Training in progress, step 10000, checkpoint

last-checkpoint/README.md

@@ -0,0 +1,202 @@
+---
+base_model: meta-llama/Meta-Llama-3-8B-Instruct
+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 0.12.0

last-checkpoint/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "meta-llama/Meta-Llama-3-8B-Instruct",
+  "bias": "none",
+  "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": 32,
+  "lora_dropout": 0.1,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "o_proj",
+    "up_proj",
+    "down_proj",
+    "k_proj",
+    "v_proj",
+    "q_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

last-checkpoint/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1474cf437e6fcee5d794193be670d628b5ab5c1f0fd155eca54aae05745396f0
+size 42002584

last-checkpoint/global_step10000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9aa83cd559642330de39ef0589c699f240311dfbf19742ae25c68194f121e3f
+size 251710672

last-checkpoint/global_step10000/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e491c98b4c6ae85bbc78fee7bbb1120024331b8b6556148937b423b5555c662d
+size 153747385

last-checkpoint/latest

@@ -0,0 +1 @@
+global_step10000

last-checkpoint/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72c3a70cd1bd8ddf95e5dbee237f75f88a88c828d3ba3a83a9ec614b92f4bae4
+size 14244

last-checkpoint/trainer_state.json

@@ -0,0 +1,1440 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.29856985041650497,
+  "eval_steps": 1000,
+  "global_step": 10000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 2.9856985041650495e-05,
+      "grad_norm": 8.064935684204102,
+      "learning_rate": 0.0,
+      "loss": 3.4849,
+      "step": 1
+    },
+    {
+      "epoch": 0.0014928492520825246,
+      "grad_norm": 3.3433420658111572,
+      "learning_rate": 0.00019998742849959144,
+      "loss": 1.9038,
+      "step": 50
+    },
+    {
+      "epoch": 0.0029856985041650493,
+      "grad_norm": 2.907883405685425,
+      "learning_rate": 0.0001999731427036726,
+      "loss": 1.608,
+      "step": 100
+    },
+    {
+      "epoch": 0.004478547756247574,
+      "grad_norm": 3.2614288330078125,
+      "learning_rate": 0.00019995885690775376,
+      "loss": 1.5505,
+      "step": 150
+    },
+    {
+      "epoch": 0.005971397008330099,
+      "grad_norm": 3.8400654792785645,
+      "learning_rate": 0.00019994457111183493,
+      "loss": 1.5737,
+      "step": 200
+    },
+    {
+      "epoch": 0.007464246260412624,
+      "grad_norm": 3.3358442783355713,
+      "learning_rate": 0.00019993028531591612,
+      "loss": 1.567,
+      "step": 250
+    },
+    {
+      "epoch": 0.008957095512495149,
+      "grad_norm": 2.131911277770996,
+      "learning_rate": 0.00019991599951999726,
+      "loss": 1.5208,
+      "step": 300
+    },
+    {
+      "epoch": 0.010449944764577673,
+      "grad_norm": 3.180992364883423,
+      "learning_rate": 0.00019990171372407845,
+      "loss": 1.5586,
+      "step": 350
+    },
+    {
+      "epoch": 0.011942794016660197,
+      "grad_norm": 3.024989128112793,
+      "learning_rate": 0.0001998874279281596,
+      "loss": 1.5267,
+      "step": 400
+    },
+    {
+      "epoch": 0.013435643268742723,
+      "grad_norm": 3.4935102462768555,
+      "learning_rate": 0.00019987314213224078,
+      "loss": 1.5609,
+      "step": 450
+    },
+    {
+      "epoch": 0.014928492520825247,
+      "grad_norm": 3.565504550933838,
+      "learning_rate": 0.00019985885633632194,
+      "loss": 1.5112,
+      "step": 500
+    },
+    {
+      "epoch": 0.01642134177290777,
+      "grad_norm": 2.0692882537841797,
+      "learning_rate": 0.00019984457054040308,
+      "loss": 1.5087,
+      "step": 550
+    },
+    {
+      "epoch": 0.017914191024990297,
+      "grad_norm": 4.110323905944824,
+      "learning_rate": 0.00019983028474448427,
+      "loss": 1.5073,
+      "step": 600
+    },
+    {
+      "epoch": 0.01940704027707282,
+      "grad_norm": 2.8632736206054688,
+      "learning_rate": 0.0001998159989485654,
+      "loss": 1.472,
+      "step": 650
+    },
+    {
+      "epoch": 0.020899889529155346,
+      "grad_norm": 2.629347324371338,
+      "learning_rate": 0.0001998017131526466,
+      "loss": 1.5236,
+      "step": 700
+    },
+    {
+      "epoch": 0.022392738781237872,
+      "grad_norm": 3.696873188018799,
+      "learning_rate": 0.00019978742735672774,
+      "loss": 1.5216,
+      "step": 750
+    },
+    {
+      "epoch": 0.023885588033320394,
+      "grad_norm": 3.253305435180664,
+      "learning_rate": 0.00019977314156080893,
+      "loss": 1.5915,
+      "step": 800
+    },
+    {
+      "epoch": 0.02537843728540292,
+      "grad_norm": 2.9587886333465576,
+      "learning_rate": 0.0001997588557648901,
+      "loss": 1.4862,
+      "step": 850
+    },
+    {
+      "epoch": 0.026871286537485446,
