+      "learning_rate": 8.981841763942932e-06,
+      "loss": 1.65,
+      "step": 345
+    },
+    {
+      "epoch": 1.112540192926045,
+      "grad_norm": 0.4363830053964839,
+      "learning_rate": 8.97560975609756e-06,
+      "loss": 1.6127,
+      "step": 346
+    },
+    {
+      "epoch": 1.1157556270096463,
+      "grad_norm": 0.6573325426179556,
+      "learning_rate": 8.969341161121983e-06,
+      "loss": 1.6575,
+      "step": 347
+    },
+    {
+      "epoch": 1.1189710610932475,
+      "grad_norm": 0.32836452628847956,
+      "learning_rate": 8.96303565587177e-06,
+      "loss": 1.6591,
+      "step": 348
+    },
+    {
+      "epoch": 1.122186495176849,
+      "grad_norm": 0.3950745313773845,
+      "learning_rate": 8.956692913385829e-06,
+      "loss": 1.7586,
+      "step": 349
+    },
+    {
+      "epoch": 1.1254019292604502,
+      "grad_norm": 0.44881182769164957,
+      "learning_rate": 8.95031260282988e-06,
+      "loss": 1.6586,
+      "step": 350
+    },
+    {
+      "epoch": 1.1286173633440515,
+      "grad_norm": 0.41574637746746984,
+      "learning_rate": 8.943894389438945e-06,
+      "loss": 1.5619,
+      "step": 351
+    },
+    {
+      "epoch": 1.1318327974276527,
+      "grad_norm": 0.5027726156090061,
+      "learning_rate": 8.937437934458789e-06,
+      "loss": 1.6758,
+      "step": 352
+    },
+    {
+      "epoch": 1.135048231511254,
+      "grad_norm": 0.47656134851226883,
+      "learning_rate": 8.930942895086322e-06,
+      "loss": 1.7921,
+      "step": 353
+    },
+    {
+      "epoch": 1.1382636655948553,
+      "grad_norm": 0.4935315407019529,
+      "learning_rate": 8.924408924408925e-06,
+      "loss": 1.7119,
+      "step": 354
+    },
+    {
+      "epoch": 1.1414790996784565,
+      "grad_norm": 0.40970084732835305,
+      "learning_rate": 8.917835671342687e-06,
+      "loss": 1.7594,
+      "step": 355
+    },
+    {
+      "epoch": 1.144694533762058,
+      "grad_norm": 0.4989653507775804,
+      "learning_rate": 8.911222780569515e-06,
+      "loss": 1.7095,
+      "step": 356
+    },
+    {
+      "epoch": 1.1479099678456592,
+      "grad_norm": 0.3357044002665187,
+      "learning_rate": 8.904569892473119e-06,
+      "loss": 1.5999,
+      "step": 357
+    },
+    {
+      "epoch": 1.1511254019292605,
+      "grad_norm": 0.521670365854549,
+      "learning_rate": 8.897876643073813e-06,
+      "loss": 1.6787,
+      "step": 358
+    },
+    {
+      "epoch": 1.1543408360128617,
+      "grad_norm": 0.6290404037058941,
+      "learning_rate": 8.891142663962136e-06,
+      "loss": 1.6908,
+      "step": 359
+    },
+    {
+      "epoch": 1.157556270096463,
+      "grad_norm": 0.57998281113531,
+      "learning_rate": 8.884367582231265e-06,
+      "loss": 1.7031,
+      "step": 360
+    },
+    {
+      "epoch": 1.1607717041800643,
+      "grad_norm": 0.45392621186353943,
+      "learning_rate": 8.877551020408163e-06,
+      "loss": 1.6294,
+      "step": 361
+    },
+    {
+      "epoch": 1.1639871382636655,
+      "grad_norm": 0.5309768238185915,
+      "learning_rate": 8.870692596383487e-06,
+      "loss": 1.7666,
+      "step": 362
+    },
+    {
+      "epoch": 1.167202572347267,
+      "grad_norm": 0.37321760200459936,
