diff --git a/save_lora_instruct_r8_2e4_s11_v2/README.md b/save_lora_instruct_r8_2e4_s11_v2/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..d380bbbf0bf0fa1f211b53410559607e5007a209
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: meta-llama/Llama-2-7b-hf
+---
+
+# 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 0.9.0
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/adapter_config.json b/save_lora_instruct_r8_2e4_s11_v2/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c81a592f0ff08374a4f61f6003029070a8429ae
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/adapter_config.json
@@ -0,0 +1,35 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "mora_type": 1,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_mora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/adapter_model.safetensors b/save_lora_instruct_r8_2e4_s11_v2/adapter_model.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..05554982ab60151866ce542b304ed8121b43add9
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/adapter_model.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a485c55c32ea5d70c88fda6d211474749d855619eb4c5b4aa225293e09d9da98
+size 564342240
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/README.md b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..d380bbbf0bf0fa1f211b53410559607e5007a209
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: meta-llama/Llama-2-7b-hf
+---
+
+# 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 0.9.0
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/adapter_config.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c81a592f0ff08374a4f61f6003029070a8429ae
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/adapter_config.json
@@ -0,0 +1,35 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "mora_type": 1,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_mora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/adapter_model.safetensors b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/adapter_model.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..e23e88143160a3d3863343dc09c9abbe2c89b14b
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/adapter_model.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc1edca5f464acc3dc6f239977259478fa8ba15acaa46569fe428bbfa9fdcc36
+size 564342240
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/global_step2000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/global_step2000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3703e0bbcd850fc1dc9604bae2c989d9cd153bb6
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/global_step2000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1bcf95c3aae5cb9697ebddd8878b6984e90a1b1280dc7aa85efc0d1494c84b96
+size 239914224
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/global_step2000/mp_rank_00_model_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/global_step2000/mp_rank_00_model_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..73542385294d522dac16ff57e8fbf084ba07c198
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/global_step2000/mp_rank_00_model_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16a382ded0b4fd0796d713dafe5721f70f4af6c9faa0d211d8502f5c13cfb1d8
+size 40303404
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/latest b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/latest
new file mode 100644
index 0000000000000000000000000000000000000000..2a79fdc19587e6bc9de060e90633f3a151b04516
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/latest
@@ -0,0 +1 @@
+global_step2000
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/rng_state.pth b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/rng_state.pth
new file mode 100644
index 0000000000000000000000000000000000000000..dc5168604114fe8873a415e695c403804f881e2d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/rng_state.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01850e1315165c5e052c036124faa2f6de6533691d8649e2233f6ae6493d461c
+size 14244
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/scheduler.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/scheduler.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4cba1ee9f732a8d9836710421bdd2f6db8a9c37d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/scheduler.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0868cf0d4c19e8673c246e79cd537d21bbd3e13298159c66bd69908f72d5c416
+size 1064
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/trainer_state.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..829bef516b28c25cdc9c5a84a96dbe3e49769815
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/trainer_state.json
@@ -0,0 +1,2421 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.7983284997037453,
+  "eval_steps": 500,
+  "global_step": 2000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.3797,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.3952,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2e-05,
+      "loss": 1.3112,
+      "step": 15
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 1.3136,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 1.2012,
+      "step": 25
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4e-05,
+      "loss": 1.1277,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 1.0832,
+      "step": 35
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.333333333333333e-05,
+      "loss": 1.0659,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-05,
+      "loss": 0.983,
+      "step": 45
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.9763,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.333333333333333e-05,
+      "loss": 0.9992,
+      "step": 55
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8e-05,
+      "loss": 0.9405,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.666666666666667e-05,
+      "loss": 0.9827,
+      "step": 65
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.333333333333334e-05,
+      "loss": 0.9061,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001,
+      "loss": 0.8921,
+      "step": 75
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010666666666666667,
+      "loss": 0.9377,
+      "step": 80
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011333333333333334,
+      "loss": 0.8528,
+      "step": 85
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012,
+      "loss": 0.8913,
+      "step": 90
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012666666666666666,
+      "loss": 0.8562,
+      "step": 95
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.877,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014,
+      "loss": 0.8781,
+      "step": 105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014666666666666666,
+      "loss": 0.8546,
+      "step": 110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015333333333333334,
+      "loss": 0.8672,
+      "step": 115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016,
+      "loss": 0.8611,
+      "step": 120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.8624,
+      "step": 125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017333333333333334,
+      "loss": 0.8477,
+      "step": 130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018,
+      "loss": 0.8227,
+      "step": 135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001866666666666667,
+      "loss": 0.8064,
+      "step": 140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019333333333333333,
+      "loss": 0.8422,
+      "step": 145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002,
+      "loss": 0.8471,
+      "step": 150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001997942386831276,
+      "loss": 0.8183,
+      "step": 155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958847736625514,
+      "loss": 0.8229,
+      "step": 160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019938271604938272,
+      "loss": 0.8114,
+      "step": 165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991769547325103,
+      "loss": 0.7861,
+      "step": 170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019897119341563788,
+      "loss": 0.841,
+      "step": 175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019876543209876543,
+      "loss": 0.8549,
+      "step": 180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000198559670781893,
+      "loss": 0.8449,
+      "step": 185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001983539094650206,
+      "loss": 0.8311,
+      "step": 190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019814814814814814,
+      "loss": 0.8296,
+      "step": 195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019794238683127572,
+      "loss": 0.8263,
+      "step": 200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001977366255144033,
+      "loss": 0.8452,
+      "step": 205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019753086419753085,
+      "loss": 0.8044,
+      "step": 210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019732510288065843,
+      "loss": 0.8119,
+      "step": 215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000197119341563786,
+      "loss": 0.8127,
+      "step": 220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001969135802469136,
+      "loss": 0.7921,
+      "step": 225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019670781893004114,
+      "loss": 0.803,
+      "step": 230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019650205761316872,
+      "loss": 0.838,
+      "step": 235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001962962962962963,
+      "loss": 0.7801,
+      "step": 240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019609053497942388,
+      "loss": 0.817,
+      "step": 245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019588477366255146,
+      "loss": 0.8211,
+      "step": 250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000195679012345679,
+      "loss": 0.8423,
+      "step": 255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001954732510288066,
+      "loss": 0.8007,
+      "step": 260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019526748971193417,
+      "loss": 0.8135,
+      "step": 265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019506172839506175,
+      "loss": 0.7961,
+      "step": 270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019485596707818933,
+      "loss": 0.797,
+      "step": 275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019465020576131688,
+      "loss": 0.793,
+      "step": 280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019444444444444446,
+      "loss": 0.7809,
+      "step": 285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019423868312757204,
+      "loss": 0.7811,
+      "step": 290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019403292181069962,
+      "loss": 0.8074,
+      "step": 295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019382716049382717,
+      "loss": 0.8028,
+      "step": 300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019362139917695475,
+      "loss": 0.7895,
+      "step": 305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019341563786008233,
+      "loss": 0.7777,
+      "step": 310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019320987654320988,
+      "loss": 0.7868,
+      "step": 315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019300411522633746,
+      "loss": 0.7921,
+      "step": 320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019279835390946504,
+      "loss": 0.8,
+      "step": 325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001925925925925926,
+      "loss": 0.7781,
+      "step": 330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019238683127572017,
+      "loss": 0.7799,
+      "step": 335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019218106995884775,
+      "loss": 0.7921,
+      "step": 340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019197530864197533,
+      "loss": 0.7851,
+      "step": 345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019176954732510288,
+      "loss": 0.8106,
+      "step": 350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019156378600823046,
+      "loss": 0.7953,
+      "step": 355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019135802469135804,
+      "loss": 0.8084,
+      "step": 360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001911522633744856,
+      "loss": 0.8298,
+      "step": 365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019094650205761317,
+      "loss": 0.8273,
+      "step": 370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019074074074074075,
+      "loss": 0.8047,
+      "step": 375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019053497942386833,
+      "loss": 0.8108,
+      "step": 380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019032921810699588,
+      "loss": 0.8207,
+      "step": 385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019012345679012346,
+      "loss": 0.7884,
+      "step": 390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018991769547325104,
+      "loss": 0.7918,
+      "step": 395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001897119341563786,
+      "loss": 0.7554,
+      "step": 400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018950617283950617,
+      "loss": 0.8099,
+      "step": 405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018930041152263375,
+      "loss": 0.807,
+      "step": 410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018909465020576133,
+      "loss": 0.7793,
+      "step": 415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018888888888888888,
+      "loss": 0.7734,
+      "step": 420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018868312757201646,
+      "loss": 0.7786,
+      "step": 425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018847736625514404,
+      "loss": 0.7548,
+      "step": 430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001882716049382716,
+      "loss": 0.8024,
+      "step": 435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018806584362139917,
+      "loss": 0.7708,
+      "step": 440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018786008230452675,
+      "loss": 0.8045,
+      "step": 445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018765432098765433,
+      "loss": 0.7477,
+      "step": 450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018744855967078189,
+      "loss": 0.7474,
+      "step": 455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018724279835390946,
+      "loss": 0.7629,
+      "step": 460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018703703703703704,
+      "loss": 0.7777,
+      "step": 465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018683127572016462,
+      "loss": 0.7823,
+      "step": 470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001866255144032922,
+      "loss": 0.7762,
+      "step": 475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018641975308641978,
+      "loss": 0.782,
+      "step": 480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018621399176954733,
+      "loss": 0.7783,
+      "step": 485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001860082304526749,
+      "loss": 0.7376,
+      "step": 490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001858024691358025,
+      "loss": 0.7683,
+      "step": 495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018559670781893007,
+      "loss": 0.7547,
+      "step": 500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018539094650205762,
+      "loss": 0.7569,
+      "step": 505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001851851851851852,
+      "loss": 0.7796,
+      "step": 510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018497942386831278,
+      "loss": 0.7888,
+      "step": 515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018477366255144033,
+      "loss": 0.7763,
+      "step": 520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018456790123456791,
+      "loss": 0.7787,
+      "step": 525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001843621399176955,
+      "loss": 0.772,
+      "step": 530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018415637860082307,
+      "loss": 0.786,
+      "step": 535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018395061728395062,
+      "loss": 0.7639,
+      "step": 540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001837448559670782,
+      "loss": 0.775,
+      "step": 545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018353909465020578,
+      "loss": 0.8002,
+      "step": 550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 0.7359,
+      "step": 555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018312757201646091,
+      "loss": 0.747,
+      "step": 560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001829218106995885,
+      "loss": 0.8234,
+      "step": 565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018271604938271605,
+      "loss": 0.7586,
+      "step": 570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018251028806584363,
+      "loss": 0.7315,
+      "step": 575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001823045267489712,
+      "loss": 0.7986,
+      "step": 580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018209876543209878,
+      "loss": 0.7758,
+      "step": 585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018189300411522634,
+      "loss": 0.7869,
+      "step": 590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018168724279835392,
+      "loss": 0.7596,
+      "step": 595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001814814814814815,
+      "loss": 0.7658,
+      "step": 600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018127572016460905,
+      "loss": 0.7851,
+      "step": 605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018106995884773663,
+      "loss": 0.7957,
+      "step": 610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001808641975308642,
+      "loss": 0.7756,
+      "step": 615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018065843621399178,
+      "loss": 0.7911,
+      "step": 620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018045267489711934,
+      "loss": 0.7864,
+      "step": 625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018024691358024692,
+      "loss": 0.7683,
+      "step": 630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001800411522633745,
+      "loss": 0.7731,
+      "step": 635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017983539094650205,
+      "loss": 0.7671,
+      "step": 640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017962962962962963,
+      "loss": 0.7367,
+      "step": 645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001794238683127572,
+      "loss": 0.8024,
+      "step": 650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017921810699588476,
+      "loss": 0.7572,
+      "step": 655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017901234567901234,
+      "loss": 0.7554,
+      "step": 660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017880658436213992,
+      "loss": 0.7677,
+      "step": 665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001786008230452675,
+      "loss": 0.7652,
+      "step": 670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017839506172839508,
+      "loss": 0.7771,
+      "step": 675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017818930041152265,
+      "loss": 0.7747,
+      "step": 680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001779835390946502,
+      "loss": 0.767,
+      "step": 685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017777777777777779,
+      "loss": 0.7738,
+      "step": 690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017757201646090537,
+      "loss": 0.7271,
+      "step": 695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017736625514403294,
+      "loss": 0.7316,
+      "step": 700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017716049382716052,
+      "loss": 0.7831,
+      "step": 705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017695473251028808,
+      "loss": 0.7835,
+      "step": 710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017674897119341566,
+      "loss": 0.7497,
+      "step": 715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017654320987654323,
+      "loss": 0.7757,
+      "step": 720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001763374485596708,
+      "loss": 0.74,
+      "step": 725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017613168724279837,
+      "loss": 0.7697,
+      "step": 730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017592592592592595,
+      "loss": 0.774,
+      "step": 735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017572016460905352,
+      "loss": 0.7668,
+      "step": 740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017551440329218108,
+      "loss": 0.7779,
+      "step": 745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017530864197530866,
+      "loss": 0.7441,
+      "step": 750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017510288065843624,
+      "loss": 0.7578,
+      "step": 755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001748971193415638,
+      "loss": 0.8358,
+      "step": 760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017469135802469137,
+      "loss": 0.7473,
+      "step": 765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017448559670781895,
+      "loss": 0.7517,
+      "step": 770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001742798353909465,
+      "loss": 0.7593,
+      "step": 775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017407407407407408,
+      "loss": 0.7251,
+      "step": 780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017386831275720166,
+      "loss": 0.7365,
+      "step": 785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017366255144032924,
+      "loss": 0.7798,
+      "step": 790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001734567901234568,
+      "loss": 0.746,
+      "step": 795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017325102880658437,
+      "loss": 0.7607,
+      "step": 800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017304526748971195,
+      "loss": 0.7694,
+      "step": 805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001728395061728395,
+      "loss": 0.7926,
+      "step": 810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017263374485596708,
+      "loss": 0.795,
+      "step": 815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017242798353909466,
+      "loss": 0.7581,
+      "step": 820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017222222222222224,
+      "loss": 0.7629,
+      "step": 825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001720164609053498,
+      "loss": 0.7526,
+      "step": 830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017181069958847737,
+      "loss": 0.7338,
+      "step": 835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017160493827160495,
+      "loss": 0.7544,
+      "step": 840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001713991769547325,
+      "loss": 0.7522,
+      "step": 845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017119341563786008,
+      "loss": 0.7639,
+      "step": 850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017098765432098766,
+      "loss": 0.7914,
+      "step": 855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017078189300411524,
+      "loss": 0.7635,
+      "step": 860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001705761316872428,
+      "loss": 0.8343,
+      "step": 865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017037037037037037,
+      "loss": 0.7299,
+      "step": 870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017016460905349795,
+      "loss": 0.7643,
+      "step": 875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016995884773662553,
+      "loss": 0.7343,
+      "step": 880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016975308641975308,
+      "loss": 0.7546,
+      "step": 885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016954732510288066,
+      "loss": 0.7435,
+      "step": 890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016934156378600824,
+      "loss": 0.8086,
+      "step": 895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016913580246913582,
+      "loss": 0.7274,
+      "step": 900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001689300411522634,
+      "loss": 0.7511,
+      "step": 905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016872427983539098,
+      "loss": 0.7689,
+      "step": 910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016851851851851853,
+      "loss": 0.754,
+      "step": 915
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001683127572016461,
+      "loss": 0.7298,
+      "step": 920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001681069958847737,
+      "loss": 0.7755,
+      "step": 925
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016790123456790124,
+      "loss": 0.7854,
+      "step": 930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016769547325102882,
+      "loss": 0.7452,
+      "step": 935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001674897119341564,
+      "loss": 0.7473,
+      "step": 940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016728395061728398,
+      "loss": 0.7855,
+      "step": 945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016707818930041153,
+      "loss": 0.7527,
+      "step": 950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001668724279835391,
+      "loss": 0.7641,
+      "step": 955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.828,
+      "step": 960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016646090534979424,
+      "loss": 0.7571,
+      "step": 965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016625514403292182,
+      "loss": 0.7768,
+      "step": 970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001660493827160494,
+      "loss": 0.7644,
+      "step": 975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016584362139917695,
+      "loss": 0.7747,
+      "step": 980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016563786008230453,
+      "loss": 0.7926,
+      "step": 985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001654320987654321,
+      "loss": 0.7645,
+      "step": 990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001652263374485597,
+      "loss": 0.7464,
+      "step": 995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016502057613168724,
+      "loss": 0.7558,
+      "step": 1000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016481481481481482,
+      "loss": 0.7574,
+      "step": 1005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001646090534979424,
+      "loss": 0.7491,
+      "step": 1010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016440329218106995,
+      "loss": 0.756,
+      "step": 1015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016419753086419753,
+      "loss": 0.7679,
+      "step": 1020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001639917695473251,
+      "loss": 0.776,
+      "step": 1025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001637860082304527,
+      "loss": 0.7235,
+      "step": 1030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016358024691358024,
+      "loss": 0.7817,
+      "step": 1035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016337448559670782,
+      "loss": 0.7333,
+      "step": 1040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001631687242798354,
+      "loss": 0.754,
+      "step": 1045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016296296296296295,
+      "loss": 0.74,
+      "step": 1050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016275720164609053,
+      "loss": 0.7857,
+      "step": 1055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001625514403292181,
+      "loss": 0.8015,
+      "step": 1060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001623456790123457,
+      "loss": 0.7508,
+      "step": 1065
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016213991769547324,
+      "loss": 0.7514,
+      "step": 1070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016193415637860082,
+      "loss": 0.7326,
+      "step": 1075
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001617283950617284,
+      "loss": 0.7322,
+      "step": 1080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016152263374485595,
+      "loss": 0.7373,
+      "step": 1085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016131687242798353,
+      "loss": 0.7637,
+      "step": 1090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001611111111111111,
+      "loss": 0.7697,
+      "step": 1095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001609053497942387,
+      "loss": 0.7505,
+      "step": 1100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016069958847736627,
+      "loss": 0.7568,
+      "step": 1105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016049382716049385,
+      "loss": 0.786,
+      "step": 1110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001602880658436214,
+      "loss": 0.7525,
+      "step": 1115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016008230452674898,
+      "loss": 0.7696,
+      "step": 1120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015987654320987656,
+      "loss": 0.7648,
+      "step": 1125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015967078189300414,
+      "loss": 0.7492,
+      "step": 1130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001594650205761317,
+      "loss": 0.809,
+      "step": 1135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015925925925925927,
+      "loss": 0.7592,
+      "step": 1140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015905349794238685,
+      "loss": 0.7874,
+      "step": 1145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015884773662551443,
+      "loss": 0.7874,
+      "step": 1150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015864197530864198,
+      "loss": 0.7374,
+      "step": 1155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015843621399176956,
+      "loss": 0.7218,
+      "step": 1160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015823045267489714,
+      "loss": 0.7494,
+      "step": 1165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001580246913580247,
+      "loss": 0.7449,
+      "step": 1170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015781893004115227,
+      "loss": 0.725,
+      "step": 1175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015761316872427985,
+      "loss": 0.739,
+      "step": 1180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015740740740740743,
+      "loss": 0.7554,
+      "step": 1185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015720164609053498,
+      "loss": 0.7702,
+      "step": 1190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015699588477366256,
+      "loss": 0.7418,
+      "step": 1195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015679012345679014,
+      "loss": 0.7217,
+      "step": 1200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001565843621399177,
+      "loss": 0.7391,
+      "step": 1205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015637860082304527,
+      "loss": 0.7696,
+      "step": 1210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015617283950617285,
+      "loss": 0.7594,
+      "step": 1215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001559670781893004,
+      "loss": 0.777,
+      "step": 1220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015576131687242798,
+      "loss": 0.7441,
+      "step": 1225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015555555555555556,
+      "loss": 0.7611,
+      "step": 1230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015534979423868314,
+      "loss": 0.7278,
+      "step": 1235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001551440329218107,
+      "loss": 0.7452,
+      "step": 1240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015493827160493827,
+      "loss": 0.7939,
+      "step": 1245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015473251028806585,
+      "loss": 0.7161,
+      "step": 1250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001545267489711934,
+      "loss": 0.7344,
+      "step": 1255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015432098765432098,
+      "loss": 0.7368,
+      "step": 1260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015411522633744856,
+      "loss": 0.7172,
+      "step": 1265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015390946502057614,
+      "loss": 0.7589,
+      "step": 1270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001537037037037037,
+      "loss": 0.7641,
+      "step": 1275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015349794238683127,
+      "loss": 0.761,
+      "step": 1280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015329218106995885,
+      "loss": 0.7414,
+      "step": 1285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001530864197530864,
+      "loss": 0.7713,
+      "step": 1290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015288065843621399,
+      "loss": 0.7543,
+      "step": 1295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015267489711934156,
+      "loss": 0.7683,
+      "step": 1300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015246913580246914,
+      "loss": 0.8106,
+      "step": 1305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015226337448559672,
+      "loss": 0.7739,
+      "step": 1310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015205761316872428,
+      "loss": 0.7585,
+      "step": 1315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015185185185185185,
+      "loss": 0.7577,
+      "step": 1320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015164609053497943,
+      "loss": 0.7608,
+      "step": 1325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000151440329218107,
+      "loss": 0.7715,
+      "step": 1330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001512345679012346,
+      "loss": 0.7681,
+      "step": 1335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015102880658436214,
+      "loss": 0.7476,
+      "step": 1340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015082304526748972,
+      "loss": 0.7539,
+      "step": 1345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001506172839506173,
+      "loss": 0.7492,
+      "step": 1350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015041152263374488,
+      "loss": 0.7491,
+      "step": 1355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015020576131687243,
+      "loss": 0.7236,
+      "step": 1360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 0.7328,
+      "step": 1365
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001497942386831276,
+      "loss": 0.7603,
+      "step": 1370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014958847736625515,
+      "loss": 0.7206,
+      "step": 1375
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014938271604938272,
+      "loss": 0.7001,
+      "step": 1380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001491769547325103,
+      "loss": 0.7241,
+      "step": 1385
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014897119341563788,
+      "loss": 0.7511,
+      "step": 1390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014876543209876544,
+      "loss": 0.7546,
+      "step": 1395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014855967078189301,
+      "loss": 0.762,
+      "step": 1400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001483539094650206,
+      "loss": 0.6979,
+      "step": 1405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014814814814814815,
+      "loss": 0.7352,
+      "step": 1410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014794238683127573,
+      "loss": 0.7604,
+      "step": 1415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001477366255144033,
+      "loss": 0.7612,
+      "step": 1420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014753086419753086,
+      "loss": 0.7618,
+      "step": 1425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014732510288065844,
+      "loss": 0.7107,
+      "step": 1430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014711934156378602,
+      "loss": 0.7247,
+      "step": 1435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001469135802469136,
+      "loss": 0.7733,
+      "step": 1440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014670781893004115,
+      "loss": 0.7627,
+      "step": 1445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014650205761316873,
+      "loss": 0.7323,
+      "step": 1450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001462962962962963,
+      "loss": 0.7728,
+      "step": 1455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014609053497942386,
+      "loss": 0.7541,
+      "step": 1460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014588477366255144,
+      "loss": 0.7781,
+      "step": 1465
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014567901234567902,
+      "loss": 0.7474,
+      "step": 1470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001454732510288066,
+      "loss": 0.7789,
+      "step": 1475
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014526748971193415,
+      "loss": 0.7497,
+      "step": 1480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014506172839506173,
+      "loss": 0.7487,
+      "step": 1485
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001448559670781893,
+      "loss": 0.7563,
+      "step": 1490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014465020576131686,
+      "loss": 0.7444,
+      "step": 1495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014444444444444444,
+      "loss": 0.7399,
+      "step": 1500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014423868312757202,
+      "loss": 0.7288,
+      "step": 1505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001440329218106996,
+      "loss": 0.7458,
+      "step": 1510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014382716049382718,
+      "loss": 0.7297,
+      "step": 1515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014362139917695473,
+      "loss": 0.7684,
+      "step": 1520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001434156378600823,
+      "loss": 0.7428,
+      "step": 1525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014320987654320989,
+      "loss": 0.705,
+      "step": 1530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014300411522633747,
+      "loss": 0.7445,
+      "step": 1535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014279835390946504,
+      "loss": 0.7109,
+      "step": 1540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001425925925925926,
+      "loss": 0.7502,
+      "step": 1545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014238683127572018,
+      "loss": 0.7485,
+      "step": 1550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014218106995884776,
+      "loss": 0.7421,
+      "step": 1555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014197530864197534,
+      "loss": 0.7369,
+      "step": 1560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001417695473251029,
+      "loss": 0.7746,
+      "step": 1565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014156378600823047,
+      "loss": 0.7334,
+      "step": 1570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014135802469135805,
+      "loss": 0.7251,
+      "step": 1575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001411522633744856,
+      "loss": 0.7146,
+      "step": 1580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014094650205761318,
+      "loss": 0.7563,
+      "step": 1585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014074074074074076,
+      "loss": 0.7247,
+      "step": 1590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014053497942386834,
+      "loss": 0.7379,
+      "step": 1595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001403292181069959,
+      "loss": 0.7684,
+      "step": 1600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014012345679012347,
+      "loss": 0.7574,
+      "step": 1605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00013991769547325105,
+      "loss": 0.7138,
+      "step": 1610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001397119341563786,
+      "loss": 0.7472,
+      "step": 1615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013950617283950618,
+      "loss": 0.7466,
+      "step": 1620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013930041152263376,
+      "loss": 0.7288,
+      "step": 1625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013909465020576134,
+      "loss": 0.7717,
+      "step": 1630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001388888888888889,
+      "loss": 0.7665,
+      "step": 1635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013868312757201647,
+      "loss": 0.7408,
+      "step": 1640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013847736625514405,
+      "loss": 0.7418,
+      "step": 1645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001382716049382716,
+      "loss": 0.7502,
+      "step": 1650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013806584362139918,
+      "loss": 0.7657,
+      "step": 1655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013786008230452676,
+      "loss": 0.728,
+      "step": 1660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001376543209876543,
+      "loss": 0.7142,
+      "step": 1665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001374485596707819,
+      "loss": 0.7971,
+      "step": 1670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013724279835390947,
+      "loss": 0.737,
+      "step": 1675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013703703703703705,
+      "loss": 0.7947,
+      "step": 1680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001368312757201646,
+      "loss": 0.7525,
+      "step": 1685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013662551440329218,
+      "loss": 0.711,
+      "step": 1690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013641975308641976,
+      "loss": 0.7459,
+      "step": 1695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001362139917695473,
+      "loss": 0.7404,
+      "step": 1700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001360082304526749,
+      "loss": 0.7458,
+      "step": 1705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013580246913580247,
+      "loss": 0.754,
+      "step": 1710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013559670781893005,
+      "loss": 0.7217,
+      "step": 1715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001353909465020576,
+      "loss": 0.7373,
+      "step": 1720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013518518518518518,
+      "loss": 0.7567,
+      "step": 1725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013497942386831276,
+      "loss": 0.7525,
+      "step": 1730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013477366255144034,
+      "loss": 0.752,
+      "step": 1735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013456790123456792,
+      "loss": 0.7342,
+      "step": 1740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001343621399176955,
+      "loss": 0.7276,
+      "step": 1745
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013415637860082305,
+      "loss": 0.6915,
+      "step": 1750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013395061728395063,
+      "loss": 0.737,
+      "step": 1755
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001337448559670782,
+      "loss": 0.7427,
+      "step": 1760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001335390946502058,
+      "loss": 0.7126,
+      "step": 1765
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.7241,
+      "step": 1770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013312757201646092,
+      "loss": 0.755,
+      "step": 1775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001329218106995885,
+      "loss": 0.7494,
+      "step": 1780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013271604938271605,
+      "loss": 0.709,
+      "step": 1785
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013251028806584363,
+      "loss": 0.7555,
+      "step": 1790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001323045267489712,
+      "loss": 0.7298,
+      "step": 1795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001320987654320988,
+      "loss": 0.7232,
+      "step": 1800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013189300411522634,
+      "loss": 0.7441,
+      "step": 1805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013168724279835392,
+      "loss": 0.7623,
+      "step": 1810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001314814814814815,
+      "loss": 0.7349,
+      "step": 1815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013127572016460905,
+      "loss": 0.7773,
+      "step": 1820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013106995884773663,
+      "loss": 0.7537,
+      "step": 1825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001308641975308642,
+      "loss": 0.7579,
+      "step": 1830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001306584362139918,
+      "loss": 0.7321,
+      "step": 1835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013045267489711934,
+      "loss": 0.7175,
+      "step": 1840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00013024691358024692,
+      "loss": 0.7553,
+      "step": 1845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001300411522633745,
+      "loss": 0.7591,
+      "step": 1850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012983539094650205,
+      "loss": 0.7581,
+      "step": 1855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012962962962962963,
+      "loss": 0.7632,
+      "step": 1860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001294238683127572,
+      "loss": 0.7434,
+      "step": 1865
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012921810699588476,
+      "loss": 0.7695,
+      "step": 1870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012901234567901234,
+      "loss": 0.7808,
+      "step": 1875
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012880658436213992,
+      "loss": 0.7428,
+      "step": 1880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001286008230452675,
+      "loss": 0.7487,
+      "step": 1885
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012839506172839505,
+      "loss": 0.7033,
+      "step": 1890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012818930041152263,
+      "loss": 0.7639,
+      "step": 1895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001279835390946502,
+      "loss": 0.7345,
+      "step": 1900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012777777777777776,
+      "loss": 0.7434,
+      "step": 1905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012757201646090534,
+      "loss": 0.7082,
+      "step": 1910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012736625514403292,
+      "loss": 0.7294,
+      "step": 1915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001271604938271605,
+      "loss": 0.74,
+      "step": 1920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012695473251028805,
+      "loss": 0.7694,
+      "step": 1925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012674897119341563,
+      "loss": 0.7339,
+      "step": 1930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001265432098765432,
+      "loss": 0.7259,
+      "step": 1935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001263374485596708,
+      "loss": 0.7307,
+      "step": 1940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012613168724279837,
+      "loss": 0.7085,
+      "step": 1945
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012592592592592592,
+      "loss": 0.7662,
+      "step": 1950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001257201646090535,
+      "loss": 0.7145,
+      "step": 1955
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012551440329218108,
+      "loss": 0.7528,
+      "step": 1960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012530864197530866,
+      "loss": 0.7308,
+      "step": 1965
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012510288065843624,
+      "loss": 0.7163,
+      "step": 1970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001248971193415638,
+      "loss": 0.7254,
+      "step": 1975
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012469135802469137,
+      "loss": 0.7467,
+      "step": 1980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012448559670781895,
+      "loss": 0.7391,
+      "step": 1985
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001242798353909465,
+      "loss": 0.7314,
+      "step": 1990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012407407407407408,
+      "loss": 0.7048,
+      "step": 1995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012386831275720166,
+      "loss": 0.7187,
+      "step": 2000
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 5010,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 500,
+  "total_flos": 8.056468541093708e+18,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/training_args.bin b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e707059a03fbe62142ac41332e145ac1ac1f36cc
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b99a1729574e0e46ee782b42dc90101a785d5b4792871920d293f44722c28cc
+size 5944
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/zero_to_fp32.py b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/zero_to_fp32.py
new file mode 100644
index 0000000000000000000000000000000000000000..c98caae31534368be22b67fc4ae906836c992a8d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2000/zero_to_fp32.py
@@ -0,0 +1,587 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+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):
+    """
+    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)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+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 _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()
+            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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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 param_shapes.items():
+
+        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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    """
+    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``
+
+    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)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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("-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_file, tag=args.tag)
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/README.md b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..d380bbbf0bf0fa1f211b53410559607e5007a209
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: meta-llama/Llama-2-7b-hf
+---
+
+# 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 0.9.0
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/adapter_config.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c81a592f0ff08374a4f61f6003029070a8429ae
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/adapter_config.json
@@ -0,0 +1,35 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "mora_type": 1,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_mora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/adapter_model.safetensors b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/adapter_model.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..bd8a663922e9e392f755995c88e506cacc1ec665
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/adapter_model.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1286cce8a1ff02fa26b8fc6175aa59320d52dc871b3c727b773e08d246798839
+size 564342240
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/global_step2500/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/global_step2500/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e2fe9a5de6dc5f01843fc4219215c38fe0cffd75
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/global_step2500/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bc5a584db3f84f71a5531be6e84ee9012f906801e236164bf1930863c96b349
+size 239914224
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/global_step2500/mp_rank_00_model_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/global_step2500/mp_rank_00_model_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..db6c59da154981a2521b71a63cb1723a6f5c12d9
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/global_step2500/mp_rank_00_model_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6714b8877a46c01dfe0c658fd930d5ff477446b11507a1bf0bf987d13c62d61
+size 40303404
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/latest b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/latest
new file mode 100644
index 0000000000000000000000000000000000000000..98f8bed9a5485ee900d9931cc06950de69499848
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/latest
@@ -0,0 +1 @@
+global_step2500
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/rng_state.pth b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/rng_state.pth
new file mode 100644
index 0000000000000000000000000000000000000000..8b083fe9709f50483130314aef7895e0a0989009
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/rng_state.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c06751c7f8c719f6b24d07fe3e53f4df2e8851764965d05ce5c2050d512804f2
+size 14244
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/scheduler.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/scheduler.pt
new file mode 100644
index 0000000000000000000000000000000000000000..5fc605059c2c25108953bfd80fb285e7ad18418c
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/scheduler.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c72b22311c7cbc36b4527a16e3812d8e5f456c956cbf82d4e8d13847a239836
+size 1064
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/trainer_state.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8358794a803a0fcbcf2aebd98f10e13052c5d9d7
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/trainer_state.json
@@ -0,0 +1,3021 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9979106246296816,
+  "eval_steps": 500,
+  "global_step": 2500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.3797,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.3952,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2e-05,
+      "loss": 1.3112,
+      "step": 15
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 1.3136,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 1.2012,
+      "step": 25
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4e-05,
+      "loss": 1.1277,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 1.0832,
+      "step": 35
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.333333333333333e-05,
+      "loss": 1.0659,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-05,
+      "loss": 0.983,
+      "step": 45
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.9763,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.333333333333333e-05,
+      "loss": 0.9992,
+      "step": 55
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8e-05,
+      "loss": 0.9405,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.666666666666667e-05,
+      "loss": 0.9827,
+      "step": 65
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.333333333333334e-05,
+      "loss": 0.9061,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001,
+      "loss": 0.8921,
+      "step": 75
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010666666666666667,
+      "loss": 0.9377,
+      "step": 80
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011333333333333334,
+      "loss": 0.8528,
+      "step": 85
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012,
+      "loss": 0.8913,
+      "step": 90
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012666666666666666,
+      "loss": 0.8562,
+      "step": 95
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.877,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014,
+      "loss": 0.8781,
+      "step": 105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014666666666666666,
+      "loss": 0.8546,
+      "step": 110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015333333333333334,
+      "loss": 0.8672,
+      "step": 115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016,
+      "loss": 0.8611,
+      "step": 120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.8624,
+      "step": 125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017333333333333334,
+      "loss": 0.8477,
+      "step": 130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018,
+      "loss": 0.8227,
+      "step": 135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001866666666666667,
+      "loss": 0.8064,
+      "step": 140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019333333333333333,
+      "loss": 0.8422,
+      "step": 145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002,
+      "loss": 0.8471,
+      "step": 150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001997942386831276,
+      "loss": 0.8183,
+      "step": 155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958847736625514,
+      "loss": 0.8229,
+      "step": 160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019938271604938272,
+      "loss": 0.8114,
+      "step": 165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991769547325103,
+      "loss": 0.7861,
+      "step": 170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019897119341563788,
+      "loss": 0.841,
+      "step": 175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019876543209876543,
+      "loss": 0.8549,
+      "step": 180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000198559670781893,
+      "loss": 0.8449,
+      "step": 185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001983539094650206,
+      "loss": 0.8311,
+      "step": 190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019814814814814814,
+      "loss": 0.8296,
+      "step": 195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019794238683127572,
+      "loss": 0.8263,
+      "step": 200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001977366255144033,
+      "loss": 0.8452,
+      "step": 205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019753086419753085,
+      "loss": 0.8044,
+      "step": 210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019732510288065843,
+      "loss": 0.8119,
+      "step": 215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000197119341563786,
+      "loss": 0.8127,
+      "step": 220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001969135802469136,
+      "loss": 0.7921,
+      "step": 225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019670781893004114,
+      "loss": 0.803,
+      "step": 230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019650205761316872,
+      "loss": 0.838,
+      "step": 235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001962962962962963,
+      "loss": 0.7801,
+      "step": 240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019609053497942388,
+      "loss": 0.817,
+      "step": 245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019588477366255146,
+      "loss": 0.8211,
+      "step": 250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000195679012345679,
+      "loss": 0.8423,
+      "step": 255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001954732510288066,
+      "loss": 0.8007,
+      "step": 260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019526748971193417,
+      "loss": 0.8135,
+      "step": 265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019506172839506175,
+      "loss": 0.7961,
+      "step": 270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019485596707818933,
+      "loss": 0.797,
+      "step": 275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019465020576131688,
+      "loss": 0.793,
+      "step": 280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019444444444444446,
+      "loss": 0.7809,
+      "step": 285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019423868312757204,
+      "loss": 0.7811,
+      "step": 290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019403292181069962,
+      "loss": 0.8074,
+      "step": 295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019382716049382717,
+      "loss": 0.8028,
+      "step": 300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019362139917695475,
+      "loss": 0.7895,
+      "step": 305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019341563786008233,
+      "loss": 0.7777,
+      "step": 310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019320987654320988,
+      "loss": 0.7868,
+      "step": 315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019300411522633746,
+      "loss": 0.7921,
+      "step": 320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019279835390946504,
+      "loss": 0.8,
+      "step": 325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001925925925925926,
+      "loss": 0.7781,
+      "step": 330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019238683127572017,
+      "loss": 0.7799,
+      "step": 335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019218106995884775,
+      "loss": 0.7921,
+      "step": 340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019197530864197533,
+      "loss": 0.7851,
+      "step": 345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019176954732510288,
+      "loss": 0.8106,
+      "step": 350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019156378600823046,
+      "loss": 0.7953,
+      "step": 355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019135802469135804,
+      "loss": 0.8084,
+      "step": 360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001911522633744856,
+      "loss": 0.8298,
+      "step": 365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019094650205761317,
+      "loss": 0.8273,
+      "step": 370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019074074074074075,
+      "loss": 0.8047,
+      "step": 375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019053497942386833,
+      "loss": 0.8108,
+      "step": 380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019032921810699588,
+      "loss": 0.8207,
+      "step": 385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019012345679012346,
+      "loss": 0.7884,
+      "step": 390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018991769547325104,
+      "loss": 0.7918,
+      "step": 395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001897119341563786,
+      "loss": 0.7554,
+      "step": 400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018950617283950617,
+      "loss": 0.8099,
+      "step": 405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018930041152263375,
+      "loss": 0.807,
+      "step": 410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018909465020576133,
+      "loss": 0.7793,
+      "step": 415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018888888888888888,
+      "loss": 0.7734,
+      "step": 420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018868312757201646,
+      "loss": 0.7786,
+      "step": 425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018847736625514404,
+      "loss": 0.7548,
+      "step": 430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001882716049382716,
+      "loss": 0.8024,
+      "step": 435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018806584362139917,
+      "loss": 0.7708,
+      "step": 440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018786008230452675,
+      "loss": 0.8045,
+      "step": 445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018765432098765433,
+      "loss": 0.7477,
+      "step": 450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018744855967078189,
+      "loss": 0.7474,
+      "step": 455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018724279835390946,
+      "loss": 0.7629,
+      "step": 460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018703703703703704,
+      "loss": 0.7777,
+      "step": 465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018683127572016462,
+      "loss": 0.7823,
+      "step": 470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001866255144032922,
+      "loss": 0.7762,
+      "step": 475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018641975308641978,
+      "loss": 0.782,
+      "step": 480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018621399176954733,
+      "loss": 0.7783,
+      "step": 485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001860082304526749,
+      "loss": 0.7376,
+      "step": 490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001858024691358025,
+      "loss": 0.7683,
+      "step": 495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018559670781893007,
+      "loss": 0.7547,
+      "step": 500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018539094650205762,
+      "loss": 0.7569,
+      "step": 505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001851851851851852,
+      "loss": 0.7796,
+      "step": 510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018497942386831278,
+      "loss": 0.7888,
+      "step": 515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018477366255144033,
+      "loss": 0.7763,
+      "step": 520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018456790123456791,
+      "loss": 0.7787,
+      "step": 525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001843621399176955,
+      "loss": 0.772,
+      "step": 530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018415637860082307,
+      "loss": 0.786,
+      "step": 535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018395061728395062,
+      "loss": 0.7639,
+      "step": 540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001837448559670782,
+      "loss": 0.775,
+      "step": 545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018353909465020578,
+      "loss": 0.8002,
+      "step": 550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 0.7359,
+      "step": 555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018312757201646091,
+      "loss": 0.747,
+      "step": 560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001829218106995885,
+      "loss": 0.8234,
+      "step": 565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018271604938271605,
+      "loss": 0.7586,
+      "step": 570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018251028806584363,
+      "loss": 0.7315,
+      "step": 575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001823045267489712,
+      "loss": 0.7986,
+      "step": 580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018209876543209878,
+      "loss": 0.7758,
+      "step": 585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018189300411522634,
+      "loss": 0.7869,
+      "step": 590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018168724279835392,
+      "loss": 0.7596,
+      "step": 595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001814814814814815,
+      "loss": 0.7658,
+      "step": 600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018127572016460905,
+      "loss": 0.7851,
+      "step": 605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018106995884773663,
+      "loss": 0.7957,
+      "step": 610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001808641975308642,
+      "loss": 0.7756,
+      "step": 615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018065843621399178,
+      "loss": 0.7911,
+      "step": 620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018045267489711934,
+      "loss": 0.7864,
+      "step": 625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018024691358024692,
+      "loss": 0.7683,
+      "step": 630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001800411522633745,
+      "loss": 0.7731,
+      "step": 635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017983539094650205,
+      "loss": 0.7671,
+      "step": 640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017962962962962963,
+      "loss": 0.7367,
+      "step": 645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001794238683127572,
+      "loss": 0.8024,
+      "step": 650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017921810699588476,
+      "loss": 0.7572,
+      "step": 655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017901234567901234,
+      "loss": 0.7554,
+      "step": 660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017880658436213992,
+      "loss": 0.7677,
+      "step": 665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001786008230452675,
+      "loss": 0.7652,
+      "step": 670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017839506172839508,
+      "loss": 0.7771,
+      "step": 675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017818930041152265,
+      "loss": 0.7747,
+      "step": 680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001779835390946502,
+      "loss": 0.767,
+      "step": 685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017777777777777779,
+      "loss": 0.7738,
+      "step": 690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017757201646090537,
+      "loss": 0.7271,
+      "step": 695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017736625514403294,
+      "loss": 0.7316,
+      "step": 700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017716049382716052,
+      "loss": 0.7831,
+      "step": 705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017695473251028808,
+      "loss": 0.7835,
+      "step": 710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017674897119341566,
+      "loss": 0.7497,
+      "step": 715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017654320987654323,
+      "loss": 0.7757,
+      "step": 720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001763374485596708,
+      "loss": 0.74,
+      "step": 725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017613168724279837,
+      "loss": 0.7697,
+      "step": 730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017592592592592595,
+      "loss": 0.774,
+      "step": 735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017572016460905352,
+      "loss": 0.7668,
+      "step": 740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017551440329218108,
+      "loss": 0.7779,
+      "step": 745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017530864197530866,
+      "loss": 0.7441,
+      "step": 750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017510288065843624,
+      "loss": 0.7578,
+      "step": 755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001748971193415638,
+      "loss": 0.8358,
+      "step": 760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017469135802469137,
+      "loss": 0.7473,
+      "step": 765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017448559670781895,
+      "loss": 0.7517,
+      "step": 770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001742798353909465,
+      "loss": 0.7593,
+      "step": 775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017407407407407408,
+      "loss": 0.7251,
+      "step": 780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017386831275720166,
+      "loss": 0.7365,
+      "step": 785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017366255144032924,
+      "loss": 0.7798,
+      "step": 790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001734567901234568,
+      "loss": 0.746,
+      "step": 795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017325102880658437,
+      "loss": 0.7607,
+      "step": 800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017304526748971195,
+      "loss": 0.7694,
+      "step": 805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001728395061728395,
+      "loss": 0.7926,
+      "step": 810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017263374485596708,
+      "loss": 0.795,
+      "step": 815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017242798353909466,
+      "loss": 0.7581,
+      "step": 820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017222222222222224,
+      "loss": 0.7629,
+      "step": 825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001720164609053498,
+      "loss": 0.7526,
+      "step": 830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017181069958847737,
+      "loss": 0.7338,
+      "step": 835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017160493827160495,
+      "loss": 0.7544,
+      "step": 840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001713991769547325,
+      "loss": 0.7522,
+      "step": 845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017119341563786008,
+      "loss": 0.7639,
+      "step": 850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017098765432098766,
+      "loss": 0.7914,
+      "step": 855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017078189300411524,
+      "loss": 0.7635,
+      "step": 860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001705761316872428,
+      "loss": 0.8343,
+      "step": 865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017037037037037037,
+      "loss": 0.7299,
+      "step": 870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017016460905349795,
+      "loss": 0.7643,
+      "step": 875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016995884773662553,
+      "loss": 0.7343,
+      "step": 880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016975308641975308,
+      "loss": 0.7546,
+      "step": 885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016954732510288066,
+      "loss": 0.7435,
+      "step": 890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016934156378600824,
+      "loss": 0.8086,