+      "grad_norm": 3.3311073780059814,
+      "learning_rate": 0.00019974456996897126,
+      "loss": 1.4703,
+      "step": 900
+    },
+    {
+      "epoch": 0.02836413578956797,
+      "grad_norm": 2.5960264205932617,
+      "learning_rate": 0.00019973028417305243,
+      "loss": 1.4818,
+      "step": 950
+    },
+    {
+      "epoch": 0.029856985041650495,
+      "grad_norm": 3.3142144680023193,
+      "learning_rate": 0.0001997159983771336,
+      "loss": 1.4746,
+      "step": 1000
+    },
+    {
+      "epoch": 0.03134983429373302,
+      "grad_norm": 3.5049827098846436,
+      "learning_rate": 0.00019970171258121476,
+      "loss": 1.442,
+      "step": 1050
+    },
+    {
+      "epoch": 0.03284268354581554,
+      "grad_norm": 3.0218605995178223,
+      "learning_rate": 0.00019968742678529592,
+      "loss": 1.5265,
+      "step": 1100
+    },
+    {
+      "epoch": 0.03433553279789807,
+      "grad_norm": 2.936182975769043,
+      "learning_rate": 0.00019967314098937709,
+      "loss": 1.5174,
+      "step": 1150
+    },
+    {
+      "epoch": 0.035828382049980595,
+      "grad_norm": 2.877253293991089,
+      "learning_rate": 0.00019965885519345825,
+      "loss": 1.4499,
+      "step": 1200
+    },
+    {
+      "epoch": 0.03732123130206312,
+      "grad_norm": 6.07016658782959,
+      "learning_rate": 0.00019964456939753941,
+      "loss": 1.4542,
+      "step": 1250
+    },
+    {
+      "epoch": 0.03881408055414564,
+      "grad_norm": 2.1618189811706543,
+      "learning_rate": 0.0001996302836016206,
+      "loss": 1.4343,
+      "step": 1300
+    },
+    {
+      "epoch": 0.040306929806228166,
+      "grad_norm": 2.8267719745635986,
+      "learning_rate": 0.00019961599780570174,
+      "loss": 1.463,
+      "step": 1350
+    },
+    {
+      "epoch": 0.04179977905831069,
+      "grad_norm": 2.6036462783813477,
+      "learning_rate": 0.00019960171200978294,
+      "loss": 1.4557,
+      "step": 1400
+    },
+    {
+      "epoch": 0.04329262831039322,
+      "grad_norm": 3.0187127590179443,
+      "learning_rate": 0.00019958742621386407,
+      "loss": 1.4472,
+      "step": 1450
+    },
+    {
+      "epoch": 0.044785477562475744,
+      "grad_norm": 3.9822633266448975,
+      "learning_rate": 0.00019957314041794526,
+      "loss": 1.4384,
+      "step": 1500
+    },
+    {
+      "epoch": 0.04627832681455826,
+      "grad_norm": 2.919654607772827,
+      "learning_rate": 0.0001995588546220264,
+      "loss": 1.4969,
+      "step": 1550
+    },
+    {
+      "epoch": 0.04777117606664079,
+      "grad_norm": 2.922963857650757,
+      "learning_rate": 0.0001995445688261076,
+      "loss": 1.4987,
+      "step": 1600
+    },
+    {
+      "epoch": 0.049264025318723315,
+      "grad_norm": 2.9638512134552,
+      "learning_rate": 0.00019953028303018876,
+      "loss": 1.4722,
+      "step": 1650
+    },
+    {
+      "epoch": 0.05075687457080584,
+      "grad_norm": 3.408391237258911,
+      "learning_rate": 0.00019951599723426992,
+      "loss": 1.4723,
+      "step": 1700
+    },
+    {
+      "epoch": 0.052249723822888366,
+      "grad_norm": 3.023597240447998,
+      "learning_rate": 0.0001995017114383511,
+      "loss": 1.4478,
+      "step": 1750
+    },
+    {
+      "epoch": 0.05374257307497089,
+      "grad_norm": 2.1655213832855225,
+      "learning_rate": 0.00019948742564243225,
+      "loss": 1.4071,
+      "step": 1800
+    },
+    {
+      "epoch": 0.05523542232705341,
+      "grad_norm": 3.796663999557495,
+      "learning_rate": 0.00019947313984651342,
+      "loss": 1.446,
+      "step": 1850
+    },
+    {
+      "epoch": 0.05672827157913594,
+      "grad_norm": 3.0415594577789307,
+      "learning_rate": 0.00019945885405059458,
+      "loss": 1.4324,
+      "step": 1900
+    },
+    {
+      "epoch": 0.05822112083121846,
+      "grad_norm": 2.524627685546875,
+      "learning_rate": 0.00019944456825467575,
+      "loss": 1.3963,
+      "step": 1950
+    },
+    {
+      "epoch": 0.05971397008330099,
+      "grad_norm": 3.2881991863250732,
+      "learning_rate": 0.0001994302824587569,
+      "loss": 1.4522,
+      "step": 2000
+    },
+    {
+      "epoch": 0.061206819335383515,
+      "grad_norm": 3.392430067062378,
+      "learning_rate": 0.00019941599666283808,
+      "loss": 1.4329,
+      "step": 2050
+    },
+    {
+      "epoch": 0.06269966858746603,
+      "grad_norm": 3.9426393508911133,
+      "learning_rate": 0.00019940171086691927,
+      "loss": 1.5203,
+      "step": 2100
+    },
+    {
+      "epoch": 0.06419251783954856,
+      "grad_norm": 3.3737235069274902,
+      "learning_rate": 0.0001993874250710004,
+      "loss": 1.3674,
+      "step": 2150
+    },
+    {
+      "epoch": 0.06568536709163109,
+      "grad_norm": 3.783085346221924,
+      "learning_rate": 0.0001993731392750816,