+      "learning_rate": 8.863791923340179e-06,
+      "loss": 1.6406,
+      "step": 363
+    },
+    {
+      "epoch": 1.1704180064308682,
+      "grad_norm": 0.36510942582873324,
+      "learning_rate": 8.856848609680742e-06,
+      "loss": 1.6532,
+      "step": 364
+    },
+    {
+      "epoch": 1.1736334405144695,
+      "grad_norm": 0.4563118581507281,
+      "learning_rate": 8.849862258953169e-06,
+      "loss": 1.5005,
+      "step": 365
+    },
+    {
+      "epoch": 1.1768488745980707,
+      "grad_norm": 0.4652354499783872,
+      "learning_rate": 8.842832469775476e-06,
+      "loss": 1.6991,
+      "step": 366
+    },
+    {
+      "epoch": 1.180064308681672,
+      "grad_norm": 0.4576168976717978,
+      "learning_rate": 8.835758835758838e-06,
+      "loss": 1.4367,
+      "step": 367
+    },
+    {
+      "epoch": 1.1832797427652733,
+      "grad_norm": 0.4513490696938803,
+      "learning_rate": 8.828640945429268e-06,
+      "loss": 1.7007,
+      "step": 368
+    },
+    {
+      "epoch": 1.1864951768488745,
+      "grad_norm": 0.4875941359520122,
+      "learning_rate": 8.82147838214784e-06,
+      "loss": 1.704,
+      "step": 369
+    },
+    {
+      "epoch": 1.189710610932476,
+      "grad_norm": 0.5119607567186321,
+      "learning_rate": 8.814270724029382e-06,
+      "loss": 1.8265,
+      "step": 370
+    },
+    {
+      "epoch": 1.1929260450160772,
+      "grad_norm": 0.4266982458686031,
+      "learning_rate": 8.807017543859649e-06,
+      "loss": 1.6714,
+      "step": 371
+    },
+    {
+      "epoch": 1.1961414790996785,
+      "grad_norm": 1.014493936774352,
+      "learning_rate": 8.799718409010911e-06,
+      "loss": 1.5943,
+      "step": 372
+    },
+    {
+      "epoch": 1.1993569131832797,
+      "grad_norm": 0.5718393100363165,
+      "learning_rate": 8.792372881355933e-06,
+      "loss": 1.7018,
+      "step": 373
+    },
+    {
+      "epoch": 1.202572347266881,
+      "grad_norm": 0.5138041344477277,
+      "learning_rate": 8.784980517180305e-06,
+      "loss": 1.6492,
+      "step": 374
+    },
+    {
+      "epoch": 1.2057877813504823,
+      "grad_norm": 0.3717081444541438,
+      "learning_rate": 8.777540867093106e-06,
+      "loss": 1.6968,
+      "step": 375
+    },
+    {
+      "epoch": 1.2090032154340835,
+      "grad_norm": 0.8842083310577157,
+      "learning_rate": 8.77005347593583e-06,
+      "loss": 1.5702,
+      "step": 376
+    },
+    {
+      "epoch": 1.212218649517685,
+      "grad_norm": 0.6554473379838456,
+      "learning_rate": 8.762517882689556e-06,
+      "loss": 1.5817,
+      "step": 377
+    },
+    {
+      "epoch": 1.2154340836012862,
+      "grad_norm": 0.5187574325471235,
+      "learning_rate": 8.754933620380337e-06,
+      "loss": 1.5754,
+      "step": 378
+    },
+    {
+      "epoch": 1.2186495176848875,
+      "grad_norm": 0.9386480775254568,
+      "learning_rate": 8.74730021598272e-06,
+      "loss": 1.6499,
+      "step": 379
+    },
+    {
+      "epoch": 1.2218649517684887,
+      "grad_norm": 0.42547480233916785,
+      "learning_rate": 8.739617190321417e-06,
+      "loss": 1.6008,
+      "step": 380
+    },
+    {
+      "epoch": 1.22508038585209,
