Upload folder using huggingface_hub

44_128_e5_3e-5/checkpoint-1458/README.md

@@ -0,0 +1,202 @@
+---
+base_model: ibm-granite/granite-3.3-8b-base
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.15.2

44_128_e5_3e-5/checkpoint-1458/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "down_proj",
+    "q_proj",
+    "gate_proj",
+    "v_proj",
+    "o_proj",
+    "k_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

44_128_e5_3e-5/checkpoint-1458/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac571f16406fb11ed91a9f0da0582975de3e137e4e3f47f404c345b5ca50ecff
+size 791751704

44_128_e5_3e-5/checkpoint-1458/latest

@@ -0,0 +1 @@
+global_step1458

44_128_e5_3e-5/checkpoint-1458/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94756fb4aec8e9dd440e044eeadf7da6d4f17c3cf51555c905686dc11d173246
+size 15920

44_128_e5_3e-5/checkpoint-1458/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:091ddc3632e8dec995cd28a5c5a355d494e24e412a2f927bea4034dc96e18793
+size 15920

44_128_e5_3e-5/checkpoint-1458/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c94c13ab7c302adb07849755b777b29e7381ca2b2112df0396ab078c237cd47
+size 15920

44_128_e5_3e-5/checkpoint-1458/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d4a797a40c818fc2585b48d9c45743da270dd30d60fb254352b03ed0558351d
+size 15920

44_128_e5_3e-5/checkpoint-1458/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28346b3764c46b60c1df2da9d8a053dd9743aeaf99f2df92c00937a3aaf083d5
+size 15920

44_128_e5_3e-5/checkpoint-1458/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ecaf12b56239a77698b3f199aacfa49e64fc2ef65512383d514951c208c010a
+size 15920

44_128_e5_3e-5/checkpoint-1458/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:415e8d3b9e9bf6c6d6e564e4e17cf92b53847f8c21e8d66612faeaf64f455717
+size 15920

44_128_e5_3e-5/checkpoint-1458/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23bc841b0f83aae6cec37bb1a91926c5a38cfa7d7c72ac6b5898d1a84ef21d7d
+size 15920

44_128_e5_3e-5/checkpoint-1458/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7c5d2b502e98153ed4dff42d0cc9ce3b7ba1695c30bdc0a2864426c7a21a707
+size 1064

44_128_e5_3e-5/checkpoint-1458/special_tokens_map.json

@@ -0,0 +1,45 @@
+{
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<reponame>",
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

44_128_e5_3e-5/checkpoint-1458/tokenizer_config.json

@@ -0,0 +1,188 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<fim_prefix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<fim_middle>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<fim_suffix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<fim_pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<filename>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<gh_stars>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<issue_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "8": {
+      "content": "<issue_comment>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "9": {
+      "content": "<issue_closed>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "10": {
+      "content": "<jupyter_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "11": {
+      "content": "<jupyter_text>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "12": {
+      "content": "<jupyter_code>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "13": {
+      "content": "<jupyter_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "14": {
+      "content": "<empty_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "15": {
+      "content": "<commit_before>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "16": {
+      "content": "<commit_msg>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "17": {
+      "content": "<commit_after>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "18": {
+      "content": "<reponame>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<reponame>",
+  "padding_side": "left",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