+      "loss": 1.4339,
+      "step": 2200
+    },
+    {
+      "epoch": 0.06717821634371361,
+      "grad_norm": 3.4819202423095703,
+      "learning_rate": 0.00019935885347916273,
+      "loss": 1.4436,
+      "step": 2250
+    },
+    {
+      "epoch": 0.06867106559579614,
+      "grad_norm": 3.141775608062744,
+      "learning_rate": 0.00019934456768324393,
+      "loss": 1.4683,
+      "step": 2300
+    },
+    {
+      "epoch": 0.07016391484787866,
+      "grad_norm": 3.2881035804748535,
+      "learning_rate": 0.00019933028188732506,
+      "loss": 1.4395,
+      "step": 2350
+    },
+    {
+      "epoch": 0.07165676409996119,
+      "grad_norm": 3.718122959136963,
+      "learning_rate": 0.00019931599609140626,
+      "loss": 1.4396,
+      "step": 2400
+    },
+    {
+      "epoch": 0.07314961335204372,
+      "grad_norm": 4.3829474449157715,
+      "learning_rate": 0.00019930171029548742,
+      "loss": 1.4477,
+      "step": 2450
+    },
+    {
+      "epoch": 0.07464246260412624,
+      "grad_norm": 3.3698525428771973,
+      "learning_rate": 0.00019928742449956858,
+      "loss": 1.3529,
+      "step": 2500
+    },
+    {
+      "epoch": 0.07613531185620875,
+      "grad_norm": 3.7569565773010254,
+      "learning_rate": 0.00019927313870364975,
+      "loss": 1.4246,
+      "step": 2550
+    },
+    {
+      "epoch": 0.07762816110829128,
+      "grad_norm": 3.1486406326293945,
+      "learning_rate": 0.00019925885290773091,
+      "loss": 1.3813,
+      "step": 2600
+    },
+    {
+      "epoch": 0.0791210103603738,
+      "grad_norm": 4.0635480880737305,
+      "learning_rate": 0.00019924456711181208,
+      "loss": 1.4259,
+      "step": 2650
+    },
+    {
+      "epoch": 0.08061385961245633,
+      "grad_norm": 3.2710611820220947,
+      "learning_rate": 0.00019923028131589324,
+      "loss": 1.3747,
+      "step": 2700
+    },
+    {
+      "epoch": 0.08210670886453886,
+      "grad_norm": 3.4968345165252686,
+      "learning_rate": 0.0001992159955199744,
+      "loss": 1.4721,
+      "step": 2750
+    },
+    {
+      "epoch": 0.08359955811662138,
+      "grad_norm": 4.274214267730713,
+      "learning_rate": 0.00019920170972405557,
+      "loss": 1.437,
+      "step": 2800
+    },
+    {
+      "epoch": 0.08509240736870391,
+      "grad_norm": 2.970602512359619,
+      "learning_rate": 0.00019918742392813674,
+      "loss": 1.3336,
+      "step": 2850
+    },
+    {
+      "epoch": 0.08658525662078644,
+      "grad_norm": 4.143342971801758,
+      "learning_rate": 0.00019917313813221793,
+      "loss": 1.4264,
+      "step": 2900
+    },
+    {
+      "epoch": 0.08807810587286896,
+      "grad_norm": 3.7546920776367188,
+      "learning_rate": 0.00019915885233629907,
+      "loss": 1.441,
+      "step": 2950
+    },
+    {
+      "epoch": 0.08957095512495149,
+      "grad_norm": 3.9160516262054443,
+      "learning_rate": 0.00019914456654038026,
+      "loss": 1.4261,
+      "step": 3000
+    },
+    {
+      "epoch": 0.09106380437703401,
+      "grad_norm": 3.842073917388916,
+      "learning_rate": 0.0001991302807444614,
+      "loss": 1.4076,
+      "step": 3050
+    },
+    {
+      "epoch": 0.09255665362911653,
+      "grad_norm": 4.392395496368408,
+      "learning_rate": 0.0001991159949485426,
+      "loss": 1.3789,
+      "step": 3100
+    },
+    {
+      "epoch": 0.09404950288119905,
+      "grad_norm": 3.822425603866577,
+      "learning_rate": 0.00019910170915262373,
+      "loss": 1.3877,
+      "step": 3150
+    },
+    {
+      "epoch": 0.09554235213328158,
+      "grad_norm": 3.1348562240600586,
+      "learning_rate": 0.0001990874233567049,
+      "loss": 1.4081,
+      "step": 3200
+    },
+    {
+      "epoch": 0.0970352013853641,
+      "grad_norm": 3.453887939453125,
+      "learning_rate": 0.00019907313756078608,
+      "loss": 1.4143,
+      "step": 3250
+    },
+    {
+      "epoch": 0.09852805063744663,
+      "grad_norm": 3.5057384967803955,
+      "learning_rate": 0.00019905885176486722,
+      "loss": 1.4264,
+      "step": 3300
+    },
+    {
+      "epoch": 0.10002089988952916,
+      "grad_norm": 3.145796060562134,
+      "learning_rate": 0.0001990445659689484,
+      "loss": 1.4368,
+      "step": 3350
+    },
+    {
+      "epoch": 0.10151374914161168,
+      "grad_norm": 3.4077043533325195,
+      "learning_rate": 0.00019903028017302955,
+      "loss": 1.388,
+      "step": 3400
+    },
+    {
+      "epoch": 0.10300659839369421,
+      "grad_norm": 3.65567946434021,
+      "learning_rate": 0.00019901599437711074,
+      "loss": 1.42,
+      "step": 3450
+    },
+    {
+      "epoch": 0.10449944764577673,
+      "grad_norm": 4.460702419281006,
+      "learning_rate": 0.0001990017085811919,
+      "loss": 1.3991,
+      "step": 3500
+    },
+    {
