Training in progress, step 179, checkpoint

checkpoint-179/README.md

@@ -0,0 +1,202 @@
+---
+base_model: hardlyworking/Noodles-Merge-12B
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.15.1

checkpoint-179/adapter_config.json

@@ -0,0 +1,42 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "hardlyworking/Noodles-Merge-12B",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": [
+    "embed_tokens",
+    "lm_head"
+  ],
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "down_proj",
+    "gate_proj",
+    "up_proj",
+    "k_proj",
+    "q_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": true
+}

checkpoint-179/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b292b880d04bb776b8c3167fca8ae800e642d797c3c8d01489639b6522dd85d
+size 3596694008

checkpoint-179/global_step179/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31f79bb3cdc8f199bb51c23b657b0dd377097ab7c5eb6371460f864f70b46e99
+size 2712459662

checkpoint-179/global_step179/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25a5cd5370a51df03d145fbbed5b76c8779368cd5398b1cc947887a25d8a834f
+size 2712459662

checkpoint-179/global_step179/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89202a61ba1c1becfdc90e59fbdb3345d983c12e2bf25d0685d5b07904fdc797
+size 2712459662

checkpoint-179/global_step179/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98b6ac4d2366b7784a9debfab37108c209a06092fdc5e717148ea58a7b4f0559
+size 2712459662

checkpoint-179/global_step179/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18aceac870f68d7941f76cfb101bf8f200b8e03f2143e2b338191e5d46bbbbb0
+size 558360

checkpoint-179/global_step179/zero_pp_rank_1_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2668fd56da0a78e07d1546ff5fc939e355559c85f7960b93ffb481ebf41353c4
+size 558360

checkpoint-179/global_step179/zero_pp_rank_2_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ffafd99e3a841cdbdd0a1475a97baeb68d388860719c4d01160e5adb235f699
+size 558360

checkpoint-179/global_step179/zero_pp_rank_3_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d15193d1ce12ab20826317163f45c510db408a2c7c08c8531f1d8ac371fdbacc
+size 558360

checkpoint-179/latest

@@ -0,0 +1 @@
+global_step179

checkpoint-179/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5b55c2673c956f8bcadf505e6f8c669a3242704c70f78654a072adc3aff9775
+size 15024

checkpoint-179/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b010356c5eb050eb79db6e5d22dd8a0ff3f866de8d2ce0103d58afbf73694ad8
+size 15024

checkpoint-179/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61d85a2b89442eb6e7d3dc16abb0f82563235ac85a5fc07ed0a2463c855c6d81
+size 15024

checkpoint-179/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c550a27f668f3d4432e465ccf1b7ca4eef2528a8a5a16b70d05021b40c41a9a
+size 15024

checkpoint-179/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ecb9a5e224389470f0749774a340fad57d06dd231cbde400e43d9eddb912010
+size 1064

checkpoint-179/special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-179/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b4c8fcd33487a449c07f423d47adb035bba8347ccf13eb074b4d1fef8acf919
+size 17078288