+      "step": 895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016913580246913582,
+      "loss": 0.7274,
+      "step": 900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001689300411522634,
+      "loss": 0.7511,
+      "step": 905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016872427983539098,
+      "loss": 0.7689,
+      "step": 910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016851851851851853,
+      "loss": 0.754,
+      "step": 915
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001683127572016461,
+      "loss": 0.7298,
+      "step": 920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001681069958847737,
+      "loss": 0.7755,
+      "step": 925
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016790123456790124,
+      "loss": 0.7854,
+      "step": 930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016769547325102882,
+      "loss": 0.7452,
+      "step": 935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001674897119341564,
+      "loss": 0.7473,
+      "step": 940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016728395061728398,
+      "loss": 0.7855,
+      "step": 945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016707818930041153,
+      "loss": 0.7527,
+      "step": 950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001668724279835391,
+      "loss": 0.7641,
+      "step": 955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.828,
+      "step": 960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016646090534979424,
+      "loss": 0.7571,
+      "step": 965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016625514403292182,
+      "loss": 0.7768,
+      "step": 970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001660493827160494,
+      "loss": 0.7644,
+      "step": 975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016584362139917695,
+      "loss": 0.7747,
+      "step": 980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016563786008230453,
+      "loss": 0.7926,
+      "step": 985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001654320987654321,
+      "loss": 0.7645,
+      "step": 990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001652263374485597,
+      "loss": 0.7464,
+      "step": 995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016502057613168724,
+      "loss": 0.7558,
+      "step": 1000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016481481481481482,
+      "loss": 0.7574,
+      "step": 1005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001646090534979424,
+      "loss": 0.7491,
+      "step": 1010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016440329218106995,
+      "loss": 0.756,
+      "step": 1015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016419753086419753,
+      "loss": 0.7679,
+      "step": 1020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001639917695473251,
+      "loss": 0.776,
+      "step": 1025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001637860082304527,
+      "loss": 0.7235,
+      "step": 1030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016358024691358024,
+      "loss": 0.7817,
+      "step": 1035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016337448559670782,
+      "loss": 0.7333,
+      "step": 1040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001631687242798354,
+      "loss": 0.754,
+      "step": 1045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016296296296296295,
+      "loss": 0.74,
+      "step": 1050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016275720164609053,
+      "loss": 0.7857,
+      "step": 1055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001625514403292181,
+      "loss": 0.8015,
+      "step": 1060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001623456790123457,
+      "loss": 0.7508,
+      "step": 1065
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016213991769547324,
+      "loss": 0.7514,
+      "step": 1070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016193415637860082,
+      "loss": 0.7326,
+      "step": 1075
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001617283950617284,
+      "loss": 0.7322,
+      "step": 1080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016152263374485595,
+      "loss": 0.7373,
+      "step": 1085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016131687242798353,
+      "loss": 0.7637,
+      "step": 1090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001611111111111111,
+      "loss": 0.7697,
+      "step": 1095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001609053497942387,
+      "loss": 0.7505,
+      "step": 1100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016069958847736627,
+      "loss": 0.7568,
+      "step": 1105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016049382716049385,
+      "loss": 0.786,
+      "step": 1110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001602880658436214,
+      "loss": 0.7525,
+      "step": 1115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016008230452674898,
+      "loss": 0.7696,
+      "step": 1120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015987654320987656,
+      "loss": 0.7648,
+      "step": 1125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015967078189300414,
+      "loss": 0.7492,
+      "step": 1130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001594650205761317,
+      "loss": 0.809,
+      "step": 1135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015925925925925927,
+      "loss": 0.7592,
+      "step": 1140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015905349794238685,
+      "loss": 0.7874,
+      "step": 1145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015884773662551443,
+      "loss": 0.7874,
+      "step": 1150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015864197530864198,
+      "loss": 0.7374,
+      "step": 1155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015843621399176956,
+      "loss": 0.7218,
+      "step": 1160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015823045267489714,
+      "loss": 0.7494,
+      "step": 1165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001580246913580247,
+      "loss": 0.7449,
+      "step": 1170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015781893004115227,
+      "loss": 0.725,
+      "step": 1175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015761316872427985,
+      "loss": 0.739,
+      "step": 1180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015740740740740743,
+      "loss": 0.7554,
+      "step": 1185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015720164609053498,
+      "loss": 0.7702,
+      "step": 1190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015699588477366256,
+      "loss": 0.7418,
+      "step": 1195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015679012345679014,
+      "loss": 0.7217,
+      "step": 1200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001565843621399177,
+      "loss": 0.7391,
+      "step": 1205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015637860082304527,
+      "loss": 0.7696,
+      "step": 1210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015617283950617285,
+      "loss": 0.7594,
+      "step": 1215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001559670781893004,
+      "loss": 0.777,
+      "step": 1220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015576131687242798,
+      "loss": 0.7441,
+      "step": 1225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015555555555555556,
+      "loss": 0.7611,
+      "step": 1230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015534979423868314,
+      "loss": 0.7278,
+      "step": 1235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001551440329218107,
+      "loss": 0.7452,
+      "step": 1240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015493827160493827,
+      "loss": 0.7939,
+      "step": 1245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015473251028806585,
+      "loss": 0.7161,
+      "step": 1250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001545267489711934,
+      "loss": 0.7344,
+      "step": 1255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015432098765432098,
+      "loss": 0.7368,
+      "step": 1260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015411522633744856,
+      "loss": 0.7172,
+      "step": 1265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015390946502057614,
+      "loss": 0.7589,
+      "step": 1270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001537037037037037,
+      "loss": 0.7641,
+      "step": 1275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015349794238683127,
+      "loss": 0.761,
+      "step": 1280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015329218106995885,
+      "loss": 0.7414,
+      "step": 1285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001530864197530864,
+      "loss": 0.7713,
+      "step": 1290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015288065843621399,
+      "loss": 0.7543,
+      "step": 1295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015267489711934156,
+      "loss": 0.7683,
+      "step": 1300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015246913580246914,
+      "loss": 0.8106,
+      "step": 1305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015226337448559672,
+      "loss": 0.7739,
+      "step": 1310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015205761316872428,
+      "loss": 0.7585,
+      "step": 1315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015185185185185185,
+      "loss": 0.7577,
+      "step": 1320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015164609053497943,
+      "loss": 0.7608,
+      "step": 1325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000151440329218107,
+      "loss": 0.7715,
+      "step": 1330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001512345679012346,
+      "loss": 0.7681,
+      "step": 1335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015102880658436214,
+      "loss": 0.7476,
+      "step": 1340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015082304526748972,
+      "loss": 0.7539,
+      "step": 1345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001506172839506173,
+      "loss": 0.7492,
+      "step": 1350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015041152263374488,
+      "loss": 0.7491,
+      "step": 1355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015020576131687243,
+      "loss": 0.7236,
+      "step": 1360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 0.7328,
+      "step": 1365
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001497942386831276,
+      "loss": 0.7603,
+      "step": 1370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014958847736625515,
+      "loss": 0.7206,
+      "step": 1375
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014938271604938272,
+      "loss": 0.7001,
+      "step": 1380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001491769547325103,
+      "loss": 0.7241,
+      "step": 1385
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014897119341563788,
+      "loss": 0.7511,
+      "step": 1390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014876543209876544,
+      "loss": 0.7546,
+      "step": 1395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014855967078189301,
+      "loss": 0.762,
+      "step": 1400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001483539094650206,
+      "loss": 0.6979,
+      "step": 1405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014814814814814815,
+      "loss": 0.7352,
+      "step": 1410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014794238683127573,
+      "loss": 0.7604,
+      "step": 1415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001477366255144033,
+      "loss": 0.7612,
+      "step": 1420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014753086419753086,
+      "loss": 0.7618,
+      "step": 1425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014732510288065844,
+      "loss": 0.7107,
+      "step": 1430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014711934156378602,
+      "loss": 0.7247,
+      "step": 1435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001469135802469136,
+      "loss": 0.7733,
+      "step": 1440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014670781893004115,
+      "loss": 0.7627,
+      "step": 1445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014650205761316873,
+      "loss": 0.7323,
+      "step": 1450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001462962962962963,
+      "loss": 0.7728,
+      "step": 1455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014609053497942386,
+      "loss": 0.7541,
+      "step": 1460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014588477366255144,
+      "loss": 0.7781,
+      "step": 1465
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014567901234567902,
+      "loss": 0.7474,
+      "step": 1470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001454732510288066,
+      "loss": 0.7789,
+      "step": 1475
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014526748971193415,
+      "loss": 0.7497,
+      "step": 1480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014506172839506173,
+      "loss": 0.7487,
+      "step": 1485
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001448559670781893,
+      "loss": 0.7563,
+      "step": 1490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014465020576131686,
+      "loss": 0.7444,
+      "step": 1495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014444444444444444,
+      "loss": 0.7399,
+      "step": 1500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014423868312757202,
+      "loss": 0.7288,
+      "step": 1505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001440329218106996,
+      "loss": 0.7458,
+      "step": 1510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014382716049382718,
+      "loss": 0.7297,
+      "step": 1515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014362139917695473,
+      "loss": 0.7684,
+      "step": 1520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001434156378600823,
+      "loss": 0.7428,
+      "step": 1525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014320987654320989,
+      "loss": 0.705,
+      "step": 1530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014300411522633747,
+      "loss": 0.7445,
+      "step": 1535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014279835390946504,
+      "loss": 0.7109,
+      "step": 1540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001425925925925926,
+      "loss": 0.7502,
+      "step": 1545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014238683127572018,
+      "loss": 0.7485,
+      "step": 1550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014218106995884776,
+      "loss": 0.7421,
+      "step": 1555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014197530864197534,
+      "loss": 0.7369,
+      "step": 1560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001417695473251029,
+      "loss": 0.7746,
+      "step": 1565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014156378600823047,
+      "loss": 0.7334,
+      "step": 1570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014135802469135805,
+      "loss": 0.7251,
+      "step": 1575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001411522633744856,
+      "loss": 0.7146,
+      "step": 1580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014094650205761318,
+      "loss": 0.7563,
+      "step": 1585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014074074074074076,
+      "loss": 0.7247,
+      "step": 1590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014053497942386834,
+      "loss": 0.7379,
+      "step": 1595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001403292181069959,
+      "loss": 0.7684,
+      "step": 1600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014012345679012347,
+      "loss": 0.7574,
+      "step": 1605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00013991769547325105,
+      "loss": 0.7138,
+      "step": 1610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001397119341563786,
+      "loss": 0.7472,
+      "step": 1615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013950617283950618,
+      "loss": 0.7466,
+      "step": 1620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013930041152263376,
+      "loss": 0.7288,
+      "step": 1625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013909465020576134,
+      "loss": 0.7717,
+      "step": 1630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001388888888888889,
+      "loss": 0.7665,
+      "step": 1635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013868312757201647,
+      "loss": 0.7408,
+      "step": 1640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013847736625514405,
+      "loss": 0.7418,
+      "step": 1645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001382716049382716,
+      "loss": 0.7502,
+      "step": 1650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013806584362139918,
+      "loss": 0.7657,
+      "step": 1655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013786008230452676,
+      "loss": 0.728,
+      "step": 1660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001376543209876543,
+      "loss": 0.7142,
+      "step": 1665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001374485596707819,
+      "loss": 0.7971,
+      "step": 1670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013724279835390947,
+      "loss": 0.737,
+      "step": 1675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013703703703703705,
+      "loss": 0.7947,
+      "step": 1680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001368312757201646,
+      "loss": 0.7525,
+      "step": 1685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013662551440329218,
+      "loss": 0.711,
+      "step": 1690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013641975308641976,
+      "loss": 0.7459,
+      "step": 1695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001362139917695473,
+      "loss": 0.7404,
+      "step": 1700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001360082304526749,
+      "loss": 0.7458,
+      "step": 1705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013580246913580247,
+      "loss": 0.754,
+      "step": 1710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013559670781893005,
+      "loss": 0.7217,
+      "step": 1715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001353909465020576,
+      "loss": 0.7373,
+      "step": 1720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013518518518518518,
+      "loss": 0.7567,
+      "step": 1725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013497942386831276,
+      "loss": 0.7525,
+      "step": 1730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013477366255144034,
+      "loss": 0.752,
+      "step": 1735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013456790123456792,
+      "loss": 0.7342,
+      "step": 1740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001343621399176955,
+      "loss": 0.7276,
+      "step": 1745
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013415637860082305,
+      "loss": 0.6915,
+      "step": 1750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013395061728395063,
+      "loss": 0.737,
+      "step": 1755
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001337448559670782,
+      "loss": 0.7427,
+      "step": 1760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001335390946502058,
+      "loss": 0.7126,
+      "step": 1765
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.7241,
+      "step": 1770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013312757201646092,
+      "loss": 0.755,
+      "step": 1775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001329218106995885,
+      "loss": 0.7494,
+      "step": 1780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013271604938271605,
+      "loss": 0.709,
+      "step": 1785
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013251028806584363,
+      "loss": 0.7555,
+      "step": 1790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001323045267489712,
+      "loss": 0.7298,
+      "step": 1795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001320987654320988,
+      "loss": 0.7232,
+      "step": 1800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013189300411522634,
+      "loss": 0.7441,
+      "step": 1805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013168724279835392,
+      "loss": 0.7623,
+      "step": 1810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001314814814814815,
+      "loss": 0.7349,
+      "step": 1815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013127572016460905,
+      "loss": 0.7773,
+      "step": 1820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013106995884773663,
+      "loss": 0.7537,
+      "step": 1825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001308641975308642,
+      "loss": 0.7579,
+      "step": 1830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001306584362139918,
+      "loss": 0.7321,
+      "step": 1835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013045267489711934,
+      "loss": 0.7175,
+      "step": 1840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00013024691358024692,
+      "loss": 0.7553,
+      "step": 1845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001300411522633745,
+      "loss": 0.7591,
+      "step": 1850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012983539094650205,
+      "loss": 0.7581,
+      "step": 1855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012962962962962963,
+      "loss": 0.7632,
+      "step": 1860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001294238683127572,
+      "loss": 0.7434,
+      "step": 1865
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012921810699588476,
+      "loss": 0.7695,
+      "step": 1870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012901234567901234,
+      "loss": 0.7808,
+      "step": 1875
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012880658436213992,
+      "loss": 0.7428,
+      "step": 1880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001286008230452675,
+      "loss": 0.7487,
+      "step": 1885
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012839506172839505,
+      "loss": 0.7033,
+      "step": 1890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012818930041152263,
+      "loss": 0.7639,
+      "step": 1895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001279835390946502,
+      "loss": 0.7345,
+      "step": 1900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012777777777777776,
+      "loss": 0.7434,
+      "step": 1905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012757201646090534,
+      "loss": 0.7082,
+      "step": 1910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012736625514403292,
+      "loss": 0.7294,
+      "step": 1915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001271604938271605,
+      "loss": 0.74,
+      "step": 1920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012695473251028805,
+      "loss": 0.7694,
+      "step": 1925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012674897119341563,
+      "loss": 0.7339,
+      "step": 1930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001265432098765432,
+      "loss": 0.7259,
+      "step": 1935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001263374485596708,
+      "loss": 0.7307,
+      "step": 1940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012613168724279837,
+      "loss": 0.7085,
+      "step": 1945
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012592592592592592,
+      "loss": 0.7662,
+      "step": 1950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001257201646090535,
+      "loss": 0.7145,
+      "step": 1955
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012551440329218108,
+      "loss": 0.7528,
+      "step": 1960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012530864197530866,
+      "loss": 0.7308,
+      "step": 1965
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012510288065843624,
+      "loss": 0.7163,
+      "step": 1970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001248971193415638,
+      "loss": 0.7254,
+      "step": 1975
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012469135802469137,
+      "loss": 0.7467,
+      "step": 1980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012448559670781895,
+      "loss": 0.7391,
+      "step": 1985
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001242798353909465,
+      "loss": 0.7314,
+      "step": 1990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012407407407407408,
+      "loss": 0.7048,
+      "step": 1995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012386831275720166,
+      "loss": 0.7187,
+      "step": 2000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012366255144032924,
+      "loss": 0.735,
+      "step": 2005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001234567901234568,
+      "loss": 0.7662,
+      "step": 2010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012325102880658437,
+      "loss": 0.7437,
+      "step": 2015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012304526748971195,
+      "loss": 0.7675,
+      "step": 2020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001228395061728395,
+      "loss": 0.7143,
+      "step": 2025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012263374485596708,
+      "loss": 0.7101,
+      "step": 2030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012242798353909466,
+      "loss": 0.7257,
+      "step": 2035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012222222222222224,
+      "loss": 0.7522,
+      "step": 2040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001220164609053498,
+      "loss": 0.7669,
+      "step": 2045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012181069958847737,
+      "loss": 0.7283,
+      "step": 2050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012160493827160495,
+      "loss": 0.7413,
+      "step": 2055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001213991769547325,
+      "loss": 0.6988,
+      "step": 2060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012119341563786008,
+      "loss": 0.7421,
+      "step": 2065
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012098765432098766,
+      "loss": 0.7152,
+      "step": 2070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012078189300411524,
+      "loss": 0.7741,
+      "step": 2075
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001205761316872428,
+      "loss": 0.7381,
+      "step": 2080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012037037037037037,
+      "loss": 0.738,
+      "step": 2085
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012016460905349795,
+      "loss": 0.7258,
+      "step": 2090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011995884773662552,
+      "loss": 0.7426,
+      "step": 2095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011975308641975308,
+      "loss": 0.7185,
+      "step": 2100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011954732510288066,
+      "loss": 0.7401,
+      "step": 2105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011934156378600823,
+      "loss": 0.751,
+      "step": 2110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011913580246913581,
+      "loss": 0.7284,
+      "step": 2115
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011893004115226339,
+      "loss": 0.7901,
+      "step": 2120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011872427983539097,
+      "loss": 0.7432,
+      "step": 2125
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011851851851851852,
+      "loss": 0.7355,
+      "step": 2130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001183127572016461,
+      "loss": 0.7333,
+      "step": 2135
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011810699588477368,
+      "loss": 0.7504,
+      "step": 2140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011790123456790123,
+      "loss": 0.7503,
+      "step": 2145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011769547325102881,
+      "loss": 0.728,
+      "step": 2150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011748971193415639,
+      "loss": 0.7485,
+      "step": 2155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011728395061728397,
+      "loss": 0.7127,
+      "step": 2160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011707818930041152,
+      "loss": 0.7278,
+      "step": 2165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001168724279835391,
+      "loss": 0.7513,
+      "step": 2170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011666666666666668,
+      "loss": 0.7193,
+      "step": 2175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011646090534979423,
+      "loss": 0.7245,
+      "step": 2180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011625514403292181,
+      "loss": 0.7266,
+      "step": 2185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011604938271604939,
+      "loss": 0.7531,
+      "step": 2190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011584362139917696,
+      "loss": 0.7637,
+      "step": 2195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011563786008230452,
+      "loss": 0.7388,
+      "step": 2200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001154320987654321,
+      "loss": 0.7222,
+      "step": 2205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011522633744855968,
+      "loss": 0.7667,
+      "step": 2210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011502057613168725,
+      "loss": 0.7382,
+      "step": 2215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011481481481481482,
+      "loss": 0.7243,
+      "step": 2220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001146090534979424,
+      "loss": 0.7192,
+      "step": 2225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011440329218106996,
+      "loss": 0.7237,
+      "step": 2230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011419753086419754,
+      "loss": 0.7498,
+      "step": 2235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011399176954732511,
+      "loss": 0.7007,
+      "step": 2240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001137860082304527,
+      "loss": 0.758,
+      "step": 2245
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011358024691358025,
+      "loss": 0.7111,
+      "step": 2250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011337448559670783,
+      "loss": 0.706,
+      "step": 2255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001131687242798354,
+      "loss": 0.7423,
+      "step": 2260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011296296296296296,
+      "loss": 0.7094,
+      "step": 2265
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011275720164609054,
+      "loss": 0.7331,
+      "step": 2270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011255144032921812,
+      "loss": 0.7164,
+      "step": 2275
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001123456790123457,
+      "loss": 0.7452,
+      "step": 2280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011213991769547325,
+      "loss": 0.7172,
+      "step": 2285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011193415637860083,
+      "loss": 0.7232,
+      "step": 2290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001117283950617284,
+      "loss": 0.7112,
+      "step": 2295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011152263374485596,
+      "loss": 0.7459,
+      "step": 2300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011131687242798354,
+      "loss": 0.7353,
+      "step": 2305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011111111111111112,
+      "loss": 0.7406,
+      "step": 2310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011090534979423868,
+      "loss": 0.7623,
+      "step": 2315
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011069958847736626,
+      "loss": 0.7567,
+      "step": 2320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011049382716049384,
+      "loss": 0.7752,
+      "step": 2325
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001102880658436214,
+      "loss": 0.7419,
+      "step": 2330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011008230452674897,
+      "loss": 0.7244,
+      "step": 2335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00010987654320987655,
+      "loss": 0.7553,
+      "step": 2340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010967078189300413,
+      "loss": 0.7389,
+      "step": 2345
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010946502057613168,
+      "loss": 0.7002,
+      "step": 2350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010925925925925926,
+      "loss": 0.7238,
+      "step": 2355
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010905349794238684,
+      "loss": 0.6923,
+      "step": 2360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010884773662551442,
+      "loss": 0.7579,
+      "step": 2365
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010864197530864197,
+      "loss": 0.7291,
+      "step": 2370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010843621399176955,
+      "loss": 0.721,
+      "step": 2375
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010823045267489713,
+      "loss": 0.7605,
+      "step": 2380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010802469135802468,
+      "loss": 0.7297,
+      "step": 2385
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010781893004115226,
+      "loss": 0.7565,
+      "step": 2390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010761316872427984,
+      "loss": 0.701,
+      "step": 2395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010740740740740742,
+      "loss": 0.736,
+      "step": 2400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010720164609053497,
+      "loss": 0.7079,
+      "step": 2405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010699588477366255,
+      "loss": 0.7349,
+      "step": 2410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010679012345679013,
+      "loss": 0.7106,
+      "step": 2415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001065843621399177,
+      "loss": 0.7396,
+      "step": 2420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010637860082304528,
+      "loss": 0.7713,
+      "step": 2425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010617283950617284,
+      "loss": 0.747,
+      "step": 2430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010596707818930041,
+      "loss": 0.7475,
+      "step": 2435
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010576131687242799,
+      "loss": 0.7449,
+      "step": 2440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010555555555555557,
+      "loss": 0.7208,
+      "step": 2445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010534979423868315,
+      "loss": 0.7317,
+      "step": 2450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001051440329218107,
+      "loss": 0.7419,
+      "step": 2455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010493827160493828,
+      "loss": 0.7148,
+      "step": 2460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010473251028806586,
+      "loss": 0.7143,
+      "step": 2465
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010452674897119341,
+      "loss": 0.747,
+      "step": 2470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010432098765432099,
+      "loss": 0.7079,
+      "step": 2475
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010411522633744857,
+      "loss": 0.7059,
+      "step": 2480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010390946502057615,
+      "loss": 0.7272,
+      "step": 2485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001037037037037037,
+      "loss": 0.7281,
+      "step": 2490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010349794238683128,
+      "loss": 0.6956,
+      "step": 2495
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010329218106995886,
+      "loss": 0.7571,
+      "step": 2500
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 5010,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 500,
+  "total_flos": 1.006726589213311e+19,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/training_args.bin b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e707059a03fbe62142ac41332e145ac1ac1f36cc
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b99a1729574e0e46ee782b42dc90101a785d5b4792871920d293f44722c28cc
+size 5944
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/zero_to_fp32.py b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/zero_to_fp32.py
new file mode 100644
index 0000000000000000000000000000000000000000..c98caae31534368be22b67fc4ae906836c992a8d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-2500/zero_to_fp32.py
@@ -0,0 +1,587 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+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):
+    """
+    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)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+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 _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()
+            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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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 param_shapes.items():
+
+        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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    """
+    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``
+
+    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)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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("-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_file, tag=args.tag)
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/README.md b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..d380bbbf0bf0fa1f211b53410559607e5007a209
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: meta-llama/Llama-2-7b-hf
+---
+
+# 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 0.9.0
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/adapter_config.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c81a592f0ff08374a4f61f6003029070a8429ae
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/adapter_config.json
@@ -0,0 +1,35 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "mora_type": 1,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_mora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/adapter_model.safetensors b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/adapter_model.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..9b54d94d95fd4d335426c8041703b751ef521862
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/adapter_model.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12bb4d2478f00de5cd40c10cae5ef5bf80ef1fec9bd44f7398d51bca636a3ddc
+size 564342240
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/global_step3000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/global_step3000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..6397f62166a7fcede76485314e2a1620fc0f2b90
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/global_step3000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0210f3f6e2ac8f9cd0b17e5408639966f58dbda6ac71bbc377f17243c37ba278
+size 239914224
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/global_step3000/mp_rank_00_model_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/global_step3000/mp_rank_00_model_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ca006784c18a3fdfe765bdb105bbfac2400d3214
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/global_step3000/mp_rank_00_model_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b60211ffd4f70597b76b44474e6e539456e92269af560e92b3dd2a1125a2312
+size 40303404
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/latest b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/latest
new file mode 100644
index 0000000000000000000000000000000000000000..1d7ba373feda3875a77d01589035020625a7c98e
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/latest
@@ -0,0 +1 @@
+global_step3000
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/rng_state.pth b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/rng_state.pth
new file mode 100644
index 0000000000000000000000000000000000000000..ef5bdcfd1c8949add7096d6bba3bbd15a996a7f4
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/rng_state.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19d07e457963fb633a8a76af2488db91a9a823d4ae9e728350053659f7550aff
+size 14244
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/scheduler.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/scheduler.pt
new file mode 100644
index 0000000000000000000000000000000000000000..cae8d8d17ef2ea198cb9db32e2e75500ed2fcb86
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/scheduler.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e52b346ba1d827109b4132f75a5a289e15d677fb11398880b39f9c40c0bdea2f
+size 1064
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/trainer_state.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0c1ad939aa7d70c81823fbc7fb66d8cbc69d5cd
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/trainer_state.json
@@ -0,0 +1,3621 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.197492749555618,
+  "eval_steps": 500,
+  "global_step": 3000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.3797,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.3952,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2e-05,
+      "loss": 1.3112,
+      "step": 15
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 1.3136,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 1.2012,
+      "step": 25
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4e-05,
+      "loss": 1.1277,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 1.0832,
+      "step": 35
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.333333333333333e-05,
+      "loss": 1.0659,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-05,
+      "loss": 0.983,
+      "step": 45
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.9763,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.333333333333333e-05,
+      "loss": 0.9992,
+      "step": 55
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8e-05,
+      "loss": 0.9405,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.666666666666667e-05,
+      "loss": 0.9827,
+      "step": 65
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.333333333333334e-05,
+      "loss": 0.9061,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001,
+      "loss": 0.8921,
+      "step": 75
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010666666666666667,
+      "loss": 0.9377,
+      "step": 80
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011333333333333334,
+      "loss": 0.8528,
+      "step": 85
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012,
+      "loss": 0.8913,
+      "step": 90
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012666666666666666,
+      "loss": 0.8562,
+      "step": 95
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.877,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014,
+      "loss": 0.8781,
+      "step": 105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014666666666666666,
+      "loss": 0.8546,
+      "step": 110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015333333333333334,
+      "loss": 0.8672,
+      "step": 115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016,
+      "loss": 0.8611,
+      "step": 120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.8624,
+      "step": 125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017333333333333334,
+      "loss": 0.8477,
+      "step": 130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018,
+      "loss": 0.8227,
+      "step": 135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001866666666666667,
+      "loss": 0.8064,
+      "step": 140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019333333333333333,
+      "loss": 0.8422,
+      "step": 145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002,
+      "loss": 0.8471,
+      "step": 150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001997942386831276,
+      "loss": 0.8183,
+      "step": 155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958847736625514,
+      "loss": 0.8229,
+      "step": 160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019938271604938272,
+      "loss": 0.8114,
+      "step": 165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991769547325103,
+      "loss": 0.7861,
+      "step": 170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019897119341563788,
+      "loss": 0.841,
+      "step": 175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019876543209876543,
+      "loss": 0.8549,
+      "step": 180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000198559670781893,
+      "loss": 0.8449,
+      "step": 185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001983539094650206,
+      "loss": 0.8311,
+      "step": 190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019814814814814814,
+      "loss": 0.8296,
+      "step": 195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019794238683127572,
+      "loss": 0.8263,
+      "step": 200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001977366255144033,
+      "loss": 0.8452,
+      "step": 205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019753086419753085,
+      "loss": 0.8044,
+      "step": 210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019732510288065843,
+      "loss": 0.8119,
+      "step": 215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000197119341563786,
+      "loss": 0.8127,
+      "step": 220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001969135802469136,
+      "loss": 0.7921,
+      "step": 225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019670781893004114,
+      "loss": 0.803,
+      "step": 230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019650205761316872,
+      "loss": 0.838,
+      "step": 235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001962962962962963,
+      "loss": 0.7801,
+      "step": 240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019609053497942388,
+      "loss": 0.817,
+      "step": 245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019588477366255146,
+      "loss": 0.8211,
+      "step": 250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000195679012345679,
+      "loss": 0.8423,
+      "step": 255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001954732510288066,
+      "loss": 0.8007,
+      "step": 260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019526748971193417,
+      "loss": 0.8135,
+      "step": 265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019506172839506175,
+      "loss": 0.7961,
+      "step": 270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019485596707818933,
+      "loss": 0.797,
+      "step": 275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019465020576131688,
+      "loss": 0.793,
+      "step": 280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019444444444444446,
+      "loss": 0.7809,
+      "step": 285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019423868312757204,
+      "loss": 0.7811,
+      "step": 290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019403292181069962,
+      "loss": 0.8074,
+      "step": 295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019382716049382717,
+      "loss": 0.8028,
+      "step": 300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019362139917695475,
+      "loss": 0.7895,
+      "step": 305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019341563786008233,
+      "loss": 0.7777,
+      "step": 310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019320987654320988,
+      "loss": 0.7868,
+      "step": 315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019300411522633746,
+      "loss": 0.7921,
+      "step": 320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019279835390946504,
+      "loss": 0.8,
+      "step": 325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001925925925925926,
+      "loss": 0.7781,
+      "step": 330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019238683127572017,
+      "loss": 0.7799,
+      "step": 335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019218106995884775,
+      "loss": 0.7921,
+      "step": 340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019197530864197533,
+      "loss": 0.7851,
+      "step": 345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019176954732510288,
+      "loss": 0.8106,
+      "step": 350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019156378600823046,
+      "loss": 0.7953,
+      "step": 355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019135802469135804,
+      "loss": 0.8084,
+      "step": 360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001911522633744856,
+      "loss": 0.8298,
+      "step": 365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019094650205761317,
+      "loss": 0.8273,
+      "step": 370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019074074074074075,
+      "loss": 0.8047,
+      "step": 375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019053497942386833,
+      "loss": 0.8108,
+      "step": 380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019032921810699588,
+      "loss": 0.8207,
+      "step": 385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019012345679012346,
+      "loss": 0.7884,
+      "step": 390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018991769547325104,
+      "loss": 0.7918,
+      "step": 395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001897119341563786,
+      "loss": 0.7554,
+      "step": 400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018950617283950617,
+      "loss": 0.8099,
+      "step": 405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018930041152263375,
+      "loss": 0.807,
+      "step": 410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018909465020576133,
+      "loss": 0.7793,
+      "step": 415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018888888888888888,
+      "loss": 0.7734,
+      "step": 420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018868312757201646,
+      "loss": 0.7786,
+      "step": 425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018847736625514404,
+      "loss": 0.7548,
+      "step": 430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001882716049382716,
+      "loss": 0.8024,
+      "step": 435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018806584362139917,
+      "loss": 0.7708,
+      "step": 440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018786008230452675,
+      "loss": 0.8045,
+      "step": 445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018765432098765433,
+      "loss": 0.7477,
+      "step": 450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018744855967078189,
+      "loss": 0.7474,
+      "step": 455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018724279835390946,
+      "loss": 0.7629,
+      "step": 460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018703703703703704,
+      "loss": 0.7777,
+      "step": 465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018683127572016462,
+      "loss": 0.7823,
+      "step": 470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001866255144032922,
+      "loss": 0.7762,
+      "step": 475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018641975308641978,
+      "loss": 0.782,
+      "step": 480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018621399176954733,
+      "loss": 0.7783,
+      "step": 485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001860082304526749,
+      "loss": 0.7376,
+      "step": 490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001858024691358025,
+      "loss": 0.7683,
+      "step": 495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018559670781893007,
+      "loss": 0.7547,
+      "step": 500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018539094650205762,
+      "loss": 0.7569,
+      "step": 505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001851851851851852,
+      "loss": 0.7796,
+      "step": 510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018497942386831278,
+      "loss": 0.7888,
+      "step": 515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018477366255144033,
+      "loss": 0.7763,
+      "step": 520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018456790123456791,
+      "loss": 0.7787,
+      "step": 525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001843621399176955,
+      "loss": 0.772,
+      "step": 530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018415637860082307,
+      "loss": 0.786,
+      "step": 535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018395061728395062,
+      "loss": 0.7639,
+      "step": 540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001837448559670782,
+      "loss": 0.775,
+      "step": 545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018353909465020578,
+      "loss": 0.8002,
+      "step": 550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 0.7359,
+      "step": 555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018312757201646091,
+      "loss": 0.747,
+      "step": 560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001829218106995885,
+      "loss": 0.8234,
+      "step": 565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018271604938271605,
+      "loss": 0.7586,
+      "step": 570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018251028806584363,
+      "loss": 0.7315,
+      "step": 575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001823045267489712,
+      "loss": 0.7986,
+      "step": 580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018209876543209878,
+      "loss": 0.7758,
+      "step": 585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018189300411522634,
+      "loss": 0.7869,
+      "step": 590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018168724279835392,
+      "loss": 0.7596,
+      "step": 595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001814814814814815,
+      "loss": 0.7658,
+      "step": 600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018127572016460905,
+      "loss": 0.7851,
+      "step": 605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018106995884773663,
+      "loss": 0.7957,
+      "step": 610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001808641975308642,
+      "loss": 0.7756,
+      "step": 615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018065843621399178,
+      "loss": 0.7911,
+      "step": 620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018045267489711934,
+      "loss": 0.7864,
+      "step": 625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018024691358024692,
+      "loss": 0.7683,
+      "step": 630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001800411522633745,
+      "loss": 0.7731,
+      "step": 635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017983539094650205,
+      "loss": 0.7671,
+      "step": 640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017962962962962963,
+      "loss": 0.7367,
+      "step": 645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001794238683127572,
+      "loss": 0.8024,
+      "step": 650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017921810699588476,
+      "loss": 0.7572,
+      "step": 655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017901234567901234,
+      "loss": 0.7554,
+      "step": 660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017880658436213992,
+      "loss": 0.7677,
+      "step": 665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001786008230452675,
+      "loss": 0.7652,
+      "step": 670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017839506172839508,
+      "loss": 0.7771,
+      "step": 675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017818930041152265,
+      "loss": 0.7747,
+      "step": 680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001779835390946502,
+      "loss": 0.767,
+      "step": 685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017777777777777779,
+      "loss": 0.7738,
+      "step": 690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017757201646090537,
+      "loss": 0.7271,
+      "step": 695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017736625514403294,
+      "loss": 0.7316,
+      "step": 700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017716049382716052,
+      "loss": 0.7831,
+      "step": 705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017695473251028808,
+      "loss": 0.7835,
+      "step": 710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017674897119341566,
+      "loss": 0.7497,
+      "step": 715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017654320987654323,
+      "loss": 0.7757,
+      "step": 720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001763374485596708,
+      "loss": 0.74,
+      "step": 725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017613168724279837,
+      "loss": 0.7697,
+      "step": 730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017592592592592595,
+      "loss": 0.774,
+      "step": 735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017572016460905352,
+      "loss": 0.7668,
+      "step": 740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017551440329218108,
+      "loss": 0.7779,
+      "step": 745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017530864197530866,
+      "loss": 0.7441,
+      "step": 750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017510288065843624,
+      "loss": 0.7578,
+      "step": 755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001748971193415638,
+      "loss": 0.8358,
+      "step": 760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017469135802469137,
+      "loss": 0.7473,
+      "step": 765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017448559670781895,
+      "loss": 0.7517,
+      "step": 770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001742798353909465,
+      "loss": 0.7593,
+      "step": 775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017407407407407408,
+      "loss": 0.7251,
+      "step": 780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017386831275720166,
+      "loss": 0.7365,
+      "step": 785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017366255144032924,
+      "loss": 0.7798,
+      "step": 790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001734567901234568,
+      "loss": 0.746,
+      "step": 795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017325102880658437,
+      "loss": 0.7607,
+      "step": 800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017304526748971195,
+      "loss": 0.7694,
+      "step": 805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001728395061728395,
+      "loss": 0.7926,
+      "step": 810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017263374485596708,
+      "loss": 0.795,
+      "step": 815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017242798353909466,
+      "loss": 0.7581,
+      "step": 820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017222222222222224,
+      "loss": 0.7629,
+      "step": 825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001720164609053498,
+      "loss": 0.7526,
+      "step": 830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017181069958847737,
+      "loss": 0.7338,
+      "step": 835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017160493827160495,
+      "loss": 0.7544,
+      "step": 840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001713991769547325,
+      "loss": 0.7522,
+      "step": 845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017119341563786008,
+      "loss": 0.7639,
+      "step": 850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017098765432098766,
+      "loss": 0.7914,
+      "step": 855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017078189300411524,
+      "loss": 0.7635,
+      "step": 860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001705761316872428,
+      "loss": 0.8343,
+      "step": 865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017037037037037037,
+      "loss": 0.7299,
+      "step": 870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017016460905349795,
+      "loss": 0.7643,
+      "step": 875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016995884773662553,
+      "loss": 0.7343,
+      "step": 880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016975308641975308,
+      "loss": 0.7546,
+      "step": 885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016954732510288066,
+      "loss": 0.7435,
+      "step": 890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016934156378600824,
+      "loss": 0.8086,
+      "step": 895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016913580246913582,
+      "loss": 0.7274,
+      "step": 900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001689300411522634,
+      "loss": 0.7511,
+      "step": 905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016872427983539098,
+      "loss": 0.7689,
+      "step": 910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016851851851851853,
+      "loss": 0.754,
+      "step": 915
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001683127572016461,
+      "loss": 0.7298,
+      "step": 920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001681069958847737,
+      "loss": 0.7755,
+      "step": 925
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016790123456790124,
+      "loss": 0.7854,
+      "step": 930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016769547325102882,
+      "loss": 0.7452,
+      "step": 935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001674897119341564,
+      "loss": 0.7473,
+      "step": 940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016728395061728398,
+      "loss": 0.7855,
+      "step": 945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016707818930041153,
+      "loss": 0.7527,
+      "step": 950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001668724279835391,
+      "loss": 0.7641,
+      "step": 955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.828,
+      "step": 960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016646090534979424,
+      "loss": 0.7571,
+      "step": 965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016625514403292182,
+      "loss": 0.7768,
+      "step": 970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001660493827160494,
+      "loss": 0.7644,
+      "step": 975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016584362139917695,
+      "loss": 0.7747,
+      "step": 980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016563786008230453,
+      "loss": 0.7926,
+      "step": 985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001654320987654321,
+      "loss": 0.7645,
+      "step": 990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001652263374485597,
+      "loss": 0.7464,
+      "step": 995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016502057613168724,
+      "loss": 0.7558,
+      "step": 1000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016481481481481482,
+      "loss": 0.7574,
+      "step": 1005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001646090534979424,
+      "loss": 0.7491,
+      "step": 1010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016440329218106995,
+      "loss": 0.756,
+      "step": 1015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016419753086419753,
+      "loss": 0.7679,
+      "step": 1020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001639917695473251,
+      "loss": 0.776,
+      "step": 1025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001637860082304527,
+      "loss": 0.7235,
+      "step": 1030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016358024691358024,
+      "loss": 0.7817,
+      "step": 1035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016337448559670782,
+      "loss": 0.7333,
+      "step": 1040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001631687242798354,
+      "loss": 0.754,
+      "step": 1045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016296296296296295,
+      "loss": 0.74,
+      "step": 1050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016275720164609053,
+      "loss": 0.7857,
+      "step": 1055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001625514403292181,
+      "loss": 0.8015,
+      "step": 1060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001623456790123457,
+      "loss": 0.7508,
+      "step": 1065
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016213991769547324,
+      "loss": 0.7514,
+      "step": 1070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016193415637860082,
+      "loss": 0.7326,
+      "step": 1075
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001617283950617284,
+      "loss": 0.7322,
+      "step": 1080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016152263374485595,
+      "loss": 0.7373,
+      "step": 1085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016131687242798353,
+      "loss": 0.7637,
+      "step": 1090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001611111111111111,
+      "loss": 0.7697,
+      "step": 1095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001609053497942387,
+      "loss": 0.7505,
+      "step": 1100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016069958847736627,
+      "loss": 0.7568,
+      "step": 1105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016049382716049385,
+      "loss": 0.786,
+      "step": 1110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001602880658436214,
+      "loss": 0.7525,
+      "step": 1115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016008230452674898,
+      "loss": 0.7696,
+      "step": 1120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015987654320987656,
+      "loss": 0.7648,
+      "step": 1125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015967078189300414,
+      "loss": 0.7492,
+      "step": 1130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001594650205761317,
+      "loss": 0.809,
+      "step": 1135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015925925925925927,
+      "loss": 0.7592,
+      "step": 1140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015905349794238685,
+      "loss": 0.7874,
+      "step": 1145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015884773662551443,
+      "loss": 0.7874,
+      "step": 1150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015864197530864198,
+      "loss": 0.7374,
+      "step": 1155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015843621399176956,
+      "loss": 0.7218,
+      "step": 1160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015823045267489714,
+      "loss": 0.7494,
+      "step": 1165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001580246913580247,
+      "loss": 0.7449,
+      "step": 1170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015781893004115227,
+      "loss": 0.725,
+      "step": 1175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015761316872427985,
+      "loss": 0.739,
+      "step": 1180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015740740740740743,
+      "loss": 0.7554,
+      "step": 1185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015720164609053498,
+      "loss": 0.7702,
+      "step": 1190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015699588477366256,
+      "loss": 0.7418,
+      "step": 1195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015679012345679014,
+      "loss": 0.7217,
+      "step": 1200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001565843621399177,
+      "loss": 0.7391,
+      "step": 1205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015637860082304527,
+      "loss": 0.7696,
+      "step": 1210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015617283950617285,
+      "loss": 0.7594,
+      "step": 1215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001559670781893004,
+      "loss": 0.777,
+      "step": 1220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015576131687242798,
+      "loss": 0.7441,
+      "step": 1225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015555555555555556,
+      "loss": 0.7611,
+      "step": 1230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015534979423868314,
+      "loss": 0.7278,
+      "step": 1235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001551440329218107,
+      "loss": 0.7452,
+      "step": 1240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015493827160493827,
+      "loss": 0.7939,
+      "step": 1245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015473251028806585,
+      "loss": 0.7161,
+      "step": 1250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001545267489711934,
+      "loss": 0.7344,
+      "step": 1255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015432098765432098,
+      "loss": 0.7368,
+      "step": 1260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015411522633744856,
+      "loss": 0.7172,
+      "step": 1265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015390946502057614,
+      "loss": 0.7589,
+      "step": 1270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001537037037037037,
+      "loss": 0.7641,
+      "step": 1275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015349794238683127,
+      "loss": 0.761,
+      "step": 1280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015329218106995885,
+      "loss": 0.7414,
+      "step": 1285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001530864197530864,
+      "loss": 0.7713,
+      "step": 1290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015288065843621399,
+      "loss": 0.7543,
+      "step": 1295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015267489711934156,
+      "loss": 0.7683,
+      "step": 1300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015246913580246914,
+      "loss": 0.8106,
+      "step": 1305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015226337448559672,
+      "loss": 0.7739,
+      "step": 1310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015205761316872428,
+      "loss": 0.7585,
+      "step": 1315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015185185185185185,
+      "loss": 0.7577,
+      "step": 1320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015164609053497943,
+      "loss": 0.7608,
+      "step": 1325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000151440329218107,
+      "loss": 0.7715,
+      "step": 1330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001512345679012346,
+      "loss": 0.7681,
+      "step": 1335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015102880658436214,
+      "loss": 0.7476,
+      "step": 1340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015082304526748972,
+      "loss": 0.7539,
+      "step": 1345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001506172839506173,
+      "loss": 0.7492,
+      "step": 1350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015041152263374488,
+      "loss": 0.7491,
+      "step": 1355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015020576131687243,
+      "loss": 0.7236,
+      "step": 1360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 0.7328,
+      "step": 1365
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001497942386831276,
+      "loss": 0.7603,
+      "step": 1370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014958847736625515,
+      "loss": 0.7206,
+      "step": 1375
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014938271604938272,
+      "loss": 0.7001,
+      "step": 1380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001491769547325103,
+      "loss": 0.7241,
+      "step": 1385
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014897119341563788,
+      "loss": 0.7511,
+      "step": 1390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014876543209876544,
+      "loss": 0.7546,
+      "step": 1395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014855967078189301,
+      "loss": 0.762,
+      "step": 1400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001483539094650206,
+      "loss": 0.6979,
+      "step": 1405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014814814814814815,
+      "loss": 0.7352,
+      "step": 1410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014794238683127573,
+      "loss": 0.7604,
+      "step": 1415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001477366255144033,
+      "loss": 0.7612,
+      "step": 1420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014753086419753086,
+      "loss": 0.7618,
+      "step": 1425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014732510288065844,
+      "loss": 0.7107,
+      "step": 1430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014711934156378602,
+      "loss": 0.7247,
+      "step": 1435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001469135802469136,
+      "loss": 0.7733,
+      "step": 1440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014670781893004115,
+      "loss": 0.7627,
+      "step": 1445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014650205761316873,
+      "loss": 0.7323,
+      "step": 1450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001462962962962963,
+      "loss": 0.7728,
+      "step": 1455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014609053497942386,
+      "loss": 0.7541,
+      "step": 1460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014588477366255144,
+      "loss": 0.7781,