44_128_e5_3e-5/checkpoint-1458/trainer_state.json

@@ -0,0 +1,2071 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1458,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.010298661174047374,
+      "grad_norm": 1.141814947128296,
+      "learning_rate": 9.836065573770493e-07,
+      "loss": 1.3338,
+      "step": 5
+    },
+    {
+      "epoch": 0.02059732234809475,
+      "grad_norm": 0.9742817878723145,
+      "learning_rate": 2.213114754098361e-06,
+      "loss": 1.3646,
+      "step": 10
+    },
+    {
+      "epoch": 0.030895983522142123,
+      "grad_norm": 0.7402164340019226,
+      "learning_rate": 3.4426229508196724e-06,
+      "loss": 1.3261,
+      "step": 15
+    },
+    {
+      "epoch": 0.0411946446961895,
+      "grad_norm": 0.5555693507194519,
+      "learning_rate": 4.672131147540984e-06,
+      "loss": 1.3035,
+      "step": 20
+    },
+    {
+      "epoch": 0.05149330587023687,
+      "grad_norm": 0.6190383434295654,
+      "learning_rate": 5.901639344262295e-06,
+      "loss": 1.3847,
+      "step": 25
+    },
+    {
+      "epoch": 0.061791967044284246,
+      "grad_norm": 0.5282045006752014,
+      "learning_rate": 7.131147540983606e-06,
+      "loss": 1.3197,
+      "step": 30
+    },
+    {
+      "epoch": 0.07209062821833162,
+      "grad_norm": 0.48918476700782776,
+      "learning_rate": 8.360655737704917e-06,
+      "loss": 1.2899,
+      "step": 35
+    },
+    {
+      "epoch": 0.082389289392379,
+      "grad_norm": 0.7961065769195557,
+      "learning_rate": 9.590163934426231e-06,
+      "loss": 1.2656,
+      "step": 40
+    },
+    {
+      "epoch": 0.09268795056642637,
+      "grad_norm": 0.4608859419822693,
+      "learning_rate": 1.0819672131147542e-05,
+      "loss": 1.26,
+      "step": 45
+    },
+    {
+      "epoch": 0.10298661174047374,
+      "grad_norm": 0.555935800075531,
+      "learning_rate": 1.2049180327868853e-05,
+      "loss": 1.2548,
+      "step": 50
+    },
+    {
+      "epoch": 0.11328527291452112,
+      "grad_norm": 0.4993849992752075,
+      "learning_rate": 1.3278688524590163e-05,
+      "loss": 1.2389,
+      "step": 55
+    },
+    {
+      "epoch": 0.12358393408856849,
+      "grad_norm": 0.5452463030815125,
+      "learning_rate": 1.4508196721311476e-05,
+      "loss": 1.2124,
+      "step": 60
+    },
+    {
+      "epoch": 0.13388259526261587,
+      "grad_norm": 0.4450685381889343,
+      "learning_rate": 1.5737704918032788e-05,
+      "loss": 1.2105,
+      "step": 65
+    },
+    {
+      "epoch": 0.14418125643666324,
+      "grad_norm": 0.5045918822288513,
+      "learning_rate": 1.6967213114754097e-05,
+      "loss": 1.2305,
+      "step": 70
+    },
+    {
+      "epoch": 0.15447991761071062,
+      "grad_norm": 0.49198681116104126,
+      "learning_rate": 1.819672131147541e-05,
+      "loss": 1.2464,
+      "step": 75
+    },
+    {
+      "epoch": 0.164778578784758,
+      "grad_norm": 0.4819914996623993,
+      "learning_rate": 1.9426229508196722e-05,
+      "loss": 1.194,
+      "step": 80
+    },
+    {
+      "epoch": 0.17507723995880536,
+      "grad_norm": 0.4840977191925049,
+      "learning_rate": 2.0655737704918034e-05,
+      "loss": 1.1949,
+      "step": 85
+    },
+    {
+      "epoch": 0.18537590113285274,
+      "grad_norm": 0.55226069688797,
+      "learning_rate": 2.1885245901639347e-05,
+      "loss": 1.1936,
+      "step": 90
+    },
+    {
+      "epoch": 0.1956745623069001,
+      "grad_norm": 0.47093257308006287,
+      "learning_rate": 2.3114754098360656e-05,
+      "loss": 1.1214,
+      "step": 95
+    },
+    {
+      "epoch": 0.2059732234809475,
+      "grad_norm": 0.5413655042648315,
+      "learning_rate": 2.434426229508197e-05,
+      "loss": 1.1661,
+      "step": 100
+    },
+    {
+      "epoch": 0.21627188465499486,
+      "grad_norm": 0.525501012802124,
+      "learning_rate": 2.5573770491803277e-05,
+      "loss": 1.1657,
+      "step": 105
+    },
+    {
+      "epoch": 0.22657054582904224,
+      "grad_norm": 0.5898400545120239,
+      "learning_rate": 2.680327868852459e-05,
+      "loss": 1.1567,
+      "step": 110
+    },
+    {
+      "epoch": 0.2368692070030896,
+      "grad_norm": 0.5453243851661682,
+      "learning_rate": 2.8032786885245902e-05,
+      "loss": 1.1429,
+      "step": 115
+    },
+    {
+      "epoch": 0.24716786817713698,
+      "grad_norm": 0.5925690531730652,
+      "learning_rate": 2.9262295081967215e-05,
+      "loss": 1.165,
+      "step": 120
+    },
+    {
+      "epoch": 0.25746652935118436,
+      "grad_norm": 0.5159111618995667,
+      "learning_rate": 2.9999944416086244e-05,
+      "loss": 1.0712,
+      "step": 125
+    },
+    {
+      "epoch": 0.26776519052523173,
+      "grad_norm": 0.5997148752212524,
+      "learning_rate": 2.999931910178738e-05,
+      "loss": 1.087,
+      "step": 130
+    },
+    {
+      "epoch": 0.2780638516992791,
+      "grad_norm": 0.5427290201187134,
+      "learning_rate": 2.9997999022358424e-05,
+      "loss": 1.1449,
+      "step": 135
+    },
+    {
+      "epoch": 0.2883625128733265,
+      "grad_norm": 0.7504834532737732,
+      "learning_rate": 2.999598423894516e-05,
+      "loss": 1.1485,
+      "step": 140
+    },
+    {
+      "epoch": 0.29866117404737386,
+      "grad_norm": 0.5923535823822021,
+      "learning_rate": 2.9993274844871917e-05,
+      "loss": 1.1358,
+      "step": 145
+    },
+    {
+      "epoch": 0.30895983522142123,
+      "grad_norm": 0.6039471626281738,
+      "learning_rate": 2.9989870965637215e-05,
+      "loss": 1.1134,
+      "step": 150
+    },
+    {
+      "epoch": 0.3192584963954686,
+      "grad_norm": 0.5884091854095459,
+      "learning_rate": 2.9985772758907992e-05,
+      "loss": 1.101,
+      "step": 155
+    },
+    {
+      "epoch": 0.329557157569516,
+      "grad_norm": 0.6560966968536377,
+      "learning_rate": 2.998098041451227e-05,
+      "loss": 1.126,
+      "step": 160
+    },
+    {
+      "epoch": 0.33985581874356335,
+      "grad_norm": 0.557138979434967,
+      "learning_rate": 2.9975494154430378e-05,
+      "loss": 1.0722,
+      "step": 165
+    },
+    {
+      "epoch": 0.35015447991761073,
+      "grad_norm": 0.7041076421737671,
+      "learning_rate": 2.996931423278467e-05,
+      "loss": 1.0913,
+      "step": 170
+    },
+    {
+      "epoch": 0.3604531410916581,
+      "grad_norm": 0.6222100257873535,
+      "learning_rate": 2.9962440935827735e-05,
+      "loss": 1.0392,
+      "step": 175
+    },
+    {
+      "epoch": 0.3707518022657055,
+      "grad_norm": 0.6287522912025452,
+      "learning_rate": 2.995487458192917e-05,
+      "loss": 1.003,
+      "step": 180
+    },
+    {
+      "epoch": 0.38105046343975285,
+      "grad_norm": 0.8934054970741272,
+      "learning_rate": 2.9946615521560805e-05,
+      "loss": 1.0176,
+      "step": 185
+    },
+    {
+      "epoch": 0.3913491246138002,
+      "grad_norm": 0.6240345239639282,
+      "learning_rate": 2.9937664137280478e-05,
+      "loss": 1.0205,
+      "step": 190
+    },
+    {
+      "epoch": 0.4016477857878476,
+      "grad_norm": 0.6924830675125122,
+      "learning_rate": 2.9928020843714323e-05,
+      "loss": 1.0218,
+      "step": 195
+    },
+    {
+      "epoch": 0.411946446961895,
+      "grad_norm": 0.7187322378158569,
+      "learning_rate": 2.9917686087537563e-05,
+      "loss": 1.0581,
+      "step": 200
+    },
+    {
+      "epoch": 0.42224510813594235,
+      "grad_norm": 0.6714951992034912,
+      "learning_rate": 2.9906660347453803e-05,
+      "loss": 0.9597,
+      "step": 205
+    },
+    {
+      "epoch": 0.4325437693099897,
+      "grad_norm": 0.758985698223114,
+      "learning_rate": 2.9894944134172876e-05,
+      "loss": 0.9869,
+      "step": 210
+    },
+    {
+      "epoch": 0.4428424304840371,
+      "grad_norm": 0.6603527069091797,
+      "learning_rate": 2.988253799038718e-05,
+      "loss": 0.9687,
+      "step": 215
+    },
+    {
+      "epoch": 0.45314109165808447,
+      "grad_norm": 0.7007045745849609,
+      "learning_rate": 2.986944249074654e-05,
+      "loss": 1.0301,
+      "step": 220
+    },
+    {
+      "epoch": 0.46343975283213185,
+      "grad_norm": 0.7556740641593933,
+      "learning_rate": 2.985565824183159e-05,
+      "loss": 1.0108,
+      "step": 225
+    },
+    {
+      "epoch": 0.4737384140061792,
+      "grad_norm": 0.7658377885818481,
+      "learning_rate": 2.9841185882125682e-05,
+      "loss": 0.9966,
+      "step": 230
+    },
+    {
+      "epoch": 0.4840370751802266,
+      "grad_norm": 0.8318830728530884,
+      "learning_rate": 2.9826026081985305e-05,
+      "loss": 0.9416,
+      "step": 235
+    },
+    {
+      "epoch": 0.49433573635427397,
+      "grad_norm": 0.7977103590965271,
+      "learning_rate": 2.9810179543609032e-05,
+      "loss": 0.9753,
+      "step": 240
+    },
+    {
+      "epoch": 0.5046343975283213,
+      "grad_norm": 0.7902314066886902,
+      "learning_rate": 2.9793647001005002e-05,
+      "loss": 0.9422,
+      "step": 245
+    },
+    {
+      "epoch": 0.5149330587023687,
+      "grad_norm": 0.8238946795463562,
+      "learning_rate": 2.9776429219956917e-05,
+      "loss": 0.9609,
+      "step": 250
+    },
+    {
+      "epoch": 0.525231719876416,
+      "grad_norm": 0.8614672422409058,
+      "learning_rate": 2.975852699798857e-05,
+      "loss": 0.9375,
+      "step": 255
+    },
+    {
+      "epoch": 0.5355303810504635,
+      "grad_norm": 0.946057915687561,
+      "learning_rate": 2.9739941164326914e-05,
+      "loss": 0.8812,
+      "step": 260
+    },
+    {
+      "epoch": 0.5458290422245108,
+      "grad_norm": 0.8593087196350098,
+      "learning_rate": 2.9720672579863633e-05,
+      "loss": 0.9091,
+      "step": 265
+    },
+    {
+      "epoch": 0.5561277033985582,
+      "grad_norm": 0.8131739497184753,
+      "learning_rate": 2.970072213711528e-05,
+      "loss": 0.9136,
+      "step": 270
+    },
+    {
+      "epoch": 0.5664263645726055,
+      "grad_norm": 0.8276655077934265,
+      "learning_rate": 2.9680090760181933e-05,
+      "loss": 0.9188,
+      "step": 275
+    },
+    {
+      "epoch": 0.576725025746653,
+      "grad_norm": 0.8448908925056458,
+      "learning_rate": 2.9658779404704387e-05,
+      "loss": 0.8959,
+      "step": 280
+    },
+    {
+      "epoch": 0.5870236869207003,
+      "grad_norm": 0.7980640530586243,
+      "learning_rate": 2.9636789057819903e-05,
+      "loss": 0.9117,
+      "step": 285
+    },
+    {
+      "epoch": 0.5973223480947477,
+      "grad_norm": 0.8234493136405945,
+      "learning_rate": 2.961412073811646e-05,
+      "loss": 0.904,
+      "step": 290
+    },
+    {
+      "epoch": 0.607621009268795,
+      "grad_norm": 1.027661919593811,
+      "learning_rate": 2.9590775495585597e-05,
+      "loss": 0.8299,
+      "step": 295
+    },
+    {
+      "epoch": 0.6179196704428425,
+      "grad_norm": 0.8117839097976685,
+      "learning_rate": 2.956675441157376e-05,
+      "loss": 0.8743,
+      "step": 300
+    },
+    {
+      "epoch": 0.6282183316168898,
+      "grad_norm": 0.8687512278556824,
+      "learning_rate": 2.954205859873223e-05,
+      "loss": 0.8539,
+      "step": 305
+    },
+    {
+      "epoch": 0.6385169927909372,
+      "grad_norm": 0.9031465649604797,
+      "learning_rate": 2.951668920096557e-05,
+      "loss": 0.8416,
+      "step": 310
+    },
+    {
+      "epoch": 0.6488156539649845,
+      "grad_norm": 1.0350013971328735,
+      "learning_rate": 2.9490647393378656e-05,
+      "loss": 0.8505,
+      "step": 315
+    },
+    {
+      "epoch": 0.659114315139032,
+      "grad_norm": 0.9072204828262329,
+      "learning_rate": 2.9463934382222226e-05,
+      "loss": 0.8328,
+      "step": 320
+    },
+    {
+      "epoch": 0.6694129763130793,
+      "grad_norm": 0.8805303573608398,
+      "learning_rate": 2.943655140483703e-05,
+      "loss": 0.8498,
+      "step": 325
+    },
+    {
+      "epoch": 0.6797116374871267,
+      "grad_norm": 0.9569805264472961,
+      "learning_rate": 2.94084997295965e-05,
+      "loss": 0.8613,
+      "step": 330
+    },
+    {
+      "epoch": 0.690010298661174,
+      "grad_norm": 0.842765748500824,
+      "learning_rate": 2.9379780655848e-05,
+      "loss": 0.9271,
+      "step": 335
+    },
+    {
+      "epoch": 0.7003089598352215,
+      "grad_norm": 1.0585700273513794,
+      "learning_rate": 2.9350395513852655e-05,
+      "loss": 0.8049,
+      "step": 340
+    },
+    {
+      "epoch": 0.7106076210092688,
+      "grad_norm": 0.9681879878044128,
+      "learning_rate": 2.9320345664723713e-05,
+      "loss": 0.8611,
+      "step": 345
+    },
+    {
+      "epoch": 0.7209062821833162,
+      "grad_norm": 0.9282833933830261,
+      "learning_rate": 2.928963250036351e-05,
+      "loss": 0.8645,
+      "step": 350
+    },
+    {
+      "epoch": 0.7312049433573635,
+      "grad_norm": 0.880826473236084,
+      "learning_rate": 2.9258257443399007e-05,
+      "loss": 0.8409,
+      "step": 355
+    },
+    {
+      "epoch": 0.741503604531411,
+      "grad_norm": 0.9096245169639587,
+      "learning_rate": 2.922622194711587e-05,
+      "loss": 0.8648,
+      "step": 360
+    },
+    {
+      "epoch": 0.7518022657054583,
+      "grad_norm": 0.9456668496131897,
+      "learning_rate": 2.919352749539117e-05,
+      "loss": 0.8045,
+      "step": 365
+    },
+    {
+      "epoch": 0.7621009268795057,
+      "grad_norm": 0.9902530312538147,
+      "learning_rate": 2.916017560262466e-05,
+      "loss": 0.84,
+      "step": 370
+    },
+    {
+      "epoch": 0.772399588053553,
+      "grad_norm": 1.0271276235580444,
+      "learning_rate": 2.91261678136686e-05,
+      "loss": 0.8052,
+      "step": 375
+    },
+    {
+      "epoch": 0.7826982492276005,
+      "grad_norm": 0.9190011024475098,
+      "learning_rate": 2.9091505703756224e-05,
+      "loss": 0.7965,
+      "step": 380
+    },
+    {
+      "epoch": 0.7929969104016478,
+      "grad_norm": 1.022260308265686,
+      "learning_rate": 2.9056190878428766e-05,
+      "loss": 0.8417,
+      "step": 385
+    },
+    {
+      "epoch": 0.8032955715756952,
+      "grad_norm": 0.9204807281494141,
+      "learning_rate": 2.9020224973461098e-05,
+      "loss": 0.7853,
+      "step": 390
+    },
+    {
+      "epoch": 0.8135942327497425,
+      "grad_norm": 0.9588913917541504,
+      "learning_rate": 2.8983609654785948e-05,
+      "loss": 0.8235,
+      "step": 395
+    },
+    {
+      "epoch": 0.82389289392379,
+      "grad_norm": 0.8754395246505737,
+      "learning_rate": 2.8946346618416742e-05,
+      "loss": 0.7409,
+      "step": 400
+    },
+    {
+      "epoch": 0.8341915550978373,
+      "grad_norm": 1.0218687057495117,
+      "learning_rate": 2.8908437590369056e-05,
+      "loss": 0.7466,
+      "step": 405
+    },
+    {
+      "epoch": 0.8444902162718847,
+      "grad_norm": 1.1078494787216187,
+      "learning_rate": 2.886988432658065e-05,
+      "loss": 0.7767,
+      "step": 410
+    },
+    {
+      "epoch": 0.854788877445932,
+      "grad_norm": 0.970990002155304,
+      "learning_rate": 2.8830688612830142e-05,
+      "loss": 0.7584,
+      "step": 415
+    },
+    {
+      "epoch": 0.8650875386199794,
+      "grad_norm": 1.0772037506103516,
+      "learning_rate": 2.8790852264654287e-05,
+      "loss": 0.8034,
+      "step": 420
+    },
+    {
+      "epoch": 0.8753861997940268,
+      "grad_norm": 0.9200147986412048,
+      "learning_rate": 2.875037712726389e-05,
+      "loss": 0.7289,
+      "step": 425
+    },
+    {
+      "epoch": 0.8856848609680742,
+      "grad_norm": 1.0537430047988892,
+      "learning_rate": 2.8709265075458324e-05,
+      "loss": 0.7462,
+      "step": 430
+    },
+    {
+      "epoch": 0.8959835221421215,
+      "grad_norm": 0.9600791335105896,
+      "learning_rate": 2.86675180135387e-05,
+      "loss": 0.68,
+      "step": 435
+    },
+    {
+      "epoch": 0.9062821833161689,
+      "grad_norm": 1.05148184299469,
+      "learning_rate": 2.8625137875219655e-05,
+      "loss": 0.7472,
+      "step": 440
+    },
+    {
+      "epoch": 0.9165808444902163,
+      "grad_norm": 0.9907565712928772,
+      "learning_rate": 2.8582126623539787e-05,
+      "loss": 0.6766,
+      "step": 445
+    },
+    {
+      "epoch": 0.9268795056642637,
+      "grad_norm": 0.9938767552375793,
+      "learning_rate": 2.8538486250770724e-05,
+      "loss": 0.7238,
+      "step": 450
+    },
+    {
+      "epoch": 0.937178166838311,
+      "grad_norm": 0.9231424331665039,
+      "learning_rate": 2.849421877832484e-05,
+      "loss": 0.7381,
+      "step": 455
+    },
+    {
+      "epoch": 0.9474768280123584,
+      "grad_norm": 1.0045597553253174,
+      "learning_rate": 2.844932625666163e-05,
+      "loss": 0.6675,
+      "step": 460
+    },
+    {
+      "epoch": 0.9577754891864058,
+      "grad_norm": 0.9766920804977417,
+      "learning_rate": 2.8403810765192727e-05,
+      "loss": 0.7332,
+      "step": 465
+    },
+    {
+      "epoch": 0.9680741503604532,
+      "grad_norm": 0.9820194244384766,
+      "learning_rate": 2.835767441218559e-05,
+      "loss": 0.726,
+      "step": 470
+    },
+    {
+      "epoch": 0.9783728115345005,
+      "grad_norm": 1.1101776361465454,
+      "learning_rate": 2.831091933466584e-05,
+      "loss": 0.7179,
+      "step": 475
+    },
+    {
+      "epoch": 0.9886714727085479,
+      "grad_norm": 1.3044633865356445,
+      "learning_rate": 2.82635476983183e-05,
+      "loss": 0.6778,
+      "step": 480
+    },
+    {
+      "epoch": 0.9989701338825953,
+      "grad_norm": 1.0158640146255493,
+      "learning_rate": 2.821556169738663e-05,
+      "loss": 0.7167,
+      "step": 485
+    },
+    {
+      "epoch": 1.008238928939238,
+      "grad_norm": 1.292767882347107,
+      "learning_rate": 2.8166963554571748e-05,
+      "loss": 0.585,
+      "step": 490
+    },
+    {
+      "epoch": 1.0185375901132854,
+      "grad_norm": 0.9503241777420044,
+      "learning_rate": 2.8117755520928834e-05,
+      "loss": 0.6277,
+      "step": 495
+    },
+    {
+      "epoch": 1.0288362512873326,
+      "grad_norm": 1.0615237951278687,
+      "learning_rate": 2.8067939875763078e-05,
+      "loss": 0.6104,
+      "step": 500
+    },
+    {
+      "epoch": 1.03913491246138,
+      "grad_norm": 1.053903341293335,
+      "learning_rate": 2.8017518926524103e-05,
+      "loss": 0.5969,
+      "step": 505
+    },
+    {
+      "epoch": 1.0494335736354274,
+      "grad_norm": 1.1858323812484741,
+      "learning_rate": 2.7966495008699083e-05,
+      "loss": 0.592,
+      "step": 510
+    },
+    {
+      "epoch": 1.0597322348094749,
+      "grad_norm": 0.9921702146530151,
+      "learning_rate": 2.7914870485704563e-05,
+      "loss": 0.5961,
+      "step": 515
+    },
+    {
+      "epoch": 1.070030895983522,
+      "grad_norm": 0.9998088479042053,
+      "learning_rate": 2.7862647748776978e-05,
+      "loss": 0.6306,
+      "step": 520
+    },
+    {
+      "epoch": 1.0803295571575695,
+      "grad_norm": 1.146559238433838,
+      "learning_rate": 2.780982921686191e-05,
+      "loss": 0.582,
+      "step": 525
+    },
+    {
+      "epoch": 1.090628218331617,
+      "grad_norm": 1.0381414890289307,
+      "learning_rate": 2.775641733650203e-05,
+      "loss": 0.5503,
+      "step": 530
+    },
+    {
+      "epoch": 1.1009268795056641,
+      "grad_norm": 1.0624871253967285,
+      "learning_rate": 2.770241458172378e-05,
+      "loss": 0.5694,
+      "step": 535
+    },
+    {
+      "epoch": 1.1112255406797116,
+      "grad_norm": 1.0869427919387817,
+      "learning_rate": 2.7647823453922757e-05,
+      "loss": 0.6295,
+      "step": 540
+    },
+    {
+      "epoch": 1.121524201853759,
+      "grad_norm": 1.2071483135223389,
+      "learning_rate": 2.7592646481747888e-05,
+      "loss": 0.5973,
+      "step": 545
+    },
+    {
+      "epoch": 1.1318228630278064,
+      "grad_norm": 1.0342060327529907,
+      "learning_rate": 2.753688622098427e-05,
+      "loss": 0.5955,
+      "step": 550
+    },
+    {
+      "epoch": 1.1421215242018539,
+      "grad_norm": 1.0878804922103882,
+      "learning_rate": 2.7480545254434798e-05,
+      "loss": 0.6152,
+      "step": 555
+    },
+    {
+      "epoch": 1.152420185375901,
+      "grad_norm": 1.1208651065826416,
+      "learning_rate": 2.7423626191800532e-05,
+      "loss": 0.586,
+      "step": 560
+    },
+    {
+      "epoch": 1.1627188465499485,
+      "grad_norm": 1.063509464263916,
+      "learning_rate": 2.7366131669559816e-05,
+      "loss": 0.5879,
+      "step": 565
+    },
+    {
+      "epoch": 1.173017507723996,
+      "grad_norm": 1.2478981018066406,
+      "learning_rate": 2.7308064350846148e-05,
+      "loss": 0.5478,
+      "step": 570
+    },
+    {
+      "epoch": 1.1833161688980431,
+      "grad_norm": 1.1873619556427002,
+      "learning_rate": 2.7249426925324844e-05,
+      "loss": 0.5383,
+      "step": 575
+    },
+    {
+      "epoch": 1.1936148300720906,
+      "grad_norm": 1.0544021129608154,
+      "learning_rate": 2.7190222109068427e-05,
+      "loss": 0.6021,
+      "step": 580
+    },
+    {
+      "epoch": 1.203913491246138,
+      "grad_norm": 1.2434903383255005,
+      "learning_rate": 2.7130452644430844e-05,
+      "loss": 0.5455,
+      "step": 585
+    },
+    {
+      "epoch": 1.2142121524201854,
+      "grad_norm": 1.1316081285476685,
+      "learning_rate": 2.7070121299920422e-05,
+      "loss": 0.5933,
+      "step": 590
+    },
+    {
+      "epoch": 1.2245108135942329,
+      "grad_norm": 0.9929481148719788,
+      "learning_rate": 2.7009230870071645e-05,
+      "loss": 0.5502,
+      "step": 595
+    },
+    {
+      "epoch": 1.23480947476828,
+      "grad_norm": 1.0308139324188232,
+      "learning_rate": 2.6947784175315694e-05,
+      "loss": 0.5177,
+      "step": 600
+    },
+    {
+      "epoch": 1.2451081359423275,
+      "grad_norm": 1.0328620672225952,
+      "learning_rate": 2.688578406184983e-05,
+      "loss": 0.5156,
+      "step": 605
+    },
+    {
+      "epoch": 1.255406797116375,
+      "grad_norm": 1.1804379224777222,
+      "learning_rate": 2.6823233401505548e-05,
+      "loss": 0.5648,
+      "step": 610
+    },
+    {
+      "epoch": 1.2657054582904221,
+      "grad_norm": 1.2127525806427002,
+      "learning_rate": 2.6760135091615535e-05,
+      "loss": 0.5654,
+      "step": 615
+    },
+    {
+      "epoch": 1.2760041194644696,
+      "grad_norm": 1.1704093217849731,
+      "learning_rate": 2.66964920548795e-05,
+      "loss": 0.5534,
+      "step": 620
+    },
+    {
+      "epoch": 1.286302780638517,
+      "grad_norm": 1.062095284461975,
+      "learning_rate": 2.663230723922877e-05,
+      "loss": 0.4969,
+      "step": 625
+    },
+    {
+      "epoch": 1.2966014418125644,
+      "grad_norm": 1.391966462135315,
+      "learning_rate": 2.6567583617689754e-05,
+      "loss": 0.5283,
+      "step": 630
+    },
+    {
+      "epoch": 1.3069001029866119,
+      "grad_norm": 1.1415766477584839,
+      "learning_rate": 2.6502324188246226e-05,
+      "loss": 0.5447,
+      "step": 635
+    },
+    {
+      "epoch": 1.317198764160659,
+      "grad_norm": 1.146954894065857,
+      "learning_rate": 2.643653197370048e-05,
+      "loss": 0.5171,
+      "step": 640
+    },
+    {
+      "epoch": 1.3274974253347065,
+      "grad_norm": 1.064420461654663,
+      "learning_rate": 2.6370210021533266e-05,
+      "loss": 0.5127,
+      "step": 645
+    },
+    {
+      "epoch": 1.337796086508754,
+      "grad_norm": 1.172444224357605,
+      "learning_rate": 2.6303361403762686e-05,
+      "loss": 0.5017,
+      "step": 650
+    },
+    {
+      "epoch": 1.3480947476828011,
+      "grad_norm": 1.0418468713760376,
+      "learning_rate": 2.623598921680188e-05,
+      "loss": 0.4846,
+      "step": 655
+    },
+    {
+      "epoch": 1.3583934088568486,
+      "grad_norm": 1.2469124794006348,
+      "learning_rate": 2.6168096581315574e-05,
+      "loss": 0.4795,
+      "step": 660
+    },
+    {
+      "epoch": 1.368692070030896,
+      "grad_norm": 1.1334832906723022,
+      "learning_rate": 2.609968664207558e-05,
+      "loss": 0.5371,
+      "step": 665
+    },
+    {
+      "epoch": 1.3789907312049434,
+      "grad_norm": 1.1697006225585938,
+      "learning_rate": 2.6030762567815093e-05,
+      "loss": 0.4977,
+      "step": 670
+    },
+    {
+      "epoch": 1.3892893923789909,
+      "grad_norm": 1.094186544418335,
+      "learning_rate": 2.596132755108191e-05,
+      "loss": 0.5173,
+      "step": 675
+    },
+    {
+      "epoch": 1.399588053553038,
+      "grad_norm": 1.1183247566223145,
+      "learning_rate": 2.5891384808090608e-05,
+      "loss": 0.5067,
+      "step": 680
+    },
+    {
+      "epoch": 1.4098867147270855,
+      "grad_norm": 1.0611745119094849,
+      "learning_rate": 2.5820937578573505e-05,
+      "loss": 0.5012,
+      "step": 685
+    },
+    {
+      "epoch": 1.420185375901133,
+      "grad_norm": 1.0735154151916504,
+      "learning_rate": 2.5749989125630625e-05,
+      "loss": 0.4825,
+      "step": 690
+    },
+    {
+      "epoch": 1.4304840370751801,
+      "grad_norm": 1.0683612823486328,
+      "learning_rate": 2.5678542735578553e-05,
+      "loss": 0.4787,
+      "step": 695
+    },
+    {
+      "epoch": 1.4407826982492276,
+      "grad_norm": 1.1233255863189697,
+      "learning_rate": 2.5606601717798212e-05,
+      "loss": 0.4897,
+      "step": 700
+    },
+    {
+      "epoch": 1.451081359423275,
+      "grad_norm": 1.1237461566925049,
+      "learning_rate": 2.5534169404581576e-05,
+      "loss": 0.4728,
+      "step": 705
+    },
+    {
+      "epoch": 1.4613800205973224,
+      "grad_norm": 1.2569210529327393,
+      "learning_rate": 2.5461249150977304e-05,
+      "loss": 0.4589,
+      "step": 710
+    },
+    {
+      "epoch": 1.4716786817713698,
+      "grad_norm": 1.130018711090088,
+      "learning_rate": 2.5387844334635356e-05,
+      "loss": 0.4912,
+      "step": 715
+    },
+    {
+      "epoch": 1.481977342945417,
+      "grad_norm": 1.1941906213760376,
+      "learning_rate": 2.5313958355650517e-05,
+      "loss": 0.5104,
+      "step": 720
+    },
+    {
+      "epoch": 1.4922760041194645,
+      "grad_norm": 1.155961036682129,
+      "learning_rate": 2.523959463640493e-05,
+      "loss": 0.458,
+      "step": 725
+    },
+    {
+      "epoch": 1.502574665293512,
+      "grad_norm": 1.408888339996338,
+      "learning_rate": 2.516475662140956e-05,