checkpoint-179/trainer_state.json

@@ -0,0 +1,1319 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.499302649930265,
+  "eval_steps": 45,
+  "global_step": 179,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.002789400278940028,
+      "grad_norm": 2.008944315843832,
+      "learning_rate": 0.0,
+      "loss": 1.7653,
+      "step": 1
+    },
+    {
+      "epoch": 0.002789400278940028,
+      "eval_loss": 1.7864996194839478,
+      "eval_runtime": 155.6922,
+      "eval_samples_per_second": 2.042,
+      "eval_steps_per_second": 0.128,
+      "step": 1
+    },
+    {
+      "epoch": 0.005578800557880056,
+      "grad_norm": 1.5986956798562522,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 1.3721,
+      "step": 2
+    },
+    {
+      "epoch": 0.008368200836820083,
+      "grad_norm": 2.487093074085383,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 1.4468,
+      "step": 3
+    },
+    {
+      "epoch": 0.011157601115760111,
+      "grad_norm": 1.9408908369837041,
+      "learning_rate": 1.5e-06,
+      "loss": 1.8555,
+      "step": 4
+    },
+    {
+      "epoch": 0.01394700139470014,
+      "grad_norm": 1.855397723025267,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 1.5173,
+      "step": 5
+    },
+    {
+      "epoch": 0.016736401673640166,
+      "grad_norm": 1.6798170163537172,
+      "learning_rate": 2.5e-06,
+      "loss": 1.3812,
+      "step": 6
+    },
+    {
+      "epoch": 0.019525801952580194,
+      "grad_norm": 2.0478066370088133,
+      "learning_rate": 3e-06,
+      "loss": 1.475,
+      "step": 7
+    },
+    {
+      "epoch": 0.022315202231520222,
+      "grad_norm": 1.6541684745930123,
+      "learning_rate": 3.5e-06,
+      "loss": 1.6239,
+      "step": 8
+    },
+    {
+      "epoch": 0.02510460251046025,
+      "grad_norm": 1.5940373225900581,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 1.4474,
+      "step": 9
+    },
+    {
+      "epoch": 0.02789400278940028,
+      "grad_norm": 1.3436762031717087,
+      "learning_rate": 4.5e-06,
+      "loss": 1.5277,
+      "step": 10
+    },
+    {
+      "epoch": 0.030683403068340307,
+      "grad_norm": 1.4820988105255548,
+      "learning_rate": 5e-06,
+      "loss": 1.3905,
+      "step": 11
+    },
+    {
+      "epoch": 0.03347280334728033,
+      "grad_norm": 1.3026823080341792,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 1.2716,
+      "step": 12
+    },
+    {
+      "epoch": 0.03626220362622036,
+      "grad_norm": 1.2786618249049673,
+      "learning_rate": 6e-06,
+      "loss": 1.4357,
+      "step": 13
+    },
+    {
+      "epoch": 0.03905160390516039,
+      "grad_norm": 1.4401448629219828,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 1.5418,
+      "step": 14
+    },
+    {
+      "epoch": 0.04184100418410042,
+      "grad_norm": 1.4108691585319235,
+      "learning_rate": 7e-06,
+      "loss": 1.419,
+      "step": 15
+    },
+    {
+      "epoch": 0.044630404463040445,
+      "grad_norm": 1.434665713671276,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 1.4042,
+      "step": 16
+    },
+    {
+      "epoch": 0.04741980474198047,
+      "grad_norm": 1.4864898541989087,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.4338,
+      "step": 17
+    },
+    {
+      "epoch": 0.0502092050209205,
+      "grad_norm": 1.7497952919365851,