+      "step": 1465
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014567901234567902,
+      "loss": 0.7474,
+      "step": 1470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001454732510288066,
+      "loss": 0.7789,
+      "step": 1475
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014526748971193415,
+      "loss": 0.7497,
+      "step": 1480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014506172839506173,
+      "loss": 0.7487,
+      "step": 1485
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001448559670781893,
+      "loss": 0.7563,
+      "step": 1490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014465020576131686,
+      "loss": 0.7444,
+      "step": 1495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014444444444444444,
+      "loss": 0.7399,
+      "step": 1500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014423868312757202,
+      "loss": 0.7288,
+      "step": 1505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001440329218106996,
+      "loss": 0.7458,
+      "step": 1510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014382716049382718,
+      "loss": 0.7297,
+      "step": 1515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014362139917695473,
+      "loss": 0.7684,
+      "step": 1520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001434156378600823,
+      "loss": 0.7428,
+      "step": 1525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014320987654320989,
+      "loss": 0.705,
+      "step": 1530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014300411522633747,
+      "loss": 0.7445,
+      "step": 1535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014279835390946504,
+      "loss": 0.7109,
+      "step": 1540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001425925925925926,
+      "loss": 0.7502,
+      "step": 1545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014238683127572018,
+      "loss": 0.7485,
+      "step": 1550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014218106995884776,
+      "loss": 0.7421,
+      "step": 1555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014197530864197534,
+      "loss": 0.7369,
+      "step": 1560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001417695473251029,
+      "loss": 0.7746,
+      "step": 1565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014156378600823047,
+      "loss": 0.7334,
+      "step": 1570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014135802469135805,
+      "loss": 0.7251,
+      "step": 1575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001411522633744856,
+      "loss": 0.7146,
+      "step": 1580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014094650205761318,
+      "loss": 0.7563,
+      "step": 1585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014074074074074076,
+      "loss": 0.7247,
+      "step": 1590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014053497942386834,
+      "loss": 0.7379,
+      "step": 1595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001403292181069959,
+      "loss": 0.7684,
+      "step": 1600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014012345679012347,
+      "loss": 0.7574,
+      "step": 1605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00013991769547325105,
+      "loss": 0.7138,
+      "step": 1610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001397119341563786,
+      "loss": 0.7472,
+      "step": 1615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013950617283950618,
+      "loss": 0.7466,
+      "step": 1620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013930041152263376,
+      "loss": 0.7288,
+      "step": 1625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013909465020576134,
+      "loss": 0.7717,
+      "step": 1630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001388888888888889,
+      "loss": 0.7665,
+      "step": 1635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013868312757201647,
+      "loss": 0.7408,
+      "step": 1640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013847736625514405,
+      "loss": 0.7418,
+      "step": 1645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001382716049382716,
+      "loss": 0.7502,
+      "step": 1650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013806584362139918,
+      "loss": 0.7657,
+      "step": 1655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013786008230452676,
+      "loss": 0.728,
+      "step": 1660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001376543209876543,
+      "loss": 0.7142,
+      "step": 1665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001374485596707819,
+      "loss": 0.7971,
+      "step": 1670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013724279835390947,
+      "loss": 0.737,
+      "step": 1675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013703703703703705,
+      "loss": 0.7947,
+      "step": 1680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001368312757201646,
+      "loss": 0.7525,
+      "step": 1685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013662551440329218,
+      "loss": 0.711,
+      "step": 1690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013641975308641976,
+      "loss": 0.7459,
+      "step": 1695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001362139917695473,
+      "loss": 0.7404,
+      "step": 1700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001360082304526749,
+      "loss": 0.7458,
+      "step": 1705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013580246913580247,
+      "loss": 0.754,
+      "step": 1710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013559670781893005,
+      "loss": 0.7217,
+      "step": 1715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001353909465020576,
+      "loss": 0.7373,
+      "step": 1720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013518518518518518,
+      "loss": 0.7567,
+      "step": 1725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013497942386831276,
+      "loss": 0.7525,
+      "step": 1730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013477366255144034,
+      "loss": 0.752,
+      "step": 1735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013456790123456792,
+      "loss": 0.7342,
+      "step": 1740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001343621399176955,
+      "loss": 0.7276,
+      "step": 1745
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013415637860082305,
+      "loss": 0.6915,
+      "step": 1750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013395061728395063,
+      "loss": 0.737,
+      "step": 1755
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001337448559670782,
+      "loss": 0.7427,
+      "step": 1760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001335390946502058,
+      "loss": 0.7126,
+      "step": 1765
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.7241,
+      "step": 1770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013312757201646092,
+      "loss": 0.755,
+      "step": 1775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001329218106995885,
+      "loss": 0.7494,
+      "step": 1780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013271604938271605,
+      "loss": 0.709,
+      "step": 1785
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013251028806584363,
+      "loss": 0.7555,
+      "step": 1790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001323045267489712,
+      "loss": 0.7298,
+      "step": 1795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001320987654320988,
+      "loss": 0.7232,
+      "step": 1800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013189300411522634,
+      "loss": 0.7441,
+      "step": 1805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013168724279835392,
+      "loss": 0.7623,
+      "step": 1810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001314814814814815,
+      "loss": 0.7349,
+      "step": 1815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013127572016460905,
+      "loss": 0.7773,
+      "step": 1820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013106995884773663,
+      "loss": 0.7537,
+      "step": 1825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001308641975308642,
+      "loss": 0.7579,
+      "step": 1830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001306584362139918,
+      "loss": 0.7321,
+      "step": 1835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013045267489711934,
+      "loss": 0.7175,
+      "step": 1840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00013024691358024692,
+      "loss": 0.7553,
+      "step": 1845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001300411522633745,
+      "loss": 0.7591,
+      "step": 1850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012983539094650205,
+      "loss": 0.7581,
+      "step": 1855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012962962962962963,
+      "loss": 0.7632,
+      "step": 1860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001294238683127572,
+      "loss": 0.7434,
+      "step": 1865
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012921810699588476,
+      "loss": 0.7695,
+      "step": 1870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012901234567901234,
+      "loss": 0.7808,
+      "step": 1875
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012880658436213992,
+      "loss": 0.7428,
+      "step": 1880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001286008230452675,
+      "loss": 0.7487,
+      "step": 1885
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012839506172839505,
+      "loss": 0.7033,
+      "step": 1890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012818930041152263,
+      "loss": 0.7639,
+      "step": 1895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001279835390946502,
+      "loss": 0.7345,
+      "step": 1900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012777777777777776,
+      "loss": 0.7434,
+      "step": 1905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012757201646090534,
+      "loss": 0.7082,
+      "step": 1910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012736625514403292,
+      "loss": 0.7294,
+      "step": 1915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001271604938271605,
+      "loss": 0.74,
+      "step": 1920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012695473251028805,
+      "loss": 0.7694,
+      "step": 1925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012674897119341563,
+      "loss": 0.7339,
+      "step": 1930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001265432098765432,
+      "loss": 0.7259,
+      "step": 1935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001263374485596708,
+      "loss": 0.7307,
+      "step": 1940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012613168724279837,
+      "loss": 0.7085,
+      "step": 1945
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012592592592592592,
+      "loss": 0.7662,
+      "step": 1950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001257201646090535,
+      "loss": 0.7145,
+      "step": 1955
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012551440329218108,
+      "loss": 0.7528,
+      "step": 1960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012530864197530866,
+      "loss": 0.7308,
+      "step": 1965
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012510288065843624,
+      "loss": 0.7163,
+      "step": 1970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001248971193415638,
+      "loss": 0.7254,
+      "step": 1975
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012469135802469137,
+      "loss": 0.7467,
+      "step": 1980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012448559670781895,
+      "loss": 0.7391,
+      "step": 1985
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001242798353909465,
+      "loss": 0.7314,
+      "step": 1990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012407407407407408,
+      "loss": 0.7048,
+      "step": 1995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012386831275720166,
+      "loss": 0.7187,
+      "step": 2000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012366255144032924,
+      "loss": 0.735,
+      "step": 2005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001234567901234568,
+      "loss": 0.7662,
+      "step": 2010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012325102880658437,
+      "loss": 0.7437,
+      "step": 2015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012304526748971195,
+      "loss": 0.7675,
+      "step": 2020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001228395061728395,
+      "loss": 0.7143,
+      "step": 2025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012263374485596708,
+      "loss": 0.7101,
+      "step": 2030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012242798353909466,
+      "loss": 0.7257,
+      "step": 2035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012222222222222224,
+      "loss": 0.7522,
+      "step": 2040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001220164609053498,
+      "loss": 0.7669,
+      "step": 2045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012181069958847737,
+      "loss": 0.7283,
+      "step": 2050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012160493827160495,
+      "loss": 0.7413,
+      "step": 2055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001213991769547325,
+      "loss": 0.6988,
+      "step": 2060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012119341563786008,
+      "loss": 0.7421,
+      "step": 2065
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012098765432098766,
+      "loss": 0.7152,
+      "step": 2070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012078189300411524,
+      "loss": 0.7741,
+      "step": 2075
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001205761316872428,
+      "loss": 0.7381,
+      "step": 2080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012037037037037037,
+      "loss": 0.738,
+      "step": 2085
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012016460905349795,
+      "loss": 0.7258,
+      "step": 2090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011995884773662552,
+      "loss": 0.7426,
+      "step": 2095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011975308641975308,
+      "loss": 0.7185,
+      "step": 2100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011954732510288066,
+      "loss": 0.7401,
+      "step": 2105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011934156378600823,
+      "loss": 0.751,
+      "step": 2110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011913580246913581,
+      "loss": 0.7284,
+      "step": 2115
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011893004115226339,
+      "loss": 0.7901,
+      "step": 2120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011872427983539097,
+      "loss": 0.7432,
+      "step": 2125
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011851851851851852,
+      "loss": 0.7355,
+      "step": 2130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001183127572016461,
+      "loss": 0.7333,
+      "step": 2135
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011810699588477368,
+      "loss": 0.7504,
+      "step": 2140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011790123456790123,
+      "loss": 0.7503,
+      "step": 2145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011769547325102881,
+      "loss": 0.728,
+      "step": 2150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011748971193415639,
+      "loss": 0.7485,
+      "step": 2155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011728395061728397,
+      "loss": 0.7127,
+      "step": 2160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011707818930041152,
+      "loss": 0.7278,
+      "step": 2165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001168724279835391,
+      "loss": 0.7513,
+      "step": 2170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011666666666666668,
+      "loss": 0.7193,
+      "step": 2175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011646090534979423,
+      "loss": 0.7245,
+      "step": 2180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011625514403292181,
+      "loss": 0.7266,
+      "step": 2185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011604938271604939,
+      "loss": 0.7531,
+      "step": 2190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011584362139917696,
+      "loss": 0.7637,
+      "step": 2195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011563786008230452,
+      "loss": 0.7388,
+      "step": 2200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001154320987654321,
+      "loss": 0.7222,
+      "step": 2205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011522633744855968,
+      "loss": 0.7667,
+      "step": 2210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011502057613168725,
+      "loss": 0.7382,
+      "step": 2215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011481481481481482,
+      "loss": 0.7243,
+      "step": 2220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001146090534979424,
+      "loss": 0.7192,
+      "step": 2225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011440329218106996,
+      "loss": 0.7237,
+      "step": 2230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011419753086419754,
+      "loss": 0.7498,
+      "step": 2235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011399176954732511,
+      "loss": 0.7007,
+      "step": 2240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001137860082304527,
+      "loss": 0.758,
+      "step": 2245
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011358024691358025,
+      "loss": 0.7111,
+      "step": 2250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011337448559670783,
+      "loss": 0.706,
+      "step": 2255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001131687242798354,
+      "loss": 0.7423,
+      "step": 2260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011296296296296296,
+      "loss": 0.7094,
+      "step": 2265
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011275720164609054,
+      "loss": 0.7331,
+      "step": 2270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011255144032921812,
+      "loss": 0.7164,
+      "step": 2275
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001123456790123457,
+      "loss": 0.7452,
+      "step": 2280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011213991769547325,
+      "loss": 0.7172,
+      "step": 2285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011193415637860083,
+      "loss": 0.7232,
+      "step": 2290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001117283950617284,
+      "loss": 0.7112,
+      "step": 2295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011152263374485596,
+      "loss": 0.7459,
+      "step": 2300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011131687242798354,
+      "loss": 0.7353,
+      "step": 2305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011111111111111112,
+      "loss": 0.7406,
+      "step": 2310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011090534979423868,
+      "loss": 0.7623,
+      "step": 2315
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011069958847736626,
+      "loss": 0.7567,
+      "step": 2320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011049382716049384,
+      "loss": 0.7752,
+      "step": 2325
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001102880658436214,
+      "loss": 0.7419,
+      "step": 2330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011008230452674897,
+      "loss": 0.7244,
+      "step": 2335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00010987654320987655,
+      "loss": 0.7553,
+      "step": 2340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010967078189300413,
+      "loss": 0.7389,
+      "step": 2345
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010946502057613168,
+      "loss": 0.7002,
+      "step": 2350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010925925925925926,
+      "loss": 0.7238,
+      "step": 2355
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010905349794238684,
+      "loss": 0.6923,
+      "step": 2360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010884773662551442,
+      "loss": 0.7579,
+      "step": 2365
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010864197530864197,
+      "loss": 0.7291,
+      "step": 2370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010843621399176955,
+      "loss": 0.721,
+      "step": 2375
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010823045267489713,
+      "loss": 0.7605,
+      "step": 2380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010802469135802468,
+      "loss": 0.7297,
+      "step": 2385
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010781893004115226,
+      "loss": 0.7565,
+      "step": 2390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010761316872427984,
+      "loss": 0.701,
+      "step": 2395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010740740740740742,
+      "loss": 0.736,
+      "step": 2400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010720164609053497,
+      "loss": 0.7079,
+      "step": 2405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010699588477366255,
+      "loss": 0.7349,
+      "step": 2410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010679012345679013,
+      "loss": 0.7106,
+      "step": 2415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001065843621399177,
+      "loss": 0.7396,
+      "step": 2420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010637860082304528,
+      "loss": 0.7713,
+      "step": 2425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010617283950617284,
+      "loss": 0.747,
+      "step": 2430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010596707818930041,
+      "loss": 0.7475,
+      "step": 2435
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010576131687242799,
+      "loss": 0.7449,
+      "step": 2440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010555555555555557,
+      "loss": 0.7208,
+      "step": 2445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010534979423868315,
+      "loss": 0.7317,
+      "step": 2450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001051440329218107,
+      "loss": 0.7419,
+      "step": 2455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010493827160493828,
+      "loss": 0.7148,
+      "step": 2460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010473251028806586,
+      "loss": 0.7143,
+      "step": 2465
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010452674897119341,
+      "loss": 0.747,
+      "step": 2470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010432098765432099,
+      "loss": 0.7079,
+      "step": 2475
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010411522633744857,
+      "loss": 0.7059,
+      "step": 2480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010390946502057615,
+      "loss": 0.7272,
+      "step": 2485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001037037037037037,
+      "loss": 0.7281,
+      "step": 2490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010349794238683128,
+      "loss": 0.6956,
+      "step": 2495
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010329218106995886,
+      "loss": 0.7571,
+      "step": 2500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010308641975308641,
+      "loss": 0.7311,
+      "step": 2505
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010288065843621399,
+      "loss": 0.6815,
+      "step": 2510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010267489711934157,
+      "loss": 0.6781,
+      "step": 2515
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010246913580246915,
+      "loss": 0.6948,
+      "step": 2520
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010226337448559671,
+      "loss": 0.7133,
+      "step": 2525
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010205761316872428,
+      "loss": 0.688,
+      "step": 2530
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010185185185185186,
+      "loss": 0.6687,
+      "step": 2535
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010164609053497942,
+      "loss": 0.7142,
+      "step": 2540
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000101440329218107,
+      "loss": 0.6885,
+      "step": 2545
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010123456790123458,
+      "loss": 0.6867,
+      "step": 2550
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010102880658436214,
+      "loss": 0.7045,
+      "step": 2555
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010082304526748971,
+      "loss": 0.7375,
+      "step": 2560
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001006172839506173,
+      "loss": 0.7157,
+      "step": 2565
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010041152263374487,
+      "loss": 0.6828,
+      "step": 2570
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010020576131687243,
+      "loss": 0.7048,
+      "step": 2575
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001,
+      "loss": 0.6944,
+      "step": 2580
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.979423868312757e-05,
+      "loss": 0.696,
+      "step": 2585
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.958847736625515e-05,
+      "loss": 0.6609,
+      "step": 2590
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.938271604938272e-05,
+      "loss": 0.674,
+      "step": 2595
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.91769547325103e-05,
+      "loss": 0.7138,
+      "step": 2600
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.897119341563786e-05,
+      "loss": 0.713,
+      "step": 2605
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.876543209876543e-05,
+      "loss": 0.7022,
+      "step": 2610
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.8559670781893e-05,
+      "loss": 0.6955,
+      "step": 2615
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.835390946502057e-05,
+      "loss": 0.6888,
+      "step": 2620
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.814814814814815e-05,
+      "loss": 0.7003,
+      "step": 2625
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.794238683127573e-05,
+      "loss": 0.7167,
+      "step": 2630
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.77366255144033e-05,
+      "loss": 0.7087,
+      "step": 2635
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.753086419753087e-05,
+      "loss": 0.7019,
+      "step": 2640
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.732510288065844e-05,
+      "loss": 0.6609,
+      "step": 2645
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.711934156378602e-05,
+      "loss": 0.7367,
+      "step": 2650
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.691358024691359e-05,
+      "loss": 0.6862,
+      "step": 2655
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.670781893004116e-05,
+      "loss": 0.7121,
+      "step": 2660
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.650205761316873e-05,
+      "loss": 0.7031,
+      "step": 2665
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.62962962962963e-05,
+      "loss": 0.7005,
+      "step": 2670
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.609053497942388e-05,
+      "loss": 0.7041,
+      "step": 2675
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.588477366255144e-05,
+      "loss": 0.6713,
+      "step": 2680
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.567901234567902e-05,
+      "loss": 0.6932,
+      "step": 2685
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.547325102880659e-05,
+      "loss": 0.6849,
+      "step": 2690
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.526748971193417e-05,
+      "loss": 0.6915,
+      "step": 2695
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.506172839506173e-05,
+      "loss": 0.6677,
+      "step": 2700
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.48559670781893e-05,
+      "loss": 0.6767,
+      "step": 2705
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.465020576131688e-05,
+      "loss": 0.6873,
+      "step": 2710
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.444444444444444e-05,
+      "loss": 0.7007,
+      "step": 2715
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.423868312757202e-05,
+      "loss": 0.6782,
+      "step": 2720
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.403292181069959e-05,
+      "loss": 0.7044,
+      "step": 2725
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.382716049382717e-05,
+      "loss": 0.6787,
+      "step": 2730
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.362139917695473e-05,
+      "loss": 0.6876,
+      "step": 2735
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.341563786008231e-05,
+      "loss": 0.7253,
+      "step": 2740
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.320987654320989e-05,
+      "loss": 0.7159,
+      "step": 2745
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.300411522633746e-05,
+      "loss": 0.704,
+      "step": 2750
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.279835390946504e-05,
+      "loss": 0.6882,
+      "step": 2755
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.25925925925926e-05,
+      "loss": 0.6777,
+      "step": 2760
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.238683127572017e-05,
+      "loss": 0.7133,
+      "step": 2765
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.218106995884775e-05,
+      "loss": 0.6988,
+      "step": 2770
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.197530864197531e-05,
+      "loss": 0.7323,
+      "step": 2775
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.176954732510289e-05,
+      "loss": 0.6854,
+      "step": 2780
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.156378600823046e-05,
+      "loss": 0.6926,
+      "step": 2785
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.135802469135802e-05,
+      "loss": 0.6985,
+      "step": 2790
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.11522633744856e-05,
+      "loss": 0.6776,
+      "step": 2795
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.094650205761317e-05,
+      "loss": 0.7374,
+      "step": 2800
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.074074074074075e-05,
+      "loss": 0.6928,
+      "step": 2805
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.053497942386831e-05,
+      "loss": 0.7129,
+      "step": 2810
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.032921810699589e-05,
+      "loss": 0.6918,
+      "step": 2815
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.012345679012346e-05,
+      "loss": 0.6641,
+      "step": 2820
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.991769547325102e-05,
+      "loss": 0.7075,
+      "step": 2825
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.97119341563786e-05,
+      "loss": 0.668,
+      "step": 2830
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.950617283950617e-05,
+      "loss": 0.6836,
+      "step": 2835
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.930041152263375e-05,
+      "loss": 0.672,
+      "step": 2840
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.909465020576133e-05,
+      "loss": 0.6827,
+      "step": 2845
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.888888888888889e-05,
+      "loss": 0.6976,
+      "step": 2850
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.868312757201647e-05,
+      "loss": 0.6719,
+      "step": 2855
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.847736625514404e-05,
+      "loss": 0.6871,
+      "step": 2860
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.827160493827162e-05,
+      "loss": 0.6975,
+      "step": 2865
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.806584362139918e-05,
+      "loss": 0.684,
+      "step": 2870
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.786008230452676e-05,
+      "loss": 0.6861,
+      "step": 2875
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.765432098765433e-05,
+      "loss": 0.7116,
+      "step": 2880
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.74485596707819e-05,
+      "loss": 0.686,
+      "step": 2885
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.724279835390947e-05,
+      "loss": 0.677,
+      "step": 2890
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.703703703703704e-05,
+      "loss": 0.6996,
+      "step": 2895
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.683127572016462e-05,
+      "loss": 0.7211,
+      "step": 2900
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.662551440329218e-05,
+      "loss": 0.6766,
+      "step": 2905
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.641975308641975e-05,
+      "loss": 0.6883,
+      "step": 2910
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.621399176954733e-05,
+      "loss": 0.6829,
+      "step": 2915
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60082304526749e-05,
+      "loss": 0.7089,
+      "step": 2920
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.580246913580247e-05,
+      "loss": 0.6711,
+      "step": 2925
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.559670781893004e-05,
+      "loss": 0.6926,
+      "step": 2930
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.539094650205762e-05,
+      "loss": 0.6513,
+      "step": 2935
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.518518518518518e-05,
+      "loss": 0.6817,
+      "step": 2940
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.497942386831276e-05,
+      "loss": 0.7005,
+      "step": 2945
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.477366255144033e-05,
+      "loss": 0.7013,
+      "step": 2950
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.456790123456791e-05,
+      "loss": 0.6562,
+      "step": 2955
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.436213991769549e-05,
+      "loss": 0.7074,
+      "step": 2960
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.415637860082305e-05,
+      "loss": 0.6901,
+      "step": 2965
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.395061728395062e-05,
+      "loss": 0.6867,
+      "step": 2970
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.37448559670782e-05,
+      "loss": 0.692,
+      "step": 2975
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.353909465020576e-05,
+      "loss": 0.6863,
+      "step": 2980
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.333333333333334e-05,
+      "loss": 0.6881,
+      "step": 2985
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.312757201646091e-05,
+      "loss": 0.7203,
+      "step": 2990
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.292181069958848e-05,
+      "loss": 0.7075,
+      "step": 2995
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.271604938271605e-05,
+      "loss": 0.7187,
+      "step": 3000
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 5010,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 500,
+  "total_flos": 1.2080552599774822e+19,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/training_args.bin b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e707059a03fbe62142ac41332e145ac1ac1f36cc
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b99a1729574e0e46ee782b42dc90101a785d5b4792871920d293f44722c28cc
+size 5944
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/zero_to_fp32.py b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/zero_to_fp32.py
new file mode 100644
index 0000000000000000000000000000000000000000..c98caae31534368be22b67fc4ae906836c992a8d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3000/zero_to_fp32.py
@@ -0,0 +1,587 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+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):
+    """
+    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)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+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 _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()
+            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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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 param_shapes.items():
+
+        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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    """
+    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``
+
+    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)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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("-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_file, tag=args.tag)
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/README.md b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..d380bbbf0bf0fa1f211b53410559607e5007a209
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: meta-llama/Llama-2-7b-hf
+---
+
+# 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 0.9.0
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/adapter_config.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c81a592f0ff08374a4f61f6003029070a8429ae
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/adapter_config.json
@@ -0,0 +1,35 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "mora_type": 1,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_mora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/adapter_model.safetensors b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/adapter_model.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..56c01ad2bac411f86a23c71238e57b117dbae71a
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/adapter_model.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f825f1c91ef88e469dc88e5c1f974f5e9863bf41613c313f3edb86f79b997d8
+size 564342240
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/global_step3500/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/global_step3500/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a1500106df4d9d6b40c10087707bfc750007695a
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/global_step3500/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:c9dd2a5a634f9136b34bafaf7ebf5ae6e1bd68bd0d6ab19c9c9f1ce2e0bf2964
+size 239914224
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/global_step3500/mp_rank_00_model_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/global_step3500/mp_rank_00_model_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..9f04d2d3589976b23f26ca624531ad5bd8191ab1
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/global_step3500/mp_rank_00_model_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce6fd22ddaaae068ca6f0d80e6862c5c460bb52fddab7d0d7c1a8bebc630eda2
+size 40303404
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/latest b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/latest
new file mode 100644
index 0000000000000000000000000000000000000000..f3af93198d9ca215e3991f85bf3781ae5dfe71f3
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/latest
@@ -0,0 +1 @@
+global_step3500
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/rng_state.pth b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/rng_state.pth
new file mode 100644
index 0000000000000000000000000000000000000000..b19f858646e4561a537cbc37fd2ba0f1ec9557d4
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/rng_state.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88769897e04293935730651f96c7d6f6c9d8615f4112479ca9a2282704b6192a
+size 14244
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/scheduler.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/scheduler.pt
new file mode 100644
index 0000000000000000000000000000000000000000..b0265f30b2483860c14312296fa7e203718592a3
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/scheduler.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e760dff0fb35054d078fc10a68a9b3f94efee1e1f7c7a5303dc355ea6cc41b4
+size 1064
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/trainer_state.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5024ff931606bcd2b2b9f378ed8d21fc375bf07
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/trainer_state.json
@@ -0,0 +1,4221 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.3970748744815542,
+  "eval_steps": 500,
+  "global_step": 3500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.3797,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.3952,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2e-05,
+      "loss": 1.3112,
+      "step": 15
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 1.3136,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 1.2012,
+      "step": 25
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4e-05,
+      "loss": 1.1277,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 1.0832,
+      "step": 35
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.333333333333333e-05,
+      "loss": 1.0659,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-05,
+      "loss": 0.983,
+      "step": 45
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.9763,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.333333333333333e-05,
+      "loss": 0.9992,
+      "step": 55
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8e-05,
+      "loss": 0.9405,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.666666666666667e-05,
+      "loss": 0.9827,
+      "step": 65
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.333333333333334e-05,
+      "loss": 0.9061,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001,
+      "loss": 0.8921,
+      "step": 75
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010666666666666667,
+      "loss": 0.9377,
+      "step": 80
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011333333333333334,
+      "loss": 0.8528,
+      "step": 85
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012,
+      "loss": 0.8913,
+      "step": 90
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012666666666666666,
+      "loss": 0.8562,
+      "step": 95
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.877,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014,
+      "loss": 0.8781,
+      "step": 105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014666666666666666,
+      "loss": 0.8546,
+      "step": 110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015333333333333334,
+      "loss": 0.8672,
+      "step": 115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016,
+      "loss": 0.8611,
+      "step": 120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.8624,
+      "step": 125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017333333333333334,
+      "loss": 0.8477,
+      "step": 130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018,
+      "loss": 0.8227,
+      "step": 135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001866666666666667,
+      "loss": 0.8064,
+      "step": 140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019333333333333333,
+      "loss": 0.8422,
+      "step": 145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002,
+      "loss": 0.8471,
+      "step": 150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001997942386831276,
+      "loss": 0.8183,
+      "step": 155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958847736625514,
+      "loss": 0.8229,
+      "step": 160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019938271604938272,
+      "loss": 0.8114,
+      "step": 165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991769547325103,
+      "loss": 0.7861,
+      "step": 170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019897119341563788,
+      "loss": 0.841,
+      "step": 175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019876543209876543,
+      "loss": 0.8549,
+      "step": 180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000198559670781893,
+      "loss": 0.8449,
+      "step": 185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001983539094650206,
+      "loss": 0.8311,
+      "step": 190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019814814814814814,
+      "loss": 0.8296,
+      "step": 195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019794238683127572,
+      "loss": 0.8263,
+      "step": 200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001977366255144033,
+      "loss": 0.8452,
+      "step": 205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019753086419753085,
+      "loss": 0.8044,
+      "step": 210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019732510288065843,
+      "loss": 0.8119,
+      "step": 215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000197119341563786,
+      "loss": 0.8127,
+      "step": 220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001969135802469136,
+      "loss": 0.7921,
+      "step": 225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019670781893004114,
+      "loss": 0.803,
+      "step": 230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019650205761316872,
+      "loss": 0.838,
+      "step": 235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001962962962962963,
+      "loss": 0.7801,
+      "step": 240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019609053497942388,
+      "loss": 0.817,
+      "step": 245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019588477366255146,
+      "loss": 0.8211,
+      "step": 250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000195679012345679,
+      "loss": 0.8423,
+      "step": 255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001954732510288066,
+      "loss": 0.8007,
+      "step": 260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019526748971193417,
+      "loss": 0.8135,
+      "step": 265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019506172839506175,
+      "loss": 0.7961,
+      "step": 270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019485596707818933,
+      "loss": 0.797,
+      "step": 275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019465020576131688,
+      "loss": 0.793,
+      "step": 280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019444444444444446,
+      "loss": 0.7809,
+      "step": 285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019423868312757204,
+      "loss": 0.7811,
+      "step": 290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019403292181069962,
+      "loss": 0.8074,
+      "step": 295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019382716049382717,
+      "loss": 0.8028,
+      "step": 300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019362139917695475,
+      "loss": 0.7895,
+      "step": 305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019341563786008233,
+      "loss": 0.7777,
+      "step": 310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019320987654320988,
+      "loss": 0.7868,
+      "step": 315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019300411522633746,
+      "loss": 0.7921,
+      "step": 320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019279835390946504,
+      "loss": 0.8,
+      "step": 325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001925925925925926,
+      "loss": 0.7781,
+      "step": 330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019238683127572017,
+      "loss": 0.7799,
+      "step": 335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019218106995884775,
+      "loss": 0.7921,
+      "step": 340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019197530864197533,
+      "loss": 0.7851,
+      "step": 345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019176954732510288,
+      "loss": 0.8106,
+      "step": 350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019156378600823046,
+      "loss": 0.7953,
+      "step": 355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019135802469135804,
+      "loss": 0.8084,
+      "step": 360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001911522633744856,
+      "loss": 0.8298,
+      "step": 365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019094650205761317,
+      "loss": 0.8273,
+      "step": 370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019074074074074075,
+      "loss": 0.8047,
+      "step": 375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019053497942386833,
+      "loss": 0.8108,
+      "step": 380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019032921810699588,
+      "loss": 0.8207,
+      "step": 385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019012345679012346,
+      "loss": 0.7884,
+      "step": 390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018991769547325104,
+      "loss": 0.7918,
+      "step": 395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001897119341563786,
+      "loss": 0.7554,
+      "step": 400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018950617283950617,
+      "loss": 0.8099,
+      "step": 405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018930041152263375,
+      "loss": 0.807,
+      "step": 410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018909465020576133,
+      "loss": 0.7793,
+      "step": 415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018888888888888888,
+      "loss": 0.7734,
+      "step": 420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018868312757201646,
+      "loss": 0.7786,
+      "step": 425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018847736625514404,
+      "loss": 0.7548,
+      "step": 430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001882716049382716,
+      "loss": 0.8024,
+      "step": 435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018806584362139917,
+      "loss": 0.7708,
+      "step": 440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018786008230452675,
+      "loss": 0.8045,
+      "step": 445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018765432098765433,
+      "loss": 0.7477,
+      "step": 450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018744855967078189,
+      "loss": 0.7474,
+      "step": 455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018724279835390946,
+      "loss": 0.7629,
+      "step": 460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018703703703703704,
+      "loss": 0.7777,
+      "step": 465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018683127572016462,
+      "loss": 0.7823,
+      "step": 470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001866255144032922,
+      "loss": 0.7762,
+      "step": 475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018641975308641978,
+      "loss": 0.782,
+      "step": 480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018621399176954733,
+      "loss": 0.7783,
+      "step": 485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001860082304526749,
+      "loss": 0.7376,
+      "step": 490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001858024691358025,
+      "loss": 0.7683,
+      "step": 495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018559670781893007,
+      "loss": 0.7547,
+      "step": 500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018539094650205762,
+      "loss": 0.7569,
+      "step": 505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001851851851851852,
+      "loss": 0.7796,
+      "step": 510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018497942386831278,
+      "loss": 0.7888,
+      "step": 515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018477366255144033,
+      "loss": 0.7763,
+      "step": 520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018456790123456791,
+      "loss": 0.7787,
+      "step": 525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001843621399176955,
+      "loss": 0.772,
+      "step": 530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018415637860082307,
+      "loss": 0.786,
+      "step": 535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018395061728395062,
+      "loss": 0.7639,
+      "step": 540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001837448559670782,
+      "loss": 0.775,
+      "step": 545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018353909465020578,
+      "loss": 0.8002,
+      "step": 550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 0.7359,
+      "step": 555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018312757201646091,
+      "loss": 0.747,
+      "step": 560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001829218106995885,
+      "loss": 0.8234,
+      "step": 565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018271604938271605,
+      "loss": 0.7586,
+      "step": 570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018251028806584363,
+      "loss": 0.7315,
+      "step": 575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001823045267489712,
+      "loss": 0.7986,
+      "step": 580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018209876543209878,
+      "loss": 0.7758,
+      "step": 585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018189300411522634,
+      "loss": 0.7869,
+      "step": 590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018168724279835392,
+      "loss": 0.7596,
+      "step": 595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001814814814814815,
+      "loss": 0.7658,
+      "step": 600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018127572016460905,
+      "loss": 0.7851,
+      "step": 605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018106995884773663,
+      "loss": 0.7957,
+      "step": 610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001808641975308642,
+      "loss": 0.7756,
+      "step": 615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018065843621399178,
+      "loss": 0.7911,
+      "step": 620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018045267489711934,
+      "loss": 0.7864,
+      "step": 625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018024691358024692,
+      "loss": 0.7683,
+      "step": 630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001800411522633745,
+      "loss": 0.7731,
+      "step": 635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017983539094650205,
+      "loss": 0.7671,
+      "step": 640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017962962962962963,
+      "loss": 0.7367,
+      "step": 645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001794238683127572,
+      "loss": 0.8024,
+      "step": 650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017921810699588476,
+      "loss": 0.7572,
+      "step": 655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017901234567901234,
+      "loss": 0.7554,
+      "step": 660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017880658436213992,
+      "loss": 0.7677,
+      "step": 665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001786008230452675,
+      "loss": 0.7652,
+      "step": 670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017839506172839508,
+      "loss": 0.7771,
+      "step": 675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017818930041152265,
+      "loss": 0.7747,
+      "step": 680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001779835390946502,
+      "loss": 0.767,
+      "step": 685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017777777777777779,
+      "loss": 0.7738,
+      "step": 690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017757201646090537,
+      "loss": 0.7271,
+      "step": 695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017736625514403294,
+      "loss": 0.7316,
+      "step": 700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017716049382716052,
+      "loss": 0.7831,
+      "step": 705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017695473251028808,
+      "loss": 0.7835,
+      "step": 710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017674897119341566,
+      "loss": 0.7497,
+      "step": 715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017654320987654323,
+      "loss": 0.7757,
+      "step": 720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001763374485596708,
+      "loss": 0.74,
+      "step": 725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017613168724279837,
+      "loss": 0.7697,
+      "step": 730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017592592592592595,
+      "loss": 0.774,
+      "step": 735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017572016460905352,
+      "loss": 0.7668,
+      "step": 740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017551440329218108,
+      "loss": 0.7779,
+      "step": 745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017530864197530866,
+      "loss": 0.7441,
+      "step": 750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017510288065843624,
+      "loss": 0.7578,
+      "step": 755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001748971193415638,
+      "loss": 0.8358,
+      "step": 760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017469135802469137,
+      "loss": 0.7473,
+      "step": 765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017448559670781895,
+      "loss": 0.7517,
+      "step": 770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001742798353909465,
+      "loss": 0.7593,
+      "step": 775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017407407407407408,
+      "loss": 0.7251,
+      "step": 780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017386831275720166,
+      "loss": 0.7365,
+      "step": 785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017366255144032924,
+      "loss": 0.7798,
+      "step": 790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001734567901234568,
+      "loss": 0.746,
+      "step": 795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017325102880658437,
+      "loss": 0.7607,
+      "step": 800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017304526748971195,
+      "loss": 0.7694,
+      "step": 805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001728395061728395,
+      "loss": 0.7926,
+      "step": 810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017263374485596708,
+      "loss": 0.795,
+      "step": 815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017242798353909466,
+      "loss": 0.7581,
+      "step": 820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017222222222222224,
+      "loss": 0.7629,
+      "step": 825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001720164609053498,
+      "loss": 0.7526,
+      "step": 830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017181069958847737,
+      "loss": 0.7338,
+      "step": 835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017160493827160495,
+      "loss": 0.7544,
+      "step": 840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001713991769547325,
+      "loss": 0.7522,
+      "step": 845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017119341563786008,
+      "loss": 0.7639,
+      "step": 850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017098765432098766,
+      "loss": 0.7914,
+      "step": 855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017078189300411524,
+      "loss": 0.7635,
+      "step": 860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001705761316872428,
+      "loss": 0.8343,
+      "step": 865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017037037037037037,
+      "loss": 0.7299,
+      "step": 870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017016460905349795,
+      "loss": 0.7643,
+      "step": 875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016995884773662553,
+      "loss": 0.7343,
+      "step": 880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016975308641975308,
+      "loss": 0.7546,
+      "step": 885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016954732510288066,
+      "loss": 0.7435,
+      "step": 890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016934156378600824,
+      "loss": 0.8086,
+      "step": 895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016913580246913582,
+      "loss": 0.7274,
+      "step": 900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001689300411522634,
+      "loss": 0.7511,
+      "step": 905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016872427983539098,
+      "loss": 0.7689,
+      "step": 910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016851851851851853,
+      "loss": 0.754,
+      "step": 915
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001683127572016461,
+      "loss": 0.7298,
+      "step": 920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001681069958847737,
+      "loss": 0.7755,
+      "step": 925
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016790123456790124,
+      "loss": 0.7854,
+      "step": 930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016769547325102882,
+      "loss": 0.7452,
+      "step": 935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001674897119341564,
+      "loss": 0.7473,
+      "step": 940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016728395061728398,
+      "loss": 0.7855,
+      "step": 945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016707818930041153,
+      "loss": 0.7527,
+      "step": 950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001668724279835391,
+      "loss": 0.7641,
+      "step": 955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.828,
+      "step": 960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016646090534979424,
+      "loss": 0.7571,
+      "step": 965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016625514403292182,
+      "loss": 0.7768,
+      "step": 970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001660493827160494,
+      "loss": 0.7644,
+      "step": 975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016584362139917695,
+      "loss": 0.7747,
+      "step": 980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016563786008230453,
+      "loss": 0.7926,
+      "step": 985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001654320987654321,
+      "loss": 0.7645,
+      "step": 990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001652263374485597,
+      "loss": 0.7464,
+      "step": 995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016502057613168724,
+      "loss": 0.7558,
+      "step": 1000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016481481481481482,
+      "loss": 0.7574,
+      "step": 1005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001646090534979424,
+      "loss": 0.7491,
+      "step": 1010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016440329218106995,
+      "loss": 0.756,
+      "step": 1015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016419753086419753,
+      "loss": 0.7679,
+      "step": 1020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001639917695473251,
+      "loss": 0.776,
+      "step": 1025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001637860082304527,
+      "loss": 0.7235,
+      "step": 1030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016358024691358024,
+      "loss": 0.7817,
+      "step": 1035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016337448559670782,
+      "loss": 0.7333,
+      "step": 1040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001631687242798354,
+      "loss": 0.754,
+      "step": 1045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016296296296296295,
+      "loss": 0.74,
+      "step": 1050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016275720164609053,
+      "loss": 0.7857,
+      "step": 1055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001625514403292181,
+      "loss": 0.8015,
+      "step": 1060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001623456790123457,
+      "loss": 0.7508,
+      "step": 1065
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016213991769547324,
+      "loss": 0.7514,
+      "step": 1070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016193415637860082,
+      "loss": 0.7326,
+      "step": 1075
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001617283950617284,
+      "loss": 0.7322,
+      "step": 1080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016152263374485595,
+      "loss": 0.7373,
+      "step": 1085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016131687242798353,
+      "loss": 0.7637,
+      "step": 1090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001611111111111111,
+      "loss": 0.7697,
+      "step": 1095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001609053497942387,
+      "loss": 0.7505,
+      "step": 1100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016069958847736627,
+      "loss": 0.7568,
+      "step": 1105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016049382716049385,
+      "loss": 0.786,
+      "step": 1110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001602880658436214,
+      "loss": 0.7525,
+      "step": 1115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016008230452674898,
+      "loss": 0.7696,
+      "step": 1120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015987654320987656,
+      "loss": 0.7648,
+      "step": 1125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015967078189300414,
+      "loss": 0.7492,
+      "step": 1130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001594650205761317,
+      "loss": 0.809,
+      "step": 1135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015925925925925927,
+      "loss": 0.7592,
+      "step": 1140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015905349794238685,
+      "loss": 0.7874,
+      "step": 1145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015884773662551443,
+      "loss": 0.7874,
+      "step": 1150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015864197530864198,
+      "loss": 0.7374,
+      "step": 1155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015843621399176956,
+      "loss": 0.7218,
+      "step": 1160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015823045267489714,
+      "loss": 0.7494,
+      "step": 1165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001580246913580247,
+      "loss": 0.7449,
+      "step": 1170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015781893004115227,
+      "loss": 0.725,
+      "step": 1175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015761316872427985,
+      "loss": 0.739,
+      "step": 1180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015740740740740743,
+      "loss": 0.7554,
+      "step": 1185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015720164609053498,
+      "loss": 0.7702,
+      "step": 1190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015699588477366256,
+      "loss": 0.7418,
+      "step": 1195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015679012345679014,
+      "loss": 0.7217,
+      "step": 1200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001565843621399177,
+      "loss": 0.7391,
+      "step": 1205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015637860082304527,
+      "loss": 0.7696,
+      "step": 1210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015617283950617285,
+      "loss": 0.7594,
+      "step": 1215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001559670781893004,
+      "loss": 0.777,
+      "step": 1220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015576131687242798,
+      "loss": 0.7441,
+      "step": 1225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015555555555555556,
+      "loss": 0.7611,
+      "step": 1230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015534979423868314,
+      "loss": 0.7278,
+      "step": 1235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001551440329218107,
+      "loss": 0.7452,
+      "step": 1240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015493827160493827,
+      "loss": 0.7939,
+      "step": 1245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015473251028806585,
+      "loss": 0.7161,
+      "step": 1250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001545267489711934,
+      "loss": 0.7344,
+      "step": 1255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015432098765432098,
+      "loss": 0.7368,
+      "step": 1260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015411522633744856,
+      "loss": 0.7172,
+      "step": 1265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015390946502057614,
+      "loss": 0.7589,
+      "step": 1270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001537037037037037,
+      "loss": 0.7641,
+      "step": 1275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015349794238683127,
+      "loss": 0.761,
+      "step": 1280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015329218106995885,
+      "loss": 0.7414,
+      "step": 1285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001530864197530864,
+      "loss": 0.7713,
+      "step": 1290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015288065843621399,
+      "loss": 0.7543,
+      "step": 1295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015267489711934156,
+      "loss": 0.7683,
+      "step": 1300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015246913580246914,
+      "loss": 0.8106,
+      "step": 1305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015226337448559672,
+      "loss": 0.7739,
+      "step": 1310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015205761316872428,
+      "loss": 0.7585,
+      "step": 1315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015185185185185185,
+      "loss": 0.7577,
+      "step": 1320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015164609053497943,
+      "loss": 0.7608,
+      "step": 1325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000151440329218107,
+      "loss": 0.7715,
+      "step": 1330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001512345679012346,
+      "loss": 0.7681,
+      "step": 1335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015102880658436214,
+      "loss": 0.7476,
+      "step": 1340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015082304526748972,
+      "loss": 0.7539,
+      "step": 1345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001506172839506173,
+      "loss": 0.7492,
+      "step": 1350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015041152263374488,
+      "loss": 0.7491,
+      "step": 1355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015020576131687243,
+      "loss": 0.7236,
+      "step": 1360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 0.7328,
+      "step": 1365
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001497942386831276,
+      "loss": 0.7603,
+      "step": 1370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014958847736625515,
+      "loss": 0.7206,
+      "step": 1375
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014938271604938272,
+      "loss": 0.7001,
+      "step": 1380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001491769547325103,
+      "loss": 0.7241,
+      "step": 1385
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014897119341563788,
+      "loss": 0.7511,
+      "step": 1390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014876543209876544,
+      "loss": 0.7546,
+      "step": 1395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014855967078189301,
+      "loss": 0.762,
+      "step": 1400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001483539094650206,
+      "loss": 0.6979,
+      "step": 1405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014814814814814815,
+      "loss": 0.7352,
+      "step": 1410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014794238683127573,
+      "loss": 0.7604,
+      "step": 1415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001477366255144033,
+      "loss": 0.7612,
+      "step": 1420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014753086419753086,
+      "loss": 0.7618,
+      "step": 1425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014732510288065844,
+      "loss": 0.7107,
+      "step": 1430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014711934156378602,
+      "loss": 0.7247,
+      "step": 1435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001469135802469136,
+      "loss": 0.7733,
+      "step": 1440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014670781893004115,
+      "loss": 0.7627,
+      "step": 1445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014650205761316873,
+      "loss": 0.7323,
+      "step": 1450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001462962962962963,
+      "loss": 0.7728,
+      "step": 1455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014609053497942386,
+      "loss": 0.7541,
+      "step": 1460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014588477366255144,
+      "loss": 0.7781,
+      "step": 1465
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014567901234567902,
+      "loss": 0.7474,
+      "step": 1470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001454732510288066,
+      "loss": 0.7789,
+      "step": 1475
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014526748971193415,
+      "loss": 0.7497,
+      "step": 1480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014506172839506173,
+      "loss": 0.7487,
+      "step": 1485
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001448559670781893,
+      "loss": 0.7563,
+      "step": 1490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014465020576131686,
+      "loss": 0.7444,
+      "step": 1495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014444444444444444,
+      "loss": 0.7399,
+      "step": 1500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014423868312757202,
+      "loss": 0.7288,
+      "step": 1505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001440329218106996,
+      "loss": 0.7458,
+      "step": 1510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014382716049382718,
+      "loss": 0.7297,
+      "step": 1515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014362139917695473,
+      "loss": 0.7684,