+      "loss": 0.4501,
+      "step": 730
+    },
+    {
+      "epoch": 1.5128733264675591,
+      "grad_norm": 1.20494544506073,
+      "learning_rate": 2.508944777714464e-05,
+      "loss": 0.4649,
+      "step": 735
+    },
+    {
+      "epoch": 1.5231719876416066,
+      "grad_norm": 1.1924827098846436,
+      "learning_rate": 2.5013671591899117e-05,
+      "loss": 0.4326,
+      "step": 740
+    },
+    {
+      "epoch": 1.533470648815654,
+      "grad_norm": 1.1093214750289917,
+      "learning_rate": 2.4937431575609063e-05,
+      "loss": 0.5026,
+      "step": 745
+    },
+    {
+      "epoch": 1.5437693099897012,
+      "grad_norm": 1.1080135107040405,
+      "learning_rate": 2.4860731259695096e-05,
+      "loss": 0.4311,
+      "step": 750
+    },
+    {
+      "epoch": 1.5540679711637488,
+      "grad_norm": 1.1348096132278442,
+      "learning_rate": 2.478357419689883e-05,
+      "loss": 0.4683,
+      "step": 755
+    },
+    {
+      "epoch": 1.564366632337796,
+      "grad_norm": 1.0978420972824097,
+      "learning_rate": 2.4705963961118272e-05,
+      "loss": 0.4527,
+      "step": 760
+    },
+    {
+      "epoch": 1.5746652935118435,
+      "grad_norm": 1.116330623626709,
+      "learning_rate": 2.4627904147242315e-05,
+      "loss": 0.4474,
+      "step": 765
+    },
+    {
+      "epoch": 1.584963954685891,
+      "grad_norm": 1.0585070848464966,
+      "learning_rate": 2.4549398370984195e-05,
+      "loss": 0.4194,
+      "step": 770
+    },
+    {
+      "epoch": 1.5952626158599381,
+      "grad_norm": 1.2468876838684082,
+      "learning_rate": 2.447045026871405e-05,
+      "loss": 0.4709,
+      "step": 775
+    },
+    {
+      "epoch": 1.6055612770339855,
+      "grad_norm": 1.1126863956451416,
+      "learning_rate": 2.439106349729043e-05,
+      "loss": 0.4817,
+      "step": 780
+    },
+    {
+      "epoch": 1.615859938208033,
+      "grad_norm": 1.093266487121582,
+      "learning_rate": 2.431124173389097e-05,
+      "loss": 0.4176,
+      "step": 785
+    },
+    {
+      "epoch": 1.6261585993820802,
+      "grad_norm": 1.2111928462982178,
+      "learning_rate": 2.423098867584202e-05,
+      "loss": 0.4628,
+      "step": 790
+    },
+    {
+      "epoch": 1.6364572605561278,
+      "grad_norm": 1.1940813064575195,
+      "learning_rate": 2.4150308040447416e-05,
+      "loss": 0.4437,
+      "step": 795
+    },
+    {
+      "epoch": 1.646755921730175,
+      "grad_norm": 1.185739278793335,
+      "learning_rate": 2.4069203564816258e-05,
+      "loss": 0.4278,
+      "step": 800
+    },
+    {
+      "epoch": 1.6570545829042225,
+      "grad_norm": 1.0990055799484253,
+      "learning_rate": 2.3987679005689866e-05,
+      "loss": 0.4641,
+      "step": 805
+    },
+    {
+      "epoch": 1.66735324407827,
+      "grad_norm": 1.137831449508667,
+      "learning_rate": 2.3905738139267705e-05,
+      "loss": 0.4336,
+      "step": 810
+    },
+    {
+      "epoch": 1.677651905252317,
+      "grad_norm": 1.1449936628341675,
+      "learning_rate": 2.3823384761032512e-05,
+      "loss": 0.4235,
+      "step": 815
+    },
+    {
+      "epoch": 1.6879505664263645,
+      "grad_norm": 1.1143118143081665,
+      "learning_rate": 2.3740622685574483e-05,
+      "loss": 0.3925,
+      "step": 820
+    },
+    {
+      "epoch": 1.698249227600412,
+      "grad_norm": 1.1763856410980225,
+      "learning_rate": 2.3657455746414576e-05,
+      "loss": 0.3921,
+      "step": 825
+    },
+    {
+      "epoch": 1.7085478887744592,
+      "grad_norm": 1.239435076713562,
+      "learning_rate": 2.3573887795826942e-05,
+      "loss": 0.414,
+      "step": 830
+    },
+    {
+      "epoch": 1.7188465499485068,
+      "grad_norm": 1.1457319259643555,
+      "learning_rate": 2.3489922704660496e-05,
+      "loss": 0.4457,
+      "step": 835
+    },
+    {
+      "epoch": 1.729145211122554,
+      "grad_norm": 1.1365303993225098,
+      "learning_rate": 2.340556436215962e-05,
+      "loss": 0.4283,
+      "step": 840
+    },
+    {
+      "epoch": 1.7394438722966015,
+      "grad_norm": 1.1952276229858398,
+      "learning_rate": 2.3320816675784e-05,
+      "loss": 0.4332,
+      "step": 845
+    },
+    {
+      "epoch": 1.749742533470649,
+      "grad_norm": 1.1562155485153198,
+      "learning_rate": 2.3235683571027656e-05,
+      "loss": 0.4064,
+      "step": 850
+    },
+    {
+      "epoch": 1.760041194644696,
+      "grad_norm": 1.1927721500396729,
+      "learning_rate": 2.3150168991237088e-05,
+      "loss": 0.3924,
+      "step": 855
+    },
+    {
+      "epoch": 1.7703398558187435,
+      "grad_norm": 1.1150102615356445,
+      "learning_rate": 2.306427689742865e-05,
+      "loss": 0.4062,
+      "step": 860
+    },
+    {
+      "epoch": 1.780638516992791,
+      "grad_norm": 1.3249040842056274,
+      "learning_rate": 2.297801126810505e-05,
+      "loss": 0.4417,
+      "step": 865
+    },
+    {
+      "epoch": 1.7909371781668382,
+      "grad_norm": 1.194503903388977,
+      "learning_rate": 2.2891376099071088e-05,
+      "loss": 0.4151,
+      "step": 870
+    },
+    {
+      "epoch": 1.8012358393408858,
+      "grad_norm": 1.1773724555969238,
+      "learning_rate": 2.2804375403248547e-05,
+      "loss": 0.4218,
+      "step": 875
+    },
+    {
+      "epoch": 1.811534500514933,
+      "grad_norm": 1.1841347217559814,
+      "learning_rate": 2.2717013210490352e-05,
+      "loss": 0.386,
+      "step": 880
+    },
+    {
+      "epoch": 1.8218331616889805,
+      "grad_norm": 1.1204098463058472,
+      "learning_rate": 2.2629293567393854e-05,
+      "loss": 0.3865,
+      "step": 885
+    },
+    {
+      "epoch": 1.832131822863028,
+      "grad_norm": 1.115275263786316,
+      "learning_rate": 2.2541220537113456e-05,
+      "loss": 0.3657,
+      "step": 890
+    },
+    {
+      "epoch": 1.842430484037075,
+      "grad_norm": 1.0938341617584229,
+      "learning_rate": 2.2452798199172358e-05,
+      "loss": 0.3616,
+      "step": 895
+    },
+    {
+      "epoch": 1.8527291452111225,
+      "grad_norm": 1.2680145502090454,
+      "learning_rate": 2.2364030649273624e-05,
+      "loss": 0.419,
+      "step": 900
+    },
+    {
+      "epoch": 1.86302780638517,
+      "grad_norm": 1.057039499282837,
+      "learning_rate": 2.227492199911045e-05,
+      "loss": 0.4071,
+      "step": 905
+    },
+    {
+      "epoch": 1.8733264675592172,
+      "grad_norm": 1.2678941488265991,
+      "learning_rate": 2.2185476376175718e-05,
+      "loss": 0.3886,
+      "step": 910
+    },
+    {
+      "epoch": 1.8836251287332648,
+      "grad_norm": 1.180358648300171,
+      "learning_rate": 2.2095697923570835e-05,
+      "loss": 0.3771,
+      "step": 915
+    },
+    {
+      "epoch": 1.893923789907312,
+      "grad_norm": 1.2452682256698608,
+      "learning_rate": 2.200559079981377e-05,
+      "loss": 0.3927,
+      "step": 920
+    },
+    {
+      "epoch": 1.9042224510813595,
+      "grad_norm": 1.1591405868530273,
+      "learning_rate": 2.191515917864651e-05,
+      "loss": 0.3586,
+      "step": 925
+    },
+    {
+      "epoch": 1.9145211122554069,
+      "grad_norm": 1.1930902004241943,
+      "learning_rate": 2.182440724884165e-05,
+      "loss": 0.3766,
+      "step": 930
+    },
+    {
+      "epoch": 1.924819773429454,
+      "grad_norm": 1.2268239259719849,
+      "learning_rate": 2.1733339214008428e-05,
+      "loss": 0.3623,
+      "step": 935
+    },
+    {
+      "epoch": 1.9351184346035015,
+      "grad_norm": 1.1663581132888794,
+      "learning_rate": 2.1641959292397998e-05,
+      "loss": 0.3534,
+      "step": 940
+    },
+    {
+      "epoch": 1.945417095777549,
+      "grad_norm": 1.1473199129104614,
+      "learning_rate": 2.1550271716708038e-05,
+      "loss": 0.3578,
+      "step": 945
+    },
+    {
+      "epoch": 1.9557157569515962,
+      "grad_norm": 1.34954833984375,
+      "learning_rate": 2.14582807338867e-05,
+      "loss": 0.3931,
+      "step": 950
+    },
+    {
+      "epoch": 1.9660144181256438,
+      "grad_norm": 1.277458667755127,
+      "learning_rate": 2.1365990604935865e-05,
+      "loss": 0.3851,
+      "step": 955
+    },
+    {
+      "epoch": 1.976313079299691,
+      "grad_norm": 1.1497230529785156,
+      "learning_rate": 2.127340560471383e-05,
+      "loss": 0.3496,
+      "step": 960
+    },
+    {
+      "epoch": 1.9866117404737385,
+      "grad_norm": 1.1924526691436768,
+      "learning_rate": 2.118053002173723e-05,
+      "loss": 0.3401,
+      "step": 965
+    },
+    {
+      "epoch": 1.9969104016477859,
+      "grad_norm": 1.373230218887329,
+      "learning_rate": 2.1087368157982456e-05,
+      "loss": 0.3393,
+      "step": 970
+    },
+    {
+      "epoch": 2.0061791967044282,
+      "grad_norm": 1.1674820184707642,
+      "learning_rate": 2.0993924328686353e-05,
+      "loss": 0.309,
+      "step": 975
+    },
+    {
+      "epoch": 2.016477857878476,
+      "grad_norm": 1.2460054159164429,
+      "learning_rate": 2.090020286214633e-05,
+      "loss": 0.2843,
+      "step": 980
+    },
+    {
+      "epoch": 2.026776519052523,
+      "grad_norm": 1.200320839881897,
+      "learning_rate": 2.0806208099519916e-05,
+      "loss": 0.3074,
+      "step": 985
+    },
+    {
+      "epoch": 2.0370751802265707,
+      "grad_norm": 1.0831384658813477,
+      "learning_rate": 2.0711944394623642e-05,
+      "loss": 0.2826,
+      "step": 990
+    },
+    {
+      "epoch": 2.047373841400618,
+      "grad_norm": 1.1945862770080566,
+      "learning_rate": 2.0617416113731385e-05,
+      "loss": 0.2629,
+      "step": 995
+    },
+    {
+      "epoch": 2.057672502574665,
+      "grad_norm": 1.1249759197235107,
+      "learning_rate": 2.0522627635372126e-05,
+      "loss": 0.2838,
+      "step": 1000
+    },
+    {
+      "epoch": 2.067971163748713,
+      "grad_norm": 1.2407805919647217,
+      "learning_rate": 2.0427583350127134e-05,
+      "loss": 0.2754,
+      "step": 1005
+    },
+    {
+      "epoch": 2.07826982492276,
+      "grad_norm": 1.1751117706298828,
+      "learning_rate": 2.0332287660426598e-05,
+      "loss": 0.2892,
+      "step": 1010
+    },
+    {
+      "epoch": 2.0885684860968072,
+      "grad_norm": 1.1912750005722046,
+      "learning_rate": 2.0236744980345704e-05,
+      "loss": 0.2832,
+      "step": 1015
+    },
+    {
+      "epoch": 2.098867147270855,
+      "grad_norm": 1.0413898229599,
+      "learning_rate": 2.014095973540017e-05,
+      "loss": 0.2862,
+      "step": 1020
+    },
+    {
+      "epoch": 2.109165808444902,
+      "grad_norm": 1.1047124862670898,
+      "learning_rate": 2.0044936362341288e-05,
+      "loss": 0.2724,
+      "step": 1025
+    },
+    {
+      "epoch": 2.1194644696189497,
+      "grad_norm": 1.1833388805389404,
+      "learning_rate": 1.9948679308950358e-05,
+      "loss": 0.2744,
+      "step": 1030
+    },
+    {
+      "epoch": 2.129763130792997,
+      "grad_norm": 1.25725519657135,
+      "learning_rate": 1.985219303383273e-05,
+      "loss": 0.2705,
+      "step": 1035
+    },
+    {
+      "epoch": 2.140061791967044,
+      "grad_norm": 1.1427794694900513,
+      "learning_rate": 1.9755482006211237e-05,
+      "loss": 0.292,
+      "step": 1040
+    },
+    {
+      "epoch": 2.150360453141092,
+      "grad_norm": 1.1591993570327759,
+      "learning_rate": 1.96585507057192e-05,
+      "loss": 0.2725,
+      "step": 1045
+    },
+    {
+      "epoch": 2.160659114315139,
+      "grad_norm": 1.155144453048706,
+      "learning_rate": 1.9561403622192948e-05,
+      "loss": 0.2865,
+      "step": 1050
+    },
+    {
+      "epoch": 2.1709577754891862,
+      "grad_norm": 1.126906156539917,
+      "learning_rate": 1.9464045255463797e-05,
+      "loss": 0.2676,
+      "step": 1055
+    },
+    {
+      "epoch": 2.181256436663234,
+      "grad_norm": 1.245829463005066,
+      "learning_rate": 1.936648011514969e-05,
+      "loss": 0.2763,
+      "step": 1060
+    },
+    {
+      "epoch": 2.191555097837281,
+      "grad_norm": 1.1624958515167236,
+      "learning_rate": 1.9268712720446255e-05,
+      "loss": 0.2503,
+      "step": 1065
+    },
+    {
+      "epoch": 2.2018537590113283,
+      "grad_norm": 1.1856951713562012,
+      "learning_rate": 1.9170747599917507e-05,
+      "loss": 0.2667,
+      "step": 1070
+    },
+    {
+      "epoch": 2.212152420185376,
+      "grad_norm": 1.1675992012023926,
+      "learning_rate": 1.9072589291286073e-05,
+      "loss": 0.272,
+      "step": 1075
+    },
+    {
+      "epoch": 2.222451081359423,
+      "grad_norm": 1.3212230205535889,
+      "learning_rate": 1.897424234122302e-05,
+      "loss": 0.2993,
+      "step": 1080
+    },
+    {
+      "epoch": 2.232749742533471,
+      "grad_norm": 1.3248764276504517,
+      "learning_rate": 1.8875711305137222e-05,
+      "loss": 0.2914,
+      "step": 1085
+    },
+    {
+      "epoch": 2.243048403707518,
+      "grad_norm": 1.18609619140625,
+      "learning_rate": 1.8777000746964397e-05,
+      "loss": 0.2416,
+      "step": 1090
+    },
+    {
+      "epoch": 2.253347064881565,
+      "grad_norm": 1.0249930620193481,
+      "learning_rate": 1.8678115238955686e-05,
+      "loss": 0.2631,
+      "step": 1095
+    },
+    {
+      "epoch": 2.263645726055613,
+      "grad_norm": 1.2961682081222534,
+      "learning_rate": 1.8579059361465863e-05,
+      "loss": 0.265,
+      "step": 1100
+    },
+    {
+      "epoch": 2.27394438722966,
+      "grad_norm": 1.1270440816879272,
+      "learning_rate": 1.8479837702741166e-05,
+      "loss": 0.2721,
+      "step": 1105
+    },
+    {
+      "epoch": 2.2842430484037077,
+      "grad_norm": 1.1455341577529907,
+      "learning_rate": 1.8380454858706798e-05,
+      "loss": 0.2722,
+      "step": 1110
+    },
+    {
+      "epoch": 2.294541709577755,
+      "grad_norm": 1.1415375471115112,
+      "learning_rate": 1.8280915432754033e-05,
+      "loss": 0.2728,
+      "step": 1115
+    },
+    {
+      "epoch": 2.304840370751802,
+      "grad_norm": 1.3010354042053223,
+      "learning_rate": 1.8181224035526968e-05,
+      "loss": 0.2744,
+      "step": 1120
+    },
+    {
+      "epoch": 2.31513903192585,
+      "grad_norm": 1.1614693403244019,
+      "learning_rate": 1.8081385284708995e-05,
+      "loss": 0.2554,
+      "step": 1125
+    },
+    {
+      "epoch": 2.325437693099897,
+      "grad_norm": 1.1690951585769653,
+      "learning_rate": 1.798140380480888e-05,
+      "loss": 0.2463,
+      "step": 1130
+    },
+    {
+      "epoch": 2.335736354273944,
+      "grad_norm": 1.2176849842071533,
+      "learning_rate": 1.7881284226946558e-05,
+      "loss": 0.2519,
+      "step": 1135
+    },
+    {
+      "epoch": 2.346035015447992,
+      "grad_norm": 1.245349407196045,
+      "learning_rate": 1.778103118863866e-05,
+      "loss": 0.2511,
+      "step": 1140
+    },
+    {
+      "epoch": 2.356333676622039,
+      "grad_norm": 1.1520276069641113,
+      "learning_rate": 1.768064933358365e-05,
+      "loss": 0.2561,
+      "step": 1145
+    },
+    {
+      "epoch": 2.3666323377960863,
+      "grad_norm": 1.1642286777496338,
+      "learning_rate": 1.7580143311446766e-05,
+      "loss": 0.2368,
+      "step": 1150
+    },
+    {
+      "epoch": 2.376930998970134,
+      "grad_norm": 1.2488750219345093,
+      "learning_rate": 1.7479517777644632e-05,
+      "loss": 0.2372,
+      "step": 1155
+    },
+    {
+      "epoch": 2.387229660144181,
+      "grad_norm": 1.1384341716766357,
+      "learning_rate": 1.7378777393129635e-05,
+      "loss": 0.2389,
+      "step": 1160
+    },
+    {
+      "epoch": 2.397528321318229,
+      "grad_norm": 1.2213577032089233,
+      "learning_rate": 1.7277926824174012e-05,
+      "loss": 0.2443,
+      "step": 1165
+    },
+    {
+      "epoch": 2.407826982492276,
+      "grad_norm": 1.1850676536560059,
+      "learning_rate": 1.7176970742153717e-05,
+      "loss": 0.2542,
+      "step": 1170
+    },
+    {
+      "epoch": 2.418125643666323,
+      "grad_norm": 1.189429759979248,
+      "learning_rate": 1.7075913823332067e-05,
+      "loss": 0.2493,
+      "step": 1175
+    },
+    {
+      "epoch": 2.428424304840371,
+      "grad_norm": 1.1571177244186401,
+      "learning_rate": 1.6974760748643084e-05,
+      "loss": 0.2105,
+      "step": 1180
+    },
+    {
+      "epoch": 2.438722966014418,
+      "grad_norm": 1.1830145120620728,
+      "learning_rate": 1.6873516203474736e-05,
+      "loss": 0.2361,
+      "step": 1185
+    },
+    {
+      "epoch": 2.4490216271884657,
+      "grad_norm": 1.3023966550827026,
+      "learning_rate": 1.677218487745187e-05,
+      "loss": 0.2498,
+      "step": 1190
+    },
+    {
+      "epoch": 2.459320288362513,
+      "grad_norm": 1.2320756912231445,
+      "learning_rate": 1.6670771464219002e-05,
+      "loss": 0.2672,
+      "step": 1195
+    },
+    {
+      "epoch": 2.46961894953656,
+      "grad_norm": 1.1318459510803223,
+      "learning_rate": 1.6569280661222924e-05,
+      "loss": 0.2394,
+      "step": 1200
+    },
+    {
+      "epoch": 2.479917610710608,
+      "grad_norm": 1.1684930324554443,
+      "learning_rate": 1.646771716949509e-05,
+      "loss": 0.2377,
+      "step": 1205
+    },
+    {
+      "epoch": 2.490216271884655,
+      "grad_norm": 1.1254520416259766,
+      "learning_rate": 1.636608569343387e-05,
+      "loss": 0.2391,
+      "step": 1210
+    },
+    {
+      "epoch": 2.500514933058702,
+      "grad_norm": 1.2523401975631714,
+      "learning_rate": 1.626439094058669e-05,
+      "loss": 0.2143,
+      "step": 1215
+    },
+    {
+      "epoch": 2.51081359423275,
+      "grad_norm": 1.1195322275161743,
+      "learning_rate": 1.6162637621431906e-05,
+      "loss": 0.2394,
+      "step": 1220
+    },
+    {
+      "epoch": 2.521112255406797,
+      "grad_norm": 1.1573140621185303,
+      "learning_rate": 1.606083044916067e-05,
+      "loss": 0.2663,
+      "step": 1225
+    },
+    {
+      "epoch": 2.5314109165808443,
+      "grad_norm": 1.1650135517120361,
+      "learning_rate": 1.595897413945859e-05,
+      "loss": 0.2473,
+      "step": 1230
+    },
+    {
+      "epoch": 2.541709577754892,
+      "grad_norm": 1.1126735210418701,
+      "learning_rate": 1.5857073410287317e-05,
+      "loss": 0.2228,
+      "step": 1235
+    },
+    {
+      "epoch": 2.552008238928939,
+      "grad_norm": 1.180673360824585,
+      "learning_rate": 1.5755132981666004e-05,
+      "loss": 0.2214,
+      "step": 1240
+    },
+    {
+      "epoch": 2.5623069001029863,
+      "grad_norm": 1.1609716415405273,
+      "learning_rate": 1.5653157575452656e-05,
+      "loss": 0.2447,
+      "step": 1245
+    },
+    {
+      "epoch": 2.572605561277034,
+      "grad_norm": 1.1662496328353882,
+      "learning_rate": 1.555115191512546e-05,
+      "loss": 0.2241,
+      "step": 1250
+    },
+    {
+      "epoch": 2.582904222451081,
+      "grad_norm": 1.25954008102417,
+      "learning_rate": 1.5449120725563932e-05,
+      "loss": 0.2011,
+      "step": 1255
+    },
+    {
+      "epoch": 2.593202883625129,
+      "grad_norm": 1.3918417692184448,
+      "learning_rate": 1.5347068732830134e-05,
+      "loss": 0.2467,
+      "step": 1260
+    },
+    {
+      "epoch": 2.603501544799176,
+      "grad_norm": 1.114081621170044,
+      "learning_rate": 1.5245000663949699e-05,
+      "loss": 0.2144,
+      "step": 1265
+    },
+    {
+      "epoch": 2.6138002059732237,
+      "grad_norm": 1.2161543369293213,
+      "learning_rate": 1.5142921246692909e-05,
+      "loss": 0.2276,
+      "step": 1270
+    },
+    {
+      "epoch": 2.624098867147271,
+      "grad_norm": 1.145684003829956,
+      "learning_rate": 1.5040835209355715e-05,
+      "loss": 0.2143,
+      "step": 1275
+    },
+    {
+      "epoch": 2.634397528321318,
+      "grad_norm": 1.1903437376022339,
+      "learning_rate": 1.4938747280540692e-05,
+      "loss": 0.2264,
+      "step": 1280
+    },
+    {
+      "epoch": 2.644696189495366,
+      "grad_norm": 1.2034133672714233,
+      "learning_rate": 1.4836662188938028e-05,
+      "loss": 0.2368,
+      "step": 1285
+    },
+    {
+      "epoch": 2.654994850669413,
+      "grad_norm": 1.1376266479492188,
+      "learning_rate": 1.4734584663106506e-05,
+      "loss": 0.2265,
+      "step": 1290
+    },
+    {
+      "epoch": 2.66529351184346,
+      "grad_norm": 1.1816309690475464,
+      "learning_rate": 1.4632519431254446e-05,
+      "loss": 0.2248,
+      "step": 1295
+    },
+    {
+      "epoch": 2.675592173017508,
+      "grad_norm": 1.2210973501205444,
+      "learning_rate": 1.453047122102073e-05,
+      "loss": 0.233,
+      "step": 1300
+    },
+    {
+      "epoch": 2.685890834191555,
+      "grad_norm": 1.2578504085540771,
+      "learning_rate": 1.44284447592558e-05,
+      "loss": 0.2171,
+      "step": 1305
+    },
+    {
+      "epoch": 2.6961894953656023,
+      "grad_norm": 1.2888798713684082,
+      "learning_rate": 1.4326444771802699e-05,
+      "loss": 0.2057,
+      "step": 1310
+    },
+    {
+      "epoch": 2.70648815653965,
+      "grad_norm": 1.0893986225128174,
+      "learning_rate": 1.4224475983278204e-05,
+      "loss": 0.2133,
+      "step": 1315
+    },
+    {
+      "epoch": 2.716786817713697,
+      "grad_norm": 1.2263572216033936,
+      "learning_rate": 1.412254311685395e-05,
+      "loss": 0.2297,
+      "step": 1320
+    },
+    {
+      "epoch": 2.7270854788877443,
+      "grad_norm": 1.117436408996582,
+      "learning_rate": 1.4020650894037673e-05,
+      "loss": 0.2203,
+      "step": 1325
+    },
+    {
+      "epoch": 2.737384140061792,
+      "grad_norm": 1.1971065998077393,
+      "learning_rate": 1.3918804034454521e-05,
+      "loss": 0.2009,
+      "step": 1330
+    },
+    {
+      "epoch": 2.747682801235839,
+      "grad_norm": 1.2015117406845093,
+      "learning_rate": 1.3817007255628402e-05,
+      "loss": 0.2123,
+      "step": 1335
+    },
+    {
+      "epoch": 2.757981462409887,
+      "grad_norm": 1.2726963758468628,
+      "learning_rate": 1.3715265272763515e-05,
+      "loss": 0.2026,
+      "step": 1340
+    },
+    {
+      "epoch": 2.768280123583934,
+      "grad_norm": 1.189307451248169,
+      "learning_rate": 1.3613582798525903e-05,
+      "loss": 0.2155,
+      "step": 1345
+    },
+    {
+      "epoch": 2.7785787847579817,
+      "grad_norm": 1.0487000942230225,
+      "learning_rate": 1.3511964542825206e-05,
+      "loss": 0.1951,
+      "step": 1350
+    },
+    {
+      "epoch": 2.788877445932029,
+      "grad_norm": 1.1836564540863037,
+      "learning_rate": 1.341041521259645e-05,
+      "loss": 0.2026,
+      "step": 1355
+    },
+    {
+      "epoch": 2.799176107106076,
+      "grad_norm": 1.145855188369751,
+      "learning_rate": 1.330893951158207e-05,
+      "loss": 0.2026,
+      "step": 1360
+    },
+    {
+      "epoch": 2.8094747682801238,
+      "grad_norm": 1.2061723470687866,
+      "learning_rate": 1.3207542140114007e-05,
+      "loss": 0.1792,
+      "step": 1365
+    },
+    {
+      "epoch": 2.819773429454171,
+      "grad_norm": 1.2622625827789307,
+      "learning_rate": 1.310622779489598e-05,
+      "loss": 0.2169,
+      "step": 1370
+    },
+    {
+      "epoch": 2.830072090628218,
+      "grad_norm": 1.1733144521713257,
+      "learning_rate": 1.3005001168785973e-05,
+      "loss": 0.1994,
+      "step": 1375
+    },
+    {
+      "epoch": 2.840370751802266,
+      "grad_norm": 1.225419282913208,
+      "learning_rate": 1.290386695057882e-05,
+      "loss": 0.1817,
+      "step": 1380
+    },
+    {
+      "epoch": 2.850669412976313,
+      "grad_norm": 1.2059926986694336,
+      "learning_rate": 1.2802829824789054e-05,
+      "loss": 0.202,
+      "step": 1385
+    },
+    {
+      "epoch": 2.8609680741503603,
+      "grad_norm": 1.2589644193649292,
+      "learning_rate": 1.2701894471433913e-05,
+      "loss": 0.2131,
+      "step": 1390
+    },
+    {
+      "epoch": 2.871266735324408,
+      "grad_norm": 1.2112151384353638,
+      "learning_rate": 1.2601065565816541e-05,
+      "loss": 0.176,
+      "step": 1395
+    },
+    {
+      "epoch": 2.881565396498455,
+      "grad_norm": 1.2503776550292969,
+      "learning_rate": 1.2500347778309463e-05,
+      "loss": 0.1838,
+      "step": 1400
+    },
+    {
+      "epoch": 2.8918640576725023,
+      "grad_norm": 1.11964750289917,
+      "learning_rate": 1.2399745774138226e-05,
+      "loss": 0.2135,
+      "step": 1405
+    },
+    {
+      "epoch": 2.90216271884655,
+      "grad_norm": 1.1664073467254639,
+      "learning_rate": 1.2299264213165329e-05,
+      "loss": 0.2044,
+      "step": 1410
+    },
+    {
+      "epoch": 2.912461380020597,
+      "grad_norm": 1.0924075841903687,
+      "learning_rate": 1.2198907749674355e-05,
+      "loss": 0.1801,
+      "step": 1415
+    },
+    {
+      "epoch": 2.922760041194645,
+      "grad_norm": 1.1321125030517578,
+      "learning_rate": 1.2098681032154417e-05,
+      "loss": 0.1783,
+      "step": 1420
+    },
+    {
+      "epoch": 2.933058702368692,
+      "grad_norm": 1.2050647735595703,
+      "learning_rate": 1.1998588703084815e-05,
+      "loss": 0.1841,
+      "step": 1425
+    },
+    {
+      "epoch": 2.9433573635427397,
+      "grad_norm": 1.1679717302322388,
+      "learning_rate": 1.1898635398720003e-05,
+      "loss": 0.1824,
+      "step": 1430
+    },
+    {
+      "epoch": 2.953656024716787,
+      "grad_norm": 1.1852076053619385,
+      "learning_rate": 1.1798825748874837e-05,
+      "loss": 0.1935,
+      "step": 1435
+    },
+    {
+      "epoch": 2.963954685890834,
+      "grad_norm": 1.0744694471359253,
+      "learning_rate": 1.1699164376710146e-05,
+      "loss": 0.1908,
+      "step": 1440
+    },
+    {
+      "epoch": 2.9742533470648818,
+      "grad_norm": 1.2433335781097412,
+      "learning_rate": 1.1599655898518544e-05,
+      "loss": 0.1792,
+      "step": 1445
+    },
+    {
+      "epoch": 2.984552008238929,
+      "grad_norm": 1.1401877403259277,
+      "learning_rate": 1.1500304923510658e-05,
+      "loss": 0.176,
+      "step": 1450
+    },
+    {
+      "epoch": 2.994850669412976,
+      "grad_norm": 1.1364681720733643,
+      "learning_rate": 1.1401116053601584e-05,
+      "loss": 0.171,
+      "step": 1455
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 2430,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.0899256700114043e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