+      "learning_rate": 8.5e-06,
+      "loss": 1.5988,
+      "step": 18
+    },
+    {
+      "epoch": 0.05299860529986053,
+      "grad_norm": 1.5825124800137,
+      "learning_rate": 9e-06,
+      "loss": 1.4817,
+      "step": 19
+    },
+    {
+      "epoch": 0.05578800557880056,
+      "grad_norm": 1.5639582937511203,
+      "learning_rate": 9.5e-06,
+      "loss": 1.3795,
+      "step": 20
+    },
+    {
+      "epoch": 0.058577405857740586,
+      "grad_norm": 1.5512627729747626,
+      "learning_rate": 1e-05,
+      "loss": 1.4833,
+      "step": 21
+    },
+    {
+      "epoch": 0.061366806136680614,
+      "grad_norm": 1.4462678614801439,
+      "learning_rate": 1.0500000000000001e-05,
+      "loss": 1.3781,
+      "step": 22
+    },
+    {
+      "epoch": 0.06415620641562064,
+      "grad_norm": 1.5995776178618928,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 1.4862,
+      "step": 23
+    },
+    {
+      "epoch": 0.06694560669456066,
+      "grad_norm": 1.2623925872690254,
+      "learning_rate": 1.15e-05,
+      "loss": 1.4222,
+      "step": 24
+    },
+    {
+      "epoch": 0.0697350069735007,
+      "grad_norm": 1.2816346894508914,
+      "learning_rate": 1.2e-05,
+      "loss": 1.3473,
+      "step": 25
+    },
+    {
+      "epoch": 0.07252440725244072,
+      "grad_norm": 1.4528035335148701,
+      "learning_rate": 1.25e-05,
+      "loss": 1.5278,
+      "step": 26
+    },
+    {
+      "epoch": 0.07531380753138076,
+      "grad_norm": 1.3415339044507064,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 1.3939,
+      "step": 27
+    },
+    {
+      "epoch": 0.07810320781032078,
+      "grad_norm": 1.6072610009319113,
+      "learning_rate": 1.3500000000000001e-05,
+      "loss": 1.5401,
+      "step": 28
+    },
+    {
+      "epoch": 0.08089260808926081,
+      "grad_norm": 1.5988215139386146,
+      "learning_rate": 1.4e-05,
+      "loss": 1.3995,
+      "step": 29
+    },
+    {
+      "epoch": 0.08368200836820083,
+      "grad_norm": 1.4507828208303803,
+      "learning_rate": 1.45e-05,
+      "loss": 1.4584,
+      "step": 30
+    },
+    {
+      "epoch": 0.08647140864714087,
+      "grad_norm": 1.4975002404633069,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 1.3834,
+      "step": 31
+    },
+    {
+      "epoch": 0.08926080892608089,
+      "grad_norm": 1.4128570529331854,
+      "learning_rate": 1.55e-05,
+      "loss": 1.4216,
+      "step": 32
+    },
+    {
+      "epoch": 0.09205020920502092,
+      "grad_norm": 1.3616495178863832,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 1.5167,
+      "step": 33
+    },
+    {
+      "epoch": 0.09483960948396095,
+      "grad_norm": 1.1988790310990909,
+      "learning_rate": 1.65e-05,
+      "loss": 1.4525,
+      "step": 34
+    },
+    {
+      "epoch": 0.09762900976290098,
+      "grad_norm": 1.4483022749153813,
+      "learning_rate": 1.7e-05,
+      "loss": 1.5741,
+      "step": 35
+    },
+    {
+      "epoch": 0.100418410041841,
+      "grad_norm": 1.394589777246842,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 1.6437,
+      "step": 36
+    },
+    {
+      "epoch": 0.10320781032078104,
+      "grad_norm": 1.4186049044929165,
+      "learning_rate": 1.8e-05,
+      "loss": 1.5162,
+      "step": 37
+    },
+    {
+      "epoch": 0.10599721059972106,
+      "grad_norm": 1.4241876007622027,