+      "step": 1520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001434156378600823,
+      "loss": 0.7428,
+      "step": 1525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014320987654320989,
+      "loss": 0.705,
+      "step": 1530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014300411522633747,
+      "loss": 0.7445,
+      "step": 1535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014279835390946504,
+      "loss": 0.7109,
+      "step": 1540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001425925925925926,
+      "loss": 0.7502,
+      "step": 1545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014238683127572018,
+      "loss": 0.7485,
+      "step": 1550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014218106995884776,
+      "loss": 0.7421,
+      "step": 1555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014197530864197534,
+      "loss": 0.7369,
+      "step": 1560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001417695473251029,
+      "loss": 0.7746,
+      "step": 1565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014156378600823047,
+      "loss": 0.7334,
+      "step": 1570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014135802469135805,
+      "loss": 0.7251,
+      "step": 1575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001411522633744856,
+      "loss": 0.7146,
+      "step": 1580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014094650205761318,
+      "loss": 0.7563,
+      "step": 1585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014074074074074076,
+      "loss": 0.7247,
+      "step": 1590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014053497942386834,
+      "loss": 0.7379,
+      "step": 1595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001403292181069959,
+      "loss": 0.7684,
+      "step": 1600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014012345679012347,
+      "loss": 0.7574,
+      "step": 1605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00013991769547325105,
+      "loss": 0.7138,
+      "step": 1610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001397119341563786,
+      "loss": 0.7472,
+      "step": 1615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013950617283950618,
+      "loss": 0.7466,
+      "step": 1620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013930041152263376,
+      "loss": 0.7288,
+      "step": 1625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013909465020576134,
+      "loss": 0.7717,
+      "step": 1630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001388888888888889,
+      "loss": 0.7665,
+      "step": 1635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013868312757201647,
+      "loss": 0.7408,
+      "step": 1640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013847736625514405,
+      "loss": 0.7418,
+      "step": 1645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001382716049382716,
+      "loss": 0.7502,
+      "step": 1650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013806584362139918,
+      "loss": 0.7657,
+      "step": 1655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013786008230452676,
+      "loss": 0.728,
+      "step": 1660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001376543209876543,
+      "loss": 0.7142,
+      "step": 1665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001374485596707819,
+      "loss": 0.7971,
+      "step": 1670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013724279835390947,
+      "loss": 0.737,
+      "step": 1675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013703703703703705,
+      "loss": 0.7947,
+      "step": 1680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001368312757201646,
+      "loss": 0.7525,
+      "step": 1685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013662551440329218,
+      "loss": 0.711,
+      "step": 1690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013641975308641976,
+      "loss": 0.7459,
+      "step": 1695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001362139917695473,
+      "loss": 0.7404,
+      "step": 1700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001360082304526749,
+      "loss": 0.7458,
+      "step": 1705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013580246913580247,
+      "loss": 0.754,
+      "step": 1710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013559670781893005,
+      "loss": 0.7217,
+      "step": 1715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001353909465020576,
+      "loss": 0.7373,
+      "step": 1720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013518518518518518,
+      "loss": 0.7567,
+      "step": 1725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013497942386831276,
+      "loss": 0.7525,
+      "step": 1730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013477366255144034,
+      "loss": 0.752,
+      "step": 1735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013456790123456792,
+      "loss": 0.7342,
+      "step": 1740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001343621399176955,
+      "loss": 0.7276,
+      "step": 1745
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013415637860082305,
+      "loss": 0.6915,
+      "step": 1750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013395061728395063,
+      "loss": 0.737,
+      "step": 1755
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001337448559670782,
+      "loss": 0.7427,
+      "step": 1760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001335390946502058,
+      "loss": 0.7126,
+      "step": 1765
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.7241,
+      "step": 1770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013312757201646092,
+      "loss": 0.755,
+      "step": 1775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001329218106995885,
+      "loss": 0.7494,
+      "step": 1780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013271604938271605,
+      "loss": 0.709,
+      "step": 1785
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013251028806584363,
+      "loss": 0.7555,
+      "step": 1790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001323045267489712,
+      "loss": 0.7298,
+      "step": 1795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001320987654320988,
+      "loss": 0.7232,
+      "step": 1800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013189300411522634,
+      "loss": 0.7441,
+      "step": 1805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013168724279835392,
+      "loss": 0.7623,
+      "step": 1810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001314814814814815,
+      "loss": 0.7349,
+      "step": 1815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013127572016460905,
+      "loss": 0.7773,
+      "step": 1820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013106995884773663,
+      "loss": 0.7537,
+      "step": 1825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001308641975308642,
+      "loss": 0.7579,
+      "step": 1830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001306584362139918,
+      "loss": 0.7321,
+      "step": 1835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013045267489711934,
+      "loss": 0.7175,
+      "step": 1840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00013024691358024692,
+      "loss": 0.7553,
+      "step": 1845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001300411522633745,
+      "loss": 0.7591,
+      "step": 1850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012983539094650205,
+      "loss": 0.7581,
+      "step": 1855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012962962962962963,
+      "loss": 0.7632,
+      "step": 1860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001294238683127572,
+      "loss": 0.7434,
+      "step": 1865
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012921810699588476,
+      "loss": 0.7695,
+      "step": 1870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012901234567901234,
+      "loss": 0.7808,
+      "step": 1875
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012880658436213992,
+      "loss": 0.7428,
+      "step": 1880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001286008230452675,
+      "loss": 0.7487,
+      "step": 1885
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012839506172839505,
+      "loss": 0.7033,
+      "step": 1890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012818930041152263,
+      "loss": 0.7639,
+      "step": 1895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001279835390946502,
+      "loss": 0.7345,
+      "step": 1900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012777777777777776,
+      "loss": 0.7434,
+      "step": 1905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012757201646090534,
+      "loss": 0.7082,
+      "step": 1910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012736625514403292,
+      "loss": 0.7294,
+      "step": 1915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001271604938271605,
+      "loss": 0.74,
+      "step": 1920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012695473251028805,
+      "loss": 0.7694,
+      "step": 1925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012674897119341563,
+      "loss": 0.7339,
+      "step": 1930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001265432098765432,
+      "loss": 0.7259,
+      "step": 1935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001263374485596708,
+      "loss": 0.7307,
+      "step": 1940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012613168724279837,
+      "loss": 0.7085,
+      "step": 1945
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012592592592592592,
+      "loss": 0.7662,
+      "step": 1950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001257201646090535,
+      "loss": 0.7145,
+      "step": 1955
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012551440329218108,
+      "loss": 0.7528,
+      "step": 1960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012530864197530866,
+      "loss": 0.7308,
+      "step": 1965
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012510288065843624,
+      "loss": 0.7163,
+      "step": 1970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001248971193415638,
+      "loss": 0.7254,
+      "step": 1975
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012469135802469137,
+      "loss": 0.7467,
+      "step": 1980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012448559670781895,
+      "loss": 0.7391,
+      "step": 1985
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001242798353909465,
+      "loss": 0.7314,
+      "step": 1990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012407407407407408,
+      "loss": 0.7048,
+      "step": 1995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012386831275720166,
+      "loss": 0.7187,
+      "step": 2000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012366255144032924,
+      "loss": 0.735,
+      "step": 2005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001234567901234568,
+      "loss": 0.7662,
+      "step": 2010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012325102880658437,
+      "loss": 0.7437,
+      "step": 2015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012304526748971195,
+      "loss": 0.7675,
+      "step": 2020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001228395061728395,
+      "loss": 0.7143,
+      "step": 2025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012263374485596708,
+      "loss": 0.7101,
+      "step": 2030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012242798353909466,
+      "loss": 0.7257,
+      "step": 2035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012222222222222224,
+      "loss": 0.7522,
+      "step": 2040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001220164609053498,
+      "loss": 0.7669,
+      "step": 2045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012181069958847737,
+      "loss": 0.7283,
+      "step": 2050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012160493827160495,
+      "loss": 0.7413,
+      "step": 2055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001213991769547325,
+      "loss": 0.6988,
+      "step": 2060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012119341563786008,
+      "loss": 0.7421,
+      "step": 2065
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012098765432098766,
+      "loss": 0.7152,
+      "step": 2070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012078189300411524,
+      "loss": 0.7741,
+      "step": 2075
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001205761316872428,
+      "loss": 0.7381,
+      "step": 2080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012037037037037037,
+      "loss": 0.738,
+      "step": 2085
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012016460905349795,
+      "loss": 0.7258,
+      "step": 2090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011995884773662552,
+      "loss": 0.7426,
+      "step": 2095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011975308641975308,
+      "loss": 0.7185,
+      "step": 2100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011954732510288066,
+      "loss": 0.7401,
+      "step": 2105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011934156378600823,
+      "loss": 0.751,
+      "step": 2110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011913580246913581,
+      "loss": 0.7284,
+      "step": 2115
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011893004115226339,
+      "loss": 0.7901,
+      "step": 2120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011872427983539097,
+      "loss": 0.7432,
+      "step": 2125
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011851851851851852,
+      "loss": 0.7355,
+      "step": 2130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001183127572016461,
+      "loss": 0.7333,
+      "step": 2135
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011810699588477368,
+      "loss": 0.7504,
+      "step": 2140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011790123456790123,
+      "loss": 0.7503,
+      "step": 2145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011769547325102881,
+      "loss": 0.728,
+      "step": 2150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011748971193415639,
+      "loss": 0.7485,
+      "step": 2155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011728395061728397,
+      "loss": 0.7127,
+      "step": 2160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011707818930041152,
+      "loss": 0.7278,
+      "step": 2165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001168724279835391,
+      "loss": 0.7513,
+      "step": 2170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011666666666666668,
+      "loss": 0.7193,
+      "step": 2175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011646090534979423,
+      "loss": 0.7245,
+      "step": 2180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011625514403292181,
+      "loss": 0.7266,
+      "step": 2185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011604938271604939,
+      "loss": 0.7531,
+      "step": 2190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011584362139917696,
+      "loss": 0.7637,
+      "step": 2195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011563786008230452,
+      "loss": 0.7388,
+      "step": 2200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001154320987654321,
+      "loss": 0.7222,
+      "step": 2205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011522633744855968,
+      "loss": 0.7667,
+      "step": 2210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011502057613168725,
+      "loss": 0.7382,
+      "step": 2215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011481481481481482,
+      "loss": 0.7243,
+      "step": 2220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001146090534979424,
+      "loss": 0.7192,
+      "step": 2225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011440329218106996,
+      "loss": 0.7237,
+      "step": 2230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011419753086419754,
+      "loss": 0.7498,
+      "step": 2235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011399176954732511,
+      "loss": 0.7007,
+      "step": 2240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001137860082304527,
+      "loss": 0.758,
+      "step": 2245
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011358024691358025,
+      "loss": 0.7111,
+      "step": 2250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011337448559670783,
+      "loss": 0.706,
+      "step": 2255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001131687242798354,
+      "loss": 0.7423,
+      "step": 2260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011296296296296296,
+      "loss": 0.7094,
+      "step": 2265
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011275720164609054,
+      "loss": 0.7331,
+      "step": 2270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011255144032921812,
+      "loss": 0.7164,
+      "step": 2275
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001123456790123457,
+      "loss": 0.7452,
+      "step": 2280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011213991769547325,
+      "loss": 0.7172,
+      "step": 2285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011193415637860083,
+      "loss": 0.7232,
+      "step": 2290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001117283950617284,
+      "loss": 0.7112,
+      "step": 2295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011152263374485596,
+      "loss": 0.7459,
+      "step": 2300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011131687242798354,
+      "loss": 0.7353,
+      "step": 2305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011111111111111112,
+      "loss": 0.7406,
+      "step": 2310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011090534979423868,
+      "loss": 0.7623,
+      "step": 2315
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011069958847736626,
+      "loss": 0.7567,
+      "step": 2320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011049382716049384,
+      "loss": 0.7752,
+      "step": 2325
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001102880658436214,
+      "loss": 0.7419,
+      "step": 2330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011008230452674897,
+      "loss": 0.7244,
+      "step": 2335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00010987654320987655,
+      "loss": 0.7553,
+      "step": 2340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010967078189300413,
+      "loss": 0.7389,
+      "step": 2345
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010946502057613168,
+      "loss": 0.7002,
+      "step": 2350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010925925925925926,
+      "loss": 0.7238,
+      "step": 2355
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010905349794238684,
+      "loss": 0.6923,
+      "step": 2360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010884773662551442,
+      "loss": 0.7579,
+      "step": 2365
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010864197530864197,
+      "loss": 0.7291,
+      "step": 2370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010843621399176955,
+      "loss": 0.721,
+      "step": 2375
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010823045267489713,
+      "loss": 0.7605,
+      "step": 2380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010802469135802468,
+      "loss": 0.7297,
+      "step": 2385
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010781893004115226,
+      "loss": 0.7565,
+      "step": 2390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010761316872427984,
+      "loss": 0.701,
+      "step": 2395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010740740740740742,
+      "loss": 0.736,
+      "step": 2400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010720164609053497,
+      "loss": 0.7079,
+      "step": 2405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010699588477366255,
+      "loss": 0.7349,
+      "step": 2410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010679012345679013,
+      "loss": 0.7106,
+      "step": 2415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001065843621399177,
+      "loss": 0.7396,
+      "step": 2420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010637860082304528,
+      "loss": 0.7713,
+      "step": 2425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010617283950617284,
+      "loss": 0.747,
+      "step": 2430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010596707818930041,
+      "loss": 0.7475,
+      "step": 2435
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010576131687242799,
+      "loss": 0.7449,
+      "step": 2440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010555555555555557,
+      "loss": 0.7208,
+      "step": 2445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010534979423868315,
+      "loss": 0.7317,
+      "step": 2450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001051440329218107,
+      "loss": 0.7419,
+      "step": 2455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010493827160493828,
+      "loss": 0.7148,
+      "step": 2460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010473251028806586,
+      "loss": 0.7143,
+      "step": 2465
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010452674897119341,
+      "loss": 0.747,
+      "step": 2470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010432098765432099,
+      "loss": 0.7079,
+      "step": 2475
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010411522633744857,
+      "loss": 0.7059,
+      "step": 2480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010390946502057615,
+      "loss": 0.7272,
+      "step": 2485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001037037037037037,
+      "loss": 0.7281,
+      "step": 2490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010349794238683128,
+      "loss": 0.6956,
+      "step": 2495
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010329218106995886,
+      "loss": 0.7571,
+      "step": 2500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010308641975308641,
+      "loss": 0.7311,
+      "step": 2505
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010288065843621399,
+      "loss": 0.6815,
+      "step": 2510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010267489711934157,
+      "loss": 0.6781,
+      "step": 2515
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010246913580246915,
+      "loss": 0.6948,
+      "step": 2520
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010226337448559671,
+      "loss": 0.7133,
+      "step": 2525
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010205761316872428,
+      "loss": 0.688,
+      "step": 2530
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010185185185185186,
+      "loss": 0.6687,
+      "step": 2535
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010164609053497942,
+      "loss": 0.7142,
+      "step": 2540
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000101440329218107,
+      "loss": 0.6885,
+      "step": 2545
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010123456790123458,
+      "loss": 0.6867,
+      "step": 2550
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010102880658436214,
+      "loss": 0.7045,
+      "step": 2555
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010082304526748971,
+      "loss": 0.7375,
+      "step": 2560
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001006172839506173,
+      "loss": 0.7157,
+      "step": 2565
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010041152263374487,
+      "loss": 0.6828,
+      "step": 2570
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010020576131687243,
+      "loss": 0.7048,
+      "step": 2575
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001,
+      "loss": 0.6944,
+      "step": 2580
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.979423868312757e-05,
+      "loss": 0.696,
+      "step": 2585
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.958847736625515e-05,
+      "loss": 0.6609,
+      "step": 2590
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.938271604938272e-05,
+      "loss": 0.674,
+      "step": 2595
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.91769547325103e-05,
+      "loss": 0.7138,
+      "step": 2600
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.897119341563786e-05,
+      "loss": 0.713,
+      "step": 2605
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.876543209876543e-05,
+      "loss": 0.7022,
+      "step": 2610
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.8559670781893e-05,
+      "loss": 0.6955,
+      "step": 2615
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.835390946502057e-05,
+      "loss": 0.6888,
+      "step": 2620
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.814814814814815e-05,
+      "loss": 0.7003,
+      "step": 2625
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.794238683127573e-05,
+      "loss": 0.7167,
+      "step": 2630
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.77366255144033e-05,
+      "loss": 0.7087,
+      "step": 2635
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.753086419753087e-05,
+      "loss": 0.7019,
+      "step": 2640
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.732510288065844e-05,
+      "loss": 0.6609,
+      "step": 2645
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.711934156378602e-05,
+      "loss": 0.7367,
+      "step": 2650
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.691358024691359e-05,
+      "loss": 0.6862,
+      "step": 2655
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.670781893004116e-05,
+      "loss": 0.7121,
+      "step": 2660
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.650205761316873e-05,
+      "loss": 0.7031,
+      "step": 2665
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.62962962962963e-05,
+      "loss": 0.7005,
+      "step": 2670
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.609053497942388e-05,
+      "loss": 0.7041,
+      "step": 2675
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.588477366255144e-05,
+      "loss": 0.6713,
+      "step": 2680
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.567901234567902e-05,
+      "loss": 0.6932,
+      "step": 2685
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.547325102880659e-05,
+      "loss": 0.6849,
+      "step": 2690
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.526748971193417e-05,
+      "loss": 0.6915,
+      "step": 2695
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.506172839506173e-05,
+      "loss": 0.6677,
+      "step": 2700
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.48559670781893e-05,
+      "loss": 0.6767,
+      "step": 2705
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.465020576131688e-05,
+      "loss": 0.6873,
+      "step": 2710
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.444444444444444e-05,
+      "loss": 0.7007,
+      "step": 2715
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.423868312757202e-05,
+      "loss": 0.6782,
+      "step": 2720
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.403292181069959e-05,
+      "loss": 0.7044,
+      "step": 2725
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.382716049382717e-05,
+      "loss": 0.6787,
+      "step": 2730
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.362139917695473e-05,
+      "loss": 0.6876,
+      "step": 2735
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.341563786008231e-05,
+      "loss": 0.7253,
+      "step": 2740
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.320987654320989e-05,
+      "loss": 0.7159,
+      "step": 2745
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.300411522633746e-05,
+      "loss": 0.704,
+      "step": 2750
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.279835390946504e-05,
+      "loss": 0.6882,
+      "step": 2755
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.25925925925926e-05,
+      "loss": 0.6777,
+      "step": 2760
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.238683127572017e-05,
+      "loss": 0.7133,
+      "step": 2765
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.218106995884775e-05,
+      "loss": 0.6988,
+      "step": 2770
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.197530864197531e-05,
+      "loss": 0.7323,
+      "step": 2775
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.176954732510289e-05,
+      "loss": 0.6854,
+      "step": 2780
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.156378600823046e-05,
+      "loss": 0.6926,
+      "step": 2785
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.135802469135802e-05,
+      "loss": 0.6985,
+      "step": 2790
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.11522633744856e-05,
+      "loss": 0.6776,
+      "step": 2795
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.094650205761317e-05,
+      "loss": 0.7374,
+      "step": 2800
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.074074074074075e-05,
+      "loss": 0.6928,
+      "step": 2805
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.053497942386831e-05,
+      "loss": 0.7129,
+      "step": 2810
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.032921810699589e-05,
+      "loss": 0.6918,
+      "step": 2815
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.012345679012346e-05,
+      "loss": 0.6641,
+      "step": 2820
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.991769547325102e-05,
+      "loss": 0.7075,
+      "step": 2825
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.97119341563786e-05,
+      "loss": 0.668,
+      "step": 2830
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.950617283950617e-05,
+      "loss": 0.6836,
+      "step": 2835
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.930041152263375e-05,
+      "loss": 0.672,
+      "step": 2840
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.909465020576133e-05,
+      "loss": 0.6827,
+      "step": 2845
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.888888888888889e-05,
+      "loss": 0.6976,
+      "step": 2850
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.868312757201647e-05,
+      "loss": 0.6719,
+      "step": 2855
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.847736625514404e-05,
+      "loss": 0.6871,
+      "step": 2860
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.827160493827162e-05,
+      "loss": 0.6975,
+      "step": 2865
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.806584362139918e-05,
+      "loss": 0.684,
+      "step": 2870
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.786008230452676e-05,
+      "loss": 0.6861,
+      "step": 2875
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.765432098765433e-05,
+      "loss": 0.7116,
+      "step": 2880
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.74485596707819e-05,
+      "loss": 0.686,
+      "step": 2885
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.724279835390947e-05,
+      "loss": 0.677,
+      "step": 2890
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.703703703703704e-05,
+      "loss": 0.6996,
+      "step": 2895
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.683127572016462e-05,
+      "loss": 0.7211,
+      "step": 2900
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.662551440329218e-05,
+      "loss": 0.6766,
+      "step": 2905
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.641975308641975e-05,
+      "loss": 0.6883,
+      "step": 2910
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.621399176954733e-05,
+      "loss": 0.6829,
+      "step": 2915
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60082304526749e-05,
+      "loss": 0.7089,
+      "step": 2920
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.580246913580247e-05,
+      "loss": 0.6711,
+      "step": 2925
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.559670781893004e-05,
+      "loss": 0.6926,
+      "step": 2930
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.539094650205762e-05,
+      "loss": 0.6513,
+      "step": 2935
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.518518518518518e-05,
+      "loss": 0.6817,
+      "step": 2940
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.497942386831276e-05,
+      "loss": 0.7005,
+      "step": 2945
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.477366255144033e-05,
+      "loss": 0.7013,
+      "step": 2950
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.456790123456791e-05,
+      "loss": 0.6562,
+      "step": 2955
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.436213991769549e-05,
+      "loss": 0.7074,
+      "step": 2960
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.415637860082305e-05,
+      "loss": 0.6901,
+      "step": 2965
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.395061728395062e-05,
+      "loss": 0.6867,
+      "step": 2970
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.37448559670782e-05,
+      "loss": 0.692,
+      "step": 2975
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.353909465020576e-05,
+      "loss": 0.6863,
+      "step": 2980
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.333333333333334e-05,
+      "loss": 0.6881,
+      "step": 2985
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.312757201646091e-05,
+      "loss": 0.7203,
+      "step": 2990
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.292181069958848e-05,
+      "loss": 0.7075,
+      "step": 2995
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.271604938271605e-05,
+      "loss": 0.7187,
+      "step": 3000
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.251028806584362e-05,
+      "loss": 0.6791,
+      "step": 3005
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.23045267489712e-05,
+      "loss": 0.6937,
+      "step": 3010
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.209876543209877e-05,
+      "loss": 0.6584,
+      "step": 3015
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.189300411522634e-05,
+      "loss": 0.6818,
+      "step": 3020
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.168724279835391e-05,
+      "loss": 0.6827,
+      "step": 3025
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.148148148148148e-05,
+      "loss": 0.6774,
+      "step": 3030
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.127572016460906e-05,
+      "loss": 0.6826,
+      "step": 3035
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.106995884773662e-05,
+      "loss": 0.6767,
+      "step": 3040
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.08641975308642e-05,
+      "loss": 0.7256,
+      "step": 3045
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.065843621399177e-05,
+      "loss": 0.6875,
+      "step": 3050
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.045267489711935e-05,
+      "loss": 0.7013,
+      "step": 3055
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.024691358024692e-05,
+      "loss": 0.6907,
+      "step": 3060
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.004115226337449e-05,
+      "loss": 0.6751,
+      "step": 3065
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.983539094650207e-05,
+      "loss": 0.6893,
+      "step": 3070
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.962962962962964e-05,
+      "loss": 0.6962,
+      "step": 3075
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.942386831275721e-05,
+      "loss": 0.6743,
+      "step": 3080
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.921810699588478e-05,
+      "loss": 0.7004,
+      "step": 3085
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.901234567901235e-05,
+      "loss": 0.6908,
+      "step": 3090
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.880658436213993e-05,
+      "loss": 0.6736,
+      "step": 3095
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.860082304526749e-05,
+      "loss": 0.7153,
+      "step": 3100
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.839506172839507e-05,
+      "loss": 0.7072,
+      "step": 3105
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.818930041152264e-05,
+      "loss": 0.7163,
+      "step": 3110
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.79835390946502e-05,
+      "loss": 0.6996,
+      "step": 3115
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.777777777777778e-05,
+      "loss": 0.7257,
+      "step": 3120
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.757201646090535e-05,
+      "loss": 0.7155,
+      "step": 3125
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.736625514403293e-05,
+      "loss": 0.7068,
+      "step": 3130
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.716049382716049e-05,
+      "loss": 0.6575,
+      "step": 3135
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.695473251028807e-05,
+      "loss": 0.6771,
+      "step": 3140
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.674897119341564e-05,
+      "loss": 0.6976,
+      "step": 3145
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.65432098765432e-05,
+      "loss": 0.6568,
+      "step": 3150
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.633744855967078e-05,
+      "loss": 0.6609,
+      "step": 3155
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.613168724279836e-05,
+      "loss": 0.697,
+      "step": 3160
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.592592592592593e-05,
+      "loss": 0.714,
+      "step": 3165
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.57201646090535e-05,
+      "loss": 0.6829,
+      "step": 3170
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.551440329218107e-05,
+      "loss": 0.6932,
+      "step": 3175
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.530864197530865e-05,
+      "loss": 0.6859,
+      "step": 3180
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.510288065843622e-05,
+      "loss": 0.7117,
+      "step": 3185
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.48971193415638e-05,
+      "loss": 0.7157,
+      "step": 3190
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.469135802469136e-05,
+      "loss": 0.6593,
+      "step": 3195
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.448559670781894e-05,
+      "loss": 0.6683,
+      "step": 3200
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.427983539094651e-05,
+      "loss": 0.6876,
+      "step": 3205
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.407407407407407e-05,
+      "loss": 0.6584,
+      "step": 3210
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.386831275720165e-05,
+      "loss": 0.6816,
+      "step": 3215
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.366255144032922e-05,
+      "loss": 0.6994,
+      "step": 3220
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.34567901234568e-05,
+      "loss": 0.6701,
+      "step": 3225
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.325102880658436e-05,
+      "loss": 0.6869,
+      "step": 3230
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.304526748971193e-05,
+      "loss": 0.6895,
+      "step": 3235
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.283950617283951e-05,
+      "loss": 0.6972,
+      "step": 3240
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.263374485596707e-05,
+      "loss": 0.6849,
+      "step": 3245
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.242798353909465e-05,
+      "loss": 0.6967,
+      "step": 3250
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.222222222222222e-05,
+      "loss": 0.6679,
+      "step": 3255
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.20164609053498e-05,
+      "loss": 0.6683,
+      "step": 3260
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.181069958847736e-05,
+      "loss": 0.6787,
+      "step": 3265
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.160493827160494e-05,
+      "loss": 0.6726,
+      "step": 3270
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.139917695473252e-05,
+      "loss": 0.721,
+      "step": 3275
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.119341563786009e-05,
+      "loss": 0.6865,
+      "step": 3280
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.098765432098767e-05,
+      "loss": 0.7031,
+      "step": 3285
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.078189300411523e-05,
+      "loss": 0.6895,
+      "step": 3290
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.05761316872428e-05,
+      "loss": 0.6983,
+      "step": 3295
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.037037037037038e-05,
+      "loss": 0.6929,
+      "step": 3300
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.016460905349794e-05,
+      "loss": 0.7117,
+      "step": 3305
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.995884773662552e-05,
+      "loss": 0.664,
+      "step": 3310
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.975308641975309e-05,
+      "loss": 0.6645,
+      "step": 3315
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.954732510288067e-05,
+      "loss": 0.6985,
+      "step": 3320
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.934156378600823e-05,
+      "loss": 0.6826,
+      "step": 3325
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.91358024691358e-05,
+      "loss": 0.7324,
+      "step": 3330
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.893004115226338e-05,
+      "loss": 0.6828,
+      "step": 3335
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.872427983539094e-05,
+      "loss": 0.6801,
+      "step": 3340
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.851851851851852e-05,
+      "loss": 0.6923,
+      "step": 3345
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.831275720164609e-05,
+      "loss": 0.6787,
+      "step": 3350
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.810699588477366e-05,
+      "loss": 0.6812,
+      "step": 3355
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.790123456790123e-05,
+      "loss": 0.7059,
+      "step": 3360
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.76954732510288e-05,
+      "loss": 0.6794,
+      "step": 3365
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.748971193415638e-05,
+      "loss": 0.6635,
+      "step": 3370
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.728395061728396e-05,
+      "loss": 0.666,
+      "step": 3375
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.707818930041152e-05,
+      "loss": 0.6909,
+      "step": 3380
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.68724279835391e-05,
+      "loss": 0.7096,
+      "step": 3385
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.7049,
+      "step": 3390
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.646090534979425e-05,
+      "loss": 0.6801,
+      "step": 3395
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.625514403292181e-05,
+      "loss": 0.7045,
+      "step": 3400
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.60493827160494e-05,
+      "loss": 0.7096,
+      "step": 3405
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.584362139917696e-05,
+      "loss": 0.7192,
+      "step": 3410
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.563786008230453e-05,
+      "loss": 0.6818,
+      "step": 3415
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.54320987654321e-05,
+      "loss": 0.6358,
+      "step": 3420
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.522633744855967e-05,
+      "loss": 0.6674,
+      "step": 3425
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.502057613168725e-05,
+      "loss": 0.684,
+      "step": 3430
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.481481481481482e-05,
+      "loss": 0.6753,
+      "step": 3435
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.460905349794238e-05,
+      "loss": 0.6902,
+      "step": 3440
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.440329218106996e-05,
+      "loss": 0.6664,
+      "step": 3445
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.419753086419753e-05,
+      "loss": 0.684,
+      "step": 3450
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.39917695473251e-05,
+      "loss": 0.682,
+      "step": 3455
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.378600823045267e-05,
+      "loss": 0.6764,
+      "step": 3460
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.358024691358025e-05,
+      "loss": 0.6783,
+      "step": 3465
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.337448559670782e-05,
+      "loss": 0.6962,
+      "step": 3470
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.31687242798354e-05,
+      "loss": 0.6819,
+      "step": 3475
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.296296296296296e-05,
+      "loss": 0.6741,
+      "step": 3480
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.275720164609054e-05,
+      "loss": 0.6754,
+      "step": 3485
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.255144032921812e-05,
+      "loss": 0.6717,
+      "step": 3490
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.234567901234569e-05,
+      "loss": 0.667,
+      "step": 3495
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.213991769547325e-05,
+      "loss": 0.6822,
+      "step": 3500
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 5010,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 500,
+  "total_flos": 1.408889273989464e+19,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/training_args.bin b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e707059a03fbe62142ac41332e145ac1ac1f36cc
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b99a1729574e0e46ee782b42dc90101a785d5b4792871920d293f44722c28cc
+size 5944
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/zero_to_fp32.py b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/zero_to_fp32.py
new file mode 100644
index 0000000000000000000000000000000000000000..c98caae31534368be22b67fc4ae906836c992a8d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-3500/zero_to_fp32.py
@@ -0,0 +1,587 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+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):
+    """
+    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)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+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 _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()
+            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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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 param_shapes.items():
+
+        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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    """
+    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``
+
+    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)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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("-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_file, tag=args.tag)
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/README.md b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..d380bbbf0bf0fa1f211b53410559607e5007a209
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: meta-llama/Llama-2-7b-hf
+---
+
+# 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 0.9.0
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/adapter_config.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c81a592f0ff08374a4f61f6003029070a8429ae
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/adapter_config.json
@@ -0,0 +1,35 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "mora_type": 1,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_mora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/adapter_model.safetensors b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/adapter_model.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..5c0192bdea98f0b3b2070bdb5e703d03a67aa8af
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/adapter_model.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:760f271674e70b10f3c0862e636e6e476374822135ffd779b5bc5e8efeeaf544
+size 564342240
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/global_step4000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/global_step4000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..334b959ec8134cad314f452285076a897f7dd7dc
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/global_step4000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:488f3b09d708dea624d9aa5a2203a60393cef34ba96a723da74967d6000addbc
+size 239914224
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/global_step4000/mp_rank_00_model_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/global_step4000/mp_rank_00_model_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f69a31863f7417e2c602314eea4546a3dc3e38fa
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/global_step4000/mp_rank_00_model_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69e0b718ee03fac24adc36a83d378c71464a92a58f16855c28b0019d44f8fa25
+size 40303404
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/latest b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/latest
new file mode 100644
index 0000000000000000000000000000000000000000..641c40ce8d5e05e04a0840d69bcc5fac045a3fbd
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/latest
@@ -0,0 +1 @@
+global_step4000
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/rng_state.pth b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/rng_state.pth
new file mode 100644
index 0000000000000000000000000000000000000000..178d8815a2b3cdca228a0041ccaf1c5aaa4fb84d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/rng_state.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7dc888883ee7f81d8fd46e862508c2a1f0ce08f0cd29a1f2a2ecef8bcdbbd555
+size 14244
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/scheduler.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/scheduler.pt
new file mode 100644
index 0000000000000000000000000000000000000000..7425eca6955a32648ed4f5d49750541bf784bb01
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/scheduler.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5c63e5be21eafeb9162ff869971752483dca22315942206614c9edfe5714c18
+size 1064
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/trainer_state.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a792ed6668e5cdb11c19ef23c3d958f87f923f03
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/trainer_state.json
@@ -0,0 +1,4821 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.5966569994074906,
+  "eval_steps": 500,
+  "global_step": 4000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.3797,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.3952,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2e-05,
+      "loss": 1.3112,
+      "step": 15
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 1.3136,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 1.2012,
+      "step": 25
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4e-05,
+      "loss": 1.1277,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 1.0832,
+      "step": 35
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.333333333333333e-05,
+      "loss": 1.0659,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-05,
+      "loss": 0.983,
+      "step": 45
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.9763,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.333333333333333e-05,
+      "loss": 0.9992,
+      "step": 55
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8e-05,
+      "loss": 0.9405,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.666666666666667e-05,
+      "loss": 0.9827,
+      "step": 65
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.333333333333334e-05,
+      "loss": 0.9061,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001,
+      "loss": 0.8921,
+      "step": 75
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010666666666666667,
+      "loss": 0.9377,
+      "step": 80
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011333333333333334,
+      "loss": 0.8528,
+      "step": 85
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012,
+      "loss": 0.8913,
+      "step": 90
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012666666666666666,
+      "loss": 0.8562,
+      "step": 95
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.877,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014,
+      "loss": 0.8781,
+      "step": 105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014666666666666666,
+      "loss": 0.8546,
+      "step": 110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015333333333333334,
+      "loss": 0.8672,
+      "step": 115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016,
+      "loss": 0.8611,
+      "step": 120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.8624,
+      "step": 125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017333333333333334,
+      "loss": 0.8477,
+      "step": 130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018,
+      "loss": 0.8227,
+      "step": 135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001866666666666667,
+      "loss": 0.8064,
+      "step": 140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019333333333333333,
+      "loss": 0.8422,
+      "step": 145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002,
+      "loss": 0.8471,
+      "step": 150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001997942386831276,
+      "loss": 0.8183,
+      "step": 155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958847736625514,
+      "loss": 0.8229,
+      "step": 160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019938271604938272,
+      "loss": 0.8114,
+      "step": 165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991769547325103,
+      "loss": 0.7861,
+      "step": 170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019897119341563788,
+      "loss": 0.841,
+      "step": 175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019876543209876543,
+      "loss": 0.8549,
+      "step": 180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000198559670781893,
+      "loss": 0.8449,
+      "step": 185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001983539094650206,
+      "loss": 0.8311,
+      "step": 190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019814814814814814,
+      "loss": 0.8296,
+      "step": 195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019794238683127572,
+      "loss": 0.8263,
+      "step": 200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001977366255144033,
+      "loss": 0.8452,
+      "step": 205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019753086419753085,
+      "loss": 0.8044,
+      "step": 210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019732510288065843,
+      "loss": 0.8119,
+      "step": 215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000197119341563786,
+      "loss": 0.8127,
+      "step": 220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001969135802469136,
+      "loss": 0.7921,
+      "step": 225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019670781893004114,
+      "loss": 0.803,
+      "step": 230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019650205761316872,
+      "loss": 0.838,
+      "step": 235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001962962962962963,
+      "loss": 0.7801,
+      "step": 240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019609053497942388,
+      "loss": 0.817,
+      "step": 245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019588477366255146,
+      "loss": 0.8211,
+      "step": 250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000195679012345679,
+      "loss": 0.8423,
+      "step": 255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001954732510288066,
+      "loss": 0.8007,
+      "step": 260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019526748971193417,
+      "loss": 0.8135,
+      "step": 265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019506172839506175,
+      "loss": 0.7961,
+      "step": 270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019485596707818933,
+      "loss": 0.797,
+      "step": 275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019465020576131688,
+      "loss": 0.793,
+      "step": 280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019444444444444446,
+      "loss": 0.7809,
+      "step": 285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019423868312757204,
+      "loss": 0.7811,
+      "step": 290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019403292181069962,
+      "loss": 0.8074,
+      "step": 295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019382716049382717,
+      "loss": 0.8028,
+      "step": 300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019362139917695475,
+      "loss": 0.7895,
+      "step": 305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019341563786008233,
+      "loss": 0.7777,
+      "step": 310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019320987654320988,
+      "loss": 0.7868,
+      "step": 315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019300411522633746,
+      "loss": 0.7921,
+      "step": 320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019279835390946504,
+      "loss": 0.8,
+      "step": 325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001925925925925926,
+      "loss": 0.7781,
+      "step": 330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019238683127572017,
+      "loss": 0.7799,
+      "step": 335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019218106995884775,
+      "loss": 0.7921,
+      "step": 340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019197530864197533,
+      "loss": 0.7851,
+      "step": 345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019176954732510288,
+      "loss": 0.8106,
+      "step": 350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019156378600823046,
+      "loss": 0.7953,
+      "step": 355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019135802469135804,
+      "loss": 0.8084,
+      "step": 360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001911522633744856,
+      "loss": 0.8298,
+      "step": 365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019094650205761317,
+      "loss": 0.8273,
+      "step": 370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019074074074074075,
+      "loss": 0.8047,
+      "step": 375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019053497942386833,
+      "loss": 0.8108,
+      "step": 380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019032921810699588,
+      "loss": 0.8207,
+      "step": 385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019012345679012346,
+      "loss": 0.7884,
+      "step": 390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018991769547325104,
+      "loss": 0.7918,
+      "step": 395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001897119341563786,
+      "loss": 0.7554,
+      "step": 400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018950617283950617,
+      "loss": 0.8099,
+      "step": 405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018930041152263375,
+      "loss": 0.807,
+      "step": 410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018909465020576133,
+      "loss": 0.7793,
+      "step": 415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018888888888888888,
+      "loss": 0.7734,
+      "step": 420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018868312757201646,
+      "loss": 0.7786,
+      "step": 425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018847736625514404,
+      "loss": 0.7548,
+      "step": 430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001882716049382716,
+      "loss": 0.8024,
+      "step": 435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018806584362139917,
+      "loss": 0.7708,
+      "step": 440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018786008230452675,
+      "loss": 0.8045,
+      "step": 445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018765432098765433,
+      "loss": 0.7477,
+      "step": 450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018744855967078189,
+      "loss": 0.7474,
+      "step": 455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018724279835390946,
+      "loss": 0.7629,
+      "step": 460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018703703703703704,
+      "loss": 0.7777,
+      "step": 465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018683127572016462,
+      "loss": 0.7823,
+      "step": 470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001866255144032922,
+      "loss": 0.7762,
+      "step": 475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018641975308641978,
+      "loss": 0.782,
+      "step": 480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018621399176954733,
+      "loss": 0.7783,
+      "step": 485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001860082304526749,
+      "loss": 0.7376,
+      "step": 490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001858024691358025,
+      "loss": 0.7683,
+      "step": 495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018559670781893007,
+      "loss": 0.7547,
+      "step": 500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018539094650205762,
+      "loss": 0.7569,
+      "step": 505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001851851851851852,
+      "loss": 0.7796,
+      "step": 510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018497942386831278,
+      "loss": 0.7888,
+      "step": 515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018477366255144033,
+      "loss": 0.7763,
+      "step": 520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018456790123456791,
+      "loss": 0.7787,
+      "step": 525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001843621399176955,
+      "loss": 0.772,
+      "step": 530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018415637860082307,
+      "loss": 0.786,
+      "step": 535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018395061728395062,
+      "loss": 0.7639,
+      "step": 540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001837448559670782,
+      "loss": 0.775,
+      "step": 545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018353909465020578,
+      "loss": 0.8002,
+      "step": 550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 0.7359,
+      "step": 555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018312757201646091,
+      "loss": 0.747,
+      "step": 560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001829218106995885,
+      "loss": 0.8234,
+      "step": 565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018271604938271605,
+      "loss": 0.7586,
+      "step": 570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018251028806584363,
+      "loss": 0.7315,
+      "step": 575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001823045267489712,
+      "loss": 0.7986,
+      "step": 580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018209876543209878,
+      "loss": 0.7758,
+      "step": 585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018189300411522634,
+      "loss": 0.7869,
+      "step": 590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018168724279835392,
+      "loss": 0.7596,
+      "step": 595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001814814814814815,
+      "loss": 0.7658,
+      "step": 600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018127572016460905,
+      "loss": 0.7851,
+      "step": 605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018106995884773663,
+      "loss": 0.7957,
+      "step": 610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001808641975308642,
+      "loss": 0.7756,
+      "step": 615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018065843621399178,
+      "loss": 0.7911,
+      "step": 620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018045267489711934,
+      "loss": 0.7864,
+      "step": 625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018024691358024692,
+      "loss": 0.7683,
+      "step": 630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001800411522633745,
+      "loss": 0.7731,
+      "step": 635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017983539094650205,
+      "loss": 0.7671,
+      "step": 640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017962962962962963,
+      "loss": 0.7367,
+      "step": 645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001794238683127572,
+      "loss": 0.8024,
+      "step": 650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017921810699588476,
+      "loss": 0.7572,
+      "step": 655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017901234567901234,
+      "loss": 0.7554,
+      "step": 660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017880658436213992,
+      "loss": 0.7677,
+      "step": 665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001786008230452675,
+      "loss": 0.7652,
+      "step": 670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017839506172839508,
+      "loss": 0.7771,
+      "step": 675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017818930041152265,
+      "loss": 0.7747,
+      "step": 680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001779835390946502,
+      "loss": 0.767,
+      "step": 685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017777777777777779,
+      "loss": 0.7738,
+      "step": 690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017757201646090537,
+      "loss": 0.7271,
+      "step": 695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017736625514403294,
+      "loss": 0.7316,
+      "step": 700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017716049382716052,
+      "loss": 0.7831,
+      "step": 705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017695473251028808,
+      "loss": 0.7835,
+      "step": 710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017674897119341566,
+      "loss": 0.7497,
+      "step": 715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017654320987654323,
+      "loss": 0.7757,
+      "step": 720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001763374485596708,
+      "loss": 0.74,
+      "step": 725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017613168724279837,
+      "loss": 0.7697,
+      "step": 730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017592592592592595,
+      "loss": 0.774,
+      "step": 735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017572016460905352,
+      "loss": 0.7668,
+      "step": 740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017551440329218108,
+      "loss": 0.7779,
+      "step": 745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017530864197530866,
+      "loss": 0.7441,
+      "step": 750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017510288065843624,
+      "loss": 0.7578,
+      "step": 755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001748971193415638,
+      "loss": 0.8358,
+      "step": 760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017469135802469137,
+      "loss": 0.7473,
+      "step": 765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017448559670781895,
+      "loss": 0.7517,
+      "step": 770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001742798353909465,
+      "loss": 0.7593,
+      "step": 775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017407407407407408,
+      "loss": 0.7251,
+      "step": 780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017386831275720166,
+      "loss": 0.7365,
+      "step": 785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017366255144032924,
+      "loss": 0.7798,
+      "step": 790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001734567901234568,
+      "loss": 0.746,
+      "step": 795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017325102880658437,
+      "loss": 0.7607,
+      "step": 800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017304526748971195,
+      "loss": 0.7694,
+      "step": 805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001728395061728395,
+      "loss": 0.7926,
+      "step": 810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017263374485596708,
+      "loss": 0.795,
+      "step": 815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017242798353909466,
+      "loss": 0.7581,
+      "step": 820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017222222222222224,
+      "loss": 0.7629,
+      "step": 825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001720164609053498,
+      "loss": 0.7526,
+      "step": 830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017181069958847737,
+      "loss": 0.7338,
+      "step": 835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017160493827160495,
+      "loss": 0.7544,
+      "step": 840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001713991769547325,
+      "loss": 0.7522,
+      "step": 845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017119341563786008,
+      "loss": 0.7639,
+      "step": 850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017098765432098766,
+      "loss": 0.7914,
+      "step": 855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017078189300411524,
+      "loss": 0.7635,
+      "step": 860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001705761316872428,
+      "loss": 0.8343,
+      "step": 865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017037037037037037,
+      "loss": 0.7299,
+      "step": 870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017016460905349795,
+      "loss": 0.7643,
+      "step": 875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016995884773662553,
+      "loss": 0.7343,
+      "step": 880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016975308641975308,
+      "loss": 0.7546,
+      "step": 885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016954732510288066,
+      "loss": 0.7435,
+      "step": 890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016934156378600824,
+      "loss": 0.8086,
+      "step": 895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016913580246913582,
+      "loss": 0.7274,
+      "step": 900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001689300411522634,
+      "loss": 0.7511,
+      "step": 905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016872427983539098,
+      "loss": 0.7689,
+      "step": 910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016851851851851853,
+      "loss": 0.754,
+      "step": 915
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001683127572016461,
+      "loss": 0.7298,
+      "step": 920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001681069958847737,
+      "loss": 0.7755,
+      "step": 925
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016790123456790124,
+      "loss": 0.7854,
+      "step": 930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016769547325102882,
+      "loss": 0.7452,
+      "step": 935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001674897119341564,
+      "loss": 0.7473,
+      "step": 940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016728395061728398,
+      "loss": 0.7855,
+      "step": 945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016707818930041153,
+      "loss": 0.7527,
+      "step": 950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001668724279835391,
+      "loss": 0.7641,
+      "step": 955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.828,
+      "step": 960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016646090534979424,
+      "loss": 0.7571,
+      "step": 965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016625514403292182,
+      "loss": 0.7768,
+      "step": 970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001660493827160494,
+      "loss": 0.7644,
+      "step": 975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016584362139917695,
+      "loss": 0.7747,
+      "step": 980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016563786008230453,
+      "loss": 0.7926,
+      "step": 985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001654320987654321,
+      "loss": 0.7645,
+      "step": 990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001652263374485597,
+      "loss": 0.7464,
+      "step": 995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016502057613168724,
+      "loss": 0.7558,
+      "step": 1000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016481481481481482,
+      "loss": 0.7574,
+      "step": 1005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001646090534979424,
+      "loss": 0.7491,
+      "step": 1010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016440329218106995,
+      "loss": 0.756,
+      "step": 1015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016419753086419753,
+      "loss": 0.7679,
+      "step": 1020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001639917695473251,
+      "loss": 0.776,
+      "step": 1025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001637860082304527,
+      "loss": 0.7235,
+      "step": 1030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016358024691358024,
+      "loss": 0.7817,
+      "step": 1035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016337448559670782,
+      "loss": 0.7333,
+      "step": 1040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001631687242798354,
+      "loss": 0.754,
+      "step": 1045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016296296296296295,
+      "loss": 0.74,
+      "step": 1050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016275720164609053,
+      "loss": 0.7857,
+      "step": 1055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001625514403292181,
+      "loss": 0.8015,
+      "step": 1060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001623456790123457,
+      "loss": 0.7508,