+      "epoch": 0.10599229689785926,
+      "grad_norm": 4.155653476715088,
+      "learning_rate": 0.00019898742278527307,
+      "loss": 1.371,
+      "step": 3550
+    },
+    {
+      "epoch": 0.10748514614994178,
+      "grad_norm": 3.8904318809509277,
+      "learning_rate": 0.00019897313698935423,
+      "loss": 1.4378,
+      "step": 3600
+    },
+    {
+      "epoch": 0.1089779954020243,
+      "grad_norm": 4.0509233474731445,
+      "learning_rate": 0.0001989588511934354,
+      "loss": 1.3945,
+      "step": 3650
+    },
+    {
+      "epoch": 0.11047084465410682,
+      "grad_norm": 3.785123109817505,
+      "learning_rate": 0.00019894456539751656,
+      "loss": 1.436,
+      "step": 3700
+    },
+    {
+      "epoch": 0.11196369390618935,
+      "grad_norm": 3.4556167125701904,
+      "learning_rate": 0.00019893027960159773,
+      "loss": 1.3794,
+      "step": 3750
+    },
+    {
+      "epoch": 0.11345654315827187,
+      "grad_norm": 4.0479559898376465,
+      "learning_rate": 0.0001989159938056789,
+      "loss": 1.4734,
+      "step": 3800
+    },
+    {
+      "epoch": 0.1149493924103544,
+      "grad_norm": 3.890805721282959,
+      "learning_rate": 0.00019890170800976006,
+      "loss": 1.4341,
+      "step": 3850
+    },
+    {
+      "epoch": 0.11644224166243693,
+      "grad_norm": 3.8178727626800537,
+      "learning_rate": 0.00019888742221384122,
+      "loss": 1.4754,
+      "step": 3900
+    },
+    {
+      "epoch": 0.11793509091451945,
+      "grad_norm": 2.456165075302124,
+      "learning_rate": 0.00019887313641792241,
+      "loss": 1.3887,
+      "step": 3950
+    },
+    {
+      "epoch": 0.11942794016660198,
+      "grad_norm": 3.5763051509857178,
+      "learning_rate": 0.00019885885062200355,
+      "loss": 1.3901,
+      "step": 4000
+    },
+    {
+      "epoch": 0.1209207894186845,
+      "grad_norm": 3.885662317276001,
+      "learning_rate": 0.00019884456482608474,
+      "loss": 1.3856,
+      "step": 4050
+    },
+    {
+      "epoch": 0.12241363867076703,
+      "grad_norm": 3.6095409393310547,
+      "learning_rate": 0.00019883027903016588,
+      "loss": 1.448,
+      "step": 4100
+    },
+    {
+      "epoch": 0.12390648792284956,
+      "grad_norm": 3.7112534046173096,
+      "learning_rate": 0.00019881599323424707,
+      "loss": 1.3537,
+      "step": 4150
+    },
+    {
+      "epoch": 0.12539933717493207,
+      "grad_norm": 3.3566672801971436,
+      "learning_rate": 0.0001988017074383282,
+      "loss": 1.4389,
+      "step": 4200
+    },
+    {
+      "epoch": 0.1268921864270146,
+      "grad_norm": 4.570401191711426,
+      "learning_rate": 0.0001987874216424094,
+      "loss": 1.4191,
+      "step": 4250
+    },
+    {
+      "epoch": 0.12838503567909712,
+      "grad_norm": 4.455029010772705,
+      "learning_rate": 0.00019877313584649057,
+      "loss": 1.3677,
+      "step": 4300
+    },
+    {
+      "epoch": 0.12987788493117966,
+      "grad_norm": 3.0861828327178955,
+      "learning_rate": 0.00019875885005057173,
+      "loss": 1.3677,
+      "step": 4350
+    },
+    {
+      "epoch": 0.13137073418326217,
+      "grad_norm": 4.419896602630615,
+      "learning_rate": 0.0001987445642546529,
+      "loss": 1.4524,
+      "step": 4400
+    },
+    {
+      "epoch": 0.1328635834353447,
+      "grad_norm": 5.187576770782471,
+      "learning_rate": 0.00019873027845873406,
+      "loss": 1.3868,
+      "step": 4450
+    },
+    {
+      "epoch": 0.13435643268742722,
+      "grad_norm": 5.111696243286133,
+      "learning_rate": 0.00019871599266281523,
+      "loss": 1.4458,
+      "step": 4500
+    },
+    {
+      "epoch": 0.13584928193950974,
+      "grad_norm": 3.2652997970581055,
+      "learning_rate": 0.0001987017068668964,
+      "loss": 1.4529,
+      "step": 4550
+    },
+    {
+      "epoch": 0.13734213119159228,
+      "grad_norm": 4.190273761749268,
+      "learning_rate": 0.00019868742107097755,
+      "loss": 1.3991,
+      "step": 4600
+    },
+    {
+      "epoch": 0.1388349804436748,
+      "grad_norm": 4.85620641708374,
+      "learning_rate": 0.00019867313527505872,
+      "loss": 1.3916,
+      "step": 4650
+    },
+    {
+      "epoch": 0.14032782969575733,
+      "grad_norm": 3.030954360961914,
+      "learning_rate": 0.00019865884947913988,
+      "loss": 1.3805,
+      "step": 4700
+    },
+    {
+      "epoch": 0.14182067894783984,
+      "grad_norm": 3.264406681060791,
+      "learning_rate": 0.00019864456368322108,
+      "loss": 1.4048,
+      "step": 4750
+    },
+    {
+      "epoch": 0.14331352819992238,
+      "grad_norm": 3.2138588428497314,
+      "learning_rate": 0.0001986302778873022,
+      "loss": 1.4092,
+      "step": 4800
+    },
+    {
+      "epoch": 0.1448063774520049,
+      "grad_norm": 3.847222328186035,