44_128_e5_3e-5/checkpoint-1458/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fe7162dc6a9b5b59d136a37e3e396521ee8c38ffa6bdafb995352d8fddcfd9b
+size 7736

44_128_e5_3e-5/checkpoint-1458/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/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, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in 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,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_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``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    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("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

44_128_e5_3e-5/checkpoint-1944/README.md

@@ -0,0 +1,202 @@
+---
+base_model: ibm-granite/granite-3.3-8b-base
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.15.2

44_128_e5_3e-5/checkpoint-1944/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "down_proj",
+    "q_proj",
+    "gate_proj",
+    "v_proj",
+    "o_proj",
+    "k_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

44_128_e5_3e-5/checkpoint-1944/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0aa89da5a6d9c4be9d4034b7adcd8efad3991d497eac62580ebd621cafbfe1d
+size 791751704

44_128_e5_3e-5/checkpoint-1944/latest

@@ -0,0 +1 @@
+global_step1944

44_128_e5_3e-5/checkpoint-1944/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38e4f19a96a51ad059ef0820f69dcf8eb6521fd1897652b5738bd7689ee31d45
+size 15920

44_128_e5_3e-5/checkpoint-1944/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e51f78065033d352ac09ec720e768713fb78684f92afd81b7c7a21ba1a80c2d3
+size 15920

44_128_e5_3e-5/checkpoint-1944/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b989aa45310c6487d76bbfc47718e2c4d23e9e9bd45b7595c2046ec56ca3d84
+size 15920

44_128_e5_3e-5/checkpoint-1944/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e653894d209a27f1caf6a000052f6b1f49a7addbbe6740b538fdd2d30feee899
+size 15920

44_128_e5_3e-5/checkpoint-1944/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:747e25dce8f04aa38b74cc40bf46d4f1437547c0c8e19aef4c9e38b27860d6bc
+size 15920

44_128_e5_3e-5/checkpoint-1944/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:556753425f316bcc9e51c29054aa6e0b06c5bc3247b4fa5cfbddc5f9a6f2d7ec
+size 15920

44_128_e5_3e-5/checkpoint-1944/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1981159db51770ac39c80acfedbc5c8e267d975ef65388b0d37ad7445996a069
+size 15920

44_128_e5_3e-5/checkpoint-1944/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:590e3e3fa1f5c3d2bf54e2682920ab37895a44f4025e5bd1077eecc63fb72d79
+size 15920

44_128_e5_3e-5/checkpoint-1944/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd9fb3e8c67722cc958e25292acd7cf1ac84f79442c0fa3e83961acf5144d790
+size 1064

44_128_e5_3e-5/checkpoint-1944/special_tokens_map.json

@@ -0,0 +1,45 @@
+{
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<reponame>",
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

44_128_e5_3e-5/checkpoint-1944/tokenizer_config.json

@@ -0,0 +1,188 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<fim_prefix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<fim_middle>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<fim_suffix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<fim_pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<filename>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<gh_stars>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<issue_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "8": {
+      "content": "<issue_comment>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "9": {
+      "content": "<issue_closed>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "10": {
+      "content": "<jupyter_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "11": {
+      "content": "<jupyter_text>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "12": {
+      "content": "<jupyter_code>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "13": {
+      "content": "<jupyter_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "14": {
+      "content": "<empty_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "15": {
+      "content": "<commit_before>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "16": {
+      "content": "<commit_msg>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "17": {
+      "content": "<commit_after>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "18": {
+      "content": "<reponame>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<reponame>",
+  "padding_side": "left",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