+      "step": 1065
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016213991769547324,
+      "loss": 0.7514,
+      "step": 1070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016193415637860082,
+      "loss": 0.7326,
+      "step": 1075
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001617283950617284,
+      "loss": 0.7322,
+      "step": 1080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016152263374485595,
+      "loss": 0.7373,
+      "step": 1085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016131687242798353,
+      "loss": 0.7637,
+      "step": 1090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001611111111111111,
+      "loss": 0.7697,
+      "step": 1095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001609053497942387,
+      "loss": 0.7505,
+      "step": 1100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016069958847736627,
+      "loss": 0.7568,
+      "step": 1105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016049382716049385,
+      "loss": 0.786,
+      "step": 1110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001602880658436214,
+      "loss": 0.7525,
+      "step": 1115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016008230452674898,
+      "loss": 0.7696,
+      "step": 1120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015987654320987656,
+      "loss": 0.7648,
+      "step": 1125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015967078189300414,
+      "loss": 0.7492,
+      "step": 1130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001594650205761317,
+      "loss": 0.809,
+      "step": 1135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015925925925925927,
+      "loss": 0.7592,
+      "step": 1140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015905349794238685,
+      "loss": 0.7874,
+      "step": 1145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015884773662551443,
+      "loss": 0.7874,
+      "step": 1150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015864197530864198,
+      "loss": 0.7374,
+      "step": 1155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015843621399176956,
+      "loss": 0.7218,
+      "step": 1160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015823045267489714,
+      "loss": 0.7494,
+      "step": 1165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001580246913580247,
+      "loss": 0.7449,
+      "step": 1170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015781893004115227,
+      "loss": 0.725,
+      "step": 1175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015761316872427985,
+      "loss": 0.739,
+      "step": 1180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015740740740740743,
+      "loss": 0.7554,
+      "step": 1185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015720164609053498,
+      "loss": 0.7702,
+      "step": 1190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015699588477366256,
+      "loss": 0.7418,
+      "step": 1195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015679012345679014,
+      "loss": 0.7217,
+      "step": 1200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001565843621399177,
+      "loss": 0.7391,
+      "step": 1205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015637860082304527,
+      "loss": 0.7696,
+      "step": 1210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015617283950617285,
+      "loss": 0.7594,
+      "step": 1215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001559670781893004,
+      "loss": 0.777,
+      "step": 1220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015576131687242798,
+      "loss": 0.7441,
+      "step": 1225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015555555555555556,
+      "loss": 0.7611,
+      "step": 1230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015534979423868314,
+      "loss": 0.7278,
+      "step": 1235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001551440329218107,
+      "loss": 0.7452,
+      "step": 1240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015493827160493827,
+      "loss": 0.7939,
+      "step": 1245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015473251028806585,
+      "loss": 0.7161,
+      "step": 1250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001545267489711934,
+      "loss": 0.7344,
+      "step": 1255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015432098765432098,
+      "loss": 0.7368,
+      "step": 1260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015411522633744856,
+      "loss": 0.7172,
+      "step": 1265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015390946502057614,
+      "loss": 0.7589,
+      "step": 1270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001537037037037037,
+      "loss": 0.7641,
+      "step": 1275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015349794238683127,
+      "loss": 0.761,
+      "step": 1280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015329218106995885,
+      "loss": 0.7414,
+      "step": 1285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001530864197530864,
+      "loss": 0.7713,
+      "step": 1290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015288065843621399,
+      "loss": 0.7543,
+      "step": 1295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015267489711934156,
+      "loss": 0.7683,
+      "step": 1300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015246913580246914,
+      "loss": 0.8106,
+      "step": 1305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015226337448559672,
+      "loss": 0.7739,
+      "step": 1310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015205761316872428,
+      "loss": 0.7585,
+      "step": 1315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015185185185185185,
+      "loss": 0.7577,
+      "step": 1320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015164609053497943,
+      "loss": 0.7608,
+      "step": 1325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000151440329218107,
+      "loss": 0.7715,
+      "step": 1330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001512345679012346,
+      "loss": 0.7681,
+      "step": 1335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015102880658436214,
+      "loss": 0.7476,
+      "step": 1340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015082304526748972,
+      "loss": 0.7539,
+      "step": 1345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001506172839506173,
+      "loss": 0.7492,
+      "step": 1350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015041152263374488,
+      "loss": 0.7491,
+      "step": 1355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015020576131687243,
+      "loss": 0.7236,
+      "step": 1360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 0.7328,
+      "step": 1365
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001497942386831276,
+      "loss": 0.7603,
+      "step": 1370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014958847736625515,
+      "loss": 0.7206,
+      "step": 1375
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014938271604938272,
+      "loss": 0.7001,
+      "step": 1380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001491769547325103,
+      "loss": 0.7241,
+      "step": 1385
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014897119341563788,
+      "loss": 0.7511,
+      "step": 1390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014876543209876544,
+      "loss": 0.7546,
+      "step": 1395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014855967078189301,
+      "loss": 0.762,
+      "step": 1400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001483539094650206,
+      "loss": 0.6979,
+      "step": 1405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014814814814814815,
+      "loss": 0.7352,
+      "step": 1410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014794238683127573,
+      "loss": 0.7604,
+      "step": 1415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001477366255144033,
+      "loss": 0.7612,
+      "step": 1420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014753086419753086,
+      "loss": 0.7618,
+      "step": 1425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014732510288065844,
+      "loss": 0.7107,
+      "step": 1430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014711934156378602,
+      "loss": 0.7247,
+      "step": 1435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001469135802469136,
+      "loss": 0.7733,
+      "step": 1440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014670781893004115,
+      "loss": 0.7627,
+      "step": 1445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014650205761316873,
+      "loss": 0.7323,
+      "step": 1450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001462962962962963,
+      "loss": 0.7728,
+      "step": 1455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014609053497942386,
+      "loss": 0.7541,
+      "step": 1460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014588477366255144,
+      "loss": 0.7781,
+      "step": 1465
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014567901234567902,
+      "loss": 0.7474,
+      "step": 1470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001454732510288066,
+      "loss": 0.7789,
+      "step": 1475
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014526748971193415,
+      "loss": 0.7497,
+      "step": 1480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014506172839506173,
+      "loss": 0.7487,
+      "step": 1485
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001448559670781893,
+      "loss": 0.7563,
+      "step": 1490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014465020576131686,
+      "loss": 0.7444,
+      "step": 1495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014444444444444444,
+      "loss": 0.7399,
+      "step": 1500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014423868312757202,
+      "loss": 0.7288,
+      "step": 1505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001440329218106996,
+      "loss": 0.7458,
+      "step": 1510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014382716049382718,
+      "loss": 0.7297,
+      "step": 1515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014362139917695473,
+      "loss": 0.7684,
+      "step": 1520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001434156378600823,
+      "loss": 0.7428,
+      "step": 1525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014320987654320989,
+      "loss": 0.705,
+      "step": 1530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014300411522633747,
+      "loss": 0.7445,
+      "step": 1535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014279835390946504,
+      "loss": 0.7109,
+      "step": 1540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001425925925925926,
+      "loss": 0.7502,
+      "step": 1545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014238683127572018,
+      "loss": 0.7485,
+      "step": 1550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014218106995884776,
+      "loss": 0.7421,
+      "step": 1555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014197530864197534,
+      "loss": 0.7369,
+      "step": 1560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001417695473251029,
+      "loss": 0.7746,
+      "step": 1565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014156378600823047,
+      "loss": 0.7334,
+      "step": 1570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014135802469135805,
+      "loss": 0.7251,
+      "step": 1575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001411522633744856,
+      "loss": 0.7146,
+      "step": 1580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014094650205761318,
+      "loss": 0.7563,
+      "step": 1585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014074074074074076,
+      "loss": 0.7247,
+      "step": 1590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014053497942386834,
+      "loss": 0.7379,
+      "step": 1595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001403292181069959,
+      "loss": 0.7684,
+      "step": 1600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014012345679012347,
+      "loss": 0.7574,
+      "step": 1605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00013991769547325105,
+      "loss": 0.7138,
+      "step": 1610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001397119341563786,
+      "loss": 0.7472,
+      "step": 1615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013950617283950618,
+      "loss": 0.7466,
+      "step": 1620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013930041152263376,
+      "loss": 0.7288,
+      "step": 1625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013909465020576134,
+      "loss": 0.7717,
+      "step": 1630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001388888888888889,
+      "loss": 0.7665,
+      "step": 1635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013868312757201647,
+      "loss": 0.7408,
+      "step": 1640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013847736625514405,
+      "loss": 0.7418,
+      "step": 1645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001382716049382716,
+      "loss": 0.7502,
+      "step": 1650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013806584362139918,
+      "loss": 0.7657,
+      "step": 1655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013786008230452676,
+      "loss": 0.728,
+      "step": 1660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001376543209876543,
+      "loss": 0.7142,
+      "step": 1665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001374485596707819,
+      "loss": 0.7971,
+      "step": 1670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013724279835390947,
+      "loss": 0.737,
+      "step": 1675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013703703703703705,
+      "loss": 0.7947,
+      "step": 1680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001368312757201646,
+      "loss": 0.7525,
+      "step": 1685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013662551440329218,
+      "loss": 0.711,
+      "step": 1690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013641975308641976,
+      "loss": 0.7459,
+      "step": 1695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001362139917695473,
+      "loss": 0.7404,
+      "step": 1700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001360082304526749,
+      "loss": 0.7458,
+      "step": 1705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013580246913580247,
+      "loss": 0.754,
+      "step": 1710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013559670781893005,
+      "loss": 0.7217,
+      "step": 1715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001353909465020576,
+      "loss": 0.7373,
+      "step": 1720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013518518518518518,
+      "loss": 0.7567,
+      "step": 1725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013497942386831276,
+      "loss": 0.7525,
+      "step": 1730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013477366255144034,
+      "loss": 0.752,
+      "step": 1735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013456790123456792,
+      "loss": 0.7342,
+      "step": 1740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001343621399176955,
+      "loss": 0.7276,
+      "step": 1745
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013415637860082305,
+      "loss": 0.6915,
+      "step": 1750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013395061728395063,
+      "loss": 0.737,
+      "step": 1755
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001337448559670782,
+      "loss": 0.7427,
+      "step": 1760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001335390946502058,
+      "loss": 0.7126,
+      "step": 1765
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.7241,
+      "step": 1770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013312757201646092,
+      "loss": 0.755,
+      "step": 1775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001329218106995885,
+      "loss": 0.7494,
+      "step": 1780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013271604938271605,
+      "loss": 0.709,
+      "step": 1785
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013251028806584363,
+      "loss": 0.7555,
+      "step": 1790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001323045267489712,
+      "loss": 0.7298,
+      "step": 1795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001320987654320988,
+      "loss": 0.7232,
+      "step": 1800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013189300411522634,
+      "loss": 0.7441,
+      "step": 1805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013168724279835392,
+      "loss": 0.7623,
+      "step": 1810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001314814814814815,
+      "loss": 0.7349,
+      "step": 1815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013127572016460905,
+      "loss": 0.7773,
+      "step": 1820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013106995884773663,
+      "loss": 0.7537,
+      "step": 1825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001308641975308642,
+      "loss": 0.7579,
+      "step": 1830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001306584362139918,
+      "loss": 0.7321,
+      "step": 1835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013045267489711934,
+      "loss": 0.7175,
+      "step": 1840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00013024691358024692,
+      "loss": 0.7553,
+      "step": 1845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001300411522633745,
+      "loss": 0.7591,
+      "step": 1850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012983539094650205,
+      "loss": 0.7581,
+      "step": 1855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012962962962962963,
+      "loss": 0.7632,
+      "step": 1860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001294238683127572,
+      "loss": 0.7434,
+      "step": 1865
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012921810699588476,
+      "loss": 0.7695,
+      "step": 1870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012901234567901234,
+      "loss": 0.7808,
+      "step": 1875
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012880658436213992,
+      "loss": 0.7428,
+      "step": 1880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001286008230452675,
+      "loss": 0.7487,
+      "step": 1885
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012839506172839505,
+      "loss": 0.7033,
+      "step": 1890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012818930041152263,
+      "loss": 0.7639,
+      "step": 1895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001279835390946502,
+      "loss": 0.7345,
+      "step": 1900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012777777777777776,
+      "loss": 0.7434,
+      "step": 1905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012757201646090534,
+      "loss": 0.7082,
+      "step": 1910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012736625514403292,
+      "loss": 0.7294,
+      "step": 1915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001271604938271605,
+      "loss": 0.74,
+      "step": 1920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012695473251028805,
+      "loss": 0.7694,
+      "step": 1925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012674897119341563,
+      "loss": 0.7339,
+      "step": 1930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001265432098765432,
+      "loss": 0.7259,
+      "step": 1935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001263374485596708,
+      "loss": 0.7307,
+      "step": 1940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012613168724279837,
+      "loss": 0.7085,
+      "step": 1945
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012592592592592592,
+      "loss": 0.7662,
+      "step": 1950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001257201646090535,
+      "loss": 0.7145,
+      "step": 1955
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012551440329218108,
+      "loss": 0.7528,
+      "step": 1960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012530864197530866,
+      "loss": 0.7308,
+      "step": 1965
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012510288065843624,
+      "loss": 0.7163,
+      "step": 1970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001248971193415638,
+      "loss": 0.7254,
+      "step": 1975
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012469135802469137,
+      "loss": 0.7467,
+      "step": 1980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012448559670781895,
+      "loss": 0.7391,
+      "step": 1985
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001242798353909465,
+      "loss": 0.7314,
+      "step": 1990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012407407407407408,
+      "loss": 0.7048,
+      "step": 1995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012386831275720166,
+      "loss": 0.7187,
+      "step": 2000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012366255144032924,
+      "loss": 0.735,
+      "step": 2005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001234567901234568,
+      "loss": 0.7662,
+      "step": 2010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012325102880658437,
+      "loss": 0.7437,
+      "step": 2015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012304526748971195,
+      "loss": 0.7675,
+      "step": 2020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001228395061728395,
+      "loss": 0.7143,
+      "step": 2025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012263374485596708,
+      "loss": 0.7101,
+      "step": 2030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012242798353909466,
+      "loss": 0.7257,
+      "step": 2035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012222222222222224,
+      "loss": 0.7522,
+      "step": 2040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001220164609053498,
+      "loss": 0.7669,
+      "step": 2045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012181069958847737,
+      "loss": 0.7283,
+      "step": 2050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012160493827160495,
+      "loss": 0.7413,
+      "step": 2055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001213991769547325,
+      "loss": 0.6988,
+      "step": 2060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012119341563786008,
+      "loss": 0.7421,
+      "step": 2065
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012098765432098766,
+      "loss": 0.7152,
+      "step": 2070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012078189300411524,
+      "loss": 0.7741,
+      "step": 2075
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001205761316872428,
+      "loss": 0.7381,
+      "step": 2080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012037037037037037,
+      "loss": 0.738,
+      "step": 2085
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012016460905349795,
+      "loss": 0.7258,
+      "step": 2090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011995884773662552,
+      "loss": 0.7426,
+      "step": 2095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011975308641975308,
+      "loss": 0.7185,
+      "step": 2100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011954732510288066,
+      "loss": 0.7401,
+      "step": 2105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011934156378600823,
+      "loss": 0.751,
+      "step": 2110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011913580246913581,
+      "loss": 0.7284,
+      "step": 2115
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011893004115226339,
+      "loss": 0.7901,
+      "step": 2120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011872427983539097,
+      "loss": 0.7432,
+      "step": 2125
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011851851851851852,
+      "loss": 0.7355,
+      "step": 2130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001183127572016461,
+      "loss": 0.7333,
+      "step": 2135
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011810699588477368,
+      "loss": 0.7504,
+      "step": 2140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011790123456790123,
+      "loss": 0.7503,
+      "step": 2145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011769547325102881,
+      "loss": 0.728,
+      "step": 2150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011748971193415639,
+      "loss": 0.7485,
+      "step": 2155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011728395061728397,
+      "loss": 0.7127,
+      "step": 2160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011707818930041152,
+      "loss": 0.7278,
+      "step": 2165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001168724279835391,
+      "loss": 0.7513,
+      "step": 2170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011666666666666668,
+      "loss": 0.7193,
+      "step": 2175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011646090534979423,
+      "loss": 0.7245,
+      "step": 2180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011625514403292181,
+      "loss": 0.7266,
+      "step": 2185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011604938271604939,
+      "loss": 0.7531,
+      "step": 2190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011584362139917696,
+      "loss": 0.7637,
+      "step": 2195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011563786008230452,
+      "loss": 0.7388,
+      "step": 2200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001154320987654321,
+      "loss": 0.7222,
+      "step": 2205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011522633744855968,
+      "loss": 0.7667,
+      "step": 2210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011502057613168725,
+      "loss": 0.7382,
+      "step": 2215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011481481481481482,
+      "loss": 0.7243,
+      "step": 2220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001146090534979424,
+      "loss": 0.7192,
+      "step": 2225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011440329218106996,
+      "loss": 0.7237,
+      "step": 2230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011419753086419754,
+      "loss": 0.7498,
+      "step": 2235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011399176954732511,
+      "loss": 0.7007,
+      "step": 2240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001137860082304527,
+      "loss": 0.758,
+      "step": 2245
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011358024691358025,
+      "loss": 0.7111,
+      "step": 2250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011337448559670783,
+      "loss": 0.706,
+      "step": 2255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001131687242798354,
+      "loss": 0.7423,
+      "step": 2260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011296296296296296,
+      "loss": 0.7094,
+      "step": 2265
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011275720164609054,
+      "loss": 0.7331,
+      "step": 2270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011255144032921812,
+      "loss": 0.7164,
+      "step": 2275
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001123456790123457,
+      "loss": 0.7452,
+      "step": 2280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011213991769547325,
+      "loss": 0.7172,
+      "step": 2285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011193415637860083,
+      "loss": 0.7232,
+      "step": 2290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001117283950617284,
+      "loss": 0.7112,
+      "step": 2295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011152263374485596,
+      "loss": 0.7459,
+      "step": 2300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011131687242798354,
+      "loss": 0.7353,
+      "step": 2305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011111111111111112,
+      "loss": 0.7406,
+      "step": 2310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011090534979423868,
+      "loss": 0.7623,
+      "step": 2315
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011069958847736626,
+      "loss": 0.7567,
+      "step": 2320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011049382716049384,
+      "loss": 0.7752,
+      "step": 2325
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001102880658436214,
+      "loss": 0.7419,
+      "step": 2330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011008230452674897,
+      "loss": 0.7244,
+      "step": 2335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00010987654320987655,
+      "loss": 0.7553,
+      "step": 2340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010967078189300413,
+      "loss": 0.7389,
+      "step": 2345
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010946502057613168,
+      "loss": 0.7002,
+      "step": 2350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010925925925925926,
+      "loss": 0.7238,
+      "step": 2355
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010905349794238684,
+      "loss": 0.6923,
+      "step": 2360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010884773662551442,
+      "loss": 0.7579,
+      "step": 2365
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010864197530864197,
+      "loss": 0.7291,
+      "step": 2370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010843621399176955,
+      "loss": 0.721,
+      "step": 2375
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010823045267489713,
+      "loss": 0.7605,
+      "step": 2380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010802469135802468,
+      "loss": 0.7297,
+      "step": 2385
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010781893004115226,
+      "loss": 0.7565,
+      "step": 2390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010761316872427984,
+      "loss": 0.701,
+      "step": 2395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010740740740740742,
+      "loss": 0.736,
+      "step": 2400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010720164609053497,
+      "loss": 0.7079,
+      "step": 2405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010699588477366255,
+      "loss": 0.7349,
+      "step": 2410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010679012345679013,
+      "loss": 0.7106,
+      "step": 2415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001065843621399177,
+      "loss": 0.7396,
+      "step": 2420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010637860082304528,
+      "loss": 0.7713,
+      "step": 2425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010617283950617284,
+      "loss": 0.747,
+      "step": 2430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010596707818930041,
+      "loss": 0.7475,
+      "step": 2435
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010576131687242799,
+      "loss": 0.7449,
+      "step": 2440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010555555555555557,
+      "loss": 0.7208,
+      "step": 2445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010534979423868315,
+      "loss": 0.7317,
+      "step": 2450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001051440329218107,
+      "loss": 0.7419,
+      "step": 2455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010493827160493828,
+      "loss": 0.7148,
+      "step": 2460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010473251028806586,
+      "loss": 0.7143,
+      "step": 2465
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010452674897119341,
+      "loss": 0.747,
+      "step": 2470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010432098765432099,
+      "loss": 0.7079,
+      "step": 2475
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010411522633744857,
+      "loss": 0.7059,
+      "step": 2480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010390946502057615,
+      "loss": 0.7272,
+      "step": 2485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001037037037037037,
+      "loss": 0.7281,
+      "step": 2490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010349794238683128,
+      "loss": 0.6956,
+      "step": 2495
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010329218106995886,
+      "loss": 0.7571,
+      "step": 2500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010308641975308641,
+      "loss": 0.7311,
+      "step": 2505
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010288065843621399,
+      "loss": 0.6815,
+      "step": 2510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010267489711934157,
+      "loss": 0.6781,
+      "step": 2515
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010246913580246915,
+      "loss": 0.6948,
+      "step": 2520
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010226337448559671,
+      "loss": 0.7133,
+      "step": 2525
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010205761316872428,
+      "loss": 0.688,
+      "step": 2530
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010185185185185186,
+      "loss": 0.6687,
+      "step": 2535
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010164609053497942,
+      "loss": 0.7142,
+      "step": 2540
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000101440329218107,
+      "loss": 0.6885,
+      "step": 2545
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010123456790123458,
+      "loss": 0.6867,
+      "step": 2550
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010102880658436214,
+      "loss": 0.7045,
+      "step": 2555
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010082304526748971,
+      "loss": 0.7375,
+      "step": 2560
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001006172839506173,
+      "loss": 0.7157,
+      "step": 2565
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010041152263374487,
+      "loss": 0.6828,
+      "step": 2570
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010020576131687243,
+      "loss": 0.7048,
+      "step": 2575
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001,
+      "loss": 0.6944,
+      "step": 2580
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.979423868312757e-05,
+      "loss": 0.696,
+      "step": 2585
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.958847736625515e-05,
+      "loss": 0.6609,
+      "step": 2590
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.938271604938272e-05,
+      "loss": 0.674,
+      "step": 2595
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.91769547325103e-05,
+      "loss": 0.7138,
+      "step": 2600
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.897119341563786e-05,
+      "loss": 0.713,
+      "step": 2605
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.876543209876543e-05,
+      "loss": 0.7022,
+      "step": 2610
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.8559670781893e-05,
+      "loss": 0.6955,
+      "step": 2615
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.835390946502057e-05,
+      "loss": 0.6888,
+      "step": 2620
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.814814814814815e-05,
+      "loss": 0.7003,
+      "step": 2625
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.794238683127573e-05,
+      "loss": 0.7167,
+      "step": 2630
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.77366255144033e-05,
+      "loss": 0.7087,
+      "step": 2635
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.753086419753087e-05,
+      "loss": 0.7019,
+      "step": 2640
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.732510288065844e-05,
+      "loss": 0.6609,
+      "step": 2645
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.711934156378602e-05,
+      "loss": 0.7367,
+      "step": 2650
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.691358024691359e-05,
+      "loss": 0.6862,
+      "step": 2655
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.670781893004116e-05,
+      "loss": 0.7121,
+      "step": 2660
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.650205761316873e-05,
+      "loss": 0.7031,
+      "step": 2665
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.62962962962963e-05,
+      "loss": 0.7005,
+      "step": 2670
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.609053497942388e-05,
+      "loss": 0.7041,
+      "step": 2675
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.588477366255144e-05,
+      "loss": 0.6713,
+      "step": 2680
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.567901234567902e-05,
+      "loss": 0.6932,
+      "step": 2685
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.547325102880659e-05,
+      "loss": 0.6849,
+      "step": 2690
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.526748971193417e-05,
+      "loss": 0.6915,
+      "step": 2695
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.506172839506173e-05,
+      "loss": 0.6677,
+      "step": 2700
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.48559670781893e-05,
+      "loss": 0.6767,
+      "step": 2705
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.465020576131688e-05,
+      "loss": 0.6873,
+      "step": 2710
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.444444444444444e-05,
+      "loss": 0.7007,
+      "step": 2715
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.423868312757202e-05,
+      "loss": 0.6782,
+      "step": 2720
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.403292181069959e-05,
+      "loss": 0.7044,
+      "step": 2725
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.382716049382717e-05,
+      "loss": 0.6787,
+      "step": 2730
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.362139917695473e-05,
+      "loss": 0.6876,
+      "step": 2735
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.341563786008231e-05,
+      "loss": 0.7253,
+      "step": 2740
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.320987654320989e-05,
+      "loss": 0.7159,
+      "step": 2745
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.300411522633746e-05,
+      "loss": 0.704,
+      "step": 2750
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.279835390946504e-05,
+      "loss": 0.6882,
+      "step": 2755
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.25925925925926e-05,
+      "loss": 0.6777,
+      "step": 2760
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.238683127572017e-05,
+      "loss": 0.7133,
+      "step": 2765
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.218106995884775e-05,
+      "loss": 0.6988,
+      "step": 2770
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.197530864197531e-05,
+      "loss": 0.7323,
+      "step": 2775
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.176954732510289e-05,
+      "loss": 0.6854,
+      "step": 2780
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.156378600823046e-05,
+      "loss": 0.6926,
+      "step": 2785
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.135802469135802e-05,
+      "loss": 0.6985,
+      "step": 2790
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.11522633744856e-05,
+      "loss": 0.6776,
+      "step": 2795
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.094650205761317e-05,
+      "loss": 0.7374,
+      "step": 2800
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.074074074074075e-05,
+      "loss": 0.6928,
+      "step": 2805
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.053497942386831e-05,
+      "loss": 0.7129,
+      "step": 2810
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.032921810699589e-05,
+      "loss": 0.6918,
+      "step": 2815
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.012345679012346e-05,
+      "loss": 0.6641,
+      "step": 2820
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.991769547325102e-05,
+      "loss": 0.7075,
+      "step": 2825
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.97119341563786e-05,
+      "loss": 0.668,
+      "step": 2830
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.950617283950617e-05,
+      "loss": 0.6836,
+      "step": 2835
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.930041152263375e-05,
+      "loss": 0.672,
+      "step": 2840
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.909465020576133e-05,
+      "loss": 0.6827,
+      "step": 2845
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.888888888888889e-05,
+      "loss": 0.6976,
+      "step": 2850
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.868312757201647e-05,
+      "loss": 0.6719,
+      "step": 2855
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.847736625514404e-05,
+      "loss": 0.6871,
+      "step": 2860
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.827160493827162e-05,
+      "loss": 0.6975,
+      "step": 2865
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.806584362139918e-05,
+      "loss": 0.684,
+      "step": 2870
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.786008230452676e-05,
+      "loss": 0.6861,
+      "step": 2875
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.765432098765433e-05,
+      "loss": 0.7116,
+      "step": 2880
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.74485596707819e-05,
+      "loss": 0.686,
+      "step": 2885
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.724279835390947e-05,
+      "loss": 0.677,
+      "step": 2890
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.703703703703704e-05,
+      "loss": 0.6996,
+      "step": 2895
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.683127572016462e-05,
+      "loss": 0.7211,
+      "step": 2900
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.662551440329218e-05,
+      "loss": 0.6766,
+      "step": 2905
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.641975308641975e-05,
+      "loss": 0.6883,
+      "step": 2910
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.621399176954733e-05,
+      "loss": 0.6829,
+      "step": 2915
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60082304526749e-05,
+      "loss": 0.7089,
+      "step": 2920
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.580246913580247e-05,
+      "loss": 0.6711,
+      "step": 2925
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.559670781893004e-05,
+      "loss": 0.6926,
+      "step": 2930
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.539094650205762e-05,
+      "loss": 0.6513,
+      "step": 2935
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.518518518518518e-05,
+      "loss": 0.6817,
+      "step": 2940
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.497942386831276e-05,
+      "loss": 0.7005,
+      "step": 2945
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.477366255144033e-05,
+      "loss": 0.7013,
+      "step": 2950
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.456790123456791e-05,
+      "loss": 0.6562,
+      "step": 2955
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.436213991769549e-05,
+      "loss": 0.7074,
+      "step": 2960
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.415637860082305e-05,
+      "loss": 0.6901,
+      "step": 2965
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.395061728395062e-05,
+      "loss": 0.6867,
+      "step": 2970
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.37448559670782e-05,
+      "loss": 0.692,
+      "step": 2975
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.353909465020576e-05,
+      "loss": 0.6863,
+      "step": 2980
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.333333333333334e-05,
+      "loss": 0.6881,
+      "step": 2985
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.312757201646091e-05,
+      "loss": 0.7203,
+      "step": 2990
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.292181069958848e-05,
+      "loss": 0.7075,
+      "step": 2995
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.271604938271605e-05,
+      "loss": 0.7187,
+      "step": 3000
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.251028806584362e-05,
+      "loss": 0.6791,
+      "step": 3005
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.23045267489712e-05,
+      "loss": 0.6937,
+      "step": 3010
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.209876543209877e-05,
+      "loss": 0.6584,
+      "step": 3015
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.189300411522634e-05,
+      "loss": 0.6818,
+      "step": 3020
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.168724279835391e-05,
+      "loss": 0.6827,
+      "step": 3025
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.148148148148148e-05,
+      "loss": 0.6774,
+      "step": 3030
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.127572016460906e-05,
+      "loss": 0.6826,
+      "step": 3035
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.106995884773662e-05,
+      "loss": 0.6767,
+      "step": 3040
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.08641975308642e-05,
+      "loss": 0.7256,
+      "step": 3045
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.065843621399177e-05,
+      "loss": 0.6875,
+      "step": 3050
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.045267489711935e-05,
+      "loss": 0.7013,
+      "step": 3055
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.024691358024692e-05,
+      "loss": 0.6907,
+      "step": 3060
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.004115226337449e-05,
+      "loss": 0.6751,
+      "step": 3065
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.983539094650207e-05,
+      "loss": 0.6893,
+      "step": 3070
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.962962962962964e-05,
+      "loss": 0.6962,
+      "step": 3075
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.942386831275721e-05,
+      "loss": 0.6743,
+      "step": 3080
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.921810699588478e-05,
+      "loss": 0.7004,
+      "step": 3085
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.901234567901235e-05,
+      "loss": 0.6908,
+      "step": 3090
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.880658436213993e-05,
+      "loss": 0.6736,
+      "step": 3095
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.860082304526749e-05,
+      "loss": 0.7153,
+      "step": 3100
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.839506172839507e-05,
+      "loss": 0.7072,
+      "step": 3105
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.818930041152264e-05,
+      "loss": 0.7163,
+      "step": 3110
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.79835390946502e-05,
+      "loss": 0.6996,
+      "step": 3115
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.777777777777778e-05,
+      "loss": 0.7257,
+      "step": 3120
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.757201646090535e-05,
+      "loss": 0.7155,
+      "step": 3125
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.736625514403293e-05,
+      "loss": 0.7068,
+      "step": 3130
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.716049382716049e-05,
+      "loss": 0.6575,
+      "step": 3135
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.695473251028807e-05,
+      "loss": 0.6771,
+      "step": 3140
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.674897119341564e-05,
+      "loss": 0.6976,
+      "step": 3145
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.65432098765432e-05,
+      "loss": 0.6568,
+      "step": 3150
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.633744855967078e-05,
+      "loss": 0.6609,
+      "step": 3155
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.613168724279836e-05,
+      "loss": 0.697,
+      "step": 3160
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.592592592592593e-05,
+      "loss": 0.714,
+      "step": 3165
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.57201646090535e-05,
+      "loss": 0.6829,
+      "step": 3170
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.551440329218107e-05,
+      "loss": 0.6932,
+      "step": 3175
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.530864197530865e-05,
+      "loss": 0.6859,
+      "step": 3180
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.510288065843622e-05,
+      "loss": 0.7117,
+      "step": 3185
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.48971193415638e-05,
+      "loss": 0.7157,
+      "step": 3190
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.469135802469136e-05,
+      "loss": 0.6593,
+      "step": 3195
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.448559670781894e-05,
+      "loss": 0.6683,
+      "step": 3200
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.427983539094651e-05,
+      "loss": 0.6876,
+      "step": 3205
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.407407407407407e-05,
+      "loss": 0.6584,
+      "step": 3210
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.386831275720165e-05,
+      "loss": 0.6816,
+      "step": 3215
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.366255144032922e-05,
+      "loss": 0.6994,
+      "step": 3220
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.34567901234568e-05,
+      "loss": 0.6701,
+      "step": 3225
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.325102880658436e-05,
+      "loss": 0.6869,
+      "step": 3230
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.304526748971193e-05,
+      "loss": 0.6895,
+      "step": 3235
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.283950617283951e-05,
+      "loss": 0.6972,
+      "step": 3240
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.263374485596707e-05,
+      "loss": 0.6849,
+      "step": 3245
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.242798353909465e-05,
+      "loss": 0.6967,
+      "step": 3250
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.222222222222222e-05,
+      "loss": 0.6679,
+      "step": 3255
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.20164609053498e-05,
+      "loss": 0.6683,
+      "step": 3260
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.181069958847736e-05,
+      "loss": 0.6787,
+      "step": 3265
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.160493827160494e-05,
+      "loss": 0.6726,
+      "step": 3270
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.139917695473252e-05,
+      "loss": 0.721,
+      "step": 3275
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.119341563786009e-05,
+      "loss": 0.6865,
+      "step": 3280
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.098765432098767e-05,
+      "loss": 0.7031,
+      "step": 3285
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.078189300411523e-05,
+      "loss": 0.6895,
+      "step": 3290
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.05761316872428e-05,
+      "loss": 0.6983,
+      "step": 3295
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.037037037037038e-05,
+      "loss": 0.6929,
+      "step": 3300
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.016460905349794e-05,
+      "loss": 0.7117,
+      "step": 3305
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.995884773662552e-05,
+      "loss": 0.664,
+      "step": 3310
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.975308641975309e-05,
+      "loss": 0.6645,
+      "step": 3315
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.954732510288067e-05,
+      "loss": 0.6985,
+      "step": 3320
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.934156378600823e-05,
+      "loss": 0.6826,
+      "step": 3325
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.91358024691358e-05,
+      "loss": 0.7324,
+      "step": 3330
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.893004115226338e-05,
+      "loss": 0.6828,
+      "step": 3335
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.872427983539094e-05,
+      "loss": 0.6801,
+      "step": 3340
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.851851851851852e-05,
+      "loss": 0.6923,
+      "step": 3345
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.831275720164609e-05,
+      "loss": 0.6787,
+      "step": 3350
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.810699588477366e-05,
+      "loss": 0.6812,
+      "step": 3355
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.790123456790123e-05,
+      "loss": 0.7059,
+      "step": 3360
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.76954732510288e-05,
+      "loss": 0.6794,
+      "step": 3365
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.748971193415638e-05,
+      "loss": 0.6635,
+      "step": 3370
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.728395061728396e-05,
+      "loss": 0.666,
+      "step": 3375
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.707818930041152e-05,
+      "loss": 0.6909,
+      "step": 3380
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.68724279835391e-05,
+      "loss": 0.7096,
+      "step": 3385
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.7049,
+      "step": 3390
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.646090534979425e-05,
+      "loss": 0.6801,
+      "step": 3395
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.625514403292181e-05,
+      "loss": 0.7045,
+      "step": 3400
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.60493827160494e-05,
+      "loss": 0.7096,
+      "step": 3405
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.584362139917696e-05,
+      "loss": 0.7192,
+      "step": 3410
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.563786008230453e-05,
+      "loss": 0.6818,
+      "step": 3415
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.54320987654321e-05,
+      "loss": 0.6358,
+      "step": 3420
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.522633744855967e-05,
+      "loss": 0.6674,
+      "step": 3425
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.502057613168725e-05,
+      "loss": 0.684,
+      "step": 3430
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.481481481481482e-05,
+      "loss": 0.6753,
+      "step": 3435
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.460905349794238e-05,
+      "loss": 0.6902,
+      "step": 3440
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.440329218106996e-05,
+      "loss": 0.6664,
+      "step": 3445
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.419753086419753e-05,
+      "loss": 0.684,
+      "step": 3450
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.39917695473251e-05,
+      "loss": 0.682,
+      "step": 3455
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.378600823045267e-05,
+      "loss": 0.6764,
+      "step": 3460
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.358024691358025e-05,
+      "loss": 0.6783,
+      "step": 3465
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.337448559670782e-05,
+      "loss": 0.6962,
+      "step": 3470
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.31687242798354e-05,
+      "loss": 0.6819,
+      "step": 3475
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.296296296296296e-05,
+      "loss": 0.6741,
+      "step": 3480
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.275720164609054e-05,
+      "loss": 0.6754,
+      "step": 3485
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.255144032921812e-05,
+      "loss": 0.6717,
+      "step": 3490
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.234567901234569e-05,
+      "loss": 0.667,
+      "step": 3495
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.213991769547325e-05,
+      "loss": 0.6822,
+      "step": 3500
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.193415637860083e-05,
+      "loss": 0.7055,
+      "step": 3505
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.17283950617284e-05,
+      "loss": 0.6988,
+      "step": 3510
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.152263374485598e-05,
+      "loss": 0.6793,
+      "step": 3515
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.131687242798354e-05,
+      "loss": 0.7093,
+      "step": 3520
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.111111111111112e-05,
+      "loss": 0.694,
+      "step": 3525
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0905349794238687e-05,
+      "loss": 0.7077,
+      "step": 3530
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.069958847736625e-05,
+      "loss": 0.6938,
+      "step": 3535
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.049382716049383e-05,
+      "loss": 0.6461,
+      "step": 3540
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.02880658436214e-05,
+      "loss": 0.6993,
+      "step": 3545
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0082304526748977e-05,
+      "loss": 0.6901,
+      "step": 3550
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.987654320987654e-05,
+      "loss": 0.6937,
+      "step": 3555
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9670781893004115e-05,
+      "loss": 0.6657,
+      "step": 3560
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9465020576131694e-05,
+      "loss": 0.6593,
+      "step": 3565
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.925925925925926e-05,
+      "loss": 0.7056,
+      "step": 3570
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.905349794238684e-05,
+      "loss": 0.6395,
+      "step": 3575
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8847736625514405e-05,
+      "loss": 0.6608,
+      "step": 3580
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8641975308641984e-05,
+      "loss": 0.7007,
+      "step": 3585
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.843621399176955e-05,
+      "loss": 0.6605,
+      "step": 3590
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8230452674897116e-05,
+      "loss": 0.6752,
+      "step": 3595
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.8024691358024695e-05,
+      "loss": 0.6736,
+      "step": 3600
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.781893004115226e-05,
+      "loss": 0.671,
+      "step": 3605
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.761316872427984e-05,
+      "loss": 0.6551,
+      "step": 3610
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.740740740740741e-05,
+      "loss": 0.7053,
+      "step": 3615
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.720164609053498e-05,
+      "loss": 0.6856,
+      "step": 3620
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.699588477366256e-05,
+      "loss": 0.671,
+      "step": 3625
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.679012345679012e-05,
+      "loss": 0.6763,
+      "step": 3630
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.65843621399177e-05,
+      "loss": 0.6812,
+      "step": 3635
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.637860082304527e-05,
+      "loss": 0.7407,
+      "step": 3640
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.617283950617285e-05,
+      "loss": 0.7436,
+      "step": 3645
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.596707818930041e-05,
+      "loss": 0.682,
+      "step": 3650
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.576131687242798e-05,
+      "loss": 0.6808,
+      "step": 3655
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.555555555555556e-05,
+      "loss": 0.6696,
+      "step": 3660
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.534979423868313e-05,
+      "loss": 0.7207,
+      "step": 3665
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.51440329218107e-05,
+      "loss": 0.6737,
+      "step": 3670
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4938271604938276e-05,
+      "loss": 0.6874,
+      "step": 3675
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.473251028806584e-05,
+      "loss": 0.6652,
+      "step": 3680
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.452674897119342e-05,
+      "loss": 0.6739,
+      "step": 3685
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4320987654320986e-05,
+      "loss": 0.7039,
+      "step": 3690
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4115226337448566e-05,
+      "loss": 0.696,
+      "step": 3695
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.390946502057613e-05,
+      "loss": 0.6887,
+      "step": 3700
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.370370370370371e-05,
+      "loss": 0.6653,
+      "step": 3705
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3497942386831277e-05,
+      "loss": 0.6987,
+      "step": 3710
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.329218106995885e-05,
+      "loss": 0.6831,
+      "step": 3715
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.308641975308642e-05,
+      "loss": 0.687,
+      "step": 3720
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2880658436213994e-05,
+      "loss": 0.6848,
+      "step": 3725
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.267489711934157e-05,
+      "loss": 0.6819,
+      "step": 3730
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.246913580246914e-05,
+      "loss": 0.662,
+      "step": 3735
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2263374485596705e-05,
+      "loss": 0.6462,
+      "step": 3740
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2057613168724284e-05,
+      "loss": 0.684,
+      "step": 3745
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.185185185185185e-05,
+      "loss": 0.721,
+      "step": 3750
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.164609053497943e-05,
+      "loss": 0.6998,
+      "step": 3755
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1440329218106995e-05,
+      "loss": 0.6922,
+      "step": 3760
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1234567901234574e-05,
+      "loss": 0.6674,
+      "step": 3765
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.102880658436214e-05,
+      "loss": 0.7022,
+      "step": 3770
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.082304526748971e-05,
+      "loss": 0.6753,
+      "step": 3775
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.061728395061729e-05,
+      "loss": 0.6853,
+      "step": 3780
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.041152263374486e-05,
+      "loss": 0.6669,
+      "step": 3785
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.020576131687244e-05,
+      "loss": 0.6678,
+      "step": 3790
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5e-05,
+      "loss": 0.6881,
+      "step": 3795
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9794238683127575e-05,
+      "loss": 0.6959,
+      "step": 3800
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.958847736625515e-05,
+      "loss": 0.695,
+      "step": 3805
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.938271604938271e-05,
+      "loss": 0.6927,
+      "step": 3810
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9176954732510286e-05,
+      "loss": 0.7039,
+      "step": 3815
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.8971193415637865e-05,
+      "loss": 0.6494,
+      "step": 3820
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.876543209876544e-05,
+      "loss": 0.7015,
+      "step": 3825
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.855967078189301e-05,
+      "loss": 0.6901,
+      "step": 3830
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.835390946502058e-05,
+      "loss": 0.6618,
+      "step": 3835
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.814814814814815e-05,
+      "loss": 0.6723,
+      "step": 3840
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.794238683127572e-05,
+      "loss": 0.7196,
+      "step": 3845
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.773662551440329e-05,
+      "loss": 0.6867,
+      "step": 3850
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7530864197530866e-05,
+      "loss": 0.6921,
+      "step": 3855
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.732510288065844e-05,
+      "loss": 0.693,
+      "step": 3860
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.711934156378601e-05,
+      "loss": 0.7031,
+      "step": 3865
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.691358024691358e-05,
+      "loss": 0.7023,
+      "step": 3870
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6707818930041156e-05,
+      "loss": 0.6754,
+      "step": 3875
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.650205761316873e-05,
+      "loss": 0.6985,
+      "step": 3880
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.62962962962963e-05,
+      "loss": 0.665,
+      "step": 3885
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.609053497942387e-05,
+      "loss": 0.6902,
+      "step": 3890
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.5884773662551446e-05,
+      "loss": 0.7173,
+      "step": 3895
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.567901234567901e-05,
+      "loss": 0.7121,
+      "step": 3900
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5473251028806584e-05,
+      "loss": 0.6939,
+      "step": 3905
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5267489711934157e-05,
+      "loss": 0.7093,
+      "step": 3910
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.506172839506173e-05,
+      "loss": 0.6792,
+      "step": 3915
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.48559670781893e-05,
+      "loss": 0.6737,
+      "step": 3920
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4650205761316874e-05,
+      "loss": 0.6858,
+      "step": 3925
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4444444444444447e-05,
+      "loss": 0.6776,
+      "step": 3930
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.423868312757202e-05,
+      "loss": 0.6772,
+      "step": 3935
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.403292181069959e-05,
+      "loss": 0.6667,
+      "step": 3940
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.3827160493827164e-05,
+      "loss": 0.6892,
+      "step": 3945
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3621399176954737e-05,
+      "loss": 0.6652,
+      "step": 3950
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.341563786008231e-05,
+      "loss": 0.6892,
+      "step": 3955
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3209876543209875e-05,
+      "loss": 0.6917,
+      "step": 3960
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.300411522633745e-05,
+      "loss": 0.6708,
+      "step": 3965
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.279835390946502e-05,
+      "loss": 0.6817,
+      "step": 3970
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.259259259259259e-05,
+      "loss": 0.6761,
+      "step": 3975
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2386831275720165e-05,
+      "loss": 0.6638,
+      "step": 3980
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2181069958847744e-05,
+      "loss": 0.6707,
+      "step": 3985
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.197530864197531e-05,
+      "loss": 0.6796,
+      "step": 3990
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.176954732510288e-05,
+      "loss": 0.6903,
+      "step": 3995
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1563786008230455e-05,
+      "loss": 0.7081,
+      "step": 4000
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 5010,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 500,
+  "total_flos": 1.6102337306911834e+19,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/training_args.bin b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e707059a03fbe62142ac41332e145ac1ac1f36cc
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b99a1729574e0e46ee782b42dc90101a785d5b4792871920d293f44722c28cc
+size 5944
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/zero_to_fp32.py b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/zero_to_fp32.py
new file mode 100644
index 0000000000000000000000000000000000000000..c98caae31534368be22b67fc4ae906836c992a8d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4000/zero_to_fp32.py
@@ -0,0 +1,587 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+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):
+    """
+    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)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+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 _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()
+            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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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 param_shapes.items():
+
+        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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    """
+    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``
+
+    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)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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("-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_file, tag=args.tag)
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/README.md b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..d380bbbf0bf0fa1f211b53410559607e5007a209
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: meta-llama/Llama-2-7b-hf
+---
+
+# 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 0.9.0
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/adapter_config.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c81a592f0ff08374a4f61f6003029070a8429ae
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/adapter_config.json
@@ -0,0 +1,35 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "mora_type": 1,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_mora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/adapter_model.safetensors b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/adapter_model.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..4e7b6a47757954321a3775bcacd88ea00df0673a
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/adapter_model.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70f45d542b9751aac50878963b44fc32c5f248fe038c74462deb7a35a752d354
+size 564342240
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/global_step4500/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/global_step4500/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8c75038a1daadd042aa2b0fbff47faf98be06720
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/global_step4500/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71432279295b06256962708fc569860b615fa9057cd3188a816efa7d765a97d3
+size 239914224
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/global_step4500/mp_rank_00_model_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/global_step4500/mp_rank_00_model_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..72c96092c73fd9ff83bffb466381a6fc29b4b997
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/global_step4500/mp_rank_00_model_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f18fb718aa658b625c69ca5508ef8e74bebeb071ae4a58db78f5578b3e55aa4a
+size 40303404
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/latest b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/latest
new file mode 100644
index 0000000000000000000000000000000000000000..982b8ddd9a759078550e0894b3ed06c6084c9e9c
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/latest
@@ -0,0 +1 @@
+global_step4500
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/rng_state.pth b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/rng_state.pth
new file mode 100644
index 0000000000000000000000000000000000000000..e016740571a6688a5f991c72573e1825d45089f4
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/rng_state.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19c63835baf588a511e53431b12916db783c534add75bfd00d5d75a5a7a999e6
+size 14244
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/scheduler.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/scheduler.pt
new file mode 100644
index 0000000000000000000000000000000000000000..20781897ef32ae8a6ca52e9b08a84851d316b4f5
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/scheduler.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57969ca0ff3a5006c6164f4d25d7aa5010c76c0025d089be4262a3e81c08ba92
+size 1064
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/trainer_state.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..10564fbeb7c02f086e8d98e1ad95ea8884a7fe4d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/trainer_state.json
@@ -0,0 +1,5421 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.796239124333427,
+  "eval_steps": 500,
+  "global_step": 4500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.3797,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.3952,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2e-05,
+      "loss": 1.3112,
+      "step": 15
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 1.3136,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 1.2012,
+      "step": 25
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4e-05,
+      "loss": 1.1277,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 1.0832,
+      "step": 35
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.333333333333333e-05,
+      "loss": 1.0659,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-05,
+      "loss": 0.983,
+      "step": 45
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.9763,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.333333333333333e-05,
+      "loss": 0.9992,
+      "step": 55
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8e-05,
+      "loss": 0.9405,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.666666666666667e-05,
+      "loss": 0.9827,
+      "step": 65
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.333333333333334e-05,
+      "loss": 0.9061,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001,
+      "loss": 0.8921,
+      "step": 75
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010666666666666667,