+      "learning_rate": 1.8500000000000002e-05,
+      "loss": 1.2395,
+      "step": 38
+    },
+    {
+      "epoch": 0.1087866108786611,
+      "grad_norm": 1.3743075911034852,
+      "learning_rate": 1.9e-05,
+      "loss": 1.4631,
+      "step": 39
+    },
+    {
+      "epoch": 0.11157601115760112,
+      "grad_norm": 1.2863497247985725,
+      "learning_rate": 1.95e-05,
+      "loss": 1.3611,
+      "step": 40
+    },
+    {
+      "epoch": 0.11436541143654114,
+      "grad_norm": 1.3054075793205129,
+      "learning_rate": 2e-05,
+      "loss": 1.2375,
+      "step": 41
+    },
+    {
+      "epoch": 0.11715481171548117,
+      "grad_norm": 1.447249035073335,
+      "learning_rate": 1.9999892011980525e-05,
+      "loss": 1.4044,
+      "step": 42
+    },
+    {
+      "epoch": 0.1199442119944212,
+      "grad_norm": 1.3461748028257197,
+      "learning_rate": 1.9999568050254373e-05,
+      "loss": 1.4081,
+      "step": 43
+    },
+    {
+      "epoch": 0.12273361227336123,
+      "grad_norm": 1.6035229132022362,
+      "learning_rate": 1.999902812181835e-05,
+      "loss": 1.4167,
+      "step": 44
+    },
+    {
+      "epoch": 0.12552301255230125,
+      "grad_norm": 1.7750306011999208,
+      "learning_rate": 1.9998272238333606e-05,
+      "loss": 1.3533,
+      "step": 45
+    },
+    {
+      "epoch": 0.12552301255230125,
+      "eval_loss": 1.6828460693359375,
+      "eval_runtime": 158.4193,
+      "eval_samples_per_second": 2.007,
+      "eval_steps_per_second": 0.126,
+      "step": 45
+    },
+    {
+      "epoch": 0.12831241283124128,
+      "grad_norm": 1.354470787639867,
+      "learning_rate": 1.9997300416125426e-05,
+      "loss": 1.3901,
+      "step": 46
+    },
+    {
+      "epoch": 0.13110181311018132,
+      "grad_norm": 1.23248602443203,
+      "learning_rate": 1.999611267618283e-05,
+      "loss": 1.3485,
+      "step": 47
+    },
+    {
+      "epoch": 0.13389121338912133,
+      "grad_norm": 1.314827114627873,
+      "learning_rate": 1.9994709044158157e-05,
+      "loss": 1.3195,
+      "step": 48
+    },
+    {
+      "epoch": 0.13668061366806136,
+      "grad_norm": 1.3741001015735983,
+      "learning_rate": 1.99930895503665e-05,
+      "loss": 1.4394,
+      "step": 49
+    },
+    {
+      "epoch": 0.1394700139470014,
+      "grad_norm": 1.1624406309888304,
+      "learning_rate": 1.9991254229785043e-05,
+      "loss": 1.2943,
+      "step": 50
+    },
+    {
+      "epoch": 0.14225941422594143,
+      "grad_norm": 1.431188759221728,
+      "learning_rate": 1.998920312205231e-05,
+      "loss": 1.4614,
+      "step": 51
+    },
+    {
+      "epoch": 0.14504881450488144,
+      "grad_norm": 1.2513371987994648,
+      "learning_rate": 1.9986936271467316e-05,
+      "loss": 1.328,
+      "step": 52
+    },
+    {
+      "epoch": 0.14783821478382148,
+      "grad_norm": 1.6084481811469094,
+      "learning_rate": 1.99844537269886e-05,
+      "loss": 1.4579,
+      "step": 53
+    },
+    {
+      "epoch": 0.1506276150627615,
+      "grad_norm": 1.3643143080926854,
+      "learning_rate": 1.9981755542233175e-05,
+      "loss": 1.4471,
+      "step": 54
+    },
+    {
+      "epoch": 0.15341701534170155,
+      "grad_norm": 1.443453372098446,
+      "learning_rate": 1.9978841775475368e-05,
+      "loss": 1.3931,
+      "step": 55
+    },
+    {