+      "learning_rate": 0.0001986159920913834,
+      "loss": 1.3871,
+      "step": 4850
+    },
+    {
+      "epoch": 0.14629922670408743,
+      "grad_norm": 4.004987716674805,
+      "learning_rate": 0.00019860170629546454,
+      "loss": 1.3845,
+      "step": 4900
+    },
+    {
+      "epoch": 0.14779207595616994,
+      "grad_norm": 3.5088725090026855,
+      "learning_rate": 0.00019858742049954573,
+      "loss": 1.379,
+      "step": 4950
+    },
+    {
+      "epoch": 0.14928492520825248,
+      "grad_norm": 3.275099277496338,
+      "learning_rate": 0.00019857313470362687,
+      "loss": 1.3628,
+      "step": 5000
+    },
+    {
+      "epoch": 0.150777774460335,
+      "grad_norm": 3.7903060913085938,
+      "learning_rate": 0.00019855884890770806,
+      "loss": 1.3804,
+      "step": 5050
+    },
+    {
+      "epoch": 0.1522706237124175,
+      "grad_norm": 4.294798374176025,
+      "learning_rate": 0.00019854456311178923,
+      "loss": 1.3988,
+      "step": 5100
+    },
+    {
+      "epoch": 0.15376347296450005,
+      "grad_norm": 3.2719295024871826,
+      "learning_rate": 0.0001985302773158704,
+      "loss": 1.387,
+      "step": 5150
+    },
+    {
+      "epoch": 0.15525632221658256,
+      "grad_norm": 4.143224239349365,
+      "learning_rate": 0.00019851599151995156,
+      "loss": 1.393,
+      "step": 5200
+    },
+    {
+      "epoch": 0.1567491714686651,
+      "grad_norm": 3.404754638671875,
+      "learning_rate": 0.00019850170572403272,
+      "loss": 1.4205,
+      "step": 5250
+    },
+    {
+      "epoch": 0.1582420207207476,
+      "grad_norm": 3.607126474380493,
+      "learning_rate": 0.0001984874199281139,
+      "loss": 1.4239,
+      "step": 5300
+    },
+    {
+      "epoch": 0.15973486997283015,
+      "grad_norm": 4.140823841094971,
+      "learning_rate": 0.00019847313413219505,
+      "loss": 1.4204,
+      "step": 5350
+    },
+    {
+      "epoch": 0.16122771922491266,
+      "grad_norm": 3.893251419067383,
+      "learning_rate": 0.00019845884833627622,
+      "loss": 1.392,
+      "step": 5400
+    },
+    {
+      "epoch": 0.1627205684769952,
+      "grad_norm": 4.304211139678955,
+      "learning_rate": 0.00019844456254035738,
+      "loss": 1.44,
+      "step": 5450
+    },
+    {
+      "epoch": 0.16421341772907772,
+      "grad_norm": 5.273501873016357,
+      "learning_rate": 0.00019843027674443855,
+      "loss": 1.445,
+      "step": 5500
+    },
+    {
+      "epoch": 0.16570626698116026,
+      "grad_norm": 4.787700176239014,
+      "learning_rate": 0.00019841599094851974,
+      "loss": 1.3668,
+      "step": 5550
+    },
+    {
+      "epoch": 0.16719911623324277,
+      "grad_norm": 3.7984108924865723,
+      "learning_rate": 0.00019840170515260087,
+      "loss": 1.3523,
+      "step": 5600
+    },
+    {
+      "epoch": 0.16869196548532528,
+      "grad_norm": 3.885608673095703,
+      "learning_rate": 0.00019838741935668207,
+      "loss": 1.3917,
+      "step": 5650
+    },
+    {
+      "epoch": 0.17018481473740782,
+      "grad_norm": 3.459803342819214,
+      "learning_rate": 0.0001983731335607632,
+      "loss": 1.3833,
+      "step": 5700
+    },
+    {
+      "epoch": 0.17167766398949033,
+      "grad_norm": 3.7103006839752197,
+      "learning_rate": 0.0001983588477648444,
+      "loss": 1.4473,
+      "step": 5750
+    },
+    {
+      "epoch": 0.17317051324157287,
+      "grad_norm": 6.645928382873535,
+      "learning_rate": 0.00019834456196892553,
+      "loss": 1.3706,
+      "step": 5800
+    },
+    {
+      "epoch": 0.17466336249365538,
+      "grad_norm": 3.7201037406921387,
+      "learning_rate": 0.0001983302761730067,
+      "loss": 1.3733,
+      "step": 5850
+    },
+    {
+      "epoch": 0.17615621174573792,
+      "grad_norm": 4.050106048583984,
+      "learning_rate": 0.0001983159903770879,
+      "loss": 1.4096,
+      "step": 5900
+    },
+    {
+      "epoch": 0.17764906099782043,
+      "grad_norm": 4.190842628479004,
+      "learning_rate": 0.00019830170458116903,
+      "loss": 1.4404,
+      "step": 5950
+    },
+    {
+      "epoch": 0.17914191024990297,
+      "grad_norm": 4.393162727355957,
+      "learning_rate": 0.00019828741878525022,
+      "loss": 1.4443,
+      "step": 6000
+    },
+    {
+      "epoch": 0.1806347595019855,
+      "grad_norm": 3.597520351409912,
+      "learning_rate": 0.00019827313298933136,
+      "loss": 1.4063,
+      "step": 6050
+    },
+    {
+      "epoch": 0.18212760875406803,
+      "grad_norm": 3.608085870742798,
+      "learning_rate": 0.00019825884719341255,
+      "loss": 1.3857,
+      "step": 6100
+    },
+    {
+      "epoch": 0.18362045800615054,
+      "grad_norm": 3.7055492401123047,
+      "learning_rate": 0.0001982445613974937,
+      "loss": 1.3997,
+      "step": 6150
+    },
+    {