+      "grad_norm": 0.7977120556800337,
+      "learning_rate": 8.731884057971014e-06,
+      "loss": 1.7798,
+      "step": 381
+    },
+    {
+      "epoch": 1.2282958199356913,
+      "grad_norm": 0.3566707305409597,
+      "learning_rate": 8.724100327153764e-06,
+      "loss": 1.6463,
+      "step": 382
+    },
+    {
+      "epoch": 1.2315112540192925,
+      "grad_norm": 0.6391377987077815,
+      "learning_rate": 8.716265499635304e-06,
+      "loss": 1.5919,
+      "step": 383
+    },
+    {
+      "epoch": 1.234726688102894,
+      "grad_norm": 0.3889680511222192,
+      "learning_rate": 8.708379070618369e-06,
+      "loss": 1.7207,
+      "step": 384
+    },
+    {
+      "epoch": 1.2379421221864952,
+      "grad_norm": 0.4540928202776,
+      "learning_rate": 8.700440528634362e-06,
+      "loss": 1.7255,
+      "step": 385
+    },
+    {
+      "epoch": 1.2411575562700965,
+      "grad_norm": 0.40457708530753206,
+      "learning_rate": 8.692449355432781e-06,
+      "loss": 1.5159,
+      "step": 386
+    },
+    {
+      "epoch": 1.2443729903536977,
+      "grad_norm": 0.5512339669704394,
+      "learning_rate": 8.68440502586844e-06,
+      "loss": 1.6499,
+      "step": 387
+    },
+    {
+      "epoch": 1.247588424437299,
+      "grad_norm": 2.776005983432967,
+      "learning_rate": 8.676307007786431e-06,
+      "loss": 1.6102,
+      "step": 388
+    },
+    {
+      "epoch": 1.2508038585209003,
+      "grad_norm": 0.364952370477986,
+      "learning_rate": 8.668154761904762e-06,
+      "loss": 1.639,
+      "step": 389
+    },
+    {
+      "epoch": 1.2540192926045015,
+      "grad_norm": 0.3811434890957348,
+      "learning_rate": 8.659947741694661e-06,
+      "loss": 1.8283,
+      "step": 390
+    },
+    {
+      "epoch": 1.257234726688103,
+      "grad_norm": 0.43050657428200295,
+      "learning_rate": 8.651685393258428e-06,
+      "loss": 1.5902,
+      "step": 391
+    },
+    {
+      "epoch": 1.2604501607717042,
+      "grad_norm": 0.5117898550286608,
+      "learning_rate": 8.64336715520481e-06,
+      "loss": 1.7092,
+      "step": 392
+    },
+    {
+      "epoch": 1.2636655948553055,
+      "grad_norm": 0.6470526145584239,
+      "learning_rate": 8.63499245852187e-06,
+      "loss": 1.7055,
+      "step": 393
+    },
+    {
+      "epoch": 1.2668810289389068,
+      "grad_norm": 0.40636513274929337,
+      "learning_rate": 8.626560726447218e-06,
+      "loss": 1.5638,
+      "step": 394
+    },
+    {
+      "epoch": 1.270096463022508,
+      "grad_norm": 5.812656682454248,
+      "learning_rate": 8.618071374335612e-06,
+      "loss": 1.692,
+      "step": 395
+    },
+    {
+      "epoch": 1.2733118971061093,
+      "grad_norm": 0.44580256068577884,
+      "learning_rate": 8.609523809523808e-06,
+      "loss": 1.5292,
+      "step": 396
+    },
+    {
+      "epoch": 1.2765273311897105,
+      "grad_norm": 0.4246595984926517,
+      "learning_rate": 8.600917431192661e-06,
+      "loss": 1.7548,
+      "step": 397
+    },
+    {
+      "epoch": 1.279742765273312,
+      "grad_norm": 0.5152992435939208,
+      "learning_rate": 8.592251630226314e-06,
+      "loss": 1.7866,
+      "step": 398
+    },
+    {
+      "epoch": 1.2829581993569132,