44_128_e5_3e-5/checkpoint-1944/trainer_state.json

@@ -0,0 +1,2750 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 1944,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.010298661174047374,
+      "grad_norm": 1.141814947128296,
+      "learning_rate": 9.836065573770493e-07,
+      "loss": 1.3338,
+      "step": 5
+    },
+    {
+      "epoch": 0.02059732234809475,
+      "grad_norm": 0.9742817878723145,
+      "learning_rate": 2.213114754098361e-06,
+      "loss": 1.3646,
+      "step": 10
+    },
+    {
+      "epoch": 0.030895983522142123,
+      "grad_norm": 0.7402164340019226,
+      "learning_rate": 3.4426229508196724e-06,
+      "loss": 1.3261,
+      "step": 15
+    },
+    {
+      "epoch": 0.0411946446961895,
+      "grad_norm": 0.5555693507194519,
+      "learning_rate": 4.672131147540984e-06,
+      "loss": 1.3035,
+      "step": 20
+    },
+    {
+      "epoch": 0.05149330587023687,
+      "grad_norm": 0.6190383434295654,
+      "learning_rate": 5.901639344262295e-06,
+      "loss": 1.3847,
+      "step": 25
+    },
+    {
+      "epoch": 0.061791967044284246,
+      "grad_norm": 0.5282045006752014,
+      "learning_rate": 7.131147540983606e-06,
+      "loss": 1.3197,
+      "step": 30
+    },
+    {
+      "epoch": 0.07209062821833162,
+      "grad_norm": 0.48918476700782776,
+      "learning_rate": 8.360655737704917e-06,
+      "loss": 1.2899,
+      "step": 35
+    },
+    {
+      "epoch": 0.082389289392379,
+      "grad_norm": 0.7961065769195557,
+      "learning_rate": 9.590163934426231e-06,
+      "loss": 1.2656,
+      "step": 40
+    },
+    {
+      "epoch": 0.09268795056642637,
+      "grad_norm": 0.4608859419822693,
+      "learning_rate": 1.0819672131147542e-05,
+      "loss": 1.26,
+      "step": 45
+    },
+    {
+      "epoch": 0.10298661174047374,
+      "grad_norm": 0.555935800075531,
+      "learning_rate": 1.2049180327868853e-05,
+      "loss": 1.2548,
+      "step": 50
+    },
+    {
+      "epoch": 0.11328527291452112,
+      "grad_norm": 0.4993849992752075,
+      "learning_rate": 1.3278688524590163e-05,
+      "loss": 1.2389,
+      "step": 55
+    },
+    {
+      "epoch": 0.12358393408856849,
+      "grad_norm": 0.5452463030815125,
+      "learning_rate": 1.4508196721311476e-05,
+      "loss": 1.2124,
+      "step": 60
+    },
+    {
+      "epoch": 0.13388259526261587,
+      "grad_norm": 0.4450685381889343,
+      "learning_rate": 1.5737704918032788e-05,
+      "loss": 1.2105,
+      "step": 65
+    },
+    {
+      "epoch": 0.14418125643666324,
+      "grad_norm": 0.5045918822288513,
+      "learning_rate": 1.6967213114754097e-05,
+      "loss": 1.2305,
+      "step": 70
+    },
+    {
+      "epoch": 0.15447991761071062,
+      "grad_norm": 0.49198681116104126,
+      "learning_rate": 1.819672131147541e-05,
+      "loss": 1.2464,
+      "step": 75
+    },
+    {
+      "epoch": 0.164778578784758,
+      "grad_norm": 0.4819914996623993,
+      "learning_rate": 1.9426229508196722e-05,
+      "loss": 1.194,
+      "step": 80
+    },
+    {
+      "epoch": 0.17507723995880536,
+      "grad_norm": 0.4840977191925049,
+      "learning_rate": 2.0655737704918034e-05,
+      "loss": 1.1949,
+      "step": 85
+    },
+    {
+      "epoch": 0.18537590113285274,
+      "grad_norm": 0.55226069688797,
+      "learning_rate": 2.1885245901639347e-05,
+      "loss": 1.1936,
+      "step": 90
+    },
+    {
+      "epoch": 0.1956745623069001,
+      "grad_norm": 0.47093257308006287,
+      "learning_rate": 2.3114754098360656e-05,
+      "loss": 1.1214,
+      "step": 95
+    },
+    {
+      "epoch": 0.2059732234809475,
+      "grad_norm": 0.5413655042648315,
+      "learning_rate": 2.434426229508197e-05,
+      "loss": 1.1661,
+      "step": 100
+    },
+    {
+      "epoch": 0.21627188465499486,
+      "grad_norm": 0.525501012802124,
+      "learning_rate": 2.5573770491803277e-05,
+      "loss": 1.1657,
+      "step": 105
+    },
+    {
+      "epoch": 0.22657054582904224,
+      "grad_norm": 0.5898400545120239,
+      "learning_rate": 2.680327868852459e-05,
+      "loss": 1.1567,
+      "step": 110
+    },
+    {
+      "epoch": 0.2368692070030896,
+      "grad_norm": 0.5453243851661682,
+      "learning_rate": 2.8032786885245902e-05,
+      "loss": 1.1429,
+      "step": 115
+    },
+    {
+      "epoch": 0.24716786817713698,
+      "grad_norm": 0.5925690531730652,
+      "learning_rate": 2.9262295081967215e-05,
+      "loss": 1.165,
+      "step": 120
+    },
+    {
+      "epoch": 0.25746652935118436,
+      "grad_norm": 0.5159111618995667,
+      "learning_rate": 2.9999944416086244e-05,
+      "loss": 1.0712,
+      "step": 125
+    },
+    {
+      "epoch": 0.26776519052523173,
+      "grad_norm": 0.5997148752212524,
+      "learning_rate": 2.999931910178738e-05,
+      "loss": 1.087,
+      "step": 130
+    },
+    {
+      "epoch": 0.2780638516992791,
+      "grad_norm": 0.5427290201187134,
+      "learning_rate": 2.9997999022358424e-05,
+      "loss": 1.1449,
+      "step": 135
+    },
+    {
+      "epoch": 0.2883625128733265,
+      "grad_norm": 0.7504834532737732,
+      "learning_rate": 2.999598423894516e-05,
+      "loss": 1.1485,
+      "step": 140
+    },
+    {
+      "epoch": 0.29866117404737386,
+      "grad_norm": 0.5923535823822021,
+      "learning_rate": 2.9993274844871917e-05,
+      "loss": 1.1358,
+      "step": 145
+    },
+    {
+      "epoch": 0.30895983522142123,
+      "grad_norm": 0.6039471626281738,
+      "learning_rate": 2.9989870965637215e-05,
+      "loss": 1.1134,
+      "step": 150
+    },
+    {
+      "epoch": 0.3192584963954686,
+      "grad_norm": 0.5884091854095459,
+      "learning_rate": 2.9985772758907992e-05,
+      "loss": 1.101,
+      "step": 155
+    },
+    {
+      "epoch": 0.329557157569516,
+      "grad_norm": 0.6560966968536377,
+      "learning_rate": 2.998098041451227e-05,
+      "loss": 1.126,
+      "step": 160
+    },
+    {
+      "epoch": 0.33985581874356335,
+      "grad_norm": 0.557138979434967,
+      "learning_rate": 2.9975494154430378e-05,
+      "loss": 1.0722,
+      "step": 165
+    },
+    {
+      "epoch": 0.35015447991761073,
+      "grad_norm": 0.7041076421737671,
+      "learning_rate": 2.996931423278467e-05,
+      "loss": 1.0913,
+      "step": 170
+    },
+    {
+      "epoch": 0.3604531410916581,
+      "grad_norm": 0.6222100257873535,
+      "learning_rate": 2.9962440935827735e-05,
+      "loss": 1.0392,
+      "step": 175
+    },
+    {
+      "epoch": 0.3707518022657055,
+      "grad_norm": 0.6287522912025452,
+      "learning_rate": 2.995487458192917e-05,
+      "loss": 1.003,
+      "step": 180
+    },
+    {
+      "epoch": 0.38105046343975285,
+      "grad_norm": 0.8934054970741272,
+      "learning_rate": 2.9946615521560805e-05,
+      "loss": 1.0176,
+      "step": 185
+    },
+    {
+      "epoch": 0.3913491246138002,
+      "grad_norm": 0.6240345239639282,
+      "learning_rate": 2.9937664137280478e-05,
+      "loss": 1.0205,
+      "step": 190
+    },
+    {
+      "epoch": 0.4016477857878476,
+      "grad_norm": 0.6924830675125122,
+      "learning_rate": 2.9928020843714323e-05,
+      "loss": 1.0218,
+      "step": 195
+    },
+    {
+      "epoch": 0.411946446961895,
+      "grad_norm": 0.7187322378158569,
+      "learning_rate": 2.9917686087537563e-05,
+      "loss": 1.0581,
+      "step": 200
+    },
+    {
+      "epoch": 0.42224510813594235,
+      "grad_norm": 0.6714951992034912,
+      "learning_rate": 2.9906660347453803e-05,
+      "loss": 0.9597,
+      "step": 205
+    },
+    {
+      "epoch": 0.4325437693099897,
+      "grad_norm": 0.758985698223114,
+      "learning_rate": 2.9894944134172876e-05,
+      "loss": 0.9869,
+      "step": 210
+    },
+    {
+      "epoch": 0.4428424304840371,
+      "grad_norm": 0.6603527069091797,
+      "learning_rate": 2.988253799038718e-05,
+      "loss": 0.9687,
+      "step": 215
+    },
+    {
+      "epoch": 0.45314109165808447,
+      "grad_norm": 0.7007045745849609,
+      "learning_rate": 2.986944249074654e-05,
+      "loss": 1.0301,
+      "step": 220
+    },
+    {
+      "epoch": 0.46343975283213185,
+      "grad_norm": 0.7556740641593933,
+      "learning_rate": 2.985565824183159e-05,
+      "loss": 1.0108,
+      "step": 225
+    },
+    {
+      "epoch": 0.4737384140061792,
+      "grad_norm": 0.7658377885818481,
+      "learning_rate": 2.9841185882125682e-05,
+      "loss": 0.9966,
+      "step": 230
+    },
+    {
+      "epoch": 0.4840370751802266,
+      "grad_norm": 0.8318830728530884,
+      "learning_rate": 2.9826026081985305e-05,
+      "loss": 0.9416,
+      "step": 235
+    },
+    {
+      "epoch": 0.49433573635427397,
+      "grad_norm": 0.7977103590965271,
+      "learning_rate": 2.9810179543609032e-05,
+      "loss": 0.9753,
+      "step": 240
+    },
+    {
+      "epoch": 0.5046343975283213,
+      "grad_norm": 0.7902314066886902,
+      "learning_rate": 2.9793647001005002e-05,
+      "loss": 0.9422,
+      "step": 245
+    },
+    {
+      "epoch": 0.5149330587023687,
+      "grad_norm": 0.8238946795463562,
+      "learning_rate": 2.9776429219956917e-05,
+      "loss": 0.9609,
+      "step": 250
+    },
+    {
+      "epoch": 0.525231719876416,
+      "grad_norm": 0.8614672422409058,
+      "learning_rate": 2.975852699798857e-05,
+      "loss": 0.9375,
+      "step": 255
+    },
+    {
+      "epoch": 0.5355303810504635,
+      "grad_norm": 0.946057915687561,
+      "learning_rate": 2.9739941164326914e-05,
+      "loss": 0.8812,
+      "step": 260
+    },
+    {
+      "epoch": 0.5458290422245108,
+      "grad_norm": 0.8593087196350098,
+      "learning_rate": 2.9720672579863633e-05,
+      "loss": 0.9091,
+      "step": 265
+    },
+    {
+      "epoch": 0.5561277033985582,
+      "grad_norm": 0.8131739497184753,
+      "learning_rate": 2.970072213711528e-05,
+      "loss": 0.9136,
+      "step": 270
+    },
+    {
+      "epoch": 0.5664263645726055,
+      "grad_norm": 0.8276655077934265,
+      "learning_rate": 2.9680090760181933e-05,
+      "loss": 0.9188,
+      "step": 275
+    },
+    {
+      "epoch": 0.576725025746653,
+      "grad_norm": 0.8448908925056458,
+      "learning_rate": 2.9658779404704387e-05,
+      "loss": 0.8959,
+      "step": 280
+    },
+    {
+      "epoch": 0.5870236869207003,
+      "grad_norm": 0.7980640530586243,
+      "learning_rate": 2.9636789057819903e-05,
+      "loss": 0.9117,
+      "step": 285
+    },
+    {
+      "epoch": 0.5973223480947477,
+      "grad_norm": 0.8234493136405945,
+      "learning_rate": 2.961412073811646e-05,
+      "loss": 0.904,
+      "step": 290
+    },
+    {
+      "epoch": 0.607621009268795,
+      "grad_norm": 1.027661919593811,
+      "learning_rate": 2.9590775495585597e-05,
+      "loss": 0.8299,
+      "step": 295
+    },
+    {
+      "epoch": 0.6179196704428425,
+      "grad_norm": 0.8117839097976685,
+      "learning_rate": 2.956675441157376e-05,
+      "loss": 0.8743,
+      "step": 300
+    },
+    {
+      "epoch": 0.6282183316168898,
+      "grad_norm": 0.8687512278556824,
+      "learning_rate": 2.954205859873223e-05,
+      "loss": 0.8539,
+      "step": 305
+    },
+    {
+      "epoch": 0.6385169927909372,
+      "grad_norm": 0.9031465649604797,
+      "learning_rate": 2.951668920096557e-05,
+      "loss": 0.8416,
+      "step": 310
+    },
+    {
+      "epoch": 0.6488156539649845,
+      "grad_norm": 1.0350013971328735,
+      "learning_rate": 2.9490647393378656e-05,
+      "loss": 0.8505,
+      "step": 315
+    },
+    {
+      "epoch": 0.659114315139032,
+      "grad_norm": 0.9072204828262329,
+      "learning_rate": 2.9463934382222226e-05,
+      "loss": 0.8328,
+      "step": 320
+    },
+    {
+      "epoch": 0.6694129763130793,
+      "grad_norm": 0.8805303573608398,
+      "learning_rate": 2.943655140483703e-05,
+      "loss": 0.8498,
+      "step": 325
+    },
+    {
+      "epoch": 0.6797116374871267,
+      "grad_norm": 0.9569805264472961,
+      "learning_rate": 2.94084997295965e-05,
+      "loss": 0.8613,
+      "step": 330
+    },
+    {
+      "epoch": 0.690010298661174,
+      "grad_norm": 0.842765748500824,
+      "learning_rate": 2.9379780655848e-05,
+      "loss": 0.9271,
+      "step": 335
+    },
+    {
+      "epoch": 0.7003089598352215,
+      "grad_norm": 1.0585700273513794,
+      "learning_rate": 2.9350395513852655e-05,
+      "loss": 0.8049,
+      "step": 340
+    },
+    {
+      "epoch": 0.7106076210092688,
+      "grad_norm": 0.9681879878044128,
+      "learning_rate": 2.9320345664723713e-05,
+      "loss": 0.8611,
+      "step": 345
+    },
+    {
+      "epoch": 0.7209062821833162,
+      "grad_norm": 0.9282833933830261,
+      "learning_rate": 2.928963250036351e-05,
+      "loss": 0.8645,
+      "step": 350
+    },
+    {
+      "epoch": 0.7312049433573635,
+      "grad_norm": 0.880826473236084,
+      "learning_rate": 2.9258257443399007e-05,
+      "loss": 0.8409,
+      "step": 355
+    },
+    {
+      "epoch": 0.741503604531411,
+      "grad_norm": 0.9096245169639587,
+      "learning_rate": 2.922622194711587e-05,
+      "loss": 0.8648,
+      "step": 360
+    },
+    {
+      "epoch": 0.7518022657054583,
+      "grad_norm": 0.9456668496131897,
+      "learning_rate": 2.919352749539117e-05,
+      "loss": 0.8045,
+      "step": 365
+    },
+    {
+      "epoch": 0.7621009268795057,
+      "grad_norm": 0.9902530312538147,
+      "learning_rate": 2.916017560262466e-05,
+      "loss": 0.84,
+      "step": 370
+    },
+    {
+      "epoch": 0.772399588053553,
+      "grad_norm": 1.0271276235580444,
+      "learning_rate": 2.91261678136686e-05,
+      "loss": 0.8052,
+      "step": 375
+    },
+    {
+      "epoch": 0.7826982492276005,
+      "grad_norm": 0.9190011024475098,
+      "learning_rate": 2.9091505703756224e-05,
+      "loss": 0.7965,
+      "step": 380
+    },
+    {
+      "epoch": 0.7929969104016478,
+      "grad_norm": 1.022260308265686,
+      "learning_rate": 2.9056190878428766e-05,
+      "loss": 0.8417,
+      "step": 385
+    },
+    {
+      "epoch": 0.8032955715756952,
+      "grad_norm": 0.9204807281494141,
+      "learning_rate": 2.9020224973461098e-05,
+      "loss": 0.7853,
+      "step": 390
+    },
+    {
+      "epoch": 0.8135942327497425,
+      "grad_norm": 0.9588913917541504,
+      "learning_rate": 2.8983609654785948e-05,
+      "loss": 0.8235,
+      "step": 395
+    },
+    {
+      "epoch": 0.82389289392379,
+      "grad_norm": 0.8754395246505737,
+      "learning_rate": 2.8946346618416742e-05,
+      "loss": 0.7409,
+      "step": 400
+    },
+    {
+      "epoch": 0.8341915550978373,
+      "grad_norm": 1.0218687057495117,
+      "learning_rate": 2.8908437590369056e-05,
+      "loss": 0.7466,
+      "step": 405
+    },
+    {
+      "epoch": 0.8444902162718847,
+      "grad_norm": 1.1078494787216187,
+      "learning_rate": 2.886988432658065e-05,
+      "loss": 0.7767,
+      "step": 410
+    },
+    {
+      "epoch": 0.854788877445932,
+      "grad_norm": 0.970990002155304,
+      "learning_rate": 2.8830688612830142e-05,
+      "loss": 0.7584,
+      "step": 415
+    },
+    {
+      "epoch": 0.8650875386199794,
+      "grad_norm": 1.0772037506103516,
+      "learning_rate": 2.8790852264654287e-05,
+      "loss": 0.8034,
+      "step": 420
+    },
+    {
+      "epoch": 0.8753861997940268,
+      "grad_norm": 0.9200147986412048,
+      "learning_rate": 2.875037712726389e-05,
+      "loss": 0.7289,
+      "step": 425
+    },
+    {
+      "epoch": 0.8856848609680742,
+      "grad_norm": 1.0537430047988892,
+      "learning_rate": 2.8709265075458324e-05,
+      "loss": 0.7462,
+      "step": 430
+    },
+    {
+      "epoch": 0.8959835221421215,
+      "grad_norm": 0.9600791335105896,
+      "learning_rate": 2.86675180135387e-05,
+      "loss": 0.68,
+      "step": 435
+    },
+    {
+      "epoch": 0.9062821833161689,
+      "grad_norm": 1.05148184299469,
+      "learning_rate": 2.8625137875219655e-05,
+      "loss": 0.7472,
+      "step": 440
+    },
+    {
+      "epoch": 0.9165808444902163,
+      "grad_norm": 0.9907565712928772,
+      "learning_rate": 2.8582126623539787e-05,
+      "loss": 0.6766,
+      "step": 445
+    },
+    {
+      "epoch": 0.9268795056642637,
+      "grad_norm": 0.9938767552375793,
+      "learning_rate": 2.8538486250770724e-05,
+      "loss": 0.7238,
+      "step": 450
+    },
+    {
+      "epoch": 0.937178166838311,
+      "grad_norm": 0.9231424331665039,
+      "learning_rate": 2.849421877832484e-05,
+      "loss": 0.7381,
+      "step": 455
+    },
+    {
+      "epoch": 0.9474768280123584,
+      "grad_norm": 1.0045597553253174,
+      "learning_rate": 2.844932625666163e-05,
+      "loss": 0.6675,
+      "step": 460
+    },
+    {
+      "epoch": 0.9577754891864058,
+      "grad_norm": 0.9766920804977417,
+      "learning_rate": 2.8403810765192727e-05,
+      "loss": 0.7332,
+      "step": 465
+    },
+    {
+      "epoch": 0.9680741503604532,
+      "grad_norm": 0.9820194244384766,
+      "learning_rate": 2.835767441218559e-05,
+      "loss": 0.726,
+      "step": 470
+    },
+    {
+      "epoch": 0.9783728115345005,
+      "grad_norm": 1.1101776361465454,
+      "learning_rate": 2.831091933466584e-05,
+      "loss": 0.7179,
+      "step": 475
+    },
+    {
+      "epoch": 0.9886714727085479,
+      "grad_norm": 1.3044633865356445,
+      "learning_rate": 2.82635476983183e-05,
+      "loss": 0.6778,
+      "step": 480
+    },
+    {
+      "epoch": 0.9989701338825953,
+      "grad_norm": 1.0158640146255493,
+      "learning_rate": 2.821556169738663e-05,
+      "loss": 0.7167,
+      "step": 485
+    },
+    {
+      "epoch": 1.008238928939238,
+      "grad_norm": 1.292767882347107,
+      "learning_rate": 2.8166963554571748e-05,
+      "loss": 0.585,
+      "step": 490
+    },
+    {
+      "epoch": 1.0185375901132854,
+      "grad_norm": 0.9503241777420044,
+      "learning_rate": 2.8117755520928834e-05,
+      "loss": 0.6277,
+      "step": 495
+    },
+    {
+      "epoch": 1.0288362512873326,
+      "grad_norm": 1.0615237951278687,
+      "learning_rate": 2.8067939875763078e-05,
+      "loss": 0.6104,
+      "step": 500
+    },
+    {
+      "epoch": 1.03913491246138,
+      "grad_norm": 1.053903341293335,
+      "learning_rate": 2.8017518926524103e-05,
+      "loss": 0.5969,
+      "step": 505
+    },
+    {
+      "epoch": 1.0494335736354274,
+      "grad_norm": 1.1858323812484741,
+      "learning_rate": 2.7966495008699083e-05,
+      "loss": 0.592,
+      "step": 510
+    },
+    {
+      "epoch": 1.0597322348094749,
+      "grad_norm": 0.9921702146530151,
+      "learning_rate": 2.7914870485704563e-05,
+      "loss": 0.5961,
+      "step": 515
+    },
+    {
+      "epoch": 1.070030895983522,
+      "grad_norm": 0.9998088479042053,
+      "learning_rate": 2.7862647748776978e-05,
+      "loss": 0.6306,
+      "step": 520
+    },
+    {
+      "epoch": 1.0803295571575695,
+      "grad_norm": 1.146559238433838,
+      "learning_rate": 2.780982921686191e-05,
+      "loss": 0.582,
+      "step": 525
+    },
+    {
+      "epoch": 1.090628218331617,
+      "grad_norm": 1.0381414890289307,
+      "learning_rate": 2.775641733650203e-05,
+      "loss": 0.5503,
+      "step": 530
+    },
+    {
+      "epoch": 1.1009268795056641,
+      "grad_norm": 1.0624871253967285,
+      "learning_rate": 2.770241458172378e-05,
+      "loss": 0.5694,
+      "step": 535
+    },
+    {
+      "epoch": 1.1112255406797116,
+      "grad_norm": 1.0869427919387817,
+      "learning_rate": 2.7647823453922757e-05,
+      "loss": 0.6295,
+      "step": 540
+    },
+    {
+      "epoch": 1.121524201853759,
+      "grad_norm": 1.2071483135223389,
+      "learning_rate": 2.7592646481747888e-05,
+      "loss": 0.5973,
+      "step": 545
+    },
+    {
+      "epoch": 1.1318228630278064,
+      "grad_norm": 1.0342060327529907,
+      "learning_rate": 2.753688622098427e-05,
+      "loss": 0.5955,
+      "step": 550
+    },
+    {
+      "epoch": 1.1421215242018539,
+      "grad_norm": 1.0878804922103882,
+      "learning_rate": 2.7480545254434798e-05,
+      "loss": 0.6152,
+      "step": 555
+    },
+    {
+      "epoch": 1.152420185375901,
+      "grad_norm": 1.1208651065826416,
+      "learning_rate": 2.7423626191800532e-05,
+      "loss": 0.586,
+      "step": 560
+    },
+    {
+      "epoch": 1.1627188465499485,
+      "grad_norm": 1.063509464263916,
+      "learning_rate": 2.7366131669559816e-05,
+      "loss": 0.5879,
+      "step": 565
+    },
+    {
+      "epoch": 1.173017507723996,
+      "grad_norm": 1.2478981018066406,
+      "learning_rate": 2.7308064350846148e-05,
+      "loss": 0.5478,
+      "step": 570
+    },
+    {
+      "epoch": 1.1833161688980431,
+      "grad_norm": 1.1873619556427002,
+      "learning_rate": 2.7249426925324844e-05,
+      "loss": 0.5383,
+      "step": 575
+    },
+    {
+      "epoch": 1.1936148300720906,
+      "grad_norm": 1.0544021129608154,
+      "learning_rate": 2.7190222109068427e-05,
+      "loss": 0.6021,
+      "step": 580
+    },
+    {
+      "epoch": 1.203913491246138,
+      "grad_norm": 1.2434903383255005,
+      "learning_rate": 2.7130452644430844e-05,
+      "loss": 0.5455,
+      "step": 585
+    },
+    {
+      "epoch": 1.2142121524201854,
+      "grad_norm": 1.1316081285476685,
+      "learning_rate": 2.7070121299920422e-05,
+      "loss": 0.5933,
+      "step": 590
+    },
+    {
+      "epoch": 1.2245108135942329,
+      "grad_norm": 0.9929481148719788,
+      "learning_rate": 2.7009230870071645e-05,
+      "loss": 0.5502,
+      "step": 595
+    },
+    {
+      "epoch": 1.23480947476828,
+      "grad_norm": 1.0308139324188232,
+      "learning_rate": 2.6947784175315694e-05,
+      "loss": 0.5177,
+      "step": 600
+    },
+    {
+      "epoch": 1.2451081359423275,
+      "grad_norm": 1.0328620672225952,
+      "learning_rate": 2.688578406184983e-05,
+      "loss": 0.5156,
+      "step": 605
+    },
+    {
+      "epoch": 1.255406797116375,
+      "grad_norm": 1.1804379224777222,
+      "learning_rate": 2.6823233401505548e-05,
+      "loss": 0.5648,
+      "step": 610
+    },
+    {
+      "epoch": 1.2657054582904221,
+      "grad_norm": 1.2127525806427002,
+      "learning_rate": 2.6760135091615535e-05,
+      "loss": 0.5654,
+      "step": 615
+    },
+    {
+      "epoch": 1.2760041194644696,
+      "grad_norm": 1.1704093217849731,
+      "learning_rate": 2.66964920548795e-05,
+      "loss": 0.5534,
+      "step": 620
+    },
+    {
+      "epoch": 1.286302780638517,
+      "grad_norm": 1.062095284461975,
+      "learning_rate": 2.663230723922877e-05,
+      "loss": 0.4969,
+      "step": 625
+    },
+    {
+      "epoch": 1.2966014418125644,
+      "grad_norm": 1.391966462135315,
+      "learning_rate": 2.6567583617689754e-05,
+      "loss": 0.5283,
+      "step": 630
+    },
+    {
+      "epoch": 1.3069001029866119,
+      "grad_norm": 1.1415766477584839,
+      "learning_rate": 2.6502324188246226e-05,
+      "loss": 0.5447,
+      "step": 635
+    },
+    {
+      "epoch": 1.317198764160659,
+      "grad_norm": 1.146954894065857,
+      "learning_rate": 2.643653197370048e-05,
+      "loss": 0.5171,
+      "step": 640
+    },
+    {
+      "epoch": 1.3274974253347065,
+      "grad_norm": 1.064420461654663,
+      "learning_rate": 2.6370210021533266e-05,
+      "loss": 0.5127,
+      "step": 645
+    },
+    {