+      "epoch": 0.15620641562064155,
+      "grad_norm": 1.186174044031903,
+      "learning_rate": 1.997571248964556e-05,
+      "loss": 1.2935,
+      "step": 56
+    },
+    {
+      "epoch": 0.1589958158995816,
+      "grad_norm": 1.3967845728055575,
+      "learning_rate": 1.9972367752328824e-05,
+      "loss": 1.3819,
+      "step": 57
+    },
+    {
+      "epoch": 0.16178521617852162,
+      "grad_norm": 1.585201984076838,
+      "learning_rate": 1.9968807635763472e-05,
+      "loss": 1.5391,
+      "step": 58
+    },
+    {
+      "epoch": 0.16457461645746166,
+      "grad_norm": 1.515327450581085,
+      "learning_rate": 1.9965032216839493e-05,
+      "loss": 1.3969,
+      "step": 59
+    },
+    {
+      "epoch": 0.16736401673640167,
+      "grad_norm": 1.3595894447160555,
+      "learning_rate": 1.996104157709689e-05,
+      "loss": 1.3598,
+      "step": 60
+    },
+    {
+      "epoch": 0.1701534170153417,
+      "grad_norm": 1.4252798961269173,
+      "learning_rate": 1.9956835802723916e-05,
+      "loss": 1.447,
+      "step": 61
+    },
+    {
+      "epoch": 0.17294281729428174,
+      "grad_norm": 1.4269613254937963,
+      "learning_rate": 1.9952414984555225e-05,
+      "loss": 1.4354,
+      "step": 62
+    },
+    {
+      "epoch": 0.17573221757322174,
+      "grad_norm": 1.5984564978735487,
+      "learning_rate": 1.994777921806989e-05,
+      "loss": 1.5493,
+      "step": 63
+    },
+    {
+      "epoch": 0.17852161785216178,
+      "grad_norm": 1.5396475514798837,
+      "learning_rate": 1.9942928603389366e-05,
+      "loss": 1.4095,
+      "step": 64
+    },
+    {
+      "epoch": 0.18131101813110181,
+      "grad_norm": 1.381818615640278,
+      "learning_rate": 1.9937863245275303e-05,
+      "loss": 1.381,
+      "step": 65
+    },
+    {
+      "epoch": 0.18410041841004185,
+      "grad_norm": 1.3411933307262995,
+      "learning_rate": 1.9932583253127302e-05,
+      "loss": 1.4514,
+      "step": 66
+    },
+    {
+      "epoch": 0.18688981868898186,
+      "grad_norm": 1.4666636741675445,
+      "learning_rate": 1.992708874098054e-05,
+      "loss": 1.378,
+      "step": 67
+    },
+    {
+      "epoch": 0.1896792189679219,
+      "grad_norm": 1.321297415086869,
+      "learning_rate": 1.9921379827503316e-05,
+      "loss": 1.3916,
+      "step": 68
+    },
+    {
+      "epoch": 0.19246861924686193,
+      "grad_norm": 1.3031514888132292,
+      "learning_rate": 1.991545663599448e-05,
+      "loss": 1.3568,
+      "step": 69
+    },
+    {
+      "epoch": 0.19525801952580196,
+      "grad_norm": 1.3252501098302665,
+      "learning_rate": 1.990931929438078e-05,
+      "loss": 1.3691,
+      "step": 70
+    },
+    {
+      "epoch": 0.19804741980474197,
+      "grad_norm": 1.3190300203004819,
+      "learning_rate": 1.990296793521408e-05,
+      "loss": 1.425,
+      "step": 71
+    },
+    {
+      "epoch": 0.200836820083682,
+      "grad_norm": 1.221728638105595,
+      "learning_rate": 1.989640269566853e-05,
+      "loss": 1.3669,
+      "step": 72
+    },
+    {
+      "epoch": 0.20362622036262204,
+      "grad_norm": 1.2519446160682244,
+      "learning_rate": 1.9889623717537564e-05,
+      "loss": 1.5065,
+      "step": 73
+    },
+    {
+      "epoch": 0.20641562064156208,
+      "grad_norm": 1.3192555240060455,
+      "learning_rate": 1.9882631147230874e-05,
+      "loss": 1.3077,