+      "epoch": 0.18511330725823305,
+      "grad_norm": 3.875457763671875,
+      "learning_rate": 0.00019823027560157488,
+      "loss": 1.4296,
+      "step": 6200
+    },
+    {
+      "epoch": 0.1866061565103156,
+      "grad_norm": 5.074592590332031,
+      "learning_rate": 0.00019821598980565604,
+      "loss": 1.3785,
+      "step": 6250
+    },
+    {
+      "epoch": 0.1880990057623981,
+      "grad_norm": 6.013392448425293,
+      "learning_rate": 0.0001982017040097372,
+      "loss": 1.4391,
+      "step": 6300
+    },
+    {
+      "epoch": 0.18959185501448064,
+      "grad_norm": 5.679958820343018,
+      "learning_rate": 0.00019818741821381837,
+      "loss": 1.367,
+      "step": 6350
+    },
+    {
+      "epoch": 0.19108470426656315,
+      "grad_norm": 3.6182546615600586,
+      "learning_rate": 0.00019817313241789954,
+      "loss": 1.4508,
+      "step": 6400
+    },
+    {
+      "epoch": 0.1925775535186457,
+      "grad_norm": 5.209213733673096,
+      "learning_rate": 0.0001981588466219807,
+      "loss": 1.3878,
+      "step": 6450
+    },
+    {
+      "epoch": 0.1940704027707282,
+      "grad_norm": 3.0043230056762695,
+      "learning_rate": 0.00019814456082606187,
+      "loss": 1.4248,
+      "step": 6500
+    },
+    {
+      "epoch": 0.19556325202281075,
+      "grad_norm": 3.157851219177246,
+      "learning_rate": 0.00019813027503014303,
+      "loss": 1.3725,
+      "step": 6550
+    },
+    {
+      "epoch": 0.19705610127489326,
+      "grad_norm": 3.5292418003082275,
+      "learning_rate": 0.0001981159892342242,
+      "loss": 1.3932,
+      "step": 6600
+    },
+    {
+      "epoch": 0.1985489505269758,
+      "grad_norm": 3.2819600105285645,
+      "learning_rate": 0.00019810170343830536,
+      "loss": 1.3495,
+      "step": 6650
+    },
+    {
+      "epoch": 0.2000417997790583,
+      "grad_norm": 3.0243399143218994,
+      "learning_rate": 0.00019808741764238655,
+      "loss": 1.3689,
+      "step": 6700
+    },
+    {
+      "epoch": 0.20153464903114082,
+      "grad_norm": 3.4495368003845215,
+      "learning_rate": 0.0001980731318464677,
+      "loss": 1.3725,
+      "step": 6750
+    },
+    {
+      "epoch": 0.20302749828322336,
+      "grad_norm": 3.538259744644165,
+      "learning_rate": 0.00019805884605054888,
+      "loss": 1.3905,
+      "step": 6800
+    },
+    {
+      "epoch": 0.20452034753530587,
+      "grad_norm": 4.162181377410889,
+      "learning_rate": 0.00019804456025463002,
+      "loss": 1.4129,
+      "step": 6850
+    },
+    {
+      "epoch": 0.20601319678738841,
+      "grad_norm": 4.592432022094727,
+      "learning_rate": 0.0001980302744587112,
+      "loss": 1.3634,
+      "step": 6900
+    },
+    {
+      "epoch": 0.20750604603947093,
+      "grad_norm": 3.45967960357666,
+      "learning_rate": 0.00019801598866279237,
+      "loss": 1.416,
+      "step": 6950
+    },
+    {
+      "epoch": 0.20899889529155347,
+      "grad_norm": 4.221930503845215,
+      "learning_rate": 0.00019800170286687354,
+      "loss": 1.4051,
+      "step": 7000
+    },
+    {
+      "epoch": 0.21049174454363598,
+      "grad_norm": 4.144239902496338,
+      "learning_rate": 0.0001979874170709547,
+      "loss": 1.4219,
+      "step": 7050
+    },
+    {
+      "epoch": 0.21198459379571852,
+      "grad_norm": 4.7492570877075195,
+      "learning_rate": 0.00019797313127503587,
+      "loss": 1.4028,
+      "step": 7100
+    },
+    {
+      "epoch": 0.21347744304780103,
+      "grad_norm": 3.5841355323791504,
+      "learning_rate": 0.00019795884547911703,
+      "loss": 1.4361,
+      "step": 7150
+    },
+    {
+      "epoch": 0.21497029229988357,
+      "grad_norm": 4.662593364715576,
+      "learning_rate": 0.0001979445596831982,
+      "loss": 1.3816,
+      "step": 7200
+    },
+    {
+      "epoch": 0.21646314155196608,
+      "grad_norm": 4.700701713562012,
+      "learning_rate": 0.00019793027388727936,
+      "loss": 1.4226,
+      "step": 7250
+    },
+    {
+      "epoch": 0.2179559908040486,
+      "grad_norm": 4.025181293487549,
+      "learning_rate": 0.00019791598809136053,
+      "loss": 1.4291,
+      "step": 7300
+    },
+    {
+      "epoch": 0.21944884005613113,
+      "grad_norm": 3.064573049545288,
+      "learning_rate": 0.0001979017022954417,
+      "loss": 1.4293,
+      "step": 7350
+    },
+    {
+      "epoch": 0.22094168930821365,
+      "grad_norm": 6.342152118682861,
+      "learning_rate": 0.00019788741649952288,
+      "loss": 1.4173,
+      "step": 7400
+    },
+    {
+      "epoch": 0.22243453856029619,
+      "grad_norm": 5.89996337890625,
+      "learning_rate": 0.00019787313070360402,
+      "loss": 1.396,
+      "step": 7450
+    },
+    {
+      "epoch": 0.2239273878123787,
+      "grad_norm": 4.462945938110352,