+      "epoch": 1.337796086508754,
+      "grad_norm": 1.172444224357605,
+      "learning_rate": 2.6303361403762686e-05,
+      "loss": 0.5017,
+      "step": 650
+    },
+    {
+      "epoch": 1.3480947476828011,
+      "grad_norm": 1.0418468713760376,
+      "learning_rate": 2.623598921680188e-05,
+      "loss": 0.4846,
+      "step": 655
+    },
+    {
+      "epoch": 1.3583934088568486,
+      "grad_norm": 1.2469124794006348,
+      "learning_rate": 2.6168096581315574e-05,
+      "loss": 0.4795,
+      "step": 660
+    },
+    {
+      "epoch": 1.368692070030896,
+      "grad_norm": 1.1334832906723022,
+      "learning_rate": 2.609968664207558e-05,
+      "loss": 0.5371,
+      "step": 665
+    },
+    {
+      "epoch": 1.3789907312049434,
+      "grad_norm": 1.1697006225585938,
+      "learning_rate": 2.6030762567815093e-05,
+      "loss": 0.4977,
+      "step": 670
+    },
+    {
+      "epoch": 1.3892893923789909,
+      "grad_norm": 1.094186544418335,
+      "learning_rate": 2.596132755108191e-05,
+      "loss": 0.5173,
+      "step": 675
+    },
+    {
+      "epoch": 1.399588053553038,
+      "grad_norm": 1.1183247566223145,
+      "learning_rate": 2.5891384808090608e-05,
+      "loss": 0.5067,
+      "step": 680
+    },
+    {
+      "epoch": 1.4098867147270855,
+      "grad_norm": 1.0611745119094849,
+      "learning_rate": 2.5820937578573505e-05,
+      "loss": 0.5012,
+      "step": 685
+    },
+    {
+      "epoch": 1.420185375901133,
+      "grad_norm": 1.0735154151916504,
+      "learning_rate": 2.5749989125630625e-05,
+      "loss": 0.4825,
+      "step": 690
+    },
+    {
+      "epoch": 1.4304840370751801,
+      "grad_norm": 1.0683612823486328,
+      "learning_rate": 2.5678542735578553e-05,
+      "loss": 0.4787,
+      "step": 695
+    },
+    {
+      "epoch": 1.4407826982492276,
+      "grad_norm": 1.1233255863189697,
+      "learning_rate": 2.5606601717798212e-05,
+      "loss": 0.4897,
+      "step": 700
+    },
+    {
+      "epoch": 1.451081359423275,
+      "grad_norm": 1.1237461566925049,
+      "learning_rate": 2.5534169404581576e-05,
+      "loss": 0.4728,
+      "step": 705
+    },
+    {
+      "epoch": 1.4613800205973224,
+      "grad_norm": 1.2569210529327393,
+      "learning_rate": 2.5461249150977304e-05,
+      "loss": 0.4589,
+      "step": 710
+    },
+    {
+      "epoch": 1.4716786817713698,
+      "grad_norm": 1.130018711090088,
+      "learning_rate": 2.5387844334635356e-05,
+      "loss": 0.4912,
+      "step": 715
+    },
+    {
+      "epoch": 1.481977342945417,
+      "grad_norm": 1.1941906213760376,
+      "learning_rate": 2.5313958355650517e-05,
+      "loss": 0.5104,
+      "step": 720
+    },
+    {
+      "epoch": 1.4922760041194645,
+      "grad_norm": 1.155961036682129,
+      "learning_rate": 2.523959463640493e-05,
+      "loss": 0.458,
+      "step": 725
+    },
+    {
+      "epoch": 1.502574665293512,
+      "grad_norm": 1.408888339996338,
+      "learning_rate": 2.516475662140956e-05,
+      "loss": 0.4501,
+      "step": 730
+    },
+    {
+      "epoch": 1.5128733264675591,
+      "grad_norm": 1.20494544506073,
+      "learning_rate": 2.508944777714464e-05,
+      "loss": 0.4649,
+      "step": 735
+    },
+    {
+      "epoch": 1.5231719876416066,
+      "grad_norm": 1.1924827098846436,
+      "learning_rate": 2.5013671591899117e-05,
+      "loss": 0.4326,
+      "step": 740
+    },
+    {
+      "epoch": 1.533470648815654,
+      "grad_norm": 1.1093214750289917,
+      "learning_rate": 2.4937431575609063e-05,
+      "loss": 0.5026,
+      "step": 745
+    },
+    {
+      "epoch": 1.5437693099897012,
+      "grad_norm": 1.1080135107040405,
+      "learning_rate": 2.4860731259695096e-05,
+      "loss": 0.4311,
+      "step": 750
+    },
+    {
+      "epoch": 1.5540679711637488,
+      "grad_norm": 1.1348096132278442,
+      "learning_rate": 2.478357419689883e-05,
+      "loss": 0.4683,
+      "step": 755
+    },
+    {
+      "epoch": 1.564366632337796,
+      "grad_norm": 1.0978420972824097,
+      "learning_rate": 2.4705963961118272e-05,
+      "loss": 0.4527,
+      "step": 760
+    },
+    {
+      "epoch": 1.5746652935118435,
+      "grad_norm": 1.116330623626709,
+      "learning_rate": 2.4627904147242315e-05,
+      "loss": 0.4474,
+      "step": 765
+    },
+    {
+      "epoch": 1.584963954685891,
+      "grad_norm": 1.0585070848464966,
+      "learning_rate": 2.4549398370984195e-05,
+      "loss": 0.4194,
+      "step": 770
+    },
+    {
+      "epoch": 1.5952626158599381,
+      "grad_norm": 1.2468876838684082,
+      "learning_rate": 2.447045026871405e-05,
+      "loss": 0.4709,
+      "step": 775
+    },
+    {
+      "epoch": 1.6055612770339855,
+      "grad_norm": 1.1126863956451416,
+      "learning_rate": 2.439106349729043e-05,
+      "loss": 0.4817,
+      "step": 780
+    },
+    {
+      "epoch": 1.615859938208033,
+      "grad_norm": 1.093266487121582,
+      "learning_rate": 2.431124173389097e-05,
+      "loss": 0.4176,
+      "step": 785
+    },
+    {
+      "epoch": 1.6261585993820802,
+      "grad_norm": 1.2111928462982178,
+      "learning_rate": 2.423098867584202e-05,
+      "loss": 0.4628,
+      "step": 790
+    },
+    {
+      "epoch": 1.6364572605561278,
+      "grad_norm": 1.1940813064575195,
+      "learning_rate": 2.4150308040447416e-05,
+      "loss": 0.4437,
+      "step": 795
+    },
+    {
+      "epoch": 1.646755921730175,
+      "grad_norm": 1.185739278793335,
+      "learning_rate": 2.4069203564816258e-05,
+      "loss": 0.4278,
+      "step": 800
+    },
+    {
+      "epoch": 1.6570545829042225,
+      "grad_norm": 1.0990055799484253,
+      "learning_rate": 2.3987679005689866e-05,
+      "loss": 0.4641,
+      "step": 805
+    },
+    {
+      "epoch": 1.66735324407827,
+      "grad_norm": 1.137831449508667,
+      "learning_rate": 2.3905738139267705e-05,
+      "loss": 0.4336,
+      "step": 810
+    },
+    {
+      "epoch": 1.677651905252317,
+      "grad_norm": 1.1449936628341675,
+      "learning_rate": 2.3823384761032512e-05,
+      "loss": 0.4235,
+      "step": 815
+    },
+    {
+      "epoch": 1.6879505664263645,
+      "grad_norm": 1.1143118143081665,
+      "learning_rate": 2.3740622685574483e-05,
+      "loss": 0.3925,
+      "step": 820
+    },
+    {
+      "epoch": 1.698249227600412,
+      "grad_norm": 1.1763856410980225,
+      "learning_rate": 2.3657455746414576e-05,
+      "loss": 0.3921,
+      "step": 825
+    },
+    {
+      "epoch": 1.7085478887744592,
+      "grad_norm": 1.239435076713562,
+      "learning_rate": 2.3573887795826942e-05,
+      "loss": 0.414,
+      "step": 830
+    },
+    {
+      "epoch": 1.7188465499485068,
+      "grad_norm": 1.1457319259643555,
+      "learning_rate": 2.3489922704660496e-05,
+      "loss": 0.4457,
+      "step": 835
+    },
+    {
+      "epoch": 1.729145211122554,
+      "grad_norm": 1.1365303993225098,
+      "learning_rate": 2.340556436215962e-05,
+      "loss": 0.4283,
+      "step": 840
+    },
+    {
+      "epoch": 1.7394438722966015,
+      "grad_norm": 1.1952276229858398,
+      "learning_rate": 2.3320816675784e-05,
+      "loss": 0.4332,
+      "step": 845
+    },
+    {
+      "epoch": 1.749742533470649,
+      "grad_norm": 1.1562155485153198,
+      "learning_rate": 2.3235683571027656e-05,
+      "loss": 0.4064,
+      "step": 850
+    },
+    {
+      "epoch": 1.760041194644696,
+      "grad_norm": 1.1927721500396729,
+      "learning_rate": 2.3150168991237088e-05,
+      "loss": 0.3924,
+      "step": 855
+    },
+    {
+      "epoch": 1.7703398558187435,
+      "grad_norm": 1.1150102615356445,
+      "learning_rate": 2.306427689742865e-05,
+      "loss": 0.4062,
+      "step": 860
+    },
+    {
+      "epoch": 1.780638516992791,
+      "grad_norm": 1.3249040842056274,
+      "learning_rate": 2.297801126810505e-05,
+      "loss": 0.4417,
+      "step": 865
+    },
+    {
+      "epoch": 1.7909371781668382,
+      "grad_norm": 1.194503903388977,
+      "learning_rate": 2.2891376099071088e-05,
+      "loss": 0.4151,
+      "step": 870
+    },
+    {
+      "epoch": 1.8012358393408858,
+      "grad_norm": 1.1773724555969238,
+      "learning_rate": 2.2804375403248547e-05,
+      "loss": 0.4218,
+      "step": 875
+    },
+    {
+      "epoch": 1.811534500514933,
+      "grad_norm": 1.1841347217559814,
+      "learning_rate": 2.2717013210490352e-05,
+      "loss": 0.386,
+      "step": 880
+    },
+    {
+      "epoch": 1.8218331616889805,
+      "grad_norm": 1.1204098463058472,
+      "learning_rate": 2.2629293567393854e-05,
+      "loss": 0.3865,
+      "step": 885
+    },
+    {
+      "epoch": 1.832131822863028,
+      "grad_norm": 1.115275263786316,
+      "learning_rate": 2.2541220537113456e-05,
+      "loss": 0.3657,
+      "step": 890
+    },
+    {
+      "epoch": 1.842430484037075,
+      "grad_norm": 1.0938341617584229,
+      "learning_rate": 2.2452798199172358e-05,
+      "loss": 0.3616,
+      "step": 895
+    },
+    {
+      "epoch": 1.8527291452111225,
+      "grad_norm": 1.2680145502090454,
+      "learning_rate": 2.2364030649273624e-05,
+      "loss": 0.419,
+      "step": 900
+    },
+    {
+      "epoch": 1.86302780638517,
+      "grad_norm": 1.057039499282837,
+      "learning_rate": 2.227492199911045e-05,
+      "loss": 0.4071,
+      "step": 905
+    },
+    {
+      "epoch": 1.8733264675592172,
+      "grad_norm": 1.2678941488265991,
+      "learning_rate": 2.2185476376175718e-05,
+      "loss": 0.3886,
+      "step": 910
+    },
+    {
+      "epoch": 1.8836251287332648,
+      "grad_norm": 1.180358648300171,
+      "learning_rate": 2.2095697923570835e-05,
+      "loss": 0.3771,
+      "step": 915
+    },
+    {
+      "epoch": 1.893923789907312,
+      "grad_norm": 1.2452682256698608,
+      "learning_rate": 2.200559079981377e-05,
+      "loss": 0.3927,
+      "step": 920
+    },
+    {
+      "epoch": 1.9042224510813595,
+      "grad_norm": 1.1591405868530273,
+      "learning_rate": 2.191515917864651e-05,
+      "loss": 0.3586,
+      "step": 925
+    },
+    {
+      "epoch": 1.9145211122554069,
+      "grad_norm": 1.1930902004241943,
+      "learning_rate": 2.182440724884165e-05,
+      "loss": 0.3766,
+      "step": 930
+    },
+    {
+      "epoch": 1.924819773429454,
+      "grad_norm": 1.2268239259719849,
+      "learning_rate": 2.1733339214008428e-05,
+      "loss": 0.3623,
+      "step": 935
+    },
+    {
+      "epoch": 1.9351184346035015,
+      "grad_norm": 1.1663581132888794,
+      "learning_rate": 2.1641959292397998e-05,
+      "loss": 0.3534,
+      "step": 940
+    },
+    {
+      "epoch": 1.945417095777549,
+      "grad_norm": 1.1473199129104614,
+      "learning_rate": 2.1550271716708038e-05,
+      "loss": 0.3578,
+      "step": 945
+    },
+    {
+      "epoch": 1.9557157569515962,
+      "grad_norm": 1.34954833984375,
+      "learning_rate": 2.14582807338867e-05,
+      "loss": 0.3931,
+      "step": 950
+    },
+    {
+      "epoch": 1.9660144181256438,
+      "grad_norm": 1.277458667755127,
+      "learning_rate": 2.1365990604935865e-05,
+      "loss": 0.3851,
+      "step": 955
+    },
+    {
+      "epoch": 1.976313079299691,
+      "grad_norm": 1.1497230529785156,
+      "learning_rate": 2.127340560471383e-05,
+      "loss": 0.3496,
+      "step": 960
+    },
+    {
+      "epoch": 1.9866117404737385,
+      "grad_norm": 1.1924526691436768,
+      "learning_rate": 2.118053002173723e-05,
+      "loss": 0.3401,
+      "step": 965
+    },
+    {
+      "epoch": 1.9969104016477859,
+      "grad_norm": 1.373230218887329,
+      "learning_rate": 2.1087368157982456e-05,
+      "loss": 0.3393,
+      "step": 970
+    },
+    {
+      "epoch": 2.0061791967044282,
+      "grad_norm": 1.1674820184707642,
+      "learning_rate": 2.0993924328686353e-05,
+      "loss": 0.309,
+      "step": 975
+    },
+    {
+      "epoch": 2.016477857878476,
+      "grad_norm": 1.2460054159164429,
+      "learning_rate": 2.090020286214633e-05,
+      "loss": 0.2843,
+      "step": 980
+    },
+    {
+      "epoch": 2.026776519052523,
+      "grad_norm": 1.200320839881897,
+      "learning_rate": 2.0806208099519916e-05,
+      "loss": 0.3074,
+      "step": 985
+    },
+    {
+      "epoch": 2.0370751802265707,
+      "grad_norm": 1.0831384658813477,
+      "learning_rate": 2.0711944394623642e-05,
+      "loss": 0.2826,
+      "step": 990
+    },
+    {
+      "epoch": 2.047373841400618,
+      "grad_norm": 1.1945862770080566,
+      "learning_rate": 2.0617416113731385e-05,
+      "loss": 0.2629,
+      "step": 995
+    },
+    {
+      "epoch": 2.057672502574665,
+      "grad_norm": 1.1249759197235107,
+      "learning_rate": 2.0522627635372126e-05,
+      "loss": 0.2838,
+      "step": 1000
+    },
+    {
+      "epoch": 2.067971163748713,
+      "grad_norm": 1.2407805919647217,
+      "learning_rate": 2.0427583350127134e-05,
+      "loss": 0.2754,
+      "step": 1005
+    },
+    {
+      "epoch": 2.07826982492276,
+      "grad_norm": 1.1751117706298828,
+      "learning_rate": 2.0332287660426598e-05,
+      "loss": 0.2892,
+      "step": 1010
+    },
+    {
+      "epoch": 2.0885684860968072,
+      "grad_norm": 1.1912750005722046,
+      "learning_rate": 2.0236744980345704e-05,
+      "loss": 0.2832,
+      "step": 1015
+    },
+    {
+      "epoch": 2.098867147270855,
+      "grad_norm": 1.0413898229599,
+      "learning_rate": 2.014095973540017e-05,
+      "loss": 0.2862,
+      "step": 1020
+    },
+    {
+      "epoch": 2.109165808444902,
+      "grad_norm": 1.1047124862670898,
+      "learning_rate": 2.0044936362341288e-05,
+      "loss": 0.2724,
+      "step": 1025
+    },
+    {
+      "epoch": 2.1194644696189497,
+      "grad_norm": 1.1833388805389404,
+      "learning_rate": 1.9948679308950358e-05,
+      "loss": 0.2744,
+      "step": 1030
+    },
+    {
+      "epoch": 2.129763130792997,
+      "grad_norm": 1.25725519657135,
+      "learning_rate": 1.985219303383273e-05,
+      "loss": 0.2705,
+      "step": 1035
+    },
+    {
+      "epoch": 2.140061791967044,
+      "grad_norm": 1.1427794694900513,
+      "learning_rate": 1.9755482006211237e-05,
+      "loss": 0.292,
+      "step": 1040
+    },
+    {
+      "epoch": 2.150360453141092,
+      "grad_norm": 1.1591993570327759,
+      "learning_rate": 1.96585507057192e-05,
+      "loss": 0.2725,
+      "step": 1045
+    },
+    {
+      "epoch": 2.160659114315139,
+      "grad_norm": 1.155144453048706,
+      "learning_rate": 1.9561403622192948e-05,
+      "loss": 0.2865,
+      "step": 1050
+    },
+    {
+      "epoch": 2.1709577754891862,
+      "grad_norm": 1.126906156539917,
+      "learning_rate": 1.9464045255463797e-05,
+      "loss": 0.2676,
+      "step": 1055
+    },
+    {
+      "epoch": 2.181256436663234,
+      "grad_norm": 1.245829463005066,
+      "learning_rate": 1.936648011514969e-05,
+      "loss": 0.2763,
+      "step": 1060
+    },
+    {
+      "epoch": 2.191555097837281,
+      "grad_norm": 1.1624958515167236,
+      "learning_rate": 1.9268712720446255e-05,
+      "loss": 0.2503,
+      "step": 1065
+    },
+    {
+      "epoch": 2.2018537590113283,
+      "grad_norm": 1.1856951713562012,
+      "learning_rate": 1.9170747599917507e-05,
+      "loss": 0.2667,
+      "step": 1070
+    },
+    {
+      "epoch": 2.212152420185376,
+      "grad_norm": 1.1675992012023926,
+      "learning_rate": 1.9072589291286073e-05,
+      "loss": 0.272,
+      "step": 1075
+    },
+    {
+      "epoch": 2.222451081359423,
+      "grad_norm": 1.3212230205535889,
+      "learning_rate": 1.897424234122302e-05,
+      "loss": 0.2993,
+      "step": 1080
+    },
+    {
+      "epoch": 2.232749742533471,
+      "grad_norm": 1.3248764276504517,
+      "learning_rate": 1.8875711305137222e-05,
+      "loss": 0.2914,
+      "step": 1085
+    },
+    {
+      "epoch": 2.243048403707518,
+      "grad_norm": 1.18609619140625,
+      "learning_rate": 1.8777000746964397e-05,
+      "loss": 0.2416,
+      "step": 1090
+    },
+    {
+      "epoch": 2.253347064881565,
+      "grad_norm": 1.0249930620193481,
+      "learning_rate": 1.8678115238955686e-05,
+      "loss": 0.2631,
+      "step": 1095
+    },
+    {
+      "epoch": 2.263645726055613,
+      "grad_norm": 1.2961682081222534,
+      "learning_rate": 1.8579059361465863e-05,
+      "loss": 0.265,
+      "step": 1100
+    },
+    {
+      "epoch": 2.27394438722966,
+      "grad_norm": 1.1270440816879272,
+      "learning_rate": 1.8479837702741166e-05,
+      "loss": 0.2721,
+      "step": 1105
+    },
+    {
+      "epoch": 2.2842430484037077,
+      "grad_norm": 1.1455341577529907,
+      "learning_rate": 1.8380454858706798e-05,
+      "loss": 0.2722,
+      "step": 1110
+    },
+    {
+      "epoch": 2.294541709577755,
+      "grad_norm": 1.1415375471115112,
+      "learning_rate": 1.8280915432754033e-05,
+      "loss": 0.2728,
+      "step": 1115
+    },
+    {
+      "epoch": 2.304840370751802,
+      "grad_norm": 1.3010354042053223,
+      "learning_rate": 1.8181224035526968e-05,
+      "loss": 0.2744,
+      "step": 1120
+    },
+    {
+      "epoch": 2.31513903192585,
+      "grad_norm": 1.1614693403244019,
+      "learning_rate": 1.8081385284708995e-05,
+      "loss": 0.2554,
+      "step": 1125
+    },
+    {
+      "epoch": 2.325437693099897,
+      "grad_norm": 1.1690951585769653,
+      "learning_rate": 1.798140380480888e-05,
+      "loss": 0.2463,
+      "step": 1130
+    },
+    {
+      "epoch": 2.335736354273944,
+      "grad_norm": 1.2176849842071533,
+      "learning_rate": 1.7881284226946558e-05,
+      "loss": 0.2519,
+      "step": 1135
+    },
+    {
+      "epoch": 2.346035015447992,
+      "grad_norm": 1.245349407196045,
+      "learning_rate": 1.778103118863866e-05,
+      "loss": 0.2511,
+      "step": 1140
+    },
+    {
+      "epoch": 2.356333676622039,
+      "grad_norm": 1.1520276069641113,
+      "learning_rate": 1.768064933358365e-05,
+      "loss": 0.2561,
+      "step": 1145
+    },
+    {
+      "epoch": 2.3666323377960863,
+      "grad_norm": 1.1642286777496338,
+      "learning_rate": 1.7580143311446766e-05,
+      "loss": 0.2368,
+      "step": 1150
+    },
+    {
+      "epoch": 2.376930998970134,
+      "grad_norm": 1.2488750219345093,
+      "learning_rate": 1.7479517777644632e-05,
+      "loss": 0.2372,
+      "step": 1155
+    },
+    {
+      "epoch": 2.387229660144181,
+      "grad_norm": 1.1384341716766357,
+      "learning_rate": 1.7378777393129635e-05,
+      "loss": 0.2389,
+      "step": 1160
+    },
+    {
+      "epoch": 2.397528321318229,
+      "grad_norm": 1.2213577032089233,
+      "learning_rate": 1.7277926824174012e-05,
+      "loss": 0.2443,
+      "step": 1165
+    },
+    {
+      "epoch": 2.407826982492276,
+      "grad_norm": 1.1850676536560059,
+      "learning_rate": 1.7176970742153717e-05,
+      "loss": 0.2542,
+      "step": 1170
+    },
+    {
+      "epoch": 2.418125643666323,
+      "grad_norm": 1.189429759979248,
+      "learning_rate": 1.7075913823332067e-05,
+      "loss": 0.2493,
+      "step": 1175
+    },
+    {
+      "epoch": 2.428424304840371,
+      "grad_norm": 1.1571177244186401,
+      "learning_rate": 1.6974760748643084e-05,
+      "loss": 0.2105,
+      "step": 1180
+    },
+    {
+      "epoch": 2.438722966014418,
+      "grad_norm": 1.1830145120620728,
+      "learning_rate": 1.6873516203474736e-05,
+      "loss": 0.2361,
+      "step": 1185
+    },
+    {
+      "epoch": 2.4490216271884657,
+      "grad_norm": 1.3023966550827026,
+      "learning_rate": 1.677218487745187e-05,
+      "loss": 0.2498,
+      "step": 1190
+    },
+    {
+      "epoch": 2.459320288362513,
+      "grad_norm": 1.2320756912231445,
+      "learning_rate": 1.6670771464219002e-05,
+      "loss": 0.2672,
+      "step": 1195
+    },
+    {
+      "epoch": 2.46961894953656,
+      "grad_norm": 1.1318459510803223,
+      "learning_rate": 1.6569280661222924e-05,
+      "loss": 0.2394,
+      "step": 1200
+    },
+    {
+      "epoch": 2.479917610710608,
+      "grad_norm": 1.1684930324554443,
+      "learning_rate": 1.646771716949509e-05,
+      "loss": 0.2377,
+      "step": 1205
+    },
+    {
+      "epoch": 2.490216271884655,
+      "grad_norm": 1.1254520416259766,
+      "learning_rate": 1.636608569343387e-05,
+      "loss": 0.2391,
+      "step": 1210
+    },
+    {
+      "epoch": 2.500514933058702,
+      "grad_norm": 1.2523401975631714,
+      "learning_rate": 1.626439094058669e-05,
+      "loss": 0.2143,
+      "step": 1215
+    },
+    {
+      "epoch": 2.51081359423275,
+      "grad_norm": 1.1195322275161743,
+      "learning_rate": 1.6162637621431906e-05,
+      "loss": 0.2394,
+      "step": 1220
+    },
+    {
+      "epoch": 2.521112255406797,
+      "grad_norm": 1.1573140621185303,
+      "learning_rate": 1.606083044916067e-05,
+      "loss": 0.2663,
+      "step": 1225
+    },
+    {
+      "epoch": 2.5314109165808443,
+      "grad_norm": 1.1650135517120361,
+      "learning_rate": 1.595897413945859e-05,
+      "loss": 0.2473,
+      "step": 1230
+    },
+    {
+      "epoch": 2.541709577754892,
+      "grad_norm": 1.1126735210418701,
+      "learning_rate": 1.5857073410287317e-05,
+      "loss": 0.2228,
+      "step": 1235
+    },
+    {
+      "epoch": 2.552008238928939,
+      "grad_norm": 1.180673360824585,
+      "learning_rate": 1.5755132981666004e-05,
+      "loss": 0.2214,
+      "step": 1240
+    },
+    {
+      "epoch": 2.5623069001029863,
+      "grad_norm": 1.1609716415405273,
+      "learning_rate": 1.5653157575452656e-05,
+      "loss": 0.2447,
+      "step": 1245
+    },
+    {
+      "epoch": 2.572605561277034,
+      "grad_norm": 1.1662496328353882,
+      "learning_rate": 1.555115191512546e-05,
+      "loss": 0.2241,
+      "step": 1250
+    },
+    {
+      "epoch": 2.582904222451081,
+      "grad_norm": 1.25954008102417,
+      "learning_rate": 1.5449120725563932e-05,
+      "loss": 0.2011,
+      "step": 1255
+    },
+    {
+      "epoch": 2.593202883625129,
+      "grad_norm": 1.3918417692184448,
+      "learning_rate": 1.5347068732830134e-05,
+      "loss": 0.2467,
+      "step": 1260
+    },
+    {
+      "epoch": 2.603501544799176,
+      "grad_norm": 1.114081621170044,
+      "learning_rate": 1.5245000663949699e-05,
+      "loss": 0.2144,
+      "step": 1265
+    },
+    {
+      "epoch": 2.6138002059732237,
+      "grad_norm": 1.2161543369293213,
+      "learning_rate": 1.5142921246692909e-05,
+      "loss": 0.2276,
+      "step": 1270
+    },
+    {
+      "epoch": 2.624098867147271,
+      "grad_norm": 1.145684003829956,
+      "learning_rate": 1.5040835209355715e-05,
+      "loss": 0.2143,
+      "step": 1275
+    },
+    {
+      "epoch": 2.634397528321318,
+      "grad_norm": 1.1903437376022339,
+      "learning_rate": 1.4938747280540692e-05,
+      "loss": 0.2264,
+      "step": 1280
+    },
+    {
+      "epoch": 2.644696189495366,
+      "grad_norm": 1.2034133672714233,
+      "learning_rate": 1.4836662188938028e-05,
+      "loss": 0.2368,
+      "step": 1285
+    },
+    {
+      "epoch": 2.654994850669413,
+      "grad_norm": 1.1376266479492188,
+      "learning_rate": 1.4734584663106506e-05,
+      "loss": 0.2265,
+      "step": 1290
+    },
+    {
+      "epoch": 2.66529351184346,
+      "grad_norm": 1.1816309690475464,
+      "learning_rate": 1.4632519431254446e-05,
+      "loss": 0.2248,
+      "step": 1295
+    },
+    {