+      "step": 74
+    },
+    {
+      "epoch": 0.20920502092050208,
+      "grad_norm": 1.811281511932427,
+      "learning_rate": 1.987542513577122e-05,
+      "loss": 1.4737,
+      "step": 75
+    },
+    {
+      "epoch": 0.21199442119944212,
+      "grad_norm": 1.2107466822843787,
+      "learning_rate": 1.9868005838791185e-05,
+      "loss": 1.3632,
+      "step": 76
+    },
+    {
+      "epoch": 0.21478382147838215,
+      "grad_norm": 1.4506886695952612,
+      "learning_rate": 1.9860373416529804e-05,
+      "loss": 1.3707,
+      "step": 77
+    },
+    {
+      "epoch": 0.2175732217573222,
+      "grad_norm": 1.129475911714987,
+      "learning_rate": 1.985252803382911e-05,
+      "loss": 1.2879,
+      "step": 78
+    },
+    {
+      "epoch": 0.2203626220362622,
+      "grad_norm": 1.2378410441125478,
+      "learning_rate": 1.984446986013057e-05,
+      "loss": 1.4352,
+      "step": 79
+    },
+    {
+      "epoch": 0.22315202231520223,
+      "grad_norm": 1.2497274548196218,
+      "learning_rate": 1.983619906947144e-05,
+      "loss": 1.2929,
+      "step": 80
+    },
+    {
+      "epoch": 0.22594142259414227,
+      "grad_norm": 1.2860272115066647,
+      "learning_rate": 1.9827715840480962e-05,
+      "loss": 1.2595,
+      "step": 81
+    },
+    {
+      "epoch": 0.22873082287308227,
+      "grad_norm": 1.3259601590310435,
+      "learning_rate": 1.9819020356376562e-05,
+      "loss": 1.5421,
+      "step": 82
+    },
+    {
+      "epoch": 0.2315202231520223,
+      "grad_norm": 1.3823289295025785,
+      "learning_rate": 1.9810112804959867e-05,
+      "loss": 1.4379,
+      "step": 83
+    },
+    {
+      "epoch": 0.23430962343096234,
+      "grad_norm": 1.2790636696723037,
+      "learning_rate": 1.980099337861264e-05,
+      "loss": 1.4045,
+      "step": 84
+    },
+    {
+      "epoch": 0.23709902370990238,
+      "grad_norm": 1.2621634213682735,
+      "learning_rate": 1.9791662274292638e-05,
+      "loss": 1.3433,
+      "step": 85
+    },
+    {
+      "epoch": 0.2398884239888424,
+      "grad_norm": 1.4041226411604562,
+      "learning_rate": 1.9782119693529358e-05,
+      "loss": 1.6292,
+      "step": 86
+    },
+    {
+      "epoch": 0.24267782426778242,
+      "grad_norm": 1.4238393471976662,
+      "learning_rate": 1.977236584241968e-05,
+      "loss": 1.398,
+      "step": 87
+    },
+    {
+      "epoch": 0.24546722454672246,
+      "grad_norm": 1.3721757997449184,
+      "learning_rate": 1.9762400931623413e-05,
+      "loss": 1.372,
+      "step": 88
+    },
+    {
+      "epoch": 0.2482566248256625,
+      "grad_norm": 1.3049726438204359,
+      "learning_rate": 1.9752225176358757e-05,
+      "loss": 1.4366,
+      "step": 89
+    },
+    {
+      "epoch": 0.2510460251046025,
+      "grad_norm": 1.319316702022308,
+      "learning_rate": 1.9741838796397638e-05,
+      "loss": 1.2807,
+      "step": 90
+    },
+    {
+      "epoch": 0.2510460251046025,
+      "eval_loss": 1.6544907093048096,
+      "eval_runtime": 157.4295,
+      "eval_samples_per_second": 2.02,
+      "eval_steps_per_second": 0.127,
+      "step": 90
+    },
+    {
+      "epoch": 0.25383542538354253,
+      "grad_norm": 1.3259568867589984,
+      "learning_rate": 1.9731242016060985e-05,
+      "loss": 1.3566,
+      "step": 91
+    },
+    {
+      "epoch": 0.25662482566248257,