+      "learning_rate": 0.0001978588449076852,
+      "loss": 1.3868,
+      "step": 7500
+    },
+    {
+      "epoch": 0.22542023706446124,
+      "grad_norm": 3.6449055671691895,
+      "learning_rate": 0.00019784455911176635,
+      "loss": 1.396,
+      "step": 7550
+    },
+    {
+      "epoch": 0.22691308631654375,
+      "grad_norm": 4.674243927001953,
+      "learning_rate": 0.00019783027331584754,
+      "loss": 1.395,
+      "step": 7600
+    },
+    {
+      "epoch": 0.2284059355686263,
+      "grad_norm": 3.6160385608673096,
+      "learning_rate": 0.00019781598751992868,
+      "loss": 1.3918,
+      "step": 7650
+    },
+    {
+      "epoch": 0.2298987848207088,
+      "grad_norm": 4.326193332672119,
+      "learning_rate": 0.00019780170172400987,
+      "loss": 1.3947,
+      "step": 7700
+    },
+    {
+      "epoch": 0.23139163407279134,
+      "grad_norm": 5.4003777503967285,
+      "learning_rate": 0.00019778741592809104,
+      "loss": 1.344,
+      "step": 7750
+    },
+    {
+      "epoch": 0.23288448332487385,
+      "grad_norm": 4.711580753326416,
+      "learning_rate": 0.0001977731301321722,
+      "loss": 1.3959,
+      "step": 7800
+    },
+    {
+      "epoch": 0.23437733257695637,
+      "grad_norm": 3.4752814769744873,
+      "learning_rate": 0.00019775884433625337,
+      "loss": 1.3722,
+      "step": 7850
+    },
+    {
+      "epoch": 0.2358701818290389,
+      "grad_norm": 4.028527736663818,
+      "learning_rate": 0.00019774455854033453,
+      "loss": 1.3683,
+      "step": 7900
+    },
+    {
+      "epoch": 0.23736303108112142,
+      "grad_norm": 4.094334602355957,
+      "learning_rate": 0.0001977302727444157,
+      "loss": 1.3607,
+      "step": 7950
+    },
+    {
+      "epoch": 0.23885588033320396,
+      "grad_norm": 5.232580661773682,
+      "learning_rate": 0.00019771598694849686,
+      "loss": 1.4354,
+      "step": 8000
+    },
+    {
+      "epoch": 0.24034872958528647,
+      "grad_norm": 4.269852161407471,
+      "learning_rate": 0.00019770170115257802,
+      "loss": 1.4372,
+      "step": 8050
+    },
+    {
+      "epoch": 0.241841578837369,
+      "grad_norm": 3.312541961669922,
+      "learning_rate": 0.0001976874153566592,
+      "loss": 1.3826,
+      "step": 8100
+    },
+    {
+      "epoch": 0.24333442808945152,
+      "grad_norm": 3.8900692462921143,
+      "learning_rate": 0.00019767312956074035,
+      "loss": 1.4189,
+      "step": 8150
+    },
+    {
+      "epoch": 0.24482727734153406,
+      "grad_norm": 3.894512414932251,
+      "learning_rate": 0.00019765884376482155,
+      "loss": 1.3365,
+      "step": 8200
+    },
+    {
+      "epoch": 0.24632012659361657,
+      "grad_norm": 4.644411563873291,
+      "learning_rate": 0.00019764455796890268,
+      "loss": 1.4311,
+      "step": 8250
+    },
+    {
+      "epoch": 0.2478129758456991,
+      "grad_norm": 8.174029350280762,
+      "learning_rate": 0.00019763027217298387,
+      "loss": 1.361,
+      "step": 8300
+    },
+    {
+      "epoch": 0.24930582509778162,
+      "grad_norm": 4.615732192993164,
+      "learning_rate": 0.000197615986377065,
+      "loss": 1.4552,
+      "step": 8350
+    },
+    {
+      "epoch": 0.25079867434986414,
+      "grad_norm": 4.421249866485596,
+      "learning_rate": 0.0001976017005811462,
+      "loss": 1.3463,
+      "step": 8400
+    },
+    {
+      "epoch": 0.2522915236019467,
+      "grad_norm": 2.8386716842651367,
+      "learning_rate": 0.00019758741478522734,
+      "loss": 1.348,
+      "step": 8450
+    },
+    {
+      "epoch": 0.2537843728540292,
+      "grad_norm": 4.3141703605651855,
+      "learning_rate": 0.0001975731289893085,
+      "loss": 1.4306,
+      "step": 8500
+    },
+    {
+      "epoch": 0.2552772221061117,
+      "grad_norm": 3.947331428527832,
+      "learning_rate": 0.0001975588431933897,
+      "loss": 1.3823,
+      "step": 8550
+    },
+    {
+      "epoch": 0.25677007135819424,
+      "grad_norm": 3.2268636226654053,
+      "learning_rate": 0.00019754455739747084,
+      "loss": 1.4199,
+      "step": 8600
+    },
+    {
+      "epoch": 0.2582629206102768,
+      "grad_norm": 4.0353102684021,
+      "learning_rate": 0.00019753027160155203,
+      "loss": 1.3927,
+      "step": 8650
+    },
+    {
+      "epoch": 0.2597557698623593,
+      "grad_norm": 3.490560293197632,
+      "learning_rate": 0.00019751598580563316,
+      "loss": 1.401,
+      "step": 8700
+    },
+    {
+      "epoch": 0.2612486191144418,
+      "grad_norm": 5.577207088470459,
+      "learning_rate": 0.00019750170000971436,
+      "loss": 1.3586,
+      "step": 8750
+    },
+    {
+      "epoch": 0.26274146836652434,
+      "grad_norm": 4.168467998504639,
+      "learning_rate": 0.0001974874142137955,
+      "loss": 1.3303,
+      "step": 8800
+    },
+    {