+      "epoch": 2.675592173017508,
+      "grad_norm": 1.2210973501205444,
+      "learning_rate": 1.453047122102073e-05,
+      "loss": 0.233,
+      "step": 1300
+    },
+    {
+      "epoch": 2.685890834191555,
+      "grad_norm": 1.2578504085540771,
+      "learning_rate": 1.44284447592558e-05,
+      "loss": 0.2171,
+      "step": 1305
+    },
+    {
+      "epoch": 2.6961894953656023,
+      "grad_norm": 1.2888798713684082,
+      "learning_rate": 1.4326444771802699e-05,
+      "loss": 0.2057,
+      "step": 1310
+    },
+    {
+      "epoch": 2.70648815653965,
+      "grad_norm": 1.0893986225128174,
+      "learning_rate": 1.4224475983278204e-05,
+      "loss": 0.2133,
+      "step": 1315
+    },
+    {
+      "epoch": 2.716786817713697,
+      "grad_norm": 1.2263572216033936,
+      "learning_rate": 1.412254311685395e-05,
+      "loss": 0.2297,
+      "step": 1320
+    },
+    {
+      "epoch": 2.7270854788877443,
+      "grad_norm": 1.117436408996582,
+      "learning_rate": 1.4020650894037673e-05,
+      "loss": 0.2203,
+      "step": 1325
+    },
+    {
+      "epoch": 2.737384140061792,
+      "grad_norm": 1.1971065998077393,
+      "learning_rate": 1.3918804034454521e-05,
+      "loss": 0.2009,
+      "step": 1330
+    },
+    {
+      "epoch": 2.747682801235839,
+      "grad_norm": 1.2015117406845093,
+      "learning_rate": 1.3817007255628402e-05,
+      "loss": 0.2123,
+      "step": 1335
+    },
+    {
+      "epoch": 2.757981462409887,
+      "grad_norm": 1.2726963758468628,
+      "learning_rate": 1.3715265272763515e-05,
+      "loss": 0.2026,
+      "step": 1340
+    },
+    {
+      "epoch": 2.768280123583934,
+      "grad_norm": 1.189307451248169,
+      "learning_rate": 1.3613582798525903e-05,
+      "loss": 0.2155,
+      "step": 1345
+    },
+    {
+      "epoch": 2.7785787847579817,
+      "grad_norm": 1.0487000942230225,
+      "learning_rate": 1.3511964542825206e-05,
+      "loss": 0.1951,
+      "step": 1350
+    },
+    {
+      "epoch": 2.788877445932029,
+      "grad_norm": 1.1836564540863037,
+      "learning_rate": 1.341041521259645e-05,
+      "loss": 0.2026,
+      "step": 1355
+    },
+    {
+      "epoch": 2.799176107106076,
+      "grad_norm": 1.145855188369751,
+      "learning_rate": 1.330893951158207e-05,
+      "loss": 0.2026,
+      "step": 1360
+    },
+    {
+      "epoch": 2.8094747682801238,
+      "grad_norm": 1.2061723470687866,
+      "learning_rate": 1.3207542140114007e-05,
+      "loss": 0.1792,
+      "step": 1365
+    },
+    {
+      "epoch": 2.819773429454171,
+      "grad_norm": 1.2622625827789307,
+      "learning_rate": 1.310622779489598e-05,
+      "loss": 0.2169,
+      "step": 1370
+    },
+    {
+      "epoch": 2.830072090628218,
+      "grad_norm": 1.1733144521713257,
+      "learning_rate": 1.3005001168785973e-05,
+      "loss": 0.1994,
+      "step": 1375
+    },
+    {
+      "epoch": 2.840370751802266,
+      "grad_norm": 1.225419282913208,
+      "learning_rate": 1.290386695057882e-05,
+      "loss": 0.1817,
+      "step": 1380
+    },
+    {
+      "epoch": 2.850669412976313,
+      "grad_norm": 1.2059926986694336,
+      "learning_rate": 1.2802829824789054e-05,
+      "loss": 0.202,
+      "step": 1385
+    },
+    {
+      "epoch": 2.8609680741503603,
+      "grad_norm": 1.2589644193649292,
+      "learning_rate": 1.2701894471433913e-05,
+      "loss": 0.2131,
+      "step": 1390
+    },
+    {
+      "epoch": 2.871266735324408,
+      "grad_norm": 1.2112151384353638,
+      "learning_rate": 1.2601065565816541e-05,
+      "loss": 0.176,
+      "step": 1395
+    },
+    {
+      "epoch": 2.881565396498455,
+      "grad_norm": 1.2503776550292969,
+      "learning_rate": 1.2500347778309463e-05,
+      "loss": 0.1838,
+      "step": 1400
+    },
+    {
+      "epoch": 2.8918640576725023,
+      "grad_norm": 1.11964750289917,
+      "learning_rate": 1.2399745774138226e-05,
+      "loss": 0.2135,
+      "step": 1405
+    },
+    {
+      "epoch": 2.90216271884655,
+      "grad_norm": 1.1664073467254639,
+      "learning_rate": 1.2299264213165329e-05,
+      "loss": 0.2044,
+      "step": 1410
+    },
+    {
+      "epoch": 2.912461380020597,
+      "grad_norm": 1.0924075841903687,
+      "learning_rate": 1.2198907749674355e-05,
+      "loss": 0.1801,
+      "step": 1415
+    },
+    {
+      "epoch": 2.922760041194645,
+      "grad_norm": 1.1321125030517578,
+      "learning_rate": 1.2098681032154417e-05,
+      "loss": 0.1783,
+      "step": 1420
+    },
+    {
+      "epoch": 2.933058702368692,
+      "grad_norm": 1.2050647735595703,
+      "learning_rate": 1.1998588703084815e-05,
+      "loss": 0.1841,
+      "step": 1425
+    },
+    {
+      "epoch": 2.9433573635427397,
+      "grad_norm": 1.1679717302322388,
+      "learning_rate": 1.1898635398720003e-05,
+      "loss": 0.1824,
+      "step": 1430
+    },
+    {
+      "epoch": 2.953656024716787,
+      "grad_norm": 1.1852076053619385,
+      "learning_rate": 1.1798825748874837e-05,
+      "loss": 0.1935,
+      "step": 1435
+    },
+    {
+      "epoch": 2.963954685890834,
+      "grad_norm": 1.0744694471359253,
+      "learning_rate": 1.1699164376710146e-05,
+      "loss": 0.1908,
+      "step": 1440
+    },
+    {
+      "epoch": 2.9742533470648818,
+      "grad_norm": 1.2433335781097412,
+      "learning_rate": 1.1599655898518544e-05,
+      "loss": 0.1792,
+      "step": 1445
+    },
+    {
+      "epoch": 2.984552008238929,
+      "grad_norm": 1.1401877403259277,
+      "learning_rate": 1.1500304923510658e-05,
+      "loss": 0.176,
+      "step": 1450
+    },
+    {
+      "epoch": 2.994850669412976,
+      "grad_norm": 1.1364681720733643,
+      "learning_rate": 1.1401116053601584e-05,
+      "loss": 0.171,
+      "step": 1455
+    },
+    {
+      "epoch": 3.004119464469619,
+      "grad_norm": 0.9721819162368774,
+      "learning_rate": 1.1302093883197748e-05,
+      "loss": 0.1583,
+      "step": 1460
+    },
+    {
+      "epoch": 3.014418125643666,
+      "grad_norm": 1.1144256591796875,
+      "learning_rate": 1.1203242998984098e-05,
+      "loss": 0.1346,
+      "step": 1465
+    },
+    {
+      "epoch": 3.024716786817714,
+      "grad_norm": 1.1284878253936768,
+      "learning_rate": 1.1104567979711636e-05,
+      "loss": 0.1356,
+      "step": 1470
+    },
+    {
+      "epoch": 3.035015447991761,
+      "grad_norm": 0.9694728851318359,
+      "learning_rate": 1.1006073395985345e-05,
+      "loss": 0.1474,
+      "step": 1475
+    },
+    {
+      "epoch": 3.0453141091658082,
+      "grad_norm": 1.1852731704711914,
+      "learning_rate": 1.0907763810052478e-05,
+      "loss": 0.1479,
+      "step": 1480
+    },
+    {
+      "epoch": 3.055612770339856,
+      "grad_norm": 1.0727461576461792,
+      "learning_rate": 1.0809643775591215e-05,
+      "loss": 0.1506,
+      "step": 1485
+    },
+    {
+      "epoch": 3.065911431513903,
+      "grad_norm": 1.1076802015304565,
+      "learning_rate": 1.0711717837499775e-05,
+      "loss": 0.1415,
+      "step": 1490
+    },
+    {
+      "epoch": 3.0762100926879508,
+      "grad_norm": 1.1464875936508179,
+      "learning_rate": 1.0613990531685854e-05,
+      "loss": 0.1282,
+      "step": 1495
+    },
+    {
+      "epoch": 3.086508753861998,
+      "grad_norm": 1.0602517127990723,
+      "learning_rate": 1.0516466384856564e-05,
+      "loss": 0.1393,
+      "step": 1500
+    },
+    {
+      "epoch": 3.096807415036045,
+      "grad_norm": 0.9436390399932861,
+      "learning_rate": 1.0419149914308723e-05,
+      "loss": 0.127,
+      "step": 1505
+    },
+    {
+      "epoch": 3.107106076210093,
+      "grad_norm": 1.116821527481079,
+      "learning_rate": 1.0322045627719643e-05,
+      "loss": 0.1447,
+      "step": 1510
+    },
+    {
+      "epoch": 3.11740473738414,
+      "grad_norm": 1.014852523803711,
+      "learning_rate": 1.0225158022938318e-05,
+      "loss": 0.1471,
+      "step": 1515
+    },
+    {
+      "epoch": 3.1277033985581872,
+      "grad_norm": 1.1146396398544312,
+      "learning_rate": 1.0128491587777082e-05,
+      "loss": 0.1509,
+      "step": 1520
+    },
+    {
+      "epoch": 3.138002059732235,
+      "grad_norm": 1.0058327913284302,
+      "learning_rate": 1.003205079980375e-05,
+      "loss": 0.1269,
+      "step": 1525
+    },
+    {
+      "epoch": 3.148300720906282,
+      "grad_norm": 1.0287681818008423,
+      "learning_rate": 9.9358401261342e-06,
+      "loss": 0.1472,
+      "step": 1530
+    },
+    {
+      "epoch": 3.1585993820803298,
+      "grad_norm": 1.0611659288406372,
+      "learning_rate": 9.83986402322548e-06,
+      "loss": 0.1666,
+      "step": 1535
+    },
+    {
+      "epoch": 3.168898043254377,
+      "grad_norm": 1.1708921194076538,
+      "learning_rate": 9.744126936669352e-06,
+      "loss": 0.1392,
+      "step": 1540
+    },
+    {
+      "epoch": 3.179196704428424,
+      "grad_norm": 0.8749961853027344,
+      "learning_rate": 9.648633300986418e-06,
+      "loss": 0.126,
+      "step": 1545
+    },
+    {
+      "epoch": 3.189495365602472,
+      "grad_norm": 1.0773943662643433,
+      "learning_rate": 9.553387539420678e-06,
+      "loss": 0.127,
+      "step": 1550
+    },
+    {
+      "epoch": 3.199794026776519,
+      "grad_norm": 0.9732959270477295,
+      "learning_rate": 9.458394063734651e-06,
+      "loss": 0.1418,
+      "step": 1555
+    },
+    {
+      "epoch": 3.2100926879505662,
+      "grad_norm": 0.9006505608558655,
+      "learning_rate": 9.36365727400504e-06,
+      "loss": 0.1276,
+      "step": 1560
+    },
+    {
+      "epoch": 3.220391349124614,
+      "grad_norm": 1.2459700107574463,
+      "learning_rate": 9.26918155841891e-06,
+      "loss": 0.1296,
+      "step": 1565
+    },
+    {
+      "epoch": 3.230690010298661,
+      "grad_norm": 1.1395372152328491,
+      "learning_rate": 9.174971293070417e-06,
+      "loss": 0.1337,
+      "step": 1570
+    },
+    {
+      "epoch": 3.2409886714727087,
+      "grad_norm": 0.9972518682479858,
+      "learning_rate": 9.08103084175815e-06,
+      "loss": 0.1262,
+      "step": 1575
+    },
+    {
+      "epoch": 3.251287332646756,
+      "grad_norm": 0.9902070164680481,
+      "learning_rate": 8.987364555782945e-06,
+      "loss": 0.1249,
+      "step": 1580
+    },
+    {
+      "epoch": 3.261585993820803,
+      "grad_norm": 1.1024582386016846,
+      "learning_rate": 8.893976773746363e-06,
+      "loss": 0.1365,
+      "step": 1585
+    },
+    {
+      "epoch": 3.271884654994851,
+      "grad_norm": 0.9224875569343567,
+      "learning_rate": 8.800871821349739e-06,
+      "loss": 0.1347,
+      "step": 1590
+    },
+    {
+      "epoch": 3.282183316168898,
+      "grad_norm": 1.1592514514923096,
+      "learning_rate": 8.708054011193794e-06,
+      "loss": 0.1479,
+      "step": 1595
+    },
+    {
+      "epoch": 3.2924819773429452,
+      "grad_norm": 1.105877161026001,
+      "learning_rate": 8.61552764257887e-06,
+      "loss": 0.1268,
+      "step": 1600
+    },
+    {
+      "epoch": 3.302780638516993,
+      "grad_norm": 1.0494933128356934,
+      "learning_rate": 8.52329700130583e-06,
+      "loss": 0.1363,
+      "step": 1605
+    },
+    {
+      "epoch": 3.31307929969104,
+      "grad_norm": 1.1177783012390137,
+      "learning_rate": 8.431366359477499e-06,
+      "loss": 0.1196,
+      "step": 1610
+    },
+    {
+      "epoch": 3.3233779608650877,
+      "grad_norm": 1.217367172241211,
+      "learning_rate": 8.339739975300779e-06,
+      "loss": 0.15,
+      "step": 1615
+    },
+    {
+      "epoch": 3.333676622039135,
+      "grad_norm": 0.9796255826950073,
+      "learning_rate": 8.248422092889449e-06,
+      "loss": 0.1213,
+      "step": 1620
+    },
+    {
+      "epoch": 3.343975283213182,
+      "grad_norm": 1.0670825242996216,
+      "learning_rate": 8.15741694206754e-06,
+      "loss": 0.1346,
+      "step": 1625
+    },
+    {
+      "epoch": 3.35427394438723,
+      "grad_norm": 1.0460721254348755,
+      "learning_rate": 8.06672873817343e-06,
+      "loss": 0.1149,
+      "step": 1630
+    },
+    {
+      "epoch": 3.364572605561277,
+      "grad_norm": 0.9851893782615662,
+      "learning_rate": 7.976361681864586e-06,
+      "loss": 0.1361,
+      "step": 1635
+    },
+    {
+      "epoch": 3.3748712667353242,
+      "grad_norm": 0.9853783249855042,
+      "learning_rate": 7.886319958922992e-06,
+      "loss": 0.1202,
+      "step": 1640
+    },
+    {
+      "epoch": 3.385169927909372,
+      "grad_norm": 1.0467960834503174,
+      "learning_rate": 7.796607740061265e-06,
+      "loss": 0.1274,
+      "step": 1645
+    },
+    {
+      "epoch": 3.395468589083419,
+      "grad_norm": 1.2147797346115112,
+      "learning_rate": 7.707229180729449e-06,
+      "loss": 0.1361,
+      "step": 1650
+    },
+    {
+      "epoch": 3.4057672502574663,
+      "grad_norm": 1.0219812393188477,
+      "learning_rate": 7.618188420922568e-06,
+      "loss": 0.1258,
+      "step": 1655
+    },
+    {
+      "epoch": 3.416065911431514,
+      "grad_norm": 0.8363818526268005,
+      "learning_rate": 7.52948958498885e-06,
+      "loss": 0.1406,
+      "step": 1660
+    },
+    {
+      "epoch": 3.426364572605561,
+      "grad_norm": 1.0719374418258667,
+      "learning_rate": 7.441136781438677e-06,
+      "loss": 0.1226,
+      "step": 1665
+    },
+    {
+      "epoch": 3.436663233779609,
+      "grad_norm": 1.0610411167144775,
+      "learning_rate": 7.353134102754289e-06,
+      "loss": 0.1423,
+      "step": 1670
+    },
+    {
+      "epoch": 3.446961894953656,
+      "grad_norm": 0.9661740660667419,
+      "learning_rate": 7.265485625200221e-06,
+      "loss": 0.1507,
+      "step": 1675
+    },
+    {
+      "epoch": 3.457260556127703,
+      "grad_norm": 1.030022382736206,
+      "learning_rate": 7.1781954086344976e-06,
+      "loss": 0.1307,
+      "step": 1680
+    },
+    {
+      "epoch": 3.467559217301751,
+      "grad_norm": 0.9682163596153259,
+      "learning_rate": 7.091267496320553e-06,
+      "loss": 0.1179,
+      "step": 1685
+    },
+    {
+      "epoch": 3.477857878475798,
+      "grad_norm": 1.1524693965911865,
+      "learning_rate": 7.004705914739986e-06,
+      "loss": 0.1246,
+      "step": 1690
+    },
+    {
+      "epoch": 3.4881565396498457,
+      "grad_norm": 1.1329081058502197,
+      "learning_rate": 6.9185146734060465e-06,
+      "loss": 0.128,
+      "step": 1695
+    },
+    {
+      "epoch": 3.498455200823893,
+      "grad_norm": 0.9930932521820068,
+      "learning_rate": 6.832697764677881e-06,
+      "loss": 0.1195,
+      "step": 1700
+    },
+    {
+      "epoch": 3.50875386199794,
+      "grad_norm": 1.0590791702270508,
+      "learning_rate": 6.747259163575648e-06,
+      "loss": 0.1345,
+      "step": 1705
+    },
+    {
+      "epoch": 3.519052523171988,
+      "grad_norm": 1.0473939180374146,
+      "learning_rate": 6.662202827596386e-06,
+      "loss": 0.1268,
+      "step": 1710
+    },
+    {
+      "epoch": 3.529351184346035,
+      "grad_norm": 1.04848313331604,
+      "learning_rate": 6.577532696530678e-06,
+      "loss": 0.1273,
+      "step": 1715
+    },
+    {
+      "epoch": 3.539649845520082,
+      "grad_norm": 0.8624801635742188,
+      "learning_rate": 6.4932526922801935e-06,
+      "loss": 0.1196,
+      "step": 1720
+    },
+    {
+      "epoch": 3.54994850669413,
+      "grad_norm": 1.0373711585998535,
+      "learning_rate": 6.409366718676031e-06,
+      "loss": 0.1225,
+      "step": 1725
+    },
+    {
+      "epoch": 3.560247167868177,
+      "grad_norm": 0.9186179041862488,
+      "learning_rate": 6.325878661297845e-06,
+      "loss": 0.1207,
+      "step": 1730
+    },
+    {
+      "epoch": 3.5705458290422243,
+      "grad_norm": 0.9465278387069702,
+      "learning_rate": 6.242792387293919e-06,
+      "loss": 0.1211,
+      "step": 1735
+    },
+    {
+      "epoch": 3.580844490216272,
+      "grad_norm": 0.9691430926322937,
+      "learning_rate": 6.1601117452020125e-06,
+      "loss": 0.1135,
+      "step": 1740
+    },
+    {
+      "epoch": 3.591143151390319,
+      "grad_norm": 0.9645276069641113,
+      "learning_rate": 6.077840564771116e-06,
+      "loss": 0.1188,
+      "step": 1745
+    },
+    {
+      "epoch": 3.601441812564367,
+      "grad_norm": 1.0281277894973755,
+      "learning_rate": 5.995982656784021e-06,
+      "loss": 0.1231,
+      "step": 1750
+    },
+    {
+      "epoch": 3.611740473738414,
+      "grad_norm": 0.992897629737854,
+      "learning_rate": 5.914541812880865e-06,
+      "loss": 0.1141,
+      "step": 1755
+    },
+    {
+      "epoch": 3.6220391349124617,
+      "grad_norm": 1.2345105409622192,
+      "learning_rate": 5.833521805383451e-06,
+      "loss": 0.121,
+      "step": 1760
+    },
+    {
+      "epoch": 3.632337796086509,
+      "grad_norm": 0.9622531533241272,
+      "learning_rate": 5.752926387120527e-06,
+      "loss": 0.1075,
+      "step": 1765
+    },
+    {
+      "epoch": 3.642636457260556,
+      "grad_norm": 0.9472911357879639,
+      "learning_rate": 5.672759291253974e-06,
+      "loss": 0.1262,
+      "step": 1770
+    },
+    {
+      "epoch": 3.6529351184346037,
+      "grad_norm": 1.1322280168533325,
+      "learning_rate": 5.593024231105872e-06,
+      "loss": 0.1241,
+      "step": 1775
+    },
+    {
+      "epoch": 3.663233779608651,
+      "grad_norm": 1.0482890605926514,
+      "learning_rate": 5.513724899986501e-06,
+      "loss": 0.1025,
+      "step": 1780
+    },
+    {
+      "epoch": 3.673532440782698,
+      "grad_norm": 1.0981827974319458,
+      "learning_rate": 5.434864971023269e-06,
+      "loss": 0.1137,
+      "step": 1785
+    },
+    {
+      "epoch": 3.683831101956746,
+      "grad_norm": 1.0103306770324707,
+      "learning_rate": 5.356448096990575e-06,
+      "loss": 0.1125,
+      "step": 1790
+    },
+    {
+      "epoch": 3.694129763130793,
+      "grad_norm": 0.9068015217781067,
+      "learning_rate": 5.278477910140615e-06,
+      "loss": 0.1161,
+      "step": 1795
+    },
+    {
+      "epoch": 3.70442842430484,
+      "grad_norm": 1.001173734664917,
+      "learning_rate": 5.200958022035119e-06,
+      "loss": 0.1022,
+      "step": 1800
+    },
+    {
+      "epoch": 3.714727085478888,
+      "grad_norm": 1.0164148807525635,
+      "learning_rate": 5.123892023378098e-06,
+      "loss": 0.1195,
+      "step": 1805
+    },
+    {
+      "epoch": 3.725025746652935,
+      "grad_norm": 0.985833466053009,
+      "learning_rate": 5.047283483849497e-06,
+      "loss": 0.1166,
+      "step": 1810
+    },
+    {
+      "epoch": 3.7353244078269823,
+      "grad_norm": 0.9610691666603088,
+      "learning_rate": 4.971135951939853e-06,
+      "loss": 0.1034,
+      "step": 1815
+    },
+    {
+      "epoch": 3.74562306900103,
+      "grad_norm": 0.7836612462997437,
+      "learning_rate": 4.895452954785936e-06,
+      "loss": 0.0992,
+      "step": 1820
+    },
+    {
+      "epoch": 3.755921730175077,
+      "grad_norm": 1.0443223714828491,
+      "learning_rate": 4.820237998007369e-06,
+      "loss": 0.1134,
+      "step": 1825
+    },
+    {
+      "epoch": 3.7662203913491243,
+      "grad_norm": 0.8954272866249084,
+      "learning_rate": 4.745494565544252e-06,
+      "loss": 0.1123,
+      "step": 1830
+    },
+    {
+      "epoch": 3.776519052523172,
+      "grad_norm": 0.89225834608078,
+      "learning_rate": 4.671226119495768e-06,
+      "loss": 0.1137,
+      "step": 1835
+    },
+    {
+      "epoch": 3.786817713697219,
+      "grad_norm": 1.049306869506836,
+      "learning_rate": 4.597436099959849e-06,
+      "loss": 0.1123,
+      "step": 1840
+    },
+    {
+      "epoch": 3.797116374871267,
+      "grad_norm": 0.9005565047264099,
+      "learning_rate": 4.5241279248738285e-06,
+      "loss": 0.1075,
+      "step": 1845
+    },
+    {
+      "epoch": 3.807415036045314,
+      "grad_norm": 0.9594177603721619,
+      "learning_rate": 4.451304989856083e-06,
+      "loss": 0.1034,
+      "step": 1850
+    },
+    {
+      "epoch": 3.8177136972193617,
+      "grad_norm": 0.9349340200424194,
+      "learning_rate": 4.378970668048795e-06,
+      "loss": 0.1062,
+      "step": 1855
+    },
+    {
+      "epoch": 3.828012358393409,
+      "grad_norm": 1.1166118383407593,
+      "learning_rate": 4.307128309961694e-06,
+      "loss": 0.1147,
+      "step": 1860
+    },
+    {
+      "epoch": 3.838311019567456,
+      "grad_norm": 0.8591933250427246,
+      "learning_rate": 4.235781243316845e-06,
+      "loss": 0.1042,
+      "step": 1865
+    },
+    {
+      "epoch": 3.848609680741504,
+      "grad_norm": 1.1003227233886719,
+      "learning_rate": 4.164932772894532e-06,
+      "loss": 0.1188,
+      "step": 1870
+    },
+    {
+      "epoch": 3.858908341915551,
+      "grad_norm": 0.982711911201477,
+      "learning_rate": 4.094586180380186e-06,
+      "loss": 0.1043,
+      "step": 1875
+    },
+    {
+      "epoch": 3.869207003089598,
+      "grad_norm": 0.9575625658035278,
+      "learning_rate": 4.024744724212335e-06,
+      "loss": 0.1078,
+      "step": 1880
+    },
+    {
+      "epoch": 3.879505664263646,
+      "grad_norm": 0.9329299926757812,
+      "learning_rate": 3.955411639431728e-06,
+      "loss": 0.112,
+      "step": 1885
+    },
+    {
+      "epoch": 3.889804325437693,
+      "grad_norm": 0.8347952365875244,
+      "learning_rate": 3.886590137531454e-06,
+      "loss": 0.1018,
+      "step": 1890
+    },
+    {
+      "epoch": 3.9001029866117403,
+      "grad_norm": 0.8512252569198608,
+      "learning_rate": 3.818283406308205e-06,
+      "loss": 0.1064,
+      "step": 1895
+    },
+    {
+      "epoch": 3.910401647785788,
+      "grad_norm": 0.9007921814918518,
+      "learning_rate": 3.7504946097145885e-06,
+      "loss": 0.1119,
+      "step": 1900
+    },
+    {
+      "epoch": 3.920700308959835,
+      "grad_norm": 0.9551131725311279,
+      "learning_rate": 3.6832268877126236e-06,
+      "loss": 0.101,
+      "step": 1905
+    },
+    {
+      "epoch": 3.9309989701338823,
+      "grad_norm": 1.0248686075210571,
+      "learning_rate": 3.616483356128256e-06,
+      "loss": 0.0891,
+      "step": 1910
+    },
+    {
+      "epoch": 3.94129763130793,
+      "grad_norm": 1.0276347398757935,
+      "learning_rate": 3.550267106507038e-06,
+      "loss": 0.1091,
+      "step": 1915
+    },
+    {
+      "epoch": 3.951596292481977,
+      "grad_norm": 1.0377405881881714,
+      "learning_rate": 3.484581205970955e-06,
+      "loss": 0.1084,
+      "step": 1920
+    },
+    {
+      "epoch": 3.961894953656025,
+      "grad_norm": 0.8449147939682007,
+      "learning_rate": 3.419428697076333e-06,
+      "loss": 0.0972,
+      "step": 1925
+    },
+    {
+      "epoch": 3.972193614830072,
+      "grad_norm": 0.8973368406295776,
+      "learning_rate": 3.354812597672914e-06,
+      "loss": 0.1074,
+      "step": 1930
+    },
+    {
+      "epoch": 3.9824922760041197,
+      "grad_norm": 0.859688937664032,
+      "learning_rate": 3.290735900764077e-06,
+      "loss": 0.0983,
+      "step": 1935
+    },
+    {
+      "epoch": 3.992790937178167,
+      "grad_norm": 0.8175042271614075,
+      "learning_rate": 3.227201574368193e-06,
+      "loss": 0.1017,
+      "step": 1940
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 2430,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.7865672598148874e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