+      "loss": 0.9377,
+      "step": 80
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011333333333333334,
+      "loss": 0.8528,
+      "step": 85
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012,
+      "loss": 0.8913,
+      "step": 90
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012666666666666666,
+      "loss": 0.8562,
+      "step": 95
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.877,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014,
+      "loss": 0.8781,
+      "step": 105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014666666666666666,
+      "loss": 0.8546,
+      "step": 110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015333333333333334,
+      "loss": 0.8672,
+      "step": 115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016,
+      "loss": 0.8611,
+      "step": 120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.8624,
+      "step": 125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017333333333333334,
+      "loss": 0.8477,
+      "step": 130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018,
+      "loss": 0.8227,
+      "step": 135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001866666666666667,
+      "loss": 0.8064,
+      "step": 140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019333333333333333,
+      "loss": 0.8422,
+      "step": 145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002,
+      "loss": 0.8471,
+      "step": 150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001997942386831276,
+      "loss": 0.8183,
+      "step": 155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958847736625514,
+      "loss": 0.8229,
+      "step": 160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019938271604938272,
+      "loss": 0.8114,
+      "step": 165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991769547325103,
+      "loss": 0.7861,
+      "step": 170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019897119341563788,
+      "loss": 0.841,
+      "step": 175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019876543209876543,
+      "loss": 0.8549,
+      "step": 180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000198559670781893,
+      "loss": 0.8449,
+      "step": 185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001983539094650206,
+      "loss": 0.8311,
+      "step": 190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019814814814814814,
+      "loss": 0.8296,
+      "step": 195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019794238683127572,
+      "loss": 0.8263,
+      "step": 200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001977366255144033,
+      "loss": 0.8452,
+      "step": 205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019753086419753085,
+      "loss": 0.8044,
+      "step": 210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019732510288065843,
+      "loss": 0.8119,
+      "step": 215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000197119341563786,
+      "loss": 0.8127,
+      "step": 220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001969135802469136,
+      "loss": 0.7921,
+      "step": 225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019670781893004114,
+      "loss": 0.803,
+      "step": 230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019650205761316872,
+      "loss": 0.838,
+      "step": 235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001962962962962963,
+      "loss": 0.7801,
+      "step": 240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019609053497942388,
+      "loss": 0.817,
+      "step": 245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019588477366255146,
+      "loss": 0.8211,
+      "step": 250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000195679012345679,
+      "loss": 0.8423,
+      "step": 255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001954732510288066,
+      "loss": 0.8007,
+      "step": 260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019526748971193417,
+      "loss": 0.8135,
+      "step": 265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019506172839506175,
+      "loss": 0.7961,
+      "step": 270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019485596707818933,
+      "loss": 0.797,
+      "step": 275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019465020576131688,
+      "loss": 0.793,
+      "step": 280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019444444444444446,
+      "loss": 0.7809,
+      "step": 285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019423868312757204,
+      "loss": 0.7811,
+      "step": 290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019403292181069962,
+      "loss": 0.8074,
+      "step": 295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019382716049382717,
+      "loss": 0.8028,
+      "step": 300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019362139917695475,
+      "loss": 0.7895,
+      "step": 305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019341563786008233,
+      "loss": 0.7777,
+      "step": 310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019320987654320988,
+      "loss": 0.7868,
+      "step": 315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019300411522633746,
+      "loss": 0.7921,
+      "step": 320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019279835390946504,
+      "loss": 0.8,
+      "step": 325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001925925925925926,
+      "loss": 0.7781,
+      "step": 330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019238683127572017,
+      "loss": 0.7799,
+      "step": 335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019218106995884775,
+      "loss": 0.7921,
+      "step": 340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019197530864197533,
+      "loss": 0.7851,
+      "step": 345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019176954732510288,
+      "loss": 0.8106,
+      "step": 350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019156378600823046,
+      "loss": 0.7953,
+      "step": 355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019135802469135804,
+      "loss": 0.8084,
+      "step": 360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001911522633744856,
+      "loss": 0.8298,
+      "step": 365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019094650205761317,
+      "loss": 0.8273,
+      "step": 370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019074074074074075,
+      "loss": 0.8047,
+      "step": 375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019053497942386833,
+      "loss": 0.8108,
+      "step": 380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019032921810699588,
+      "loss": 0.8207,
+      "step": 385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019012345679012346,
+      "loss": 0.7884,
+      "step": 390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018991769547325104,
+      "loss": 0.7918,
+      "step": 395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001897119341563786,
+      "loss": 0.7554,
+      "step": 400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018950617283950617,
+      "loss": 0.8099,
+      "step": 405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018930041152263375,
+      "loss": 0.807,
+      "step": 410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018909465020576133,
+      "loss": 0.7793,
+      "step": 415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018888888888888888,
+      "loss": 0.7734,
+      "step": 420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018868312757201646,
+      "loss": 0.7786,
+      "step": 425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018847736625514404,
+      "loss": 0.7548,
+      "step": 430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001882716049382716,
+      "loss": 0.8024,
+      "step": 435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018806584362139917,
+      "loss": 0.7708,
+      "step": 440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018786008230452675,
+      "loss": 0.8045,
+      "step": 445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018765432098765433,
+      "loss": 0.7477,
+      "step": 450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018744855967078189,
+      "loss": 0.7474,
+      "step": 455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018724279835390946,
+      "loss": 0.7629,
+      "step": 460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018703703703703704,
+      "loss": 0.7777,
+      "step": 465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018683127572016462,
+      "loss": 0.7823,
+      "step": 470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001866255144032922,
+      "loss": 0.7762,
+      "step": 475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018641975308641978,
+      "loss": 0.782,
+      "step": 480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018621399176954733,
+      "loss": 0.7783,
+      "step": 485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001860082304526749,
+      "loss": 0.7376,
+      "step": 490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001858024691358025,
+      "loss": 0.7683,
+      "step": 495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018559670781893007,
+      "loss": 0.7547,
+      "step": 500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018539094650205762,
+      "loss": 0.7569,
+      "step": 505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001851851851851852,
+      "loss": 0.7796,
+      "step": 510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018497942386831278,
+      "loss": 0.7888,
+      "step": 515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018477366255144033,
+      "loss": 0.7763,
+      "step": 520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018456790123456791,
+      "loss": 0.7787,
+      "step": 525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001843621399176955,
+      "loss": 0.772,
+      "step": 530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018415637860082307,
+      "loss": 0.786,
+      "step": 535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018395061728395062,
+      "loss": 0.7639,
+      "step": 540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001837448559670782,
+      "loss": 0.775,
+      "step": 545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018353909465020578,
+      "loss": 0.8002,
+      "step": 550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 0.7359,
+      "step": 555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018312757201646091,
+      "loss": 0.747,
+      "step": 560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001829218106995885,
+      "loss": 0.8234,
+      "step": 565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018271604938271605,
+      "loss": 0.7586,
+      "step": 570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018251028806584363,
+      "loss": 0.7315,
+      "step": 575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001823045267489712,
+      "loss": 0.7986,
+      "step": 580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018209876543209878,
+      "loss": 0.7758,
+      "step": 585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018189300411522634,
+      "loss": 0.7869,
+      "step": 590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018168724279835392,
+      "loss": 0.7596,
+      "step": 595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001814814814814815,
+      "loss": 0.7658,
+      "step": 600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018127572016460905,
+      "loss": 0.7851,
+      "step": 605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018106995884773663,
+      "loss": 0.7957,
+      "step": 610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001808641975308642,
+      "loss": 0.7756,
+      "step": 615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018065843621399178,
+      "loss": 0.7911,
+      "step": 620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018045267489711934,
+      "loss": 0.7864,
+      "step": 625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018024691358024692,
+      "loss": 0.7683,
+      "step": 630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001800411522633745,
+      "loss": 0.7731,
+      "step": 635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017983539094650205,
+      "loss": 0.7671,
+      "step": 640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017962962962962963,
+      "loss": 0.7367,
+      "step": 645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001794238683127572,
+      "loss": 0.8024,
+      "step": 650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017921810699588476,
+      "loss": 0.7572,
+      "step": 655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017901234567901234,
+      "loss": 0.7554,
+      "step": 660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017880658436213992,
+      "loss": 0.7677,
+      "step": 665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001786008230452675,
+      "loss": 0.7652,
+      "step": 670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017839506172839508,
+      "loss": 0.7771,
+      "step": 675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017818930041152265,
+      "loss": 0.7747,
+      "step": 680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001779835390946502,
+      "loss": 0.767,
+      "step": 685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017777777777777779,
+      "loss": 0.7738,
+      "step": 690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017757201646090537,
+      "loss": 0.7271,
+      "step": 695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017736625514403294,
+      "loss": 0.7316,
+      "step": 700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017716049382716052,
+      "loss": 0.7831,
+      "step": 705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017695473251028808,
+      "loss": 0.7835,
+      "step": 710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017674897119341566,
+      "loss": 0.7497,
+      "step": 715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017654320987654323,
+      "loss": 0.7757,
+      "step": 720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001763374485596708,
+      "loss": 0.74,
+      "step": 725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017613168724279837,
+      "loss": 0.7697,
+      "step": 730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017592592592592595,
+      "loss": 0.774,
+      "step": 735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017572016460905352,
+      "loss": 0.7668,
+      "step": 740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017551440329218108,
+      "loss": 0.7779,
+      "step": 745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017530864197530866,
+      "loss": 0.7441,
+      "step": 750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017510288065843624,
+      "loss": 0.7578,
+      "step": 755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001748971193415638,
+      "loss": 0.8358,
+      "step": 760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017469135802469137,
+      "loss": 0.7473,
+      "step": 765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017448559670781895,
+      "loss": 0.7517,
+      "step": 770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001742798353909465,
+      "loss": 0.7593,
+      "step": 775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017407407407407408,
+      "loss": 0.7251,
+      "step": 780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017386831275720166,
+      "loss": 0.7365,
+      "step": 785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017366255144032924,
+      "loss": 0.7798,
+      "step": 790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001734567901234568,
+      "loss": 0.746,
+      "step": 795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017325102880658437,
+      "loss": 0.7607,
+      "step": 800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017304526748971195,
+      "loss": 0.7694,
+      "step": 805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001728395061728395,
+      "loss": 0.7926,
+      "step": 810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017263374485596708,
+      "loss": 0.795,
+      "step": 815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017242798353909466,
+      "loss": 0.7581,
+      "step": 820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017222222222222224,
+      "loss": 0.7629,
+      "step": 825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001720164609053498,
+      "loss": 0.7526,
+      "step": 830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017181069958847737,
+      "loss": 0.7338,
+      "step": 835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017160493827160495,
+      "loss": 0.7544,
+      "step": 840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001713991769547325,
+      "loss": 0.7522,
+      "step": 845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017119341563786008,
+      "loss": 0.7639,
+      "step": 850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017098765432098766,
+      "loss": 0.7914,
+      "step": 855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017078189300411524,
+      "loss": 0.7635,
+      "step": 860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001705761316872428,
+      "loss": 0.8343,
+      "step": 865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017037037037037037,
+      "loss": 0.7299,
+      "step": 870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017016460905349795,
+      "loss": 0.7643,
+      "step": 875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016995884773662553,
+      "loss": 0.7343,
+      "step": 880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016975308641975308,
+      "loss": 0.7546,
+      "step": 885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016954732510288066,
+      "loss": 0.7435,
+      "step": 890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016934156378600824,
+      "loss": 0.8086,
+      "step": 895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016913580246913582,
+      "loss": 0.7274,
+      "step": 900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001689300411522634,
+      "loss": 0.7511,
+      "step": 905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016872427983539098,
+      "loss": 0.7689,
+      "step": 910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016851851851851853,
+      "loss": 0.754,
+      "step": 915
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001683127572016461,
+      "loss": 0.7298,
+      "step": 920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001681069958847737,
+      "loss": 0.7755,
+      "step": 925
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016790123456790124,
+      "loss": 0.7854,
+      "step": 930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016769547325102882,
+      "loss": 0.7452,
+      "step": 935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001674897119341564,
+      "loss": 0.7473,
+      "step": 940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016728395061728398,
+      "loss": 0.7855,
+      "step": 945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016707818930041153,
+      "loss": 0.7527,
+      "step": 950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001668724279835391,
+      "loss": 0.7641,
+      "step": 955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.828,
+      "step": 960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016646090534979424,
+      "loss": 0.7571,
+      "step": 965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016625514403292182,
+      "loss": 0.7768,
+      "step": 970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001660493827160494,
+      "loss": 0.7644,
+      "step": 975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016584362139917695,
+      "loss": 0.7747,
+      "step": 980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016563786008230453,
+      "loss": 0.7926,
+      "step": 985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001654320987654321,
+      "loss": 0.7645,
+      "step": 990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001652263374485597,
+      "loss": 0.7464,
+      "step": 995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016502057613168724,
+      "loss": 0.7558,
+      "step": 1000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016481481481481482,
+      "loss": 0.7574,
+      "step": 1005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001646090534979424,
+      "loss": 0.7491,
+      "step": 1010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016440329218106995,
+      "loss": 0.756,
+      "step": 1015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016419753086419753,
+      "loss": 0.7679,
+      "step": 1020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001639917695473251,
+      "loss": 0.776,
+      "step": 1025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001637860082304527,
+      "loss": 0.7235,
+      "step": 1030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016358024691358024,
+      "loss": 0.7817,
+      "step": 1035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016337448559670782,
+      "loss": 0.7333,
+      "step": 1040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001631687242798354,
+      "loss": 0.754,
+      "step": 1045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016296296296296295,
+      "loss": 0.74,
+      "step": 1050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016275720164609053,
+      "loss": 0.7857,
+      "step": 1055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001625514403292181,
+      "loss": 0.8015,
+      "step": 1060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001623456790123457,
+      "loss": 0.7508,
+      "step": 1065
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016213991769547324,
+      "loss": 0.7514,
+      "step": 1070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016193415637860082,
+      "loss": 0.7326,
+      "step": 1075
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001617283950617284,
+      "loss": 0.7322,
+      "step": 1080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016152263374485595,
+      "loss": 0.7373,
+      "step": 1085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016131687242798353,
+      "loss": 0.7637,
+      "step": 1090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001611111111111111,
+      "loss": 0.7697,
+      "step": 1095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001609053497942387,
+      "loss": 0.7505,
+      "step": 1100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016069958847736627,
+      "loss": 0.7568,
+      "step": 1105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016049382716049385,
+      "loss": 0.786,
+      "step": 1110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001602880658436214,
+      "loss": 0.7525,
+      "step": 1115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016008230452674898,
+      "loss": 0.7696,
+      "step": 1120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015987654320987656,
+      "loss": 0.7648,
+      "step": 1125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015967078189300414,
+      "loss": 0.7492,
+      "step": 1130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001594650205761317,
+      "loss": 0.809,
+      "step": 1135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015925925925925927,
+      "loss": 0.7592,
+      "step": 1140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015905349794238685,
+      "loss": 0.7874,
+      "step": 1145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015884773662551443,
+      "loss": 0.7874,
+      "step": 1150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015864197530864198,
+      "loss": 0.7374,
+      "step": 1155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015843621399176956,
+      "loss": 0.7218,
+      "step": 1160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015823045267489714,
+      "loss": 0.7494,
+      "step": 1165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001580246913580247,
+      "loss": 0.7449,
+      "step": 1170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015781893004115227,
+      "loss": 0.725,
+      "step": 1175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015761316872427985,
+      "loss": 0.739,
+      "step": 1180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015740740740740743,
+      "loss": 0.7554,
+      "step": 1185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015720164609053498,
+      "loss": 0.7702,
+      "step": 1190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015699588477366256,
+      "loss": 0.7418,
+      "step": 1195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015679012345679014,
+      "loss": 0.7217,
+      "step": 1200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001565843621399177,
+      "loss": 0.7391,
+      "step": 1205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015637860082304527,
+      "loss": 0.7696,
+      "step": 1210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015617283950617285,
+      "loss": 0.7594,
+      "step": 1215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001559670781893004,
+      "loss": 0.777,
+      "step": 1220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015576131687242798,
+      "loss": 0.7441,
+      "step": 1225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015555555555555556,
+      "loss": 0.7611,
+      "step": 1230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015534979423868314,
+      "loss": 0.7278,
+      "step": 1235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001551440329218107,
+      "loss": 0.7452,
+      "step": 1240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015493827160493827,
+      "loss": 0.7939,
+      "step": 1245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015473251028806585,
+      "loss": 0.7161,
+      "step": 1250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001545267489711934,
+      "loss": 0.7344,
+      "step": 1255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015432098765432098,
+      "loss": 0.7368,
+      "step": 1260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015411522633744856,
+      "loss": 0.7172,
+      "step": 1265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015390946502057614,
+      "loss": 0.7589,
+      "step": 1270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001537037037037037,
+      "loss": 0.7641,
+      "step": 1275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015349794238683127,
+      "loss": 0.761,
+      "step": 1280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015329218106995885,
+      "loss": 0.7414,
+      "step": 1285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001530864197530864,
+      "loss": 0.7713,
+      "step": 1290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015288065843621399,
+      "loss": 0.7543,
+      "step": 1295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015267489711934156,
+      "loss": 0.7683,
+      "step": 1300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015246913580246914,
+      "loss": 0.8106,
+      "step": 1305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015226337448559672,
+      "loss": 0.7739,
+      "step": 1310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015205761316872428,
+      "loss": 0.7585,
+      "step": 1315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015185185185185185,
+      "loss": 0.7577,
+      "step": 1320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015164609053497943,
+      "loss": 0.7608,
+      "step": 1325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000151440329218107,
+      "loss": 0.7715,
+      "step": 1330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001512345679012346,
+      "loss": 0.7681,
+      "step": 1335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015102880658436214,
+      "loss": 0.7476,
+      "step": 1340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015082304526748972,
+      "loss": 0.7539,
+      "step": 1345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001506172839506173,
+      "loss": 0.7492,
+      "step": 1350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015041152263374488,
+      "loss": 0.7491,
+      "step": 1355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015020576131687243,
+      "loss": 0.7236,
+      "step": 1360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 0.7328,
+      "step": 1365
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001497942386831276,
+      "loss": 0.7603,
+      "step": 1370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014958847736625515,
+      "loss": 0.7206,
+      "step": 1375
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014938271604938272,
+      "loss": 0.7001,
+      "step": 1380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001491769547325103,
+      "loss": 0.7241,
+      "step": 1385
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014897119341563788,
+      "loss": 0.7511,
+      "step": 1390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014876543209876544,
+      "loss": 0.7546,
+      "step": 1395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014855967078189301,
+      "loss": 0.762,
+      "step": 1400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001483539094650206,
+      "loss": 0.6979,
+      "step": 1405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014814814814814815,
+      "loss": 0.7352,
+      "step": 1410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014794238683127573,
+      "loss": 0.7604,
+      "step": 1415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001477366255144033,
+      "loss": 0.7612,
+      "step": 1420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014753086419753086,
+      "loss": 0.7618,
+      "step": 1425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014732510288065844,
+      "loss": 0.7107,
+      "step": 1430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014711934156378602,
+      "loss": 0.7247,
+      "step": 1435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001469135802469136,
+      "loss": 0.7733,
+      "step": 1440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014670781893004115,
+      "loss": 0.7627,
+      "step": 1445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014650205761316873,
+      "loss": 0.7323,
+      "step": 1450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001462962962962963,
+      "loss": 0.7728,
+      "step": 1455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014609053497942386,
+      "loss": 0.7541,
+      "step": 1460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014588477366255144,
+      "loss": 0.7781,
+      "step": 1465
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014567901234567902,
+      "loss": 0.7474,
+      "step": 1470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001454732510288066,
+      "loss": 0.7789,
+      "step": 1475
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014526748971193415,
+      "loss": 0.7497,
+      "step": 1480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014506172839506173,
+      "loss": 0.7487,
+      "step": 1485
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001448559670781893,
+      "loss": 0.7563,
+      "step": 1490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014465020576131686,
+      "loss": 0.7444,
+      "step": 1495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014444444444444444,
+      "loss": 0.7399,
+      "step": 1500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014423868312757202,
+      "loss": 0.7288,
+      "step": 1505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001440329218106996,
+      "loss": 0.7458,
+      "step": 1510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014382716049382718,
+      "loss": 0.7297,
+      "step": 1515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014362139917695473,
+      "loss": 0.7684,
+      "step": 1520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001434156378600823,
+      "loss": 0.7428,
+      "step": 1525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014320987654320989,
+      "loss": 0.705,
+      "step": 1530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014300411522633747,
+      "loss": 0.7445,
+      "step": 1535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014279835390946504,
+      "loss": 0.7109,
+      "step": 1540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001425925925925926,
+      "loss": 0.7502,
+      "step": 1545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014238683127572018,
+      "loss": 0.7485,
+      "step": 1550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014218106995884776,
+      "loss": 0.7421,
+      "step": 1555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014197530864197534,
+      "loss": 0.7369,
+      "step": 1560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001417695473251029,
+      "loss": 0.7746,
+      "step": 1565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014156378600823047,
+      "loss": 0.7334,
+      "step": 1570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014135802469135805,
+      "loss": 0.7251,
+      "step": 1575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001411522633744856,
+      "loss": 0.7146,
+      "step": 1580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014094650205761318,
+      "loss": 0.7563,
+      "step": 1585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014074074074074076,
+      "loss": 0.7247,
+      "step": 1590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014053497942386834,
+      "loss": 0.7379,
+      "step": 1595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001403292181069959,
+      "loss": 0.7684,
+      "step": 1600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014012345679012347,
+      "loss": 0.7574,
+      "step": 1605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00013991769547325105,
+      "loss": 0.7138,
+      "step": 1610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001397119341563786,
+      "loss": 0.7472,
+      "step": 1615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013950617283950618,
+      "loss": 0.7466,
+      "step": 1620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013930041152263376,
+      "loss": 0.7288,
+      "step": 1625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013909465020576134,
+      "loss": 0.7717,
+      "step": 1630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001388888888888889,
+      "loss": 0.7665,
+      "step": 1635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013868312757201647,
+      "loss": 0.7408,
+      "step": 1640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013847736625514405,
+      "loss": 0.7418,
+      "step": 1645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001382716049382716,
+      "loss": 0.7502,
+      "step": 1650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013806584362139918,
+      "loss": 0.7657,
+      "step": 1655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013786008230452676,
+      "loss": 0.728,
+      "step": 1660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001376543209876543,
+      "loss": 0.7142,
+      "step": 1665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001374485596707819,
+      "loss": 0.7971,
+      "step": 1670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013724279835390947,
+      "loss": 0.737,
+      "step": 1675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013703703703703705,
+      "loss": 0.7947,
+      "step": 1680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001368312757201646,
+      "loss": 0.7525,
+      "step": 1685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013662551440329218,
+      "loss": 0.711,
+      "step": 1690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013641975308641976,
+      "loss": 0.7459,
+      "step": 1695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001362139917695473,
+      "loss": 0.7404,
+      "step": 1700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001360082304526749,
+      "loss": 0.7458,
+      "step": 1705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013580246913580247,
+      "loss": 0.754,
+      "step": 1710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013559670781893005,
+      "loss": 0.7217,
+      "step": 1715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001353909465020576,
+      "loss": 0.7373,
+      "step": 1720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013518518518518518,
+      "loss": 0.7567,
+      "step": 1725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013497942386831276,
+      "loss": 0.7525,
+      "step": 1730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013477366255144034,
+      "loss": 0.752,
+      "step": 1735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013456790123456792,
+      "loss": 0.7342,
+      "step": 1740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001343621399176955,
+      "loss": 0.7276,
+      "step": 1745
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013415637860082305,
+      "loss": 0.6915,
+      "step": 1750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013395061728395063,
+      "loss": 0.737,
+      "step": 1755
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001337448559670782,
+      "loss": 0.7427,
+      "step": 1760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001335390946502058,
+      "loss": 0.7126,
+      "step": 1765
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.7241,
+      "step": 1770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013312757201646092,
+      "loss": 0.755,
+      "step": 1775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001329218106995885,
+      "loss": 0.7494,
+      "step": 1780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013271604938271605,
+      "loss": 0.709,
+      "step": 1785
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013251028806584363,
+      "loss": 0.7555,
+      "step": 1790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001323045267489712,
+      "loss": 0.7298,
+      "step": 1795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001320987654320988,
+      "loss": 0.7232,
+      "step": 1800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013189300411522634,
+      "loss": 0.7441,
+      "step": 1805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013168724279835392,
+      "loss": 0.7623,
+      "step": 1810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001314814814814815,
+      "loss": 0.7349,
+      "step": 1815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013127572016460905,
+      "loss": 0.7773,
+      "step": 1820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013106995884773663,
+      "loss": 0.7537,
+      "step": 1825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001308641975308642,
+      "loss": 0.7579,
+      "step": 1830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001306584362139918,
+      "loss": 0.7321,
+      "step": 1835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013045267489711934,
+      "loss": 0.7175,
+      "step": 1840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00013024691358024692,
+      "loss": 0.7553,
+      "step": 1845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001300411522633745,
+      "loss": 0.7591,
+      "step": 1850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012983539094650205,
+      "loss": 0.7581,
+      "step": 1855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012962962962962963,
+      "loss": 0.7632,
+      "step": 1860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001294238683127572,
+      "loss": 0.7434,
+      "step": 1865
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012921810699588476,
+      "loss": 0.7695,
+      "step": 1870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012901234567901234,
+      "loss": 0.7808,
+      "step": 1875
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012880658436213992,
+      "loss": 0.7428,
+      "step": 1880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001286008230452675,
+      "loss": 0.7487,
+      "step": 1885
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012839506172839505,
+      "loss": 0.7033,
+      "step": 1890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012818930041152263,
+      "loss": 0.7639,
+      "step": 1895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001279835390946502,
+      "loss": 0.7345,
+      "step": 1900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012777777777777776,
+      "loss": 0.7434,
+      "step": 1905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012757201646090534,
+      "loss": 0.7082,
+      "step": 1910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012736625514403292,
+      "loss": 0.7294,
+      "step": 1915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001271604938271605,
+      "loss": 0.74,
+      "step": 1920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012695473251028805,
+      "loss": 0.7694,
+      "step": 1925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012674897119341563,
+      "loss": 0.7339,
+      "step": 1930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001265432098765432,
+      "loss": 0.7259,
+      "step": 1935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001263374485596708,
+      "loss": 0.7307,
+      "step": 1940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012613168724279837,
+      "loss": 0.7085,
+      "step": 1945
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012592592592592592,
+      "loss": 0.7662,
+      "step": 1950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001257201646090535,
+      "loss": 0.7145,
+      "step": 1955
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012551440329218108,
+      "loss": 0.7528,
+      "step": 1960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012530864197530866,
+      "loss": 0.7308,
+      "step": 1965
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012510288065843624,
+      "loss": 0.7163,
+      "step": 1970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001248971193415638,
+      "loss": 0.7254,
+      "step": 1975
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012469135802469137,
+      "loss": 0.7467,
+      "step": 1980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012448559670781895,
+      "loss": 0.7391,
+      "step": 1985
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001242798353909465,
+      "loss": 0.7314,
+      "step": 1990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012407407407407408,
+      "loss": 0.7048,
+      "step": 1995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012386831275720166,
+      "loss": 0.7187,
+      "step": 2000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012366255144032924,
+      "loss": 0.735,
+      "step": 2005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001234567901234568,
+      "loss": 0.7662,
+      "step": 2010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012325102880658437,
+      "loss": 0.7437,
+      "step": 2015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012304526748971195,
+      "loss": 0.7675,
+      "step": 2020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001228395061728395,
+      "loss": 0.7143,
+      "step": 2025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012263374485596708,
+      "loss": 0.7101,
+      "step": 2030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012242798353909466,
+      "loss": 0.7257,
+      "step": 2035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012222222222222224,
+      "loss": 0.7522,
+      "step": 2040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001220164609053498,
+      "loss": 0.7669,
+      "step": 2045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012181069958847737,
+      "loss": 0.7283,
+      "step": 2050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012160493827160495,
+      "loss": 0.7413,
+      "step": 2055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001213991769547325,
+      "loss": 0.6988,
+      "step": 2060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012119341563786008,
+      "loss": 0.7421,
+      "step": 2065
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012098765432098766,
+      "loss": 0.7152,
+      "step": 2070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012078189300411524,
+      "loss": 0.7741,
+      "step": 2075
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001205761316872428,
+      "loss": 0.7381,
+      "step": 2080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012037037037037037,
+      "loss": 0.738,
+      "step": 2085
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012016460905349795,
+      "loss": 0.7258,
+      "step": 2090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011995884773662552,
+      "loss": 0.7426,
+      "step": 2095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011975308641975308,
+      "loss": 0.7185,
+      "step": 2100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011954732510288066,
+      "loss": 0.7401,
+      "step": 2105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011934156378600823,
+      "loss": 0.751,
+      "step": 2110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011913580246913581,
+      "loss": 0.7284,
+      "step": 2115
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011893004115226339,
+      "loss": 0.7901,
+      "step": 2120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011872427983539097,
+      "loss": 0.7432,
+      "step": 2125
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011851851851851852,
+      "loss": 0.7355,
+      "step": 2130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001183127572016461,
+      "loss": 0.7333,
+      "step": 2135
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011810699588477368,
+      "loss": 0.7504,
+      "step": 2140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011790123456790123,
+      "loss": 0.7503,
+      "step": 2145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011769547325102881,
+      "loss": 0.728,
+      "step": 2150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011748971193415639,
+      "loss": 0.7485,
+      "step": 2155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011728395061728397,
+      "loss": 0.7127,
+      "step": 2160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011707818930041152,
+      "loss": 0.7278,
+      "step": 2165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001168724279835391,
+      "loss": 0.7513,
+      "step": 2170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011666666666666668,
+      "loss": 0.7193,
+      "step": 2175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011646090534979423,
+      "loss": 0.7245,
+      "step": 2180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011625514403292181,
+      "loss": 0.7266,
+      "step": 2185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011604938271604939,
+      "loss": 0.7531,
+      "step": 2190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011584362139917696,
+      "loss": 0.7637,
+      "step": 2195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011563786008230452,
+      "loss": 0.7388,
+      "step": 2200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001154320987654321,
+      "loss": 0.7222,
+      "step": 2205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011522633744855968,
+      "loss": 0.7667,
+      "step": 2210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011502057613168725,
+      "loss": 0.7382,
+      "step": 2215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011481481481481482,
+      "loss": 0.7243,
+      "step": 2220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001146090534979424,
+      "loss": 0.7192,
+      "step": 2225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011440329218106996,
+      "loss": 0.7237,
+      "step": 2230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011419753086419754,
+      "loss": 0.7498,
+      "step": 2235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011399176954732511,
+      "loss": 0.7007,
+      "step": 2240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001137860082304527,
+      "loss": 0.758,
+      "step": 2245
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011358024691358025,
+      "loss": 0.7111,
+      "step": 2250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011337448559670783,
+      "loss": 0.706,
+      "step": 2255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001131687242798354,
+      "loss": 0.7423,
+      "step": 2260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011296296296296296,
+      "loss": 0.7094,
+      "step": 2265
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011275720164609054,
+      "loss": 0.7331,
+      "step": 2270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011255144032921812,
+      "loss": 0.7164,
+      "step": 2275
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001123456790123457,
+      "loss": 0.7452,
+      "step": 2280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011213991769547325,
+      "loss": 0.7172,
+      "step": 2285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011193415637860083,
+      "loss": 0.7232,
+      "step": 2290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001117283950617284,
+      "loss": 0.7112,
+      "step": 2295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011152263374485596,
+      "loss": 0.7459,
+      "step": 2300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011131687242798354,
+      "loss": 0.7353,
+      "step": 2305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011111111111111112,
+      "loss": 0.7406,
+      "step": 2310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011090534979423868,
+      "loss": 0.7623,
+      "step": 2315
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011069958847736626,
+      "loss": 0.7567,
+      "step": 2320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011049382716049384,
+      "loss": 0.7752,
+      "step": 2325
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001102880658436214,
+      "loss": 0.7419,
+      "step": 2330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011008230452674897,
+      "loss": 0.7244,
+      "step": 2335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00010987654320987655,
+      "loss": 0.7553,
+      "step": 2340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010967078189300413,
+      "loss": 0.7389,
+      "step": 2345
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010946502057613168,
+      "loss": 0.7002,
+      "step": 2350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010925925925925926,
+      "loss": 0.7238,
+      "step": 2355
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010905349794238684,
+      "loss": 0.6923,
+      "step": 2360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010884773662551442,
+      "loss": 0.7579,
+      "step": 2365
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010864197530864197,
+      "loss": 0.7291,
+      "step": 2370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010843621399176955,
+      "loss": 0.721,
+      "step": 2375
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010823045267489713,
+      "loss": 0.7605,
+      "step": 2380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010802469135802468,
+      "loss": 0.7297,
+      "step": 2385
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010781893004115226,
+      "loss": 0.7565,
+      "step": 2390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010761316872427984,
+      "loss": 0.701,
+      "step": 2395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010740740740740742,
+      "loss": 0.736,
+      "step": 2400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010720164609053497,
+      "loss": 0.7079,
+      "step": 2405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010699588477366255,
+      "loss": 0.7349,
+      "step": 2410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010679012345679013,
+      "loss": 0.7106,
+      "step": 2415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001065843621399177,
+      "loss": 0.7396,
+      "step": 2420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010637860082304528,
+      "loss": 0.7713,
+      "step": 2425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010617283950617284,
+      "loss": 0.747,
+      "step": 2430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010596707818930041,
+      "loss": 0.7475,
+      "step": 2435
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010576131687242799,
+      "loss": 0.7449,
+      "step": 2440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010555555555555557,
+      "loss": 0.7208,
+      "step": 2445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010534979423868315,
+      "loss": 0.7317,
+      "step": 2450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001051440329218107,
+      "loss": 0.7419,
+      "step": 2455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010493827160493828,
+      "loss": 0.7148,
+      "step": 2460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010473251028806586,
+      "loss": 0.7143,
+      "step": 2465
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010452674897119341,
+      "loss": 0.747,
+      "step": 2470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010432098765432099,
+      "loss": 0.7079,
+      "step": 2475
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010411522633744857,
+      "loss": 0.7059,
+      "step": 2480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010390946502057615,
+      "loss": 0.7272,
+      "step": 2485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001037037037037037,
+      "loss": 0.7281,
+      "step": 2490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010349794238683128,
+      "loss": 0.6956,
+      "step": 2495
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010329218106995886,
+      "loss": 0.7571,
+      "step": 2500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010308641975308641,
+      "loss": 0.7311,
+      "step": 2505
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010288065843621399,
+      "loss": 0.6815,
+      "step": 2510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010267489711934157,
+      "loss": 0.6781,
+      "step": 2515
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010246913580246915,
+      "loss": 0.6948,
+      "step": 2520
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010226337448559671,
+      "loss": 0.7133,
+      "step": 2525
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010205761316872428,
+      "loss": 0.688,
+      "step": 2530
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010185185185185186,
+      "loss": 0.6687,
+      "step": 2535
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010164609053497942,
+      "loss": 0.7142,
+      "step": 2540
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000101440329218107,
+      "loss": 0.6885,
+      "step": 2545
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010123456790123458,
+      "loss": 0.6867,
+      "step": 2550
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010102880658436214,
+      "loss": 0.7045,
+      "step": 2555
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010082304526748971,
+      "loss": 0.7375,
+      "step": 2560
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001006172839506173,
+      "loss": 0.7157,
+      "step": 2565
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010041152263374487,
+      "loss": 0.6828,
+      "step": 2570
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010020576131687243,
+      "loss": 0.7048,
+      "step": 2575
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001,
+      "loss": 0.6944,
+      "step": 2580
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.979423868312757e-05,
+      "loss": 0.696,
+      "step": 2585
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.958847736625515e-05,
+      "loss": 0.6609,
+      "step": 2590
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.938271604938272e-05,
+      "loss": 0.674,
+      "step": 2595
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.91769547325103e-05,
+      "loss": 0.7138,
+      "step": 2600
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.897119341563786e-05,
+      "loss": 0.713,
+      "step": 2605
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.876543209876543e-05,
+      "loss": 0.7022,
+      "step": 2610
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.8559670781893e-05,
+      "loss": 0.6955,
+      "step": 2615
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.835390946502057e-05,
+      "loss": 0.6888,
+      "step": 2620
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.814814814814815e-05,
+      "loss": 0.7003,
+      "step": 2625
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.794238683127573e-05,
+      "loss": 0.7167,
+      "step": 2630
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.77366255144033e-05,
+      "loss": 0.7087,
+      "step": 2635
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.753086419753087e-05,
+      "loss": 0.7019,
+      "step": 2640
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.732510288065844e-05,
+      "loss": 0.6609,
+      "step": 2645
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.711934156378602e-05,
+      "loss": 0.7367,
+      "step": 2650
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.691358024691359e-05,
+      "loss": 0.6862,
+      "step": 2655
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.670781893004116e-05,
+      "loss": 0.7121,
+      "step": 2660
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.650205761316873e-05,
+      "loss": 0.7031,
+      "step": 2665
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.62962962962963e-05,
+      "loss": 0.7005,
+      "step": 2670
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.609053497942388e-05,
+      "loss": 0.7041,
+      "step": 2675
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.588477366255144e-05,
+      "loss": 0.6713,
+      "step": 2680
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.567901234567902e-05,
+      "loss": 0.6932,
+      "step": 2685
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.547325102880659e-05,
+      "loss": 0.6849,
+      "step": 2690
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.526748971193417e-05,
+      "loss": 0.6915,
+      "step": 2695
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.506172839506173e-05,
+      "loss": 0.6677,
+      "step": 2700
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.48559670781893e-05,
+      "loss": 0.6767,
+      "step": 2705
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.465020576131688e-05,
+      "loss": 0.6873,
+      "step": 2710
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.444444444444444e-05,
+      "loss": 0.7007,
+      "step": 2715
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.423868312757202e-05,
+      "loss": 0.6782,
+      "step": 2720
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.403292181069959e-05,
+      "loss": 0.7044,
+      "step": 2725
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.382716049382717e-05,
+      "loss": 0.6787,
+      "step": 2730
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.362139917695473e-05,
+      "loss": 0.6876,
+      "step": 2735
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.341563786008231e-05,
+      "loss": 0.7253,
+      "step": 2740
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.320987654320989e-05,
+      "loss": 0.7159,
+      "step": 2745
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.300411522633746e-05,
+      "loss": 0.704,
+      "step": 2750
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.279835390946504e-05,
+      "loss": 0.6882,
+      "step": 2755
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.25925925925926e-05,
+      "loss": 0.6777,
+      "step": 2760
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.238683127572017e-05,
+      "loss": 0.7133,
+      "step": 2765
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.218106995884775e-05,
+      "loss": 0.6988,
+      "step": 2770
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.197530864197531e-05,
+      "loss": 0.7323,
+      "step": 2775
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.176954732510289e-05,
+      "loss": 0.6854,
+      "step": 2780
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.156378600823046e-05,
+      "loss": 0.6926,
+      "step": 2785
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.135802469135802e-05,
+      "loss": 0.6985,
+      "step": 2790
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.11522633744856e-05,
+      "loss": 0.6776,
+      "step": 2795
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.094650205761317e-05,
+      "loss": 0.7374,
+      "step": 2800
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.074074074074075e-05,
+      "loss": 0.6928,
+      "step": 2805
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.053497942386831e-05,
+      "loss": 0.7129,
+      "step": 2810
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.032921810699589e-05,
+      "loss": 0.6918,
+      "step": 2815
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.012345679012346e-05,
+      "loss": 0.6641,
+      "step": 2820
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.991769547325102e-05,
+      "loss": 0.7075,
+      "step": 2825
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.97119341563786e-05,
+      "loss": 0.668,
+      "step": 2830
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.950617283950617e-05,
+      "loss": 0.6836,
+      "step": 2835
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.930041152263375e-05,
+      "loss": 0.672,
+      "step": 2840
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.909465020576133e-05,
+      "loss": 0.6827,
+      "step": 2845
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.888888888888889e-05,
+      "loss": 0.6976,
+      "step": 2850
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.868312757201647e-05,
+      "loss": 0.6719,
+      "step": 2855
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.847736625514404e-05,
+      "loss": 0.6871,
+      "step": 2860
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.827160493827162e-05,
+      "loss": 0.6975,
+      "step": 2865
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.806584362139918e-05,
+      "loss": 0.684,
+      "step": 2870
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.786008230452676e-05,
+      "loss": 0.6861,
+      "step": 2875
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.765432098765433e-05,
+      "loss": 0.7116,
+      "step": 2880
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.74485596707819e-05,
+      "loss": 0.686,
+      "step": 2885
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.724279835390947e-05,
+      "loss": 0.677,
+      "step": 2890
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.703703703703704e-05,
+      "loss": 0.6996,
+      "step": 2895
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.683127572016462e-05,
+      "loss": 0.7211,
+      "step": 2900
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.662551440329218e-05,
+      "loss": 0.6766,
+      "step": 2905
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.641975308641975e-05,
+      "loss": 0.6883,
+      "step": 2910
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.621399176954733e-05,
+      "loss": 0.6829,
+      "step": 2915
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60082304526749e-05,
+      "loss": 0.7089,
+      "step": 2920
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.580246913580247e-05,
+      "loss": 0.6711,
+      "step": 2925
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.559670781893004e-05,
+      "loss": 0.6926,
+      "step": 2930
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.539094650205762e-05,
+      "loss": 0.6513,
+      "step": 2935
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.518518518518518e-05,
+      "loss": 0.6817,
+      "step": 2940
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.497942386831276e-05,
+      "loss": 0.7005,
+      "step": 2945
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.477366255144033e-05,
+      "loss": 0.7013,
+      "step": 2950
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.456790123456791e-05,
+      "loss": 0.6562,
+      "step": 2955
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.436213991769549e-05,
+      "loss": 0.7074,
+      "step": 2960
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.415637860082305e-05,
+      "loss": 0.6901,
+      "step": 2965
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.395061728395062e-05,
+      "loss": 0.6867,
+      "step": 2970
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.37448559670782e-05,
+      "loss": 0.692,
+      "step": 2975
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.353909465020576e-05,
+      "loss": 0.6863,
+      "step": 2980
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.333333333333334e-05,
+      "loss": 0.6881,
+      "step": 2985
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.312757201646091e-05,
+      "loss": 0.7203,
+      "step": 2990
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.292181069958848e-05,
+      "loss": 0.7075,
+      "step": 2995
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.271604938271605e-05,
+      "loss": 0.7187,
+      "step": 3000
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.251028806584362e-05,
+      "loss": 0.6791,
+      "step": 3005
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.23045267489712e-05,
+      "loss": 0.6937,
+      "step": 3010
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.209876543209877e-05,
+      "loss": 0.6584,
+      "step": 3015
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.189300411522634e-05,
+      "loss": 0.6818,
+      "step": 3020
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.168724279835391e-05,
+      "loss": 0.6827,
+      "step": 3025
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.148148148148148e-05,
+      "loss": 0.6774,
+      "step": 3030
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.127572016460906e-05,
+      "loss": 0.6826,
+      "step": 3035
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.106995884773662e-05,
+      "loss": 0.6767,
+      "step": 3040
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.08641975308642e-05,
+      "loss": 0.7256,
+      "step": 3045
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.065843621399177e-05,
+      "loss": 0.6875,
+      "step": 3050
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.045267489711935e-05,
+      "loss": 0.7013,
+      "step": 3055
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.024691358024692e-05,
+      "loss": 0.6907,
+      "step": 3060
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.004115226337449e-05,
+      "loss": 0.6751,
+      "step": 3065
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.983539094650207e-05,
+      "loss": 0.6893,
+      "step": 3070
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.962962962962964e-05,
+      "loss": 0.6962,
+      "step": 3075
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.942386831275721e-05,
+      "loss": 0.6743,
+      "step": 3080
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.921810699588478e-05,
+      "loss": 0.7004,
+      "step": 3085
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.901234567901235e-05,
+      "loss": 0.6908,
+      "step": 3090
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.880658436213993e-05,
+      "loss": 0.6736,
+      "step": 3095
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.860082304526749e-05,
+      "loss": 0.7153,
+      "step": 3100
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.839506172839507e-05,
+      "loss": 0.7072,
+      "step": 3105
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.818930041152264e-05,
+      "loss": 0.7163,
+      "step": 3110
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.79835390946502e-05,
+      "loss": 0.6996,
+      "step": 3115
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.777777777777778e-05,
+      "loss": 0.7257,
+      "step": 3120
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.757201646090535e-05,
+      "loss": 0.7155,
+      "step": 3125
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.736625514403293e-05,
+      "loss": 0.7068,
+      "step": 3130
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.716049382716049e-05,
+      "loss": 0.6575,
+      "step": 3135
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.695473251028807e-05,
+      "loss": 0.6771,
+      "step": 3140
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.674897119341564e-05,
+      "loss": 0.6976,
+      "step": 3145
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.65432098765432e-05,
+      "loss": 0.6568,
+      "step": 3150
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.633744855967078e-05,
+      "loss": 0.6609,
+      "step": 3155
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.613168724279836e-05,
+      "loss": 0.697,
+      "step": 3160
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.592592592592593e-05,
+      "loss": 0.714,
+      "step": 3165
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.57201646090535e-05,
+      "loss": 0.6829,
+      "step": 3170
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.551440329218107e-05,
+      "loss": 0.6932,
+      "step": 3175
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.530864197530865e-05,
+      "loss": 0.6859,
+      "step": 3180
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.510288065843622e-05,
+      "loss": 0.7117,
+      "step": 3185
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.48971193415638e-05,
+      "loss": 0.7157,
+      "step": 3190
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.469135802469136e-05,
+      "loss": 0.6593,
+      "step": 3195
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.448559670781894e-05,
+      "loss": 0.6683,
+      "step": 3200
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.427983539094651e-05,
+      "loss": 0.6876,
+      "step": 3205
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.407407407407407e-05,
+      "loss": 0.6584,
+      "step": 3210
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.386831275720165e-05,
+      "loss": 0.6816,
+      "step": 3215
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.366255144032922e-05,
+      "loss": 0.6994,
+      "step": 3220
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.34567901234568e-05,
+      "loss": 0.6701,
+      "step": 3225
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.325102880658436e-05,
+      "loss": 0.6869,
+      "step": 3230
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.304526748971193e-05,
+      "loss": 0.6895,
+      "step": 3235
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.283950617283951e-05,
+      "loss": 0.6972,
+      "step": 3240
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.263374485596707e-05,
+      "loss": 0.6849,
+      "step": 3245
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.242798353909465e-05,
+      "loss": 0.6967,
+      "step": 3250
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.222222222222222e-05,
+      "loss": 0.6679,
+      "step": 3255
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.20164609053498e-05,
+      "loss": 0.6683,
+      "step": 3260
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.181069958847736e-05,
+      "loss": 0.6787,
+      "step": 3265
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.160493827160494e-05,
+      "loss": 0.6726,
+      "step": 3270
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.139917695473252e-05,
+      "loss": 0.721,
+      "step": 3275
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.119341563786009e-05,
+      "loss": 0.6865,
+      "step": 3280
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.098765432098767e-05,
+      "loss": 0.7031,
+      "step": 3285
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.078189300411523e-05,
+      "loss": 0.6895,
+      "step": 3290
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.05761316872428e-05,
+      "loss": 0.6983,
+      "step": 3295
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.037037037037038e-05,
+      "loss": 0.6929,
+      "step": 3300
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.016460905349794e-05,
+      "loss": 0.7117,
+      "step": 3305
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.995884773662552e-05,
+      "loss": 0.664,
+      "step": 3310
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.975308641975309e-05,
+      "loss": 0.6645,
+      "step": 3315
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.954732510288067e-05,
+      "loss": 0.6985,
+      "step": 3320
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.934156378600823e-05,
+      "loss": 0.6826,
+      "step": 3325
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.91358024691358e-05,
+      "loss": 0.7324,
+      "step": 3330
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.893004115226338e-05,
+      "loss": 0.6828,
+      "step": 3335
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.872427983539094e-05,
+      "loss": 0.6801,
+      "step": 3340