+      "epoch": 0.2642343176186069,
+      "grad_norm": 3.812627077102661,
+      "learning_rate": 0.00019747312841787669,
+      "loss": 1.3717,
+      "step": 8850
+    },
+    {
+      "epoch": 0.2657271668706894,
+      "grad_norm": 4.875237464904785,
+      "learning_rate": 0.00019745884262195785,
+      "loss": 1.3873,
+      "step": 8900
+    },
+    {
+      "epoch": 0.2672200161227719,
+      "grad_norm": 4.048189163208008,
+      "learning_rate": 0.00019744455682603902,
+      "loss": 1.3775,
+      "step": 8950
+    },
+    {
+      "epoch": 0.26871286537485445,
+      "grad_norm": 3.9090261459350586,
+      "learning_rate": 0.00019743027103012018,
+      "loss": 1.4296,
+      "step": 9000
+    },
+    {
+      "epoch": 0.270205714626937,
+      "grad_norm": 2.8476953506469727,
+      "learning_rate": 0.00019741598523420134,
+      "loss": 1.4175,
+      "step": 9050
+    },
+    {
+      "epoch": 0.2716985638790195,
+      "grad_norm": 5.782102584838867,
+      "learning_rate": 0.0001974016994382825,
+      "loss": 1.3835,
+      "step": 9100
+    },
+    {
+      "epoch": 0.273191413131102,
+      "grad_norm": 4.640264987945557,
+      "learning_rate": 0.00019738741364236367,
+      "loss": 1.4524,
+      "step": 9150
+    },
+    {
+      "epoch": 0.27468426238318455,
+      "grad_norm": 4.81790828704834,
+      "learning_rate": 0.00019737312784644484,
+      "loss": 1.3183,
+      "step": 9200
+    },
+    {
+      "epoch": 0.2761771116352671,
+      "grad_norm": 2.685009717941284,
+      "learning_rate": 0.000197358842050526,
+      "loss": 1.3243,
+      "step": 9250
+    },
+    {
+      "epoch": 0.2776699608873496,
+      "grad_norm": 5.321321487426758,
+      "learning_rate": 0.00019734455625460717,
+      "loss": 1.4086,
+      "step": 9300
+    },
+    {
+      "epoch": 0.2791628101394321,
+      "grad_norm": 3.065791368484497,
+      "learning_rate": 0.00019733027045868836,
+      "loss": 1.337,
+      "step": 9350
+    },
+    {
+      "epoch": 0.28065565939151466,
+      "grad_norm": 4.3569817543029785,
+      "learning_rate": 0.0001973159846627695,
+      "loss": 1.4082,
+      "step": 9400
+    },
+    {
+      "epoch": 0.2821485086435972,
+      "grad_norm": 4.67582368850708,
+      "learning_rate": 0.0001973016988668507,
+      "loss": 1.3832,
+      "step": 9450
+    },
+    {
+      "epoch": 0.2836413578956797,
+      "grad_norm": 4.942144870758057,
+      "learning_rate": 0.00019728741307093183,
+      "loss": 1.3734,
+      "step": 9500
+    },
+    {
+      "epoch": 0.2851342071477622,
+      "grad_norm": 4.853246688842773,
+      "learning_rate": 0.00019727312727501302,
+      "loss": 1.4111,
+      "step": 9550
+    },
+    {
+      "epoch": 0.28662705639984476,
+      "grad_norm": 3.071237325668335,
+      "learning_rate": 0.00019725884147909418,
+      "loss": 1.3746,
+      "step": 9600
+    },
+    {
+      "epoch": 0.28811990565192724,
+      "grad_norm": 4.844615459442139,
+      "learning_rate": 0.00019724455568317535,
+      "loss": 1.3051,
+      "step": 9650
+    },
+    {
+      "epoch": 0.2896127549040098,
+      "grad_norm": 5.954223155975342,
+      "learning_rate": 0.0001972302698872565,
+      "loss": 1.4131,
+      "step": 9700
+    },
+    {
+      "epoch": 0.2911056041560923,
+      "grad_norm": 3.6717801094055176,
+      "learning_rate": 0.00019721598409133768,
+      "loss": 1.4166,
+      "step": 9750
+    },
+    {
+      "epoch": 0.29259845340817486,
+      "grad_norm": 3.6257095336914062,
+      "learning_rate": 0.00019720169829541884,
+      "loss": 1.3679,
+      "step": 9800
+    },
+    {
+      "epoch": 0.29409130266025735,
+      "grad_norm": 4.245635032653809,
+      "learning_rate": 0.0001971874124995,
+      "loss": 1.3171,
+      "step": 9850
+    },
+    {
+      "epoch": 0.2955841519123399,
+      "grad_norm": 5.362602710723877,
+      "learning_rate": 0.00019717312670358117,
+      "loss": 1.3932,
+      "step": 9900
+    },
+    {
+      "epoch": 0.2970770011644224,
+      "grad_norm": 4.6283721923828125,
+      "learning_rate": 0.00019715884090766234,
+      "loss": 1.3757,
+      "step": 9950
+    },
+    {
+      "epoch": 0.29856985041650497,
+      "grad_norm": 4.299574851989746,
+      "learning_rate": 0.0001971445551117435,
+      "loss": 1.4018,
+      "step": 10000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 700001,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 21,
+  "save_steps": 10000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.535695686786089e+17,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

last-checkpoint/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0f04c18880a48291999c17738bf0f62f2b6e23d1307ab6d608fabd8b061c2af
+size 6968

last-checkpoint/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)