+      "grad_norm": 0.4963075673414894,
+      "learning_rate": 8.583525789068516e-06,
+      "loss": 1.6988,
+      "step": 399
+    },
+    {
+      "epoch": 1.2861736334405145,
+      "grad_norm": 0.45125649284081243,
+      "learning_rate": 8.574739281575898e-06,
+      "loss": 1.5701,
+      "step": 400
+    },
+    {
+      "epoch": 1.2893890675241158,
+      "grad_norm": 0.53317806439095,
+      "learning_rate": 8.565891472868219e-06,
+      "loss": 1.7179,
+      "step": 401
+    },
+    {
+      "epoch": 1.292604501607717,
+      "grad_norm": 0.6275759768767125,
+      "learning_rate": 8.55698171917542e-06,
+      "loss": 1.6918,
+      "step": 402
+    },
+    {
+      "epoch": 1.2958199356913183,
+      "grad_norm": 0.38316459378318246,
+      "learning_rate": 8.5480093676815e-06,
+      "loss": 1.7194,
+      "step": 403
+    },
+    {
+      "epoch": 1.2990353697749195,
+      "grad_norm": 0.42281492046011937,
+      "learning_rate": 8.53897375636506e-06,
+      "loss": 1.6056,
+      "step": 404
+    },
+    {
+      "epoch": 1.302250803858521,
+      "grad_norm": 0.5423212471954062,
+      "learning_rate": 8.52987421383648e-06,
+      "loss": 1.6599,
+      "step": 405
+    },
+    {
+      "epoch": 1.3054662379421222,
+      "grad_norm": 0.3926198800543086,
+      "learning_rate": 8.520710059171598e-06,
+      "loss": 1.7273,
+      "step": 406
+    },
+    {
+      "epoch": 1.3086816720257235,
+      "grad_norm": 0.41111535425621437,
+      "learning_rate": 8.511480601741886e-06,
+      "loss": 1.6712,
+      "step": 407
+    },
+    {
+      "epoch": 1.3118971061093248,
+      "grad_norm": 0.4164950928775684,
+      "learning_rate": 8.502185141040922e-06,
+      "loss": 1.6356,
+      "step": 408
+    },
+    {
+      "epoch": 1.315112540192926,
+      "grad_norm": 0.42331481781783037,
+      "learning_rate": 8.492822966507178e-06,
+      "loss": 1.7424,
+      "step": 409
+    },
+    {
+      "epoch": 1.3183279742765273,
+      "grad_norm": 0.5375647747454178,
+      "learning_rate": 8.483393357342938e-06,
+      "loss": 1.727,
+      "step": 410
+    },
+    {
+      "epoch": 1.3215434083601285,
+      "grad_norm": 0.3511881681775179,
+      "learning_rate": 8.473895582329317e-06,
+      "loss": 1.6824,
+      "step": 411
+    },
+    {
+      "epoch": 1.32475884244373,
+      "grad_norm": 0.3745512197529067,
+      "learning_rate": 8.464328899637244e-06,
+      "loss": 1.6541,
+      "step": 412
+    },
+    {
+      "epoch": 1.3279742765273312,
+      "grad_norm": 0.36035208515341016,
+      "learning_rate": 8.454692556634303e-06,
+      "loss": 1.5663,
+      "step": 413
+    },
+    {
+      "epoch": 1.3311897106109325,
+      "grad_norm": 0.5271629572900195,
+      "learning_rate": 8.444985789687374e-06,
+      "loss": 1.6666,
+      "step": 414
+    },
+    {
+      "epoch": 1.3344051446945338,
+      "grad_norm": 0.4035052849324303,
+      "learning_rate": 8.435207823960882e-06,
+      "loss": 1.666,
+      "step": 415
+    },
+    {
+      "epoch": 1.337620578778135,
+      "grad_norm": 0.5141622387219502,
+      "learning_rate": 8.425357873210634e-06,
+      "loss": 1.6095,
+      "step": 416
+    },
+    {
+      "epoch": 1.3408360128617363,