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.851851851851852e-05,
+      "loss": 0.6923,
+      "step": 3345
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.831275720164609e-05,
+      "loss": 0.6787,
+      "step": 3350
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.810699588477366e-05,
+      "loss": 0.6812,
+      "step": 3355
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.790123456790123e-05,
+      "loss": 0.7059,
+      "step": 3360
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.76954732510288e-05,
+      "loss": 0.6794,
+      "step": 3365
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.748971193415638e-05,
+      "loss": 0.6635,
+      "step": 3370
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.728395061728396e-05,
+      "loss": 0.666,
+      "step": 3375
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.707818930041152e-05,
+      "loss": 0.6909,
+      "step": 3380
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.68724279835391e-05,
+      "loss": 0.7096,
+      "step": 3385
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.7049,
+      "step": 3390
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.646090534979425e-05,
+      "loss": 0.6801,
+      "step": 3395
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.625514403292181e-05,
+      "loss": 0.7045,
+      "step": 3400
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.60493827160494e-05,
+      "loss": 0.7096,
+      "step": 3405
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.584362139917696e-05,
+      "loss": 0.7192,
+      "step": 3410
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.563786008230453e-05,
+      "loss": 0.6818,
+      "step": 3415
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.54320987654321e-05,
+      "loss": 0.6358,
+      "step": 3420
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.522633744855967e-05,
+      "loss": 0.6674,
+      "step": 3425
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.502057613168725e-05,
+      "loss": 0.684,
+      "step": 3430
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.481481481481482e-05,
+      "loss": 0.6753,
+      "step": 3435
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.460905349794238e-05,
+      "loss": 0.6902,
+      "step": 3440
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.440329218106996e-05,
+      "loss": 0.6664,
+      "step": 3445
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.419753086419753e-05,
+      "loss": 0.684,
+      "step": 3450
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.39917695473251e-05,
+      "loss": 0.682,
+      "step": 3455
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.378600823045267e-05,
+      "loss": 0.6764,
+      "step": 3460
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.358024691358025e-05,
+      "loss": 0.6783,
+      "step": 3465
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.337448559670782e-05,
+      "loss": 0.6962,
+      "step": 3470
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.31687242798354e-05,
+      "loss": 0.6819,
+      "step": 3475
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.296296296296296e-05,
+      "loss": 0.6741,
+      "step": 3480
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.275720164609054e-05,
+      "loss": 0.6754,
+      "step": 3485
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.255144032921812e-05,
+      "loss": 0.6717,
+      "step": 3490
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.234567901234569e-05,
+      "loss": 0.667,
+      "step": 3495
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.213991769547325e-05,
+      "loss": 0.6822,
+      "step": 3500
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.193415637860083e-05,
+      "loss": 0.7055,
+      "step": 3505
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.17283950617284e-05,
+      "loss": 0.6988,
+      "step": 3510
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.152263374485598e-05,
+      "loss": 0.6793,
+      "step": 3515
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.131687242798354e-05,
+      "loss": 0.7093,
+      "step": 3520
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.111111111111112e-05,
+      "loss": 0.694,
+      "step": 3525
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0905349794238687e-05,
+      "loss": 0.7077,
+      "step": 3530
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.069958847736625e-05,
+      "loss": 0.6938,
+      "step": 3535
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.049382716049383e-05,
+      "loss": 0.6461,
+      "step": 3540
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.02880658436214e-05,
+      "loss": 0.6993,
+      "step": 3545
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0082304526748977e-05,
+      "loss": 0.6901,
+      "step": 3550
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.987654320987654e-05,
+      "loss": 0.6937,
+      "step": 3555
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9670781893004115e-05,
+      "loss": 0.6657,
+      "step": 3560
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9465020576131694e-05,
+      "loss": 0.6593,
+      "step": 3565
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.925925925925926e-05,
+      "loss": 0.7056,
+      "step": 3570
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.905349794238684e-05,
+      "loss": 0.6395,
+      "step": 3575
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8847736625514405e-05,
+      "loss": 0.6608,
+      "step": 3580
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8641975308641984e-05,
+      "loss": 0.7007,
+      "step": 3585
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.843621399176955e-05,
+      "loss": 0.6605,
+      "step": 3590
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8230452674897116e-05,
+      "loss": 0.6752,
+      "step": 3595
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.8024691358024695e-05,
+      "loss": 0.6736,
+      "step": 3600
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.781893004115226e-05,
+      "loss": 0.671,
+      "step": 3605
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.761316872427984e-05,
+      "loss": 0.6551,
+      "step": 3610
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.740740740740741e-05,
+      "loss": 0.7053,
+      "step": 3615
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.720164609053498e-05,
+      "loss": 0.6856,
+      "step": 3620
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.699588477366256e-05,
+      "loss": 0.671,
+      "step": 3625
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.679012345679012e-05,
+      "loss": 0.6763,
+      "step": 3630
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.65843621399177e-05,
+      "loss": 0.6812,
+      "step": 3635
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.637860082304527e-05,
+      "loss": 0.7407,
+      "step": 3640
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.617283950617285e-05,
+      "loss": 0.7436,
+      "step": 3645
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.596707818930041e-05,
+      "loss": 0.682,
+      "step": 3650
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.576131687242798e-05,
+      "loss": 0.6808,
+      "step": 3655
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.555555555555556e-05,
+      "loss": 0.6696,
+      "step": 3660
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.534979423868313e-05,
+      "loss": 0.7207,
+      "step": 3665
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.51440329218107e-05,
+      "loss": 0.6737,
+      "step": 3670
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4938271604938276e-05,
+      "loss": 0.6874,
+      "step": 3675
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.473251028806584e-05,
+      "loss": 0.6652,
+      "step": 3680
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.452674897119342e-05,
+      "loss": 0.6739,
+      "step": 3685
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4320987654320986e-05,
+      "loss": 0.7039,
+      "step": 3690
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4115226337448566e-05,
+      "loss": 0.696,
+      "step": 3695
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.390946502057613e-05,
+      "loss": 0.6887,
+      "step": 3700
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.370370370370371e-05,
+      "loss": 0.6653,
+      "step": 3705
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3497942386831277e-05,
+      "loss": 0.6987,
+      "step": 3710
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.329218106995885e-05,
+      "loss": 0.6831,
+      "step": 3715
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.308641975308642e-05,
+      "loss": 0.687,
+      "step": 3720
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2880658436213994e-05,
+      "loss": 0.6848,
+      "step": 3725
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.267489711934157e-05,
+      "loss": 0.6819,
+      "step": 3730
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.246913580246914e-05,
+      "loss": 0.662,
+      "step": 3735
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2263374485596705e-05,
+      "loss": 0.6462,
+      "step": 3740
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2057613168724284e-05,
+      "loss": 0.684,
+      "step": 3745
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.185185185185185e-05,
+      "loss": 0.721,
+      "step": 3750
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.164609053497943e-05,
+      "loss": 0.6998,
+      "step": 3755
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1440329218106995e-05,
+      "loss": 0.6922,
+      "step": 3760
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1234567901234574e-05,
+      "loss": 0.6674,
+      "step": 3765
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.102880658436214e-05,
+      "loss": 0.7022,
+      "step": 3770
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.082304526748971e-05,
+      "loss": 0.6753,
+      "step": 3775
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.061728395061729e-05,
+      "loss": 0.6853,
+      "step": 3780
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.041152263374486e-05,
+      "loss": 0.6669,
+      "step": 3785
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.020576131687244e-05,
+      "loss": 0.6678,
+      "step": 3790
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5e-05,
+      "loss": 0.6881,
+      "step": 3795
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9794238683127575e-05,
+      "loss": 0.6959,
+      "step": 3800
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.958847736625515e-05,
+      "loss": 0.695,
+      "step": 3805
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.938271604938271e-05,
+      "loss": 0.6927,
+      "step": 3810
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9176954732510286e-05,
+      "loss": 0.7039,
+      "step": 3815
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.8971193415637865e-05,
+      "loss": 0.6494,
+      "step": 3820
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.876543209876544e-05,
+      "loss": 0.7015,
+      "step": 3825
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.855967078189301e-05,
+      "loss": 0.6901,
+      "step": 3830
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.835390946502058e-05,
+      "loss": 0.6618,
+      "step": 3835
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.814814814814815e-05,
+      "loss": 0.6723,
+      "step": 3840
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.794238683127572e-05,
+      "loss": 0.7196,
+      "step": 3845
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.773662551440329e-05,
+      "loss": 0.6867,
+      "step": 3850
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7530864197530866e-05,
+      "loss": 0.6921,
+      "step": 3855
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.732510288065844e-05,
+      "loss": 0.693,
+      "step": 3860
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.711934156378601e-05,
+      "loss": 0.7031,
+      "step": 3865
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.691358024691358e-05,
+      "loss": 0.7023,
+      "step": 3870
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6707818930041156e-05,
+      "loss": 0.6754,
+      "step": 3875
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.650205761316873e-05,
+      "loss": 0.6985,
+      "step": 3880
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.62962962962963e-05,
+      "loss": 0.665,
+      "step": 3885
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.609053497942387e-05,
+      "loss": 0.6902,
+      "step": 3890
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.5884773662551446e-05,
+      "loss": 0.7173,
+      "step": 3895
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.567901234567901e-05,
+      "loss": 0.7121,
+      "step": 3900
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5473251028806584e-05,
+      "loss": 0.6939,
+      "step": 3905
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5267489711934157e-05,
+      "loss": 0.7093,
+      "step": 3910
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.506172839506173e-05,
+      "loss": 0.6792,
+      "step": 3915
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.48559670781893e-05,
+      "loss": 0.6737,
+      "step": 3920
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4650205761316874e-05,
+      "loss": 0.6858,
+      "step": 3925
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4444444444444447e-05,
+      "loss": 0.6776,
+      "step": 3930
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.423868312757202e-05,
+      "loss": 0.6772,
+      "step": 3935
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.403292181069959e-05,
+      "loss": 0.6667,
+      "step": 3940
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.3827160493827164e-05,
+      "loss": 0.6892,
+      "step": 3945
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3621399176954737e-05,
+      "loss": 0.6652,
+      "step": 3950
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.341563786008231e-05,
+      "loss": 0.6892,
+      "step": 3955
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3209876543209875e-05,
+      "loss": 0.6917,
+      "step": 3960
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.300411522633745e-05,
+      "loss": 0.6708,
+      "step": 3965
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.279835390946502e-05,
+      "loss": 0.6817,
+      "step": 3970
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.259259259259259e-05,
+      "loss": 0.6761,
+      "step": 3975
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2386831275720165e-05,
+      "loss": 0.6638,
+      "step": 3980
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2181069958847744e-05,
+      "loss": 0.6707,
+      "step": 3985
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.197530864197531e-05,
+      "loss": 0.6796,
+      "step": 3990
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.176954732510288e-05,
+      "loss": 0.6903,
+      "step": 3995
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1563786008230455e-05,
+      "loss": 0.7081,
+      "step": 4000
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.135802469135803e-05,
+      "loss": 0.6528,
+      "step": 4005
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.11522633744856e-05,
+      "loss": 0.6613,
+      "step": 4010
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.094650205761317e-05,
+      "loss": 0.6694,
+      "step": 4015
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.074074074074074e-05,
+      "loss": 0.6982,
+      "step": 4020
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.053497942386831e-05,
+      "loss": 0.6722,
+      "step": 4025
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.032921810699588e-05,
+      "loss": 0.6505,
+      "step": 4030
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.012345679012346e-05,
+      "loss": 0.6747,
+      "step": 4035
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.9917695473251035e-05,
+      "loss": 0.6459,
+      "step": 4040
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.971193415637861e-05,
+      "loss": 0.6859,
+      "step": 4045
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.950617283950617e-05,
+      "loss": 0.6774,
+      "step": 4050
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9300411522633746e-05,
+      "loss": 0.678,
+      "step": 4055
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.909465020576132e-05,
+      "loss": 0.678,
+      "step": 4060
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.888888888888889e-05,
+      "loss": 0.7204,
+      "step": 4065
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.868312757201646e-05,
+      "loss": 0.68,
+      "step": 4070
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8477366255144036e-05,
+      "loss": 0.6797,
+      "step": 4075
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.82716049382716e-05,
+      "loss": 0.6884,
+      "step": 4080
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.806584362139918e-05,
+      "loss": 0.6749,
+      "step": 4085
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.786008230452675e-05,
+      "loss": 0.7162,
+      "step": 4090
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7654320987654326e-05,
+      "loss": 0.6702,
+      "step": 4095
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.74485596707819e-05,
+      "loss": 0.7014,
+      "step": 4100
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.724279835390947e-05,
+      "loss": 0.7031,
+      "step": 4105
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7037037037037037e-05,
+      "loss": 0.6613,
+      "step": 4110
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.683127572016461e-05,
+      "loss": 0.652,
+      "step": 4115
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.662551440329218e-05,
+      "loss": 0.6591,
+      "step": 4120
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6419753086419754e-05,
+      "loss": 0.6688,
+      "step": 4125
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6213991769547327e-05,
+      "loss": 0.6843,
+      "step": 4130
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.60082304526749e-05,
+      "loss": 0.6706,
+      "step": 4135
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.580246913580247e-05,
+      "loss": 0.6974,
+      "step": 4140
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5596707818930044e-05,
+      "loss": 0.6739,
+      "step": 4145
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.539094650205762e-05,
+      "loss": 0.6787,
+      "step": 4150
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.518518518518519e-05,
+      "loss": 0.695,
+      "step": 4155
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.497942386831276e-05,
+      "loss": 0.7012,
+      "step": 4160
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4773662551440334e-05,
+      "loss": 0.7137,
+      "step": 4165
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.45679012345679e-05,
+      "loss": 0.6919,
+      "step": 4170
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.436213991769547e-05,
+      "loss": 0.7045,
+      "step": 4175
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4156378600823045e-05,
+      "loss": 0.6638,
+      "step": 4180
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.395061728395062e-05,
+      "loss": 0.6918,
+      "step": 4185
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.374485596707819e-05,
+      "loss": 0.67,
+      "step": 4190
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.353909465020576e-05,
+      "loss": 0.6835,
+      "step": 4195
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 0.6854,
+      "step": 4200
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.312757201646091e-05,
+      "loss": 0.6846,
+      "step": 4205
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.292181069958848e-05,
+      "loss": 0.6707,
+      "step": 4210
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.271604938271605e-05,
+      "loss": 0.7155,
+      "step": 4215
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2510288065843625e-05,
+      "loss": 0.6985,
+      "step": 4220
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.230452674897119e-05,
+      "loss": 0.6842,
+      "step": 4225
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.209876543209876e-05,
+      "loss": 0.6898,
+      "step": 4230
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1893004115226336e-05,
+      "loss": 0.6707,
+      "step": 4235
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.168724279835391e-05,
+      "loss": 0.6979,
+      "step": 4240
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.148148148148148e-05,
+      "loss": 0.6899,
+      "step": 4245
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.127572016460906e-05,
+      "loss": 0.6697,
+      "step": 4250
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1069958847736626e-05,
+      "loss": 0.6819,
+      "step": 4255
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.08641975308642e-05,
+      "loss": 0.6913,
+      "step": 4260
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.065843621399177e-05,
+      "loss": 0.6894,
+      "step": 4265
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0452674897119343e-05,
+      "loss": 0.6801,
+      "step": 4270
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0246913580246916e-05,
+      "loss": 0.6932,
+      "step": 4275
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0041152263374488e-05,
+      "loss": 0.6567,
+      "step": 4280
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9835390946502057e-05,
+      "loss": 0.6416,
+      "step": 4285
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.962962962962963e-05,
+      "loss": 0.6619,
+      "step": 4290
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9423868312757202e-05,
+      "loss": 0.6746,
+      "step": 4295
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9218106995884775e-05,
+      "loss": 0.6725,
+      "step": 4300
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9012345679012347e-05,
+      "loss": 0.695,
+      "step": 4305
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.880658436213992e-05,
+      "loss": 0.68,
+      "step": 4310
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.860082304526749e-05,
+      "loss": 0.6669,
+      "step": 4315
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.839506172839506e-05,
+      "loss": 0.6626,
+      "step": 4320
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8189300411522634e-05,
+      "loss": 0.6871,
+      "step": 4325
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7983539094650207e-05,
+      "loss": 0.6719,
+      "step": 4330
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.777777777777778e-05,
+      "loss": 0.6603,
+      "step": 4335
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.757201646090535e-05,
+      "loss": 0.6378,
+      "step": 4340
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.736625514403292e-05,
+      "loss": 0.6759,
+      "step": 4345
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7160493827160493e-05,
+      "loss": 0.6704,
+      "step": 4350
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6954732510288066e-05,
+      "loss": 0.6725,
+      "step": 4355
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6748971193415638e-05,
+      "loss": 0.6659,
+      "step": 4360
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.654320987654321e-05,
+      "loss": 0.664,
+      "step": 4365
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6337448559670787e-05,
+      "loss": 0.7316,
+      "step": 4370
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6131687242798352e-05,
+      "loss": 0.6621,
+      "step": 4375
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5925925925925925e-05,
+      "loss": 0.6822,
+      "step": 4380
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5720164609053497e-05,
+      "loss": 0.6694,
+      "step": 4385
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.551440329218107e-05,
+      "loss": 0.6719,
+      "step": 4390
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5308641975308646e-05,
+      "loss": 0.6687,
+      "step": 4395
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.510288065843622e-05,
+      "loss": 0.6878,
+      "step": 4400
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4897119341563787e-05,
+      "loss": 0.6956,
+      "step": 4405
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4691358024691357e-05,
+      "loss": 0.6976,
+      "step": 4410
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4485596707818932e-05,
+      "loss": 0.6969,
+      "step": 4415
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4279835390946505e-05,
+      "loss": 0.685,
+      "step": 4420
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4074074074074074e-05,
+      "loss": 0.6835,
+      "step": 4425
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3868312757201647e-05,
+      "loss": 0.7002,
+      "step": 4430
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.366255144032922e-05,
+      "loss": 0.6657,
+      "step": 4435
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.345679012345679e-05,
+      "loss": 0.6524,
+      "step": 4440
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3251028806584364e-05,
+      "loss": 0.6672,
+      "step": 4445
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3045267489711937e-05,
+      "loss": 0.7033,
+      "step": 4450
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2839506172839506e-05,
+      "loss": 0.6922,
+      "step": 4455
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2633744855967078e-05,
+      "loss": 0.6567,
+      "step": 4460
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.242798353909465e-05,
+      "loss": 0.6764,
+      "step": 4465
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2222222222222223e-05,
+      "loss": 0.7066,
+      "step": 4470
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2016460905349796e-05,
+      "loss": 0.7036,
+      "step": 4475
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1810699588477368e-05,
+      "loss": 0.6716,
+      "step": 4480
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1604938271604937e-05,
+      "loss": 0.6954,
+      "step": 4485
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.139917695473251e-05,
+      "loss": 0.6752,
+      "step": 4490
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1193415637860082e-05,
+      "loss": 0.6845,
+      "step": 4495
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0987654320987655e-05,
+      "loss": 0.6928,
+      "step": 4500
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 5010,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 500,
+  "total_flos": 1.8116033382918914e+19,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/training_args.bin b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e707059a03fbe62142ac41332e145ac1ac1f36cc
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b99a1729574e0e46ee782b42dc90101a785d5b4792871920d293f44722c28cc
+size 5944
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/zero_to_fp32.py b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/zero_to_fp32.py
new file mode 100644
index 0000000000000000000000000000000000000000..c98caae31534368be22b67fc4ae906836c992a8d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-4500/zero_to_fp32.py
@@ -0,0 +1,587 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+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):
+    """
+    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)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+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 _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()
+            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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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 param_shapes.items():
+
+        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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    """
+    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``
+
+    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)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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("-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_file, tag=args.tag)
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/README.md b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..d380bbbf0bf0fa1f211b53410559607e5007a209
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: meta-llama/Llama-2-7b-hf
+---
+
+# 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 0.9.0
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/adapter_config.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c81a592f0ff08374a4f61f6003029070a8429ae
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/adapter_config.json
@@ -0,0 +1,35 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "mora_type": 1,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_mora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/adapter_model.safetensors b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/adapter_model.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..876400acb0137ac2304de8d478a34740f4c3803d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/adapter_model.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c7157a76c242cf0483703166583aa0d8133f197345c3660dfd11e4b2e8cf41d
+size 564342240
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/global_step5000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/global_step5000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4d95b3bf605efa71f4663fab1f4df31f4be9ea93
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/global_step5000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:486f9c160d20739fe22883feaf8ff4a47a3c7bddc605711f3c46543b4fb68d46
+size 239914224
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/global_step5000/mp_rank_00_model_states.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/global_step5000/mp_rank_00_model_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e124cbe23c4f228afc0751ee1f58356eabb23fd3
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/global_step5000/mp_rank_00_model_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19ee8111ae0a3b07fd43f7ced68cb83fe11d3452acbf650b938e31e6403c6f7d
+size 40303404
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/latest b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/latest
new file mode 100644
index 0000000000000000000000000000000000000000..f805186fa43374540c3fa51dfd3cca9ac06e56a5
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/latest
@@ -0,0 +1 @@
+global_step5000
\ No newline at end of file
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/rng_state.pth b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/rng_state.pth
new file mode 100644
index 0000000000000000000000000000000000000000..57d9e7c2f8da97786cef99bf6be53018174ed306
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/rng_state.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24b21beccd215decd3c69d041eadc01afc9d1794ccbd322384ceb50140c747f8
+size 14244
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/scheduler.pt b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/scheduler.pt
new file mode 100644
index 0000000000000000000000000000000000000000..dbd2f165904e47158cbeec4cb8a92d3d34c7a507
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/scheduler.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae247e5c98944f52ac569fb613fd9d0e95335f4fce7c104f2e6337707f118840
+size 1064
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/trainer_state.json b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7bac2a8c22be5cc599360a780cb492c2837c300b
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/trainer_state.json
@@ -0,0 +1,6021 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9958212492593632,
+  "eval_steps": 500,
+  "global_step": 5000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.3797,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.3952,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2e-05,
+      "loss": 1.3112,
+      "step": 15
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 1.3136,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 1.2012,
+      "step": 25
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4e-05,
+      "loss": 1.1277,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 1.0832,
+      "step": 35
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.333333333333333e-05,
+      "loss": 1.0659,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-05,
+      "loss": 0.983,
+      "step": 45
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.9763,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.333333333333333e-05,
+      "loss": 0.9992,
+      "step": 55
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8e-05,
+      "loss": 0.9405,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.666666666666667e-05,
+      "loss": 0.9827,
+      "step": 65
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.333333333333334e-05,
+      "loss": 0.9061,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001,
+      "loss": 0.8921,
+      "step": 75
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010666666666666667,
+      "loss": 0.9377,
+      "step": 80
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011333333333333334,
+      "loss": 0.8528,
+      "step": 85
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012,
+      "loss": 0.8913,
+      "step": 90
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00012666666666666666,
+      "loss": 0.8562,
+      "step": 95
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.877,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014,
+      "loss": 0.8781,
+      "step": 105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014666666666666666,
+      "loss": 0.8546,
+      "step": 110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015333333333333334,
+      "loss": 0.8672,
+      "step": 115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00016,
+      "loss": 0.8611,
+      "step": 120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.8624,
+      "step": 125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017333333333333334,
+      "loss": 0.8477,
+      "step": 130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018,
+      "loss": 0.8227,
+      "step": 135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001866666666666667,
+      "loss": 0.8064,
+      "step": 140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019333333333333333,
+      "loss": 0.8422,
+      "step": 145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002,
+      "loss": 0.8471,
+      "step": 150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001997942386831276,
+      "loss": 0.8183,
+      "step": 155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958847736625514,
+      "loss": 0.8229,
+      "step": 160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019938271604938272,
+      "loss": 0.8114,
+      "step": 165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991769547325103,
+      "loss": 0.7861,
+      "step": 170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019897119341563788,
+      "loss": 0.841,
+      "step": 175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019876543209876543,
+      "loss": 0.8549,
+      "step": 180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000198559670781893,
+      "loss": 0.8449,
+      "step": 185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001983539094650206,
+      "loss": 0.8311,
+      "step": 190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019814814814814814,
+      "loss": 0.8296,
+      "step": 195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019794238683127572,
+      "loss": 0.8263,
+      "step": 200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001977366255144033,
+      "loss": 0.8452,
+      "step": 205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019753086419753085,
+      "loss": 0.8044,
+      "step": 210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019732510288065843,
+      "loss": 0.8119,
+      "step": 215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000197119341563786,
+      "loss": 0.8127,
+      "step": 220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001969135802469136,
+      "loss": 0.7921,
+      "step": 225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019670781893004114,
+      "loss": 0.803,
+      "step": 230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019650205761316872,
+      "loss": 0.838,
+      "step": 235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001962962962962963,
+      "loss": 0.7801,
+      "step": 240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019609053497942388,
+      "loss": 0.817,
+      "step": 245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019588477366255146,
+      "loss": 0.8211,
+      "step": 250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000195679012345679,
+      "loss": 0.8423,
+      "step": 255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001954732510288066,
+      "loss": 0.8007,
+      "step": 260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019526748971193417,
+      "loss": 0.8135,
+      "step": 265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019506172839506175,
+      "loss": 0.7961,
+      "step": 270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019485596707818933,
+      "loss": 0.797,
+      "step": 275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019465020576131688,
+      "loss": 0.793,
+      "step": 280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019444444444444446,
+      "loss": 0.7809,
+      "step": 285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019423868312757204,
+      "loss": 0.7811,
+      "step": 290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019403292181069962,
+      "loss": 0.8074,
+      "step": 295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019382716049382717,
+      "loss": 0.8028,
+      "step": 300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019362139917695475,
+      "loss": 0.7895,
+      "step": 305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019341563786008233,
+      "loss": 0.7777,
+      "step": 310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019320987654320988,
+      "loss": 0.7868,
+      "step": 315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019300411522633746,
+      "loss": 0.7921,
+      "step": 320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019279835390946504,
+      "loss": 0.8,
+      "step": 325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001925925925925926,
+      "loss": 0.7781,
+      "step": 330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019238683127572017,
+      "loss": 0.7799,
+      "step": 335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019218106995884775,
+      "loss": 0.7921,
+      "step": 340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019197530864197533,
+      "loss": 0.7851,
+      "step": 345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019176954732510288,
+      "loss": 0.8106,
+      "step": 350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019156378600823046,
+      "loss": 0.7953,
+      "step": 355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019135802469135804,
+      "loss": 0.8084,
+      "step": 360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001911522633744856,
+      "loss": 0.8298,
+      "step": 365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019094650205761317,
+      "loss": 0.8273,
+      "step": 370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019074074074074075,
+      "loss": 0.8047,
+      "step": 375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019053497942386833,
+      "loss": 0.8108,
+      "step": 380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019032921810699588,
+      "loss": 0.8207,
+      "step": 385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019012345679012346,
+      "loss": 0.7884,
+      "step": 390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018991769547325104,
+      "loss": 0.7918,
+      "step": 395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001897119341563786,
+      "loss": 0.7554,
+      "step": 400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018950617283950617,
+      "loss": 0.8099,
+      "step": 405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00018930041152263375,
+      "loss": 0.807,
+      "step": 410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018909465020576133,
+      "loss": 0.7793,
+      "step": 415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018888888888888888,
+      "loss": 0.7734,
+      "step": 420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018868312757201646,
+      "loss": 0.7786,
+      "step": 425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018847736625514404,
+      "loss": 0.7548,
+      "step": 430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001882716049382716,
+      "loss": 0.8024,
+      "step": 435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018806584362139917,
+      "loss": 0.7708,
+      "step": 440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018786008230452675,
+      "loss": 0.8045,
+      "step": 445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018765432098765433,
+      "loss": 0.7477,
+      "step": 450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018744855967078189,
+      "loss": 0.7474,
+      "step": 455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018724279835390946,
+      "loss": 0.7629,
+      "step": 460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018703703703703704,
+      "loss": 0.7777,
+      "step": 465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018683127572016462,
+      "loss": 0.7823,
+      "step": 470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001866255144032922,
+      "loss": 0.7762,
+      "step": 475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018641975308641978,
+      "loss": 0.782,
+      "step": 480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018621399176954733,
+      "loss": 0.7783,
+      "step": 485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001860082304526749,
+      "loss": 0.7376,
+      "step": 490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001858024691358025,
+      "loss": 0.7683,
+      "step": 495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018559670781893007,
+      "loss": 0.7547,
+      "step": 500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018539094650205762,
+      "loss": 0.7569,
+      "step": 505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001851851851851852,
+      "loss": 0.7796,
+      "step": 510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018497942386831278,
+      "loss": 0.7888,
+      "step": 515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018477366255144033,
+      "loss": 0.7763,
+      "step": 520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018456790123456791,
+      "loss": 0.7787,
+      "step": 525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001843621399176955,
+      "loss": 0.772,
+      "step": 530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018415637860082307,
+      "loss": 0.786,
+      "step": 535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018395061728395062,
+      "loss": 0.7639,
+      "step": 540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001837448559670782,
+      "loss": 0.775,
+      "step": 545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018353909465020578,
+      "loss": 0.8002,
+      "step": 550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 0.7359,
+      "step": 555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018312757201646091,
+      "loss": 0.747,
+      "step": 560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001829218106995885,
+      "loss": 0.8234,
+      "step": 565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018271604938271605,
+      "loss": 0.7586,
+      "step": 570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018251028806584363,
+      "loss": 0.7315,
+      "step": 575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001823045267489712,
+      "loss": 0.7986,
+      "step": 580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018209876543209878,
+      "loss": 0.7758,
+      "step": 585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018189300411522634,
+      "loss": 0.7869,
+      "step": 590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018168724279835392,
+      "loss": 0.7596,
+      "step": 595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001814814814814815,
+      "loss": 0.7658,
+      "step": 600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018127572016460905,
+      "loss": 0.7851,
+      "step": 605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018106995884773663,
+      "loss": 0.7957,
+      "step": 610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001808641975308642,
+      "loss": 0.7756,
+      "step": 615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018065843621399178,
+      "loss": 0.7911,
+      "step": 620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018045267489711934,
+      "loss": 0.7864,
+      "step": 625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00018024691358024692,
+      "loss": 0.7683,
+      "step": 630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001800411522633745,
+      "loss": 0.7731,
+      "step": 635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017983539094650205,
+      "loss": 0.7671,
+      "step": 640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017962962962962963,
+      "loss": 0.7367,
+      "step": 645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001794238683127572,
+      "loss": 0.8024,
+      "step": 650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017921810699588476,
+      "loss": 0.7572,
+      "step": 655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017901234567901234,
+      "loss": 0.7554,
+      "step": 660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017880658436213992,
+      "loss": 0.7677,
+      "step": 665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001786008230452675,
+      "loss": 0.7652,
+      "step": 670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017839506172839508,
+      "loss": 0.7771,
+      "step": 675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017818930041152265,
+      "loss": 0.7747,
+      "step": 680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001779835390946502,
+      "loss": 0.767,
+      "step": 685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017777777777777779,
+      "loss": 0.7738,
+      "step": 690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017757201646090537,
+      "loss": 0.7271,
+      "step": 695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017736625514403294,
+      "loss": 0.7316,
+      "step": 700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017716049382716052,
+      "loss": 0.7831,
+      "step": 705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017695473251028808,
+      "loss": 0.7835,
+      "step": 710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017674897119341566,
+      "loss": 0.7497,
+      "step": 715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017654320987654323,
+      "loss": 0.7757,
+      "step": 720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001763374485596708,
+      "loss": 0.74,
+      "step": 725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017613168724279837,
+      "loss": 0.7697,
+      "step": 730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017592592592592595,
+      "loss": 0.774,
+      "step": 735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017572016460905352,
+      "loss": 0.7668,
+      "step": 740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017551440329218108,
+      "loss": 0.7779,
+      "step": 745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017530864197530866,
+      "loss": 0.7441,
+      "step": 750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00017510288065843624,
+      "loss": 0.7578,
+      "step": 755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001748971193415638,
+      "loss": 0.8358,
+      "step": 760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017469135802469137,
+      "loss": 0.7473,
+      "step": 765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017448559670781895,
+      "loss": 0.7517,
+      "step": 770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001742798353909465,
+      "loss": 0.7593,
+      "step": 775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017407407407407408,
+      "loss": 0.7251,
+      "step": 780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017386831275720166,
+      "loss": 0.7365,
+      "step": 785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017366255144032924,
+      "loss": 0.7798,
+      "step": 790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001734567901234568,
+      "loss": 0.746,
+      "step": 795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017325102880658437,
+      "loss": 0.7607,
+      "step": 800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00017304526748971195,
+      "loss": 0.7694,
+      "step": 805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001728395061728395,
+      "loss": 0.7926,
+      "step": 810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017263374485596708,
+      "loss": 0.795,
+      "step": 815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017242798353909466,
+      "loss": 0.7581,
+      "step": 820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017222222222222224,
+      "loss": 0.7629,
+      "step": 825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001720164609053498,
+      "loss": 0.7526,
+      "step": 830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00017181069958847737,
+      "loss": 0.7338,
+      "step": 835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017160493827160495,
+      "loss": 0.7544,
+      "step": 840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001713991769547325,
+      "loss": 0.7522,
+      "step": 845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017119341563786008,
+      "loss": 0.7639,
+      "step": 850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017098765432098766,
+      "loss": 0.7914,
+      "step": 855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00017078189300411524,
+      "loss": 0.7635,
+      "step": 860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001705761316872428,
+      "loss": 0.8343,
+      "step": 865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017037037037037037,
+      "loss": 0.7299,
+      "step": 870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00017016460905349795,
+      "loss": 0.7643,
+      "step": 875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016995884773662553,
+      "loss": 0.7343,
+      "step": 880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00016975308641975308,
+      "loss": 0.7546,
+      "step": 885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016954732510288066,
+      "loss": 0.7435,
+      "step": 890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016934156378600824,
+      "loss": 0.8086,
+      "step": 895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016913580246913582,
+      "loss": 0.7274,
+      "step": 900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001689300411522634,
+      "loss": 0.7511,
+      "step": 905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00016872427983539098,
+      "loss": 0.7689,
+      "step": 910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016851851851851853,
+      "loss": 0.754,
+      "step": 915
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001683127572016461,
+      "loss": 0.7298,
+      "step": 920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001681069958847737,
+      "loss": 0.7755,
+      "step": 925
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016790123456790124,
+      "loss": 0.7854,
+      "step": 930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00016769547325102882,
+      "loss": 0.7452,
+      "step": 935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001674897119341564,
+      "loss": 0.7473,
+      "step": 940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016728395061728398,
+      "loss": 0.7855,
+      "step": 945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00016707818930041153,
+      "loss": 0.7527,
+      "step": 950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001668724279835391,
+      "loss": 0.7641,
+      "step": 955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001666666666666667,
+      "loss": 0.828,
+      "step": 960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016646090534979424,
+      "loss": 0.7571,
+      "step": 965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016625514403292182,
+      "loss": 0.7768,
+      "step": 970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001660493827160494,
+      "loss": 0.7644,
+      "step": 975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016584362139917695,
+      "loss": 0.7747,
+      "step": 980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00016563786008230453,
+      "loss": 0.7926,
+      "step": 985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001654320987654321,
+      "loss": 0.7645,
+      "step": 990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001652263374485597,
+      "loss": 0.7464,
+      "step": 995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016502057613168724,
+      "loss": 0.7558,
+      "step": 1000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00016481481481481482,
+      "loss": 0.7574,
+      "step": 1005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001646090534979424,
+      "loss": 0.7491,
+      "step": 1010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016440329218106995,
+      "loss": 0.756,
+      "step": 1015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016419753086419753,
+      "loss": 0.7679,
+      "step": 1020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001639917695473251,
+      "loss": 0.776,
+      "step": 1025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001637860082304527,
+      "loss": 0.7235,
+      "step": 1030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00016358024691358024,
+      "loss": 0.7817,
+      "step": 1035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016337448559670782,
+      "loss": 0.7333,
+      "step": 1040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001631687242798354,
+      "loss": 0.754,
+      "step": 1045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016296296296296295,
+      "loss": 0.74,
+      "step": 1050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00016275720164609053,
+      "loss": 0.7857,
+      "step": 1055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001625514403292181,
+      "loss": 0.8015,
+      "step": 1060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001623456790123457,
+      "loss": 0.7508,
+      "step": 1065
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016213991769547324,
+      "loss": 0.7514,
+      "step": 1070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016193415637860082,
+      "loss": 0.7326,
+      "step": 1075
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001617283950617284,
+      "loss": 0.7322,
+      "step": 1080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00016152263374485595,
+      "loss": 0.7373,
+      "step": 1085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016131687242798353,
+      "loss": 0.7637,
+      "step": 1090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001611111111111111,
+      "loss": 0.7697,
+      "step": 1095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001609053497942387,
+      "loss": 0.7505,
+      "step": 1100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016069958847736627,
+      "loss": 0.7568,
+      "step": 1105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00016049382716049385,
+      "loss": 0.786,
+      "step": 1110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001602880658436214,
+      "loss": 0.7525,
+      "step": 1115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016008230452674898,
+      "loss": 0.7696,
+      "step": 1120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015987654320987656,
+      "loss": 0.7648,
+      "step": 1125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015967078189300414,
+      "loss": 0.7492,
+      "step": 1130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001594650205761317,
+      "loss": 0.809,
+      "step": 1135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015925925925925927,
+      "loss": 0.7592,
+      "step": 1140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015905349794238685,
+      "loss": 0.7874,
+      "step": 1145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015884773662551443,
+      "loss": 0.7874,
+      "step": 1150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015864197530864198,
+      "loss": 0.7374,
+      "step": 1155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015843621399176956,
+      "loss": 0.7218,
+      "step": 1160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015823045267489714,
+      "loss": 0.7494,
+      "step": 1165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001580246913580247,
+      "loss": 0.7449,
+      "step": 1170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015781893004115227,
+      "loss": 0.725,
+      "step": 1175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015761316872427985,
+      "loss": 0.739,
+      "step": 1180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015740740740740743,
+      "loss": 0.7554,
+      "step": 1185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015720164609053498,
+      "loss": 0.7702,
+      "step": 1190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015699588477366256,
+      "loss": 0.7418,
+      "step": 1195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015679012345679014,
+      "loss": 0.7217,
+      "step": 1200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001565843621399177,
+      "loss": 0.7391,
+      "step": 1205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015637860082304527,
+      "loss": 0.7696,
+      "step": 1210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015617283950617285,
+      "loss": 0.7594,
+      "step": 1215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001559670781893004,
+      "loss": 0.777,
+      "step": 1220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015576131687242798,
+      "loss": 0.7441,
+      "step": 1225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015555555555555556,
+      "loss": 0.7611,
+      "step": 1230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015534979423868314,
+      "loss": 0.7278,
+      "step": 1235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001551440329218107,
+      "loss": 0.7452,
+      "step": 1240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015493827160493827,
+      "loss": 0.7939,
+      "step": 1245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015473251028806585,
+      "loss": 0.7161,
+      "step": 1250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001545267489711934,
+      "loss": 0.7344,
+      "step": 1255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015432098765432098,
+      "loss": 0.7368,
+      "step": 1260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015411522633744856,
+      "loss": 0.7172,
+      "step": 1265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015390946502057614,
+      "loss": 0.7589,
+      "step": 1270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001537037037037037,
+      "loss": 0.7641,
+      "step": 1275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015349794238683127,
+      "loss": 0.761,
+      "step": 1280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00015329218106995885,
+      "loss": 0.7414,
+      "step": 1285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001530864197530864,
+      "loss": 0.7713,
+      "step": 1290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015288065843621399,
+      "loss": 0.7543,
+      "step": 1295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015267489711934156,
+      "loss": 0.7683,
+      "step": 1300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015246913580246914,
+      "loss": 0.8106,
+      "step": 1305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015226337448559672,
+      "loss": 0.7739,
+      "step": 1310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015205761316872428,
+      "loss": 0.7585,
+      "step": 1315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015185185185185185,
+      "loss": 0.7577,
+      "step": 1320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015164609053497943,
+      "loss": 0.7608,
+      "step": 1325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000151440329218107,
+      "loss": 0.7715,
+      "step": 1330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001512345679012346,
+      "loss": 0.7681,
+      "step": 1335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015102880658436214,
+      "loss": 0.7476,
+      "step": 1340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015082304526748972,
+      "loss": 0.7539,
+      "step": 1345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001506172839506173,
+      "loss": 0.7492,
+      "step": 1350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015041152263374488,
+      "loss": 0.7491,
+      "step": 1355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015020576131687243,
+      "loss": 0.7236,
+      "step": 1360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 0.7328,
+      "step": 1365
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001497942386831276,
+      "loss": 0.7603,
+      "step": 1370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014958847736625515,
+      "loss": 0.7206,
+      "step": 1375
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014938271604938272,
+      "loss": 0.7001,
+      "step": 1380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001491769547325103,
+      "loss": 0.7241,
+      "step": 1385
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00014897119341563788,
+      "loss": 0.7511,
+      "step": 1390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014876543209876544,
+      "loss": 0.7546,
+      "step": 1395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014855967078189301,
+      "loss": 0.762,
+      "step": 1400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001483539094650206,
+      "loss": 0.6979,
+      "step": 1405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014814814814814815,
+      "loss": 0.7352,
+      "step": 1410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00014794238683127573,
+      "loss": 0.7604,
+      "step": 1415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001477366255144033,
+      "loss": 0.7612,
+      "step": 1420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014753086419753086,
+      "loss": 0.7618,
+      "step": 1425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014732510288065844,
+      "loss": 0.7107,
+      "step": 1430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00014711934156378602,
+      "loss": 0.7247,
+      "step": 1435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001469135802469136,
+      "loss": 0.7733,
+      "step": 1440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014670781893004115,
+      "loss": 0.7627,
+      "step": 1445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014650205761316873,
+      "loss": 0.7323,
+      "step": 1450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001462962962962963,
+      "loss": 0.7728,
+      "step": 1455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014609053497942386,
+      "loss": 0.7541,
+      "step": 1460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00014588477366255144,
+      "loss": 0.7781,
+      "step": 1465
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014567901234567902,
+      "loss": 0.7474,
+      "step": 1470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001454732510288066,
+      "loss": 0.7789,
+      "step": 1475
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014526748971193415,
+      "loss": 0.7497,
+      "step": 1480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014506172839506173,
+      "loss": 0.7487,
+      "step": 1485
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001448559670781893,
+      "loss": 0.7563,
+      "step": 1490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014465020576131686,
+      "loss": 0.7444,
+      "step": 1495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014444444444444444,
+      "loss": 0.7399,
+      "step": 1500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014423868312757202,
+      "loss": 0.7288,
+      "step": 1505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001440329218106996,
+      "loss": 0.7458,
+      "step": 1510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014382716049382718,
+      "loss": 0.7297,
+      "step": 1515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014362139917695473,
+      "loss": 0.7684,
+      "step": 1520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001434156378600823,
+      "loss": 0.7428,
+      "step": 1525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014320987654320989,
+      "loss": 0.705,
+      "step": 1530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014300411522633747,
+      "loss": 0.7445,
+      "step": 1535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00014279835390946504,
+      "loss": 0.7109,
+      "step": 1540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001425925925925926,
+      "loss": 0.7502,
+      "step": 1545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014238683127572018,
+      "loss": 0.7485,
+      "step": 1550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014218106995884776,
+      "loss": 0.7421,
+      "step": 1555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00014197530864197534,
+      "loss": 0.7369,
+      "step": 1560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001417695473251029,
+      "loss": 0.7746,
+      "step": 1565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014156378600823047,
+      "loss": 0.7334,
+      "step": 1570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014135802469135805,
+      "loss": 0.7251,
+      "step": 1575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001411522633744856,
+      "loss": 0.7146,
+      "step": 1580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014094650205761318,
+      "loss": 0.7563,
+      "step": 1585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00014074074074074076,
+      "loss": 0.7247,
+      "step": 1590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014053497942386834,
+      "loss": 0.7379,
+      "step": 1595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001403292181069959,
+      "loss": 0.7684,
+      "step": 1600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00014012345679012347,
+      "loss": 0.7574,
+      "step": 1605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00013991769547325105,
+      "loss": 0.7138,
+      "step": 1610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001397119341563786,
+      "loss": 0.7472,
+      "step": 1615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013950617283950618,
+      "loss": 0.7466,
+      "step": 1620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013930041152263376,
+      "loss": 0.7288,
+      "step": 1625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013909465020576134,
+      "loss": 0.7717,
+      "step": 1630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001388888888888889,
+      "loss": 0.7665,
+      "step": 1635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00013868312757201647,
+      "loss": 0.7408,
+      "step": 1640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013847736625514405,
+      "loss": 0.7418,
+      "step": 1645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001382716049382716,
+      "loss": 0.7502,
+      "step": 1650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013806584362139918,
+      "loss": 0.7657,
+      "step": 1655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00013786008230452676,
+      "loss": 0.728,
+      "step": 1660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001376543209876543,
+      "loss": 0.7142,
+      "step": 1665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001374485596707819,
+      "loss": 0.7971,
+      "step": 1670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013724279835390947,
+      "loss": 0.737,
+      "step": 1675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013703703703703705,
+      "loss": 0.7947,
+      "step": 1680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001368312757201646,
+      "loss": 0.7525,
+      "step": 1685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013662551440329218,
+      "loss": 0.711,
+      "step": 1690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013641975308641976,
+      "loss": 0.7459,
+      "step": 1695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001362139917695473,
+      "loss": 0.7404,
+      "step": 1700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001360082304526749,
+      "loss": 0.7458,
+      "step": 1705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013580246913580247,
+      "loss": 0.754,
+      "step": 1710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013559670781893005,
+      "loss": 0.7217,
+      "step": 1715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001353909465020576,
+      "loss": 0.7373,
+      "step": 1720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013518518518518518,
+      "loss": 0.7567,
+      "step": 1725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013497942386831276,
+      "loss": 0.7525,
+      "step": 1730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013477366255144034,
+      "loss": 0.752,
+      "step": 1735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00013456790123456792,
+      "loss": 0.7342,
+      "step": 1740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001343621399176955,
+      "loss": 0.7276,
+      "step": 1745
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013415637860082305,
+      "loss": 0.6915,
+      "step": 1750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00013395061728395063,
+      "loss": 0.737,
+      "step": 1755
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001337448559670782,
+      "loss": 0.7427,
+      "step": 1760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001335390946502058,
+      "loss": 0.7126,
+      "step": 1765
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.7241,
+      "step": 1770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013312757201646092,
+      "loss": 0.755,
+      "step": 1775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001329218106995885,
+      "loss": 0.7494,
+      "step": 1780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013271604938271605,
+      "loss": 0.709,
+      "step": 1785
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00013251028806584363,
+      "loss": 0.7555,
+      "step": 1790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001323045267489712,
+      "loss": 0.7298,
+      "step": 1795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001320987654320988,
+      "loss": 0.7232,
+      "step": 1800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013189300411522634,
+      "loss": 0.7441,
+      "step": 1805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00013168724279835392,
+      "loss": 0.7623,
+      "step": 1810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001314814814814815,
+      "loss": 0.7349,
+      "step": 1815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013127572016460905,
+      "loss": 0.7773,
+      "step": 1820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013106995884773663,
+      "loss": 0.7537,
+      "step": 1825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001308641975308642,
+      "loss": 0.7579,
+      "step": 1830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001306584362139918,
+      "loss": 0.7321,
+      "step": 1835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00013045267489711934,
+      "loss": 0.7175,
+      "step": 1840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00013024691358024692,
+      "loss": 0.7553,
+      "step": 1845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001300411522633745,
+      "loss": 0.7591,
+      "step": 1850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012983539094650205,
+      "loss": 0.7581,
+      "step": 1855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00012962962962962963,
+      "loss": 0.7632,
+      "step": 1860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001294238683127572,
+      "loss": 0.7434,
+      "step": 1865
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012921810699588476,
+      "loss": 0.7695,
+      "step": 1870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012901234567901234,
+      "loss": 0.7808,
+      "step": 1875