44_128_e5_3e-5/checkpoint-1944/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fe7162dc6a9b5b59d136a37e3e396521ee8c38ffa6bdafb995352d8fddcfd9b
+size 7736

44_128_e5_3e-5/checkpoint-1944/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/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, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in 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,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_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``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    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("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

44_128_e5_3e-5/checkpoint-2430/README.md

@@ -0,0 +1,202 @@
+---
+base_model: ibm-granite/granite-3.3-8b-base
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.15.2

44_128_e5_3e-5/checkpoint-2430/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "down_proj",
+    "q_proj",
+    "gate_proj",
+    "v_proj",
+    "o_proj",
+    "k_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

44_128_e5_3e-5/checkpoint-2430/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d985c7eb38b7b8dfe859ddbbefc45d669ade6681301e16b5c9235520fe4fd497
+size 791751704

44_128_e5_3e-5/checkpoint-2430/latest

@@ -0,0 +1 @@
+global_step2430

44_128_e5_3e-5/checkpoint-2430/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53608c16b8dbc802169cb7e0c0109df2e38173100f868a365a598e2232ce4c4d
+size 15920

44_128_e5_3e-5/checkpoint-2430/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76b7d95311873d8fa62d4c2c601e99d08418ea3a7a664242837b1207db948932
+size 15920

44_128_e5_3e-5/checkpoint-2430/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c8067bd9b0c8d91aca9c93e100dd38be6fc18ffedfc0f4ef188472a3083921e
+size 15920