+      "grad_norm": 0.3903148174211921,
+      "learning_rate": 8.41543513957307e-06,
+      "loss": 1.7249,
+      "step": 417
+    },
+    {
+      "epoch": 1.3440514469453375,
+      "grad_norm": 0.3837027760154029,
+      "learning_rate": 8.405438813349816e-06,
+      "loss": 1.6892,
+      "step": 418
+    },
+    {
+      "epoch": 1.347266881028939,
+      "grad_norm": 0.5991756282596358,
+      "learning_rate": 8.395368072787428e-06,
+      "loss": 1.4517,
+      "step": 419
+    },
+    {
+      "epoch": 1.3504823151125402,
+      "grad_norm": 0.8352176814664566,
+      "learning_rate": 8.38522208385222e-06,
+      "loss": 1.7663,
+      "step": 420
+    },
+    {
+      "epoch": 1.3536977491961415,
+      "grad_norm": 0.39312344645907427,
+      "learning_rate": 8.375e-06,
+      "loss": 1.779,
+      "step": 421
+    },
+    {
+      "epoch": 1.3569131832797428,
+      "grad_norm": 0.6153100653488431,
+      "learning_rate": 8.364700961940612e-06,
+      "loss": 1.603,
+      "step": 422
+    },
+    {
+      "epoch": 1.360128617363344,
+      "grad_norm": 0.4917240270173188,
+      "learning_rate": 8.354324097397145e-06,
+      "loss": 1.8095,
+      "step": 423
+    },
+    {
+      "epoch": 1.3633440514469453,
+      "grad_norm": 0.45200112908044476,
+      "learning_rate": 8.34386852085967e-06,
+      "loss": 1.6269,
+      "step": 424
+    },
+    {
+      "epoch": 1.3665594855305465,
+      "grad_norm": 0.49866971451151026,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 1.5597,
+      "step": 425
+    },
+    {
+      "epoch": 1.369774919614148,
+      "grad_norm": 0.4813276516774304,
+      "learning_rate": 8.322717622080679e-06,
+      "loss": 1.6371,
+      "step": 426
+    },
+    {
+      "epoch": 1.3729903536977492,
+      "grad_norm": 0.6706487993975442,
+      "learning_rate": 8.312020460358056e-06,
+      "loss": 1.6942,
+      "step": 427
+    },
+    {
+      "epoch": 1.3762057877813505,
+      "grad_norm": 0.4410278961954752,
+      "learning_rate": 8.301240907145914e-06,
+      "loss": 1.6795,
+      "step": 428
+    },
+    {
+      "epoch": 1.3794212218649518,
+      "grad_norm": 0.6619090011753649,
+      "learning_rate": 8.290378006872852e-06,
+      "loss": 1.6588,
+      "step": 429
+    },
+    {
+      "epoch": 1.382636655948553,
+      "grad_norm": 0.3835351559503038,
+      "learning_rate": 8.279430789133249e-06,
+      "loss": 1.6239,
+      "step": 430
+    },
+    {
+      "epoch": 1.3858520900321543,
+      "grad_norm": 0.3636590588929943,
+      "learning_rate": 8.268398268398268e-06,
+      "loss": 1.6285,
+      "step": 431
+    },
+    {
+      "epoch": 1.3890675241157555,
+      "grad_norm": 0.44522127662825833,
+      "learning_rate": 8.257279443720123e-06,
+      "loss": 1.5958,
+      "step": 432
+    },
+    {
+      "epoch": 1.392282958199357,
+      "grad_norm": 0.4367527889875373,
+      "learning_rate": 8.24607329842932e-06,
+      "loss": 1.639,
+      "step": 433
+    },
+    {
+      "epoch": 1.3954983922829582,
+      "grad_norm": 0.4182542142164959,
+      "learning_rate": 8.234778799824793e-06,
+      "loss": 1.6411,
+      "step": 434
+    },
+    {
+      "epoch": 1.3987138263665595,