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012880658436213992,
+      "loss": 0.7428,
+      "step": 1880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001286008230452675,
+      "loss": 0.7487,
+      "step": 1885
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00012839506172839505,
+      "loss": 0.7033,
+      "step": 1890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012818930041152263,
+      "loss": 0.7639,
+      "step": 1895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001279835390946502,
+      "loss": 0.7345,
+      "step": 1900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012777777777777776,
+      "loss": 0.7434,
+      "step": 1905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012757201646090534,
+      "loss": 0.7082,
+      "step": 1910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00012736625514403292,
+      "loss": 0.7294,
+      "step": 1915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001271604938271605,
+      "loss": 0.74,
+      "step": 1920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012695473251028805,
+      "loss": 0.7694,
+      "step": 1925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012674897119341563,
+      "loss": 0.7339,
+      "step": 1930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001265432098765432,
+      "loss": 0.7259,
+      "step": 1935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001263374485596708,
+      "loss": 0.7307,
+      "step": 1940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012613168724279837,
+      "loss": 0.7085,
+      "step": 1945
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012592592592592592,
+      "loss": 0.7662,
+      "step": 1950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001257201646090535,
+      "loss": 0.7145,
+      "step": 1955
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012551440329218108,
+      "loss": 0.7528,
+      "step": 1960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00012530864197530866,
+      "loss": 0.7308,
+      "step": 1965
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012510288065843624,
+      "loss": 0.7163,
+      "step": 1970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001248971193415638,
+      "loss": 0.7254,
+      "step": 1975
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012469135802469137,
+      "loss": 0.7467,
+      "step": 1980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00012448559670781895,
+      "loss": 0.7391,
+      "step": 1985
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001242798353909465,
+      "loss": 0.7314,
+      "step": 1990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012407407407407408,
+      "loss": 0.7048,
+      "step": 1995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012386831275720166,
+      "loss": 0.7187,
+      "step": 2000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012366255144032924,
+      "loss": 0.735,
+      "step": 2005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001234567901234568,
+      "loss": 0.7662,
+      "step": 2010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00012325102880658437,
+      "loss": 0.7437,
+      "step": 2015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012304526748971195,
+      "loss": 0.7675,
+      "step": 2020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001228395061728395,
+      "loss": 0.7143,
+      "step": 2025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012263374485596708,
+      "loss": 0.7101,
+      "step": 2030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012242798353909466,
+      "loss": 0.7257,
+      "step": 2035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00012222222222222224,
+      "loss": 0.7522,
+      "step": 2040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001220164609053498,
+      "loss": 0.7669,
+      "step": 2045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012181069958847737,
+      "loss": 0.7283,
+      "step": 2050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012160493827160495,
+      "loss": 0.7413,
+      "step": 2055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001213991769547325,
+      "loss": 0.6988,
+      "step": 2060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00012119341563786008,
+      "loss": 0.7421,
+      "step": 2065
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012098765432098766,
+      "loss": 0.7152,
+      "step": 2070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012078189300411524,
+      "loss": 0.7741,
+      "step": 2075
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001205761316872428,
+      "loss": 0.7381,
+      "step": 2080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012037037037037037,
+      "loss": 0.738,
+      "step": 2085
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00012016460905349795,
+      "loss": 0.7258,
+      "step": 2090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011995884773662552,
+      "loss": 0.7426,
+      "step": 2095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011975308641975308,
+      "loss": 0.7185,
+      "step": 2100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011954732510288066,
+      "loss": 0.7401,
+      "step": 2105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011934156378600823,
+      "loss": 0.751,
+      "step": 2110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00011913580246913581,
+      "loss": 0.7284,
+      "step": 2115
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011893004115226339,
+      "loss": 0.7901,
+      "step": 2120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011872427983539097,
+      "loss": 0.7432,
+      "step": 2125
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011851851851851852,
+      "loss": 0.7355,
+      "step": 2130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001183127572016461,
+      "loss": 0.7333,
+      "step": 2135
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00011810699588477368,
+      "loss": 0.7504,
+      "step": 2140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011790123456790123,
+      "loss": 0.7503,
+      "step": 2145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011769547325102881,
+      "loss": 0.728,
+      "step": 2150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011748971193415639,
+      "loss": 0.7485,
+      "step": 2155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011728395061728397,
+      "loss": 0.7127,
+      "step": 2160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00011707818930041152,
+      "loss": 0.7278,
+      "step": 2165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001168724279835391,
+      "loss": 0.7513,
+      "step": 2170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011666666666666668,
+      "loss": 0.7193,
+      "step": 2175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011646090534979423,
+      "loss": 0.7245,
+      "step": 2180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011625514403292181,
+      "loss": 0.7266,
+      "step": 2185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00011604938271604939,
+      "loss": 0.7531,
+      "step": 2190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011584362139917696,
+      "loss": 0.7637,
+      "step": 2195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011563786008230452,
+      "loss": 0.7388,
+      "step": 2200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001154320987654321,
+      "loss": 0.7222,
+      "step": 2205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011522633744855968,
+      "loss": 0.7667,
+      "step": 2210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00011502057613168725,
+      "loss": 0.7382,
+      "step": 2215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011481481481481482,
+      "loss": 0.7243,
+      "step": 2220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001146090534979424,
+      "loss": 0.7192,
+      "step": 2225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011440329218106996,
+      "loss": 0.7237,
+      "step": 2230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011419753086419754,
+      "loss": 0.7498,
+      "step": 2235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00011399176954732511,
+      "loss": 0.7007,
+      "step": 2240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001137860082304527,
+      "loss": 0.758,
+      "step": 2245
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011358024691358025,
+      "loss": 0.7111,
+      "step": 2250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011337448559670783,
+      "loss": 0.706,
+      "step": 2255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001131687242798354,
+      "loss": 0.7423,
+      "step": 2260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00011296296296296296,
+      "loss": 0.7094,
+      "step": 2265
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011275720164609054,
+      "loss": 0.7331,
+      "step": 2270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011255144032921812,
+      "loss": 0.7164,
+      "step": 2275
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001123456790123457,
+      "loss": 0.7452,
+      "step": 2280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011213991769547325,
+      "loss": 0.7172,
+      "step": 2285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00011193415637860083,
+      "loss": 0.7232,
+      "step": 2290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001117283950617284,
+      "loss": 0.7112,
+      "step": 2295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011152263374485596,
+      "loss": 0.7459,
+      "step": 2300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011131687242798354,
+      "loss": 0.7353,
+      "step": 2305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011111111111111112,
+      "loss": 0.7406,
+      "step": 2310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00011090534979423868,
+      "loss": 0.7623,
+      "step": 2315
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011069958847736626,
+      "loss": 0.7567,
+      "step": 2320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011049382716049384,
+      "loss": 0.7752,
+      "step": 2325
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001102880658436214,
+      "loss": 0.7419,
+      "step": 2330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00011008230452674897,
+      "loss": 0.7244,
+      "step": 2335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00010987654320987655,
+      "loss": 0.7553,
+      "step": 2340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010967078189300413,
+      "loss": 0.7389,
+      "step": 2345
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010946502057613168,
+      "loss": 0.7002,
+      "step": 2350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010925925925925926,
+      "loss": 0.7238,
+      "step": 2355
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010905349794238684,
+      "loss": 0.6923,
+      "step": 2360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00010884773662551442,
+      "loss": 0.7579,
+      "step": 2365
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010864197530864197,
+      "loss": 0.7291,
+      "step": 2370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010843621399176955,
+      "loss": 0.721,
+      "step": 2375
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010823045267489713,
+      "loss": 0.7605,
+      "step": 2380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010802469135802468,
+      "loss": 0.7297,
+      "step": 2385
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00010781893004115226,
+      "loss": 0.7565,
+      "step": 2390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010761316872427984,
+      "loss": 0.701,
+      "step": 2395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010740740740740742,
+      "loss": 0.736,
+      "step": 2400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010720164609053497,
+      "loss": 0.7079,
+      "step": 2405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010699588477366255,
+      "loss": 0.7349,
+      "step": 2410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00010679012345679013,
+      "loss": 0.7106,
+      "step": 2415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001065843621399177,
+      "loss": 0.7396,
+      "step": 2420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010637860082304528,
+      "loss": 0.7713,
+      "step": 2425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010617283950617284,
+      "loss": 0.747,
+      "step": 2430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010596707818930041,
+      "loss": 0.7475,
+      "step": 2435
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00010576131687242799,
+      "loss": 0.7449,
+      "step": 2440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010555555555555557,
+      "loss": 0.7208,
+      "step": 2445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010534979423868315,
+      "loss": 0.7317,
+      "step": 2450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001051440329218107,
+      "loss": 0.7419,
+      "step": 2455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010493827160493828,
+      "loss": 0.7148,
+      "step": 2460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00010473251028806586,
+      "loss": 0.7143,
+      "step": 2465
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010452674897119341,
+      "loss": 0.747,
+      "step": 2470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010432098765432099,
+      "loss": 0.7079,
+      "step": 2475
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010411522633744857,
+      "loss": 0.7059,
+      "step": 2480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00010390946502057615,
+      "loss": 0.7272,
+      "step": 2485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001037037037037037,
+      "loss": 0.7281,
+      "step": 2490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010349794238683128,
+      "loss": 0.6956,
+      "step": 2495
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010329218106995886,
+      "loss": 0.7571,
+      "step": 2500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010308641975308641,
+      "loss": 0.7311,
+      "step": 2505
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010288065843621399,
+      "loss": 0.6815,
+      "step": 2510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00010267489711934157,
+      "loss": 0.6781,
+      "step": 2515
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010246913580246915,
+      "loss": 0.6948,
+      "step": 2520
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010226337448559671,
+      "loss": 0.7133,
+      "step": 2525
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010205761316872428,
+      "loss": 0.688,
+      "step": 2530
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010185185185185186,
+      "loss": 0.6687,
+      "step": 2535
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00010164609053497942,
+      "loss": 0.7142,
+      "step": 2540
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000101440329218107,
+      "loss": 0.6885,
+      "step": 2545
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010123456790123458,
+      "loss": 0.6867,
+      "step": 2550
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010102880658436214,
+      "loss": 0.7045,
+      "step": 2555
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00010082304526748971,
+      "loss": 0.7375,
+      "step": 2560
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001006172839506173,
+      "loss": 0.7157,
+      "step": 2565
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010041152263374487,
+      "loss": 0.6828,
+      "step": 2570
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00010020576131687243,
+      "loss": 0.7048,
+      "step": 2575
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001,
+      "loss": 0.6944,
+      "step": 2580
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.979423868312757e-05,
+      "loss": 0.696,
+      "step": 2585
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.958847736625515e-05,
+      "loss": 0.6609,
+      "step": 2590
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.938271604938272e-05,
+      "loss": 0.674,
+      "step": 2595
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.91769547325103e-05,
+      "loss": 0.7138,
+      "step": 2600
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.897119341563786e-05,
+      "loss": 0.713,
+      "step": 2605
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.876543209876543e-05,
+      "loss": 0.7022,
+      "step": 2610
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.8559670781893e-05,
+      "loss": 0.6955,
+      "step": 2615
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.835390946502057e-05,
+      "loss": 0.6888,
+      "step": 2620
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.814814814814815e-05,
+      "loss": 0.7003,
+      "step": 2625
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.794238683127573e-05,
+      "loss": 0.7167,
+      "step": 2630
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.77366255144033e-05,
+      "loss": 0.7087,
+      "step": 2635
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.753086419753087e-05,
+      "loss": 0.7019,
+      "step": 2640
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.732510288065844e-05,
+      "loss": 0.6609,
+      "step": 2645
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.711934156378602e-05,
+      "loss": 0.7367,
+      "step": 2650
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.691358024691359e-05,
+      "loss": 0.6862,
+      "step": 2655
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.670781893004116e-05,
+      "loss": 0.7121,
+      "step": 2660
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.650205761316873e-05,
+      "loss": 0.7031,
+      "step": 2665
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.62962962962963e-05,
+      "loss": 0.7005,
+      "step": 2670
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.609053497942388e-05,
+      "loss": 0.7041,
+      "step": 2675
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.588477366255144e-05,
+      "loss": 0.6713,
+      "step": 2680
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.567901234567902e-05,
+      "loss": 0.6932,
+      "step": 2685
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.547325102880659e-05,
+      "loss": 0.6849,
+      "step": 2690
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.526748971193417e-05,
+      "loss": 0.6915,
+      "step": 2695
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.506172839506173e-05,
+      "loss": 0.6677,
+      "step": 2700
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.48559670781893e-05,
+      "loss": 0.6767,
+      "step": 2705
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.465020576131688e-05,
+      "loss": 0.6873,
+      "step": 2710
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.444444444444444e-05,
+      "loss": 0.7007,
+      "step": 2715
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.423868312757202e-05,
+      "loss": 0.6782,
+      "step": 2720
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.403292181069959e-05,
+      "loss": 0.7044,
+      "step": 2725
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.382716049382717e-05,
+      "loss": 0.6787,
+      "step": 2730
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.362139917695473e-05,
+      "loss": 0.6876,
+      "step": 2735
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.341563786008231e-05,
+      "loss": 0.7253,
+      "step": 2740
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.320987654320989e-05,
+      "loss": 0.7159,
+      "step": 2745
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.300411522633746e-05,
+      "loss": 0.704,
+      "step": 2750
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.279835390946504e-05,
+      "loss": 0.6882,
+      "step": 2755
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.25925925925926e-05,
+      "loss": 0.6777,
+      "step": 2760
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.238683127572017e-05,
+      "loss": 0.7133,
+      "step": 2765
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.218106995884775e-05,
+      "loss": 0.6988,
+      "step": 2770
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.197530864197531e-05,
+      "loss": 0.7323,
+      "step": 2775
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.176954732510289e-05,
+      "loss": 0.6854,
+      "step": 2780
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.156378600823046e-05,
+      "loss": 0.6926,
+      "step": 2785
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.135802469135802e-05,
+      "loss": 0.6985,
+      "step": 2790
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.11522633744856e-05,
+      "loss": 0.6776,
+      "step": 2795
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.094650205761317e-05,
+      "loss": 0.7374,
+      "step": 2800
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.074074074074075e-05,
+      "loss": 0.6928,
+      "step": 2805
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.053497942386831e-05,
+      "loss": 0.7129,
+      "step": 2810
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.032921810699589e-05,
+      "loss": 0.6918,
+      "step": 2815
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.012345679012346e-05,
+      "loss": 0.6641,
+      "step": 2820
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.991769547325102e-05,
+      "loss": 0.7075,
+      "step": 2825
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.97119341563786e-05,
+      "loss": 0.668,
+      "step": 2830
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.950617283950617e-05,
+      "loss": 0.6836,
+      "step": 2835
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.930041152263375e-05,
+      "loss": 0.672,
+      "step": 2840
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.909465020576133e-05,
+      "loss": 0.6827,
+      "step": 2845
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.888888888888889e-05,
+      "loss": 0.6976,
+      "step": 2850
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.868312757201647e-05,
+      "loss": 0.6719,
+      "step": 2855
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.847736625514404e-05,
+      "loss": 0.6871,
+      "step": 2860
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.827160493827162e-05,
+      "loss": 0.6975,
+      "step": 2865
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.806584362139918e-05,
+      "loss": 0.684,
+      "step": 2870
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.786008230452676e-05,
+      "loss": 0.6861,
+      "step": 2875
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.765432098765433e-05,
+      "loss": 0.7116,
+      "step": 2880
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.74485596707819e-05,
+      "loss": 0.686,
+      "step": 2885
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.724279835390947e-05,
+      "loss": 0.677,
+      "step": 2890
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.703703703703704e-05,
+      "loss": 0.6996,
+      "step": 2895
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.683127572016462e-05,
+      "loss": 0.7211,
+      "step": 2900
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.662551440329218e-05,
+      "loss": 0.6766,
+      "step": 2905
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.641975308641975e-05,
+      "loss": 0.6883,
+      "step": 2910
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.621399176954733e-05,
+      "loss": 0.6829,
+      "step": 2915
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60082304526749e-05,
+      "loss": 0.7089,
+      "step": 2920
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.580246913580247e-05,
+      "loss": 0.6711,
+      "step": 2925
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.559670781893004e-05,
+      "loss": 0.6926,
+      "step": 2930
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.539094650205762e-05,
+      "loss": 0.6513,
+      "step": 2935
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.518518518518518e-05,
+      "loss": 0.6817,
+      "step": 2940
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.497942386831276e-05,
+      "loss": 0.7005,
+      "step": 2945
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.477366255144033e-05,
+      "loss": 0.7013,
+      "step": 2950
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.456790123456791e-05,
+      "loss": 0.6562,
+      "step": 2955
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.436213991769549e-05,
+      "loss": 0.7074,
+      "step": 2960
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.415637860082305e-05,
+      "loss": 0.6901,
+      "step": 2965
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.395061728395062e-05,
+      "loss": 0.6867,
+      "step": 2970
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.37448559670782e-05,
+      "loss": 0.692,
+      "step": 2975
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.353909465020576e-05,
+      "loss": 0.6863,
+      "step": 2980
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.333333333333334e-05,
+      "loss": 0.6881,
+      "step": 2985
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.312757201646091e-05,
+      "loss": 0.7203,
+      "step": 2990
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.292181069958848e-05,
+      "loss": 0.7075,
+      "step": 2995
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.271604938271605e-05,
+      "loss": 0.7187,
+      "step": 3000
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.251028806584362e-05,
+      "loss": 0.6791,
+      "step": 3005
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.23045267489712e-05,
+      "loss": 0.6937,
+      "step": 3010
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.209876543209877e-05,
+      "loss": 0.6584,
+      "step": 3015
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.189300411522634e-05,
+      "loss": 0.6818,
+      "step": 3020
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.168724279835391e-05,
+      "loss": 0.6827,
+      "step": 3025
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.148148148148148e-05,
+      "loss": 0.6774,
+      "step": 3030
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.127572016460906e-05,
+      "loss": 0.6826,
+      "step": 3035
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.106995884773662e-05,
+      "loss": 0.6767,
+      "step": 3040
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.08641975308642e-05,
+      "loss": 0.7256,
+      "step": 3045
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.065843621399177e-05,
+      "loss": 0.6875,
+      "step": 3050
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.045267489711935e-05,
+      "loss": 0.7013,
+      "step": 3055
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.024691358024692e-05,
+      "loss": 0.6907,
+      "step": 3060
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.004115226337449e-05,
+      "loss": 0.6751,
+      "step": 3065
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.983539094650207e-05,
+      "loss": 0.6893,
+      "step": 3070
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.962962962962964e-05,
+      "loss": 0.6962,
+      "step": 3075
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.942386831275721e-05,
+      "loss": 0.6743,
+      "step": 3080
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.921810699588478e-05,
+      "loss": 0.7004,
+      "step": 3085
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.901234567901235e-05,
+      "loss": 0.6908,
+      "step": 3090
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.880658436213993e-05,
+      "loss": 0.6736,
+      "step": 3095
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.860082304526749e-05,
+      "loss": 0.7153,
+      "step": 3100
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.839506172839507e-05,
+      "loss": 0.7072,
+      "step": 3105
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.818930041152264e-05,
+      "loss": 0.7163,
+      "step": 3110
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.79835390946502e-05,
+      "loss": 0.6996,
+      "step": 3115
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.777777777777778e-05,
+      "loss": 0.7257,
+      "step": 3120
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.757201646090535e-05,
+      "loss": 0.7155,
+      "step": 3125
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.736625514403293e-05,
+      "loss": 0.7068,
+      "step": 3130
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.716049382716049e-05,
+      "loss": 0.6575,
+      "step": 3135
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.695473251028807e-05,
+      "loss": 0.6771,
+      "step": 3140
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.674897119341564e-05,
+      "loss": 0.6976,
+      "step": 3145
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.65432098765432e-05,
+      "loss": 0.6568,
+      "step": 3150
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.633744855967078e-05,
+      "loss": 0.6609,
+      "step": 3155
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.613168724279836e-05,
+      "loss": 0.697,
+      "step": 3160
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.592592592592593e-05,
+      "loss": 0.714,
+      "step": 3165
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.57201646090535e-05,
+      "loss": 0.6829,
+      "step": 3170
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.551440329218107e-05,
+      "loss": 0.6932,
+      "step": 3175
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.530864197530865e-05,
+      "loss": 0.6859,
+      "step": 3180
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.510288065843622e-05,
+      "loss": 0.7117,
+      "step": 3185
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.48971193415638e-05,
+      "loss": 0.7157,
+      "step": 3190
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.469135802469136e-05,
+      "loss": 0.6593,
+      "step": 3195
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.448559670781894e-05,
+      "loss": 0.6683,
+      "step": 3200
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.427983539094651e-05,
+      "loss": 0.6876,
+      "step": 3205
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.407407407407407e-05,
+      "loss": 0.6584,
+      "step": 3210
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.386831275720165e-05,
+      "loss": 0.6816,
+      "step": 3215
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.366255144032922e-05,
+      "loss": 0.6994,
+      "step": 3220
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.34567901234568e-05,
+      "loss": 0.6701,
+      "step": 3225
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.325102880658436e-05,
+      "loss": 0.6869,
+      "step": 3230
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.304526748971193e-05,
+      "loss": 0.6895,
+      "step": 3235
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.283950617283951e-05,
+      "loss": 0.6972,
+      "step": 3240
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.263374485596707e-05,
+      "loss": 0.6849,
+      "step": 3245
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.242798353909465e-05,
+      "loss": 0.6967,
+      "step": 3250
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.222222222222222e-05,
+      "loss": 0.6679,
+      "step": 3255
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.20164609053498e-05,
+      "loss": 0.6683,
+      "step": 3260
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.181069958847736e-05,
+      "loss": 0.6787,
+      "step": 3265
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.160493827160494e-05,
+      "loss": 0.6726,
+      "step": 3270
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.139917695473252e-05,
+      "loss": 0.721,
+      "step": 3275
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.119341563786009e-05,
+      "loss": 0.6865,
+      "step": 3280
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.098765432098767e-05,
+      "loss": 0.7031,
+      "step": 3285
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.078189300411523e-05,
+      "loss": 0.6895,
+      "step": 3290
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.05761316872428e-05,
+      "loss": 0.6983,
+      "step": 3295
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.037037037037038e-05,
+      "loss": 0.6929,
+      "step": 3300
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.016460905349794e-05,
+      "loss": 0.7117,
+      "step": 3305
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.995884773662552e-05,
+      "loss": 0.664,
+      "step": 3310
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.975308641975309e-05,
+      "loss": 0.6645,
+      "step": 3315
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.954732510288067e-05,
+      "loss": 0.6985,
+      "step": 3320
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.934156378600823e-05,
+      "loss": 0.6826,
+      "step": 3325
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.91358024691358e-05,
+      "loss": 0.7324,
+      "step": 3330
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.893004115226338e-05,
+      "loss": 0.6828,
+      "step": 3335
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.872427983539094e-05,
+      "loss": 0.6801,
+      "step": 3340
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.851851851851852e-05,
+      "loss": 0.6923,
+      "step": 3345
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.831275720164609e-05,
+      "loss": 0.6787,
+      "step": 3350
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.810699588477366e-05,
+      "loss": 0.6812,
+      "step": 3355
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.790123456790123e-05,
+      "loss": 0.7059,
+      "step": 3360
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.76954732510288e-05,
+      "loss": 0.6794,
+      "step": 3365
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.748971193415638e-05,
+      "loss": 0.6635,
+      "step": 3370
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.728395061728396e-05,
+      "loss": 0.666,
+      "step": 3375
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.707818930041152e-05,
+      "loss": 0.6909,
+      "step": 3380
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.68724279835391e-05,
+      "loss": 0.7096,
+      "step": 3385
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.7049,
+      "step": 3390
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.646090534979425e-05,
+      "loss": 0.6801,
+      "step": 3395
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.625514403292181e-05,
+      "loss": 0.7045,
+      "step": 3400
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.60493827160494e-05,
+      "loss": 0.7096,
+      "step": 3405
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.584362139917696e-05,
+      "loss": 0.7192,
+      "step": 3410
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.563786008230453e-05,
+      "loss": 0.6818,
+      "step": 3415
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.54320987654321e-05,
+      "loss": 0.6358,
+      "step": 3420
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.522633744855967e-05,
+      "loss": 0.6674,
+      "step": 3425
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.502057613168725e-05,
+      "loss": 0.684,
+      "step": 3430
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.481481481481482e-05,
+      "loss": 0.6753,
+      "step": 3435
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.460905349794238e-05,
+      "loss": 0.6902,
+      "step": 3440
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.440329218106996e-05,
+      "loss": 0.6664,
+      "step": 3445
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.419753086419753e-05,
+      "loss": 0.684,
+      "step": 3450
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.39917695473251e-05,
+      "loss": 0.682,
+      "step": 3455
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.378600823045267e-05,
+      "loss": 0.6764,
+      "step": 3460
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.358024691358025e-05,
+      "loss": 0.6783,
+      "step": 3465
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.337448559670782e-05,
+      "loss": 0.6962,
+      "step": 3470
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.31687242798354e-05,
+      "loss": 0.6819,
+      "step": 3475
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.296296296296296e-05,
+      "loss": 0.6741,
+      "step": 3480
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.275720164609054e-05,
+      "loss": 0.6754,
+      "step": 3485
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.255144032921812e-05,
+      "loss": 0.6717,
+      "step": 3490
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.234567901234569e-05,
+      "loss": 0.667,
+      "step": 3495
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.213991769547325e-05,
+      "loss": 0.6822,
+      "step": 3500
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.193415637860083e-05,
+      "loss": 0.7055,
+      "step": 3505
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.17283950617284e-05,
+      "loss": 0.6988,
+      "step": 3510
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.152263374485598e-05,
+      "loss": 0.6793,
+      "step": 3515
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.131687242798354e-05,
+      "loss": 0.7093,
+      "step": 3520
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.111111111111112e-05,
+      "loss": 0.694,
+      "step": 3525
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0905349794238687e-05,
+      "loss": 0.7077,
+      "step": 3530
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.069958847736625e-05,
+      "loss": 0.6938,
+      "step": 3535
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.049382716049383e-05,
+      "loss": 0.6461,
+      "step": 3540
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.02880658436214e-05,
+      "loss": 0.6993,
+      "step": 3545
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0082304526748977e-05,
+      "loss": 0.6901,
+      "step": 3550
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.987654320987654e-05,
+      "loss": 0.6937,
+      "step": 3555
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9670781893004115e-05,
+      "loss": 0.6657,
+      "step": 3560
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9465020576131694e-05,
+      "loss": 0.6593,
+      "step": 3565
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.925925925925926e-05,
+      "loss": 0.7056,
+      "step": 3570
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.905349794238684e-05,
+      "loss": 0.6395,
+      "step": 3575
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8847736625514405e-05,
+      "loss": 0.6608,
+      "step": 3580
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8641975308641984e-05,
+      "loss": 0.7007,
+      "step": 3585
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.843621399176955e-05,
+      "loss": 0.6605,
+      "step": 3590
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8230452674897116e-05,
+      "loss": 0.6752,
+      "step": 3595
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.8024691358024695e-05,
+      "loss": 0.6736,
+      "step": 3600
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.781893004115226e-05,
+      "loss": 0.671,
+      "step": 3605
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.761316872427984e-05,
+      "loss": 0.6551,
+      "step": 3610
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.740740740740741e-05,
+      "loss": 0.7053,
+      "step": 3615
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.720164609053498e-05,
+      "loss": 0.6856,
+      "step": 3620
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.699588477366256e-05,
+      "loss": 0.671,
+      "step": 3625
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.679012345679012e-05,
+      "loss": 0.6763,
+      "step": 3630
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.65843621399177e-05,
+      "loss": 0.6812,
+      "step": 3635
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.637860082304527e-05,
+      "loss": 0.7407,
+      "step": 3640
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.617283950617285e-05,
+      "loss": 0.7436,
+      "step": 3645
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.596707818930041e-05,
+      "loss": 0.682,
+      "step": 3650
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.576131687242798e-05,
+      "loss": 0.6808,
+      "step": 3655
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.555555555555556e-05,
+      "loss": 0.6696,
+      "step": 3660
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.534979423868313e-05,
+      "loss": 0.7207,
+      "step": 3665
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.51440329218107e-05,
+      "loss": 0.6737,
+      "step": 3670
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4938271604938276e-05,
+      "loss": 0.6874,
+      "step": 3675
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.473251028806584e-05,
+      "loss": 0.6652,
+      "step": 3680
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.452674897119342e-05,
+      "loss": 0.6739,
+      "step": 3685
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4320987654320986e-05,
+      "loss": 0.7039,
+      "step": 3690
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4115226337448566e-05,
+      "loss": 0.696,
+      "step": 3695
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.390946502057613e-05,
+      "loss": 0.6887,
+      "step": 3700
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.370370370370371e-05,
+      "loss": 0.6653,
+      "step": 3705
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3497942386831277e-05,
+      "loss": 0.6987,
+      "step": 3710
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.329218106995885e-05,
+      "loss": 0.6831,
+      "step": 3715
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.308641975308642e-05,
+      "loss": 0.687,
+      "step": 3720
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2880658436213994e-05,
+      "loss": 0.6848,
+      "step": 3725
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.267489711934157e-05,
+      "loss": 0.6819,
+      "step": 3730
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.246913580246914e-05,
+      "loss": 0.662,
+      "step": 3735
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2263374485596705e-05,
+      "loss": 0.6462,
+      "step": 3740
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2057613168724284e-05,
+      "loss": 0.684,
+      "step": 3745
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.185185185185185e-05,
+      "loss": 0.721,
+      "step": 3750
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.164609053497943e-05,
+      "loss": 0.6998,
+      "step": 3755
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1440329218106995e-05,
+      "loss": 0.6922,
+      "step": 3760
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1234567901234574e-05,
+      "loss": 0.6674,
+      "step": 3765
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.102880658436214e-05,
+      "loss": 0.7022,
+      "step": 3770
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.082304526748971e-05,
+      "loss": 0.6753,
+      "step": 3775
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.061728395061729e-05,
+      "loss": 0.6853,
+      "step": 3780
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.041152263374486e-05,
+      "loss": 0.6669,
+      "step": 3785
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.020576131687244e-05,
+      "loss": 0.6678,
+      "step": 3790
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5e-05,
+      "loss": 0.6881,
+      "step": 3795
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9794238683127575e-05,
+      "loss": 0.6959,
+      "step": 3800
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.958847736625515e-05,
+      "loss": 0.695,
+      "step": 3805
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.938271604938271e-05,
+      "loss": 0.6927,
+      "step": 3810
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9176954732510286e-05,
+      "loss": 0.7039,
+      "step": 3815
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.8971193415637865e-05,
+      "loss": 0.6494,
+      "step": 3820
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.876543209876544e-05,
+      "loss": 0.7015,
+      "step": 3825
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.855967078189301e-05,
+      "loss": 0.6901,
+      "step": 3830
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.835390946502058e-05,
+      "loss": 0.6618,
+      "step": 3835
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.814814814814815e-05,
+      "loss": 0.6723,
+      "step": 3840
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.794238683127572e-05,
+      "loss": 0.7196,
+      "step": 3845
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.773662551440329e-05,
+      "loss": 0.6867,
+      "step": 3850
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7530864197530866e-05,
+      "loss": 0.6921,
+      "step": 3855
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.732510288065844e-05,
+      "loss": 0.693,
+      "step": 3860
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.711934156378601e-05,
+      "loss": 0.7031,
+      "step": 3865
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.691358024691358e-05,
+      "loss": 0.7023,
+      "step": 3870
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6707818930041156e-05,
+      "loss": 0.6754,
+      "step": 3875
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.650205761316873e-05,
+      "loss": 0.6985,
+      "step": 3880
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.62962962962963e-05,
+      "loss": 0.665,
+      "step": 3885
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.609053497942387e-05,
+      "loss": 0.6902,
+      "step": 3890
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.5884773662551446e-05,
+      "loss": 0.7173,
+      "step": 3895
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.567901234567901e-05,
+      "loss": 0.7121,
+      "step": 3900
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5473251028806584e-05,
+      "loss": 0.6939,
+      "step": 3905
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5267489711934157e-05,
+      "loss": 0.7093,
+      "step": 3910
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.506172839506173e-05,
+      "loss": 0.6792,
+      "step": 3915
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.48559670781893e-05,
+      "loss": 0.6737,
+      "step": 3920
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4650205761316874e-05,
+      "loss": 0.6858,
+      "step": 3925
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4444444444444447e-05,
+      "loss": 0.6776,
+      "step": 3930
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.423868312757202e-05,
+      "loss": 0.6772,
+      "step": 3935
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.403292181069959e-05,
+      "loss": 0.6667,
+      "step": 3940
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.3827160493827164e-05,
+      "loss": 0.6892,
+      "step": 3945
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3621399176954737e-05,
+      "loss": 0.6652,
+      "step": 3950
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.341563786008231e-05,
+      "loss": 0.6892,
+      "step": 3955
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3209876543209875e-05,
+      "loss": 0.6917,
+      "step": 3960
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.300411522633745e-05,
+      "loss": 0.6708,
+      "step": 3965
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.279835390946502e-05,
+      "loss": 0.6817,
+      "step": 3970
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.259259259259259e-05,
+      "loss": 0.6761,
+      "step": 3975
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2386831275720165e-05,
+      "loss": 0.6638,
+      "step": 3980
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2181069958847744e-05,
+      "loss": 0.6707,
+      "step": 3985
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.197530864197531e-05,
+      "loss": 0.6796,
+      "step": 3990
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.176954732510288e-05,
+      "loss": 0.6903,
+      "step": 3995
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1563786008230455e-05,
+      "loss": 0.7081,
+      "step": 4000
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.135802469135803e-05,
+      "loss": 0.6528,
+      "step": 4005
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.11522633744856e-05,
+      "loss": 0.6613,
+      "step": 4010
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.094650205761317e-05,
+      "loss": 0.6694,
+      "step": 4015
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.074074074074074e-05,
+      "loss": 0.6982,
+      "step": 4020
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.053497942386831e-05,
+      "loss": 0.6722,
+      "step": 4025
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.032921810699588e-05,
+      "loss": 0.6505,
+      "step": 4030
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.012345679012346e-05,
+      "loss": 0.6747,
+      "step": 4035
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.9917695473251035e-05,
+      "loss": 0.6459,
+      "step": 4040
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.971193415637861e-05,
+      "loss": 0.6859,
+      "step": 4045
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.950617283950617e-05,
+      "loss": 0.6774,
+      "step": 4050
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9300411522633746e-05,
+      "loss": 0.678,
+      "step": 4055
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.909465020576132e-05,
+      "loss": 0.678,
+      "step": 4060
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.888888888888889e-05,
+      "loss": 0.7204,
+      "step": 4065
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.868312757201646e-05,
+      "loss": 0.68,
+      "step": 4070
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8477366255144036e-05,
+      "loss": 0.6797,
+      "step": 4075
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.82716049382716e-05,
+      "loss": 0.6884,
+      "step": 4080
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.806584362139918e-05,
+      "loss": 0.6749,
+      "step": 4085
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.786008230452675e-05,
+      "loss": 0.7162,
+      "step": 4090
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7654320987654326e-05,
+      "loss": 0.6702,
+      "step": 4095
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.74485596707819e-05,
+      "loss": 0.7014,
+      "step": 4100
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.724279835390947e-05,
+      "loss": 0.7031,
+      "step": 4105
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7037037037037037e-05,
+      "loss": 0.6613,
+      "step": 4110
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.683127572016461e-05,
+      "loss": 0.652,
+      "step": 4115
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.662551440329218e-05,
+      "loss": 0.6591,
+      "step": 4120
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6419753086419754e-05,
+      "loss": 0.6688,
+      "step": 4125
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6213991769547327e-05,
+      "loss": 0.6843,
+      "step": 4130
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.60082304526749e-05,
+      "loss": 0.6706,
+      "step": 4135
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.580246913580247e-05,
+      "loss": 0.6974,
+      "step": 4140
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5596707818930044e-05,
+      "loss": 0.6739,
+      "step": 4145
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.539094650205762e-05,
+      "loss": 0.6787,
+      "step": 4150
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.518518518518519e-05,
+      "loss": 0.695,
+      "step": 4155
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.497942386831276e-05,
+      "loss": 0.7012,
+      "step": 4160
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4773662551440334e-05,
+      "loss": 0.7137,
+      "step": 4165
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.45679012345679e-05,
+      "loss": 0.6919,
+      "step": 4170
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.436213991769547e-05,
+      "loss": 0.7045,
+      "step": 4175
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4156378600823045e-05,
+      "loss": 0.6638,
+      "step": 4180
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.395061728395062e-05,
+      "loss": 0.6918,
+      "step": 4185
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.374485596707819e-05,
+      "loss": 0.67,
+      "step": 4190
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.353909465020576e-05,
+      "loss": 0.6835,
+      "step": 4195
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 0.6854,
+      "step": 4200
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.312757201646091e-05,
+      "loss": 0.6846,
+      "step": 4205
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.292181069958848e-05,
+      "loss": 0.6707,
+      "step": 4210
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.271604938271605e-05,
+      "loss": 0.7155,
+      "step": 4215
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2510288065843625e-05,
+      "loss": 0.6985,
+      "step": 4220
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.230452674897119e-05,
+      "loss": 0.6842,
+      "step": 4225
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.209876543209876e-05,
+      "loss": 0.6898,
+      "step": 4230
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1893004115226336e-05,
+      "loss": 0.6707,
+      "step": 4235
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.168724279835391e-05,
+      "loss": 0.6979,
+      "step": 4240
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.148148148148148e-05,
+      "loss": 0.6899,
+      "step": 4245
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.127572016460906e-05,
+      "loss": 0.6697,
+      "step": 4250
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1069958847736626e-05,
+      "loss": 0.6819,
+      "step": 4255
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.08641975308642e-05,
+      "loss": 0.6913,
+      "step": 4260
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.065843621399177e-05,
+      "loss": 0.6894,
+      "step": 4265
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0452674897119343e-05,
+      "loss": 0.6801,
+      "step": 4270
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0246913580246916e-05,
+      "loss": 0.6932,
+      "step": 4275
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0041152263374488e-05,
+      "loss": 0.6567,
+      "step": 4280
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9835390946502057e-05,
+      "loss": 0.6416,
+      "step": 4285
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.962962962962963e-05,
+      "loss": 0.6619,
+      "step": 4290
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9423868312757202e-05,
+      "loss": 0.6746,
+      "step": 4295
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9218106995884775e-05,
+      "loss": 0.6725,
+      "step": 4300
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9012345679012347e-05,
+      "loss": 0.695,
+      "step": 4305
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.880658436213992e-05,
+      "loss": 0.68,
+      "step": 4310
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.860082304526749e-05,
+      "loss": 0.6669,
+      "step": 4315
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.839506172839506e-05,
+      "loss": 0.6626,
+      "step": 4320
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8189300411522634e-05,
+      "loss": 0.6871,
+      "step": 4325
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7983539094650207e-05,
+      "loss": 0.6719,
+      "step": 4330
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.777777777777778e-05,
+      "loss": 0.6603,
+      "step": 4335
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.757201646090535e-05,
+      "loss": 0.6378,
+      "step": 4340
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.736625514403292e-05,
+      "loss": 0.6759,
+      "step": 4345
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7160493827160493e-05,
+      "loss": 0.6704,
+      "step": 4350
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6954732510288066e-05,
+      "loss": 0.6725,
+      "step": 4355
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6748971193415638e-05,
+      "loss": 0.6659,
+      "step": 4360
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.654320987654321e-05,
+      "loss": 0.664,
+      "step": 4365
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6337448559670787e-05,
+      "loss": 0.7316,
+      "step": 4370
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6131687242798352e-05,
+      "loss": 0.6621,
+      "step": 4375
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5925925925925925e-05,
+      "loss": 0.6822,
+      "step": 4380
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5720164609053497e-05,
+      "loss": 0.6694,
+      "step": 4385
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.551440329218107e-05,
+      "loss": 0.6719,
+      "step": 4390
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5308641975308646e-05,
+      "loss": 0.6687,
+      "step": 4395
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.510288065843622e-05,
+      "loss": 0.6878,
+      "step": 4400
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4897119341563787e-05,
+      "loss": 0.6956,
+      "step": 4405
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4691358024691357e-05,
+      "loss": 0.6976,
+      "step": 4410
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4485596707818932e-05,
+      "loss": 0.6969,
+      "step": 4415
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4279835390946505e-05,
+      "loss": 0.685,
+      "step": 4420
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4074074074074074e-05,
+      "loss": 0.6835,
+      "step": 4425
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3868312757201647e-05,
+      "loss": 0.7002,
+      "step": 4430
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.366255144032922e-05,
+      "loss": 0.6657,
+      "step": 4435
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.345679012345679e-05,
+      "loss": 0.6524,
+      "step": 4440
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3251028806584364e-05,
+      "loss": 0.6672,
+      "step": 4445
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3045267489711937e-05,
+      "loss": 0.7033,
+      "step": 4450
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2839506172839506e-05,
+      "loss": 0.6922,
+      "step": 4455
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2633744855967078e-05,
+      "loss": 0.6567,
+      "step": 4460
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.242798353909465e-05,
+      "loss": 0.6764,
+      "step": 4465
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2222222222222223e-05,
+      "loss": 0.7066,
+      "step": 4470
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2016460905349796e-05,
+      "loss": 0.7036,
+      "step": 4475
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1810699588477368e-05,
+      "loss": 0.6716,
+      "step": 4480
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1604938271604937e-05,
+      "loss": 0.6954,
+      "step": 4485
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.139917695473251e-05,
+      "loss": 0.6752,
+      "step": 4490
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1193415637860082e-05,
+      "loss": 0.6845,
+      "step": 4495
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0987654320987655e-05,
+      "loss": 0.6928,
+      "step": 4500
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0781893004115227e-05,
+      "loss": 0.6722,
+      "step": 4505
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.05761316872428e-05,
+      "loss": 0.7046,
+      "step": 4510
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.037037037037037e-05,
+      "loss": 0.6874,
+      "step": 4515
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.016460905349794e-05,
+      "loss": 0.6317,
+      "step": 4520
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9958847736625517e-05,
+      "loss": 0.6956,
+      "step": 4525
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9753086419753087e-05,
+      "loss": 0.6954,
+      "step": 4530
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.954732510288066e-05,
+      "loss": 0.6806,
+      "step": 4535
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.934156378600823e-05,
+      "loss": 0.6831,
+      "step": 4540
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.91358024691358e-05,
+      "loss": 0.6942,
+      "step": 4545
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8930041152263377e-05,
+      "loss": 0.6806,
+      "step": 4550
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.872427983539095e-05,
+      "loss": 0.6711,
+      "step": 4555
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8518518518518518e-05,
+      "loss": 0.7262,
+      "step": 4560
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.831275720164609e-05,
+      "loss": 0.6546,
+      "step": 4565
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8106995884773663e-05,
+      "loss": 0.6538,
+      "step": 4570
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7901234567901236e-05,
+      "loss": 0.7062,
+      "step": 4575
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.769547325102881e-05,
+      "loss": 0.6776,
+      "step": 4580
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.748971193415638e-05,
+      "loss": 0.663,
+      "step": 4585
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.728395061728395e-05,
+      "loss": 0.6985,
+      "step": 4590
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7078189300411522e-05,
+      "loss": 0.6924,
+      "step": 4595
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6872427983539095e-05,
+      "loss": 0.7159,
+      "step": 4600
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 0.6842,
+      "step": 4605
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.646090534979424e-05,
+      "loss": 0.6683,
+      "step": 4610
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6255144032921812e-05,
+      "loss": 0.6516,
+      "step": 4615
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.604938271604938e-05,
+      "loss": 0.6851,
+      "step": 4620
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5843621399176954e-05,
+      "loss": 0.6949,
+      "step": 4625
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.563786008230453e-05,
+      "loss": 0.6618,
+      "step": 4630
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.54320987654321e-05,
+      "loss": 0.6703,
+      "step": 4635
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5226337448559672e-05,
+      "loss": 0.673,
+      "step": 4640
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5020576131687244e-05,
+      "loss": 0.6818,
+      "step": 4645
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4814814814814815e-05,
+      "loss": 0.6987,
+      "step": 4650
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4609053497942387e-05,
+      "loss": 0.6996,
+      "step": 4655
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.440329218106996e-05,
+      "loss": 0.6698,
+      "step": 4660
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.419753086419753e-05,
+      "loss": 0.6889,
+      "step": 4665
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.3991769547325103e-05,
+      "loss": 0.69,
+      "step": 4670
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3786008230452676e-05,
+      "loss": 0.657,
+      "step": 4675
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3580246913580247e-05,
+      "loss": 0.6906,
+      "step": 4680
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3374485596707819e-05,
+      "loss": 0.6922,
+      "step": 4685
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3168724279835393e-05,
+      "loss": 0.671,
+      "step": 4690
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2962962962962962e-05,
+      "loss": 0.7035,
+      "step": 4695
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2757201646090535e-05,
+      "loss": 0.6989,
+      "step": 4700
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.255144032921811e-05,
+      "loss": 0.7081,
+      "step": 4705
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2345679012345678e-05,
+      "loss": 0.6791,
+      "step": 4710
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2139917695473252e-05,
+      "loss": 0.6594,
+      "step": 4715
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1934156378600823e-05,
+      "loss": 0.6868,
+      "step": 4720
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1728395061728396e-05,
+      "loss": 0.6884,
+      "step": 4725
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1522633744855968e-05,
+      "loss": 0.7087,
+      "step": 4730
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1316872427983539e-05,
+      "loss": 0.6753,
+      "step": 4735
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1111111111111112e-05,
+      "loss": 0.6763,
+      "step": 4740
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0905349794238684e-05,
+      "loss": 0.7185,
+      "step": 4745
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0699588477366255e-05,
+      "loss": 0.7175,
+      "step": 4750
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0493827160493827e-05,
+      "loss": 0.6795,
+      "step": 4755
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.02880658436214e-05,
+      "loss": 0.697,
+      "step": 4760
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.008230452674897e-05,
+      "loss": 0.6923,
+      "step": 4765
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.876543209876543e-06,
+      "loss": 0.6987,
+      "step": 4770
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.670781893004116e-06,
+      "loss": 0.6644,
+      "step": 4775
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.465020576131688e-06,
+      "loss": 0.6703,
+      "step": 4780
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.259259259259259e-06,
+      "loss": 0.6659,
+      "step": 4785
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.053497942386832e-06,
+      "loss": 0.6619,
+      "step": 4790
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.847736625514404e-06,
+      "loss": 0.694,
+      "step": 4795
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.641975308641975e-06,
+      "loss": 0.6688,
+      "step": 4800
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.436213991769547e-06,
+      "loss": 0.7193,
+      "step": 4805
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.23045267489712e-06,
+      "loss": 0.6711,
+      "step": 4810
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.02469135802469e-06,
+      "loss": 0.6951,
+      "step": 4815
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.818930041152265e-06,
+      "loss": 0.6955,
+      "step": 4820
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.613168724279836e-06,
+      "loss": 0.6694,
+      "step": 4825
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.4074074074074075e-06,
+      "loss": 0.6918,
+      "step": 4830
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.20164609053498e-06,
+      "loss": 0.6723,
+      "step": 4835
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.995884773662552e-06,
+      "loss": 0.6961,
+      "step": 4840
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.790123456790123e-06,
+      "loss": 0.6864,
+      "step": 4845
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 6.584362139917697e-06,
+      "loss": 0.6814,
+      "step": 4850
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 6.3786008230452675e-06,
+      "loss": 0.6628,
+      "step": 4855
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 6.172839506172839e-06,
+      "loss": 0.6854,
+      "step": 4860
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.967078189300412e-06,
+      "loss": 0.6722,
+      "step": 4865
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.761316872427984e-06,
+      "loss": 0.6889,
+      "step": 4870
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.555555555555556e-06,
+      "loss": 0.6729,
+      "step": 4875
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.3497942386831275e-06,
+      "loss": 0.684,
+      "step": 4880
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.1440329218107e-06,
+      "loss": 0.6878,
+      "step": 4885
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.938271604938272e-06,
+      "loss": 0.7092,
+      "step": 4890
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.732510288065844e-06,
+      "loss": 0.6773,
+      "step": 4895
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.526748971193416e-06,
+      "loss": 0.7095,
+      "step": 4900
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.3209876543209875e-06,
+      "loss": 0.6731,
+      "step": 4905
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.11522633744856e-06,
+      "loss": 0.682,
+      "step": 4910
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.9094650205761325e-06,
+      "loss": 0.6868,
+      "step": 4915
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.7037037037037037e-06,
+      "loss": 0.6764,
+      "step": 4920
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.497942386831276e-06,
+      "loss": 0.6804,
+      "step": 4925
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.2921810699588483e-06,
+      "loss": 0.6904,
+      "step": 4930
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.0864197530864196e-06,
+      "loss": 0.6742,
+      "step": 4935
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.880658436213992e-06,
+      "loss": 0.674,
+      "step": 4940
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.6748971193415637e-06,
+      "loss": 0.7115,
+      "step": 4945
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.469135802469136e-06,
+      "loss": 0.6989,
+      "step": 4950
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.263374485596708e-06,
+      "loss": 0.6408,
+      "step": 4955
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.05761316872428e-06,
+      "loss": 0.6561,
+      "step": 4960
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.8518518518518519e-06,
+      "loss": 0.6743,
+      "step": 4965
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.6460905349794242e-06,
+      "loss": 0.6878,
+      "step": 4970
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.440329218106996e-06,
+      "loss": 0.7012,
+      "step": 4975
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.234567901234568e-06,
+      "loss": 0.6747,
+      "step": 4980
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.02880658436214e-06,
+      "loss": 0.712,
+      "step": 4985
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.230452674897121e-07,
+      "loss": 0.6959,
+      "step": 4990
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.17283950617284e-07,
+      "loss": 0.6961,
+      "step": 4995
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.1152263374485604e-07,
+      "loss": 0.7222,
+      "step": 5000
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 5010,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 500,
+  "total_flos": 2.0132950426699956e+19,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/training_args.bin b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..e707059a03fbe62142ac41332e145ac1ac1f36cc
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b99a1729574e0e46ee782b42dc90101a785d5b4792871920d293f44722c28cc
+size 5944
diff --git a/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/zero_to_fp32.py b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/zero_to_fp32.py
new file mode 100644
index 0000000000000000000000000000000000000000..c98caae31534368be22b67fc4ae906836c992a8d
--- /dev/null
+++ b/save_lora_instruct_r8_2e4_s11_v2/checkpoint-5000/zero_to_fp32.py
@@ -0,0 +1,587 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+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):
+    """
+    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)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+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 _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()
+            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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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 param_shapes.items():
+
+        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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    """
+    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``
+
+    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)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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("-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_file, tag=args.tag)