+      "grad_norm": 0.4220733115803737,
+      "learning_rate": 8.223394898856639e-06,
+      "loss": 1.6993,
+      "step": 435
+    },
+    {
+      "epoch": 1.4019292604501608,
+      "grad_norm": 0.5095746374942131,
+      "learning_rate": 8.211920529801326e-06,
+      "loss": 1.5549,
+      "step": 436
+    },
+    {
+      "epoch": 1.405144694533762,
+      "grad_norm": 0.4003936984745165,
+      "learning_rate": 8.200354609929079e-06,
+      "loss": 1.6424,
+      "step": 437
+    },
+    {
+      "epoch": 1.4083601286173635,
+      "grad_norm": 0.35448019224992106,
+      "learning_rate": 8.18869603916333e-06,
+      "loss": 1.7231,
+      "step": 438
+    },
+    {
+      "epoch": 1.4115755627009645,
+      "grad_norm": 0.3958358149507715,
+      "learning_rate": 8.176943699731904e-06,
+      "loss": 1.5607,
+      "step": 439
+    },
+    {
+      "epoch": 1.414790996784566,
+      "grad_norm": 0.4572398550528086,
+      "learning_rate": 8.16509645580978e-06,
+      "loss": 1.7685,
+      "step": 440
+    },
+    {
+      "epoch": 1.4180064308681672,
+      "grad_norm": 0.4108744685040581,
+      "learning_rate": 8.153153153153154e-06,
+      "loss": 1.7027,
+      "step": 441
+    },
+    {
+      "epoch": 1.4212218649517685,
+      "grad_norm": 0.39339819264830506,
+      "learning_rate": 8.141112618724561e-06,
+      "loss": 1.5663,
+      "step": 442
+    },
+    {
+      "epoch": 1.4244372990353698,
+      "grad_norm": 0.555394722524567,
+      "learning_rate": 8.12897366030881e-06,
+      "loss": 1.7123,
+      "step": 443
+    },
+    {
+      "epoch": 1.427652733118971,
+      "grad_norm": 0.32604297722404485,
+      "learning_rate": 8.11673506611947e-06,
+      "loss": 1.7202,
+      "step": 444
+    },
+    {
+      "epoch": 1.4308681672025725,
+      "grad_norm": 0.57079829329324,
+      "learning_rate": 8.104395604395604e-06,
+      "loss": 1.655,
+      "step": 445
+    },
+    {
+      "epoch": 1.4340836012861735,
+      "grad_norm": 0.4490468868427245,
+      "learning_rate": 8.091954022988505e-06,
+      "loss": 1.5641,
+      "step": 446
+    },
+    {
+      "epoch": 1.437299035369775,
+      "grad_norm": 0.43447954022994156,
+      "learning_rate": 8.079409048938135e-06,
+      "loss": 1.6025,
+      "step": 447
+    },
+    {
+      "epoch": 1.4405144694533762,
+      "grad_norm": 0.5540793015292118,
+      "learning_rate": 8.066759388038942e-06,
+      "loss": 1.7527,
+      "step": 448
+    },
+    {
+      "epoch": 1.4437299035369775,
+      "grad_norm": 0.3563614235496872,
+      "learning_rate": 8.054003724394785e-06,
+      "loss": 1.5934,
+      "step": 449
+    },
+    {
+      "epoch": 1.4469453376205788,
+      "grad_norm": 0.913224093708608,
+      "learning_rate": 8.041140719962602e-06,
+      "loss": 1.6911,
+      "step": 450
+    },
+    {
+      "epoch": 1.45016077170418,
+      "grad_norm": 0.43504913816692775,
+      "learning_rate": 8.028169014084509e-06,
+      "loss": 1.5177,
+      "step": 451
+    },
+    {
+      "epoch": 1.4533762057877815,
+      "grad_norm": 0.48590476141217886,
+      "learning_rate": 8.015087223008017e-06,
+      "loss": 1.7909,
+      "step": 452
+    },
+    {
+      "epoch": 1.4565916398713825,
+      "grad_norm": 0.5080378493045535,
+      "learning_rate": 8.00189393939394e-06,
+      "loss": 1.5838,
+      "step": 453
+    },
+    {
+      "epoch": 1.459807073954984,
+      "grad_norm": 1.64509386470527,
+      "learning_rate": 7.988587731811699e-06,
+      "loss": 1.6472,
+      "step": 454
+    },
+    {
+      "epoch": 1.4630225080385852,
+      "grad_norm": 0.3673483192332043,
+      "learning_rate": 7.975167144221587e-06,
+      "loss": 1.6316,
+      "step": 455
+    },
+    {
+      "epoch": 1.4662379421221865,
+      "grad_norm": 0.41136311504318224,
+      "learning_rate": 7.961630695443646e-06,
+      "loss": 1.761,
+      "step": 456
+    },
+    {
+      "epoch": 1.4694533762057878,
+      "grad_norm": 0.35308464122017075,
+      "learning_rate": 7.947976878612718e-06,
+      "loss": 1.536,
+      "step": 457
+    },
+    {
+      "epoch": 1.472668810289389,
+      "grad_norm": 0.41734252770432045,
+      "learning_rate": 7.934204160619255e-06,
+      "loss": 1.6562,
+      "step": 458
+    },
+    {
+      "epoch": 1.4758842443729905,
+      "grad_norm": 0.44696724275948946,
+      "learning_rate": 7.920310981535472e-06,
+      "loss": 1.6589,
+      "step": 459
+    },
+    {
+      "epoch": 1.4790996784565915,
+      "grad_norm": 1.740141532341783,
+      "learning_rate": 7.906295754026353e-06,
+      "loss": 1.5841,
+      "step": 460
+    },
+    {
+      "epoch": 1.482315112540193,
+      "grad_norm": 0.38345032126320727,
+      "learning_rate": 7.892156862745098e-06,
+      "loss": 1.5586,
+      "step": 461
+    },
+    {
+      "epoch": 1.4855305466237942,
+      "grad_norm": 0.3505893213058495,
+      "learning_rate": 7.877892663712457e-06,
+      "loss": 1.6964,
+      "step": 462
+    },
+    {
+      "epoch": 1.4887459807073955,
+      "grad_norm": 0.34378052648599977,
+      "learning_rate": 7.863501483679526e-06,
+      "loss": 1.6686,
+      "step": 463
+    },
+    {
+      "epoch": 1.4919614147909968,
+      "grad_norm": 0.43787743191378786,
+      "learning_rate": 7.848981619473423e-06,
+      "loss": 1.7217,
+      "step": 464
+    },
+    {
+      "epoch": 1.495176848874598,
+      "grad_norm": 0.4599462355427195,
+      "learning_rate": 7.834331337325349e-06,
+      "loss": 1.5408,
+      "step": 465
+    },
+    {
+      "epoch": 1.4983922829581995,
+      "grad_norm": 0.47641665252632553,
+      "learning_rate": 7.819548872180452e-06,
+      "loss": 1.6856,
+      "step": 466
+    },
+    {
+      "epoch": 1.5016077170418005,
+      "grad_norm": 0.5146461637653273,
+      "learning_rate": 7.804632426988925e-06,
+      "loss": 1.6533,
+      "step": 467
+    },
+    {
+      "epoch": 1.504823151125402,
+      "grad_norm": 0.44826649985888556,
+      "learning_rate": 7.789580171977745e-06,
+      "loss": 1.587,
+      "step": 468
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 622,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 156,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.162683181480018e+17,
+  "train_batch_size": 3,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a86bf8d9477e72976e0a27b0e142493e8cb3d748490c2a974cb9f77bd67d914
+size 8888

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)