+      "grad_norm": 1.3318698765554446,
+      "learning_rate": 1.972043506421386e-05,
+      "loss": 1.3304,
+      "step": 92
+    },
+    {
+      "epoch": 0.2594142259414226,
+      "grad_norm": 1.210762982291795,
+      "learning_rate": 1.9709418174260523e-05,
+      "loss": 1.3663,
+      "step": 93
+    },
+    {
+      "epoch": 0.26220362622036264,
+      "grad_norm": 1.322366639833671,
+      "learning_rate": 1.9698191584139402e-05,
+      "loss": 1.4099,
+      "step": 94
+    },
+    {
+      "epoch": 0.2649930264993027,
+      "grad_norm": 1.286954377653034,
+      "learning_rate": 1.9686755536317945e-05,
+      "loss": 1.3539,
+      "step": 95
+    },
+    {
+      "epoch": 0.26778242677824265,
+      "grad_norm": 1.1590651579123286,
+      "learning_rate": 1.967511027778738e-05,
+      "loss": 1.2147,
+      "step": 96
+    },
+    {
+      "epoch": 0.2705718270571827,
+      "grad_norm": 1.3100829080105303,
+      "learning_rate": 1.9663256060057395e-05,
+      "loss": 1.4093,
+      "step": 97
+    },
+    {
+      "epoch": 0.2733612273361227,
+      "grad_norm": 1.346265752760783,
+      "learning_rate": 1.965119313915068e-05,
+      "loss": 1.491,
+      "step": 98
+    },
+    {
+      "epoch": 0.27615062761506276,
+      "grad_norm": 1.378168477135006,
+      "learning_rate": 1.9638921775597428e-05,
+      "loss": 1.5049,
+      "step": 99
+    },
+    {
+      "epoch": 0.2789400278940028,
+      "grad_norm": 1.4182107155054933,
+      "learning_rate": 1.9626442234429684e-05,
+      "loss": 1.5013,
+      "step": 100
+    },
+    {
+      "epoch": 0.28172942817294283,
+      "grad_norm": 1.3165711654218712,
+      "learning_rate": 1.961375478517564e-05,
+      "loss": 1.5105,
+      "step": 101
+    },
+    {
+      "epoch": 0.28451882845188287,
+      "grad_norm": 1.364780296172199,
+      "learning_rate": 1.9600859701853796e-05,
+      "loss": 1.5207,
+      "step": 102
+    },
+    {
+      "epoch": 0.28730822873082285,
+      "grad_norm": 1.2988922679076684,
+      "learning_rate": 1.958775726296706e-05,
+      "loss": 1.2881,
+      "step": 103
+    },
+    {
+      "epoch": 0.2900976290097629,
+      "grad_norm": 1.3450528168801636,
+      "learning_rate": 1.9574447751496706e-05,
+      "loss": 1.2877,
+      "step": 104
+    },
+    {
+      "epoch": 0.2928870292887029,
+      "grad_norm": 1.4142239844479165,
+      "learning_rate": 1.95609314548963e-05,
+      "loss": 1.5604,
+      "step": 105
+    },
+    {
+      "epoch": 0.29567642956764295,
+      "grad_norm": 1.3452993293539794,
+      "learning_rate": 1.954720866508546e-05,
+      "loss": 1.3447,
+      "step": 106
+    },
+    {
+      "epoch": 0.298465829846583,
+      "grad_norm": 1.4534491830895881,
+      "learning_rate": 1.953327967844356e-05,
+      "loss": 1.3643,
+      "step": 107
+    },
+    {
+      "epoch": 0.301255230125523,
+      "grad_norm": 1.350966370835706,
+      "learning_rate": 1.9519144795803342e-05,
+      "loss": 1.5196,
+      "step": 108
+    },
+    {
+      "epoch": 0.30404463040446306,
+      "grad_norm": 1.3487778511447879,
+      "learning_rate": 1.95048043224444e-05,
+      "loss": 1.4243,
+      "step": 109
+    },
+    {
+      "epoch": 0.3068340306834031,
+      "grad_norm": 1.3479508491141639,
+      "learning_rate": 1.94902585680866e-05,
+      "loss": 1.2869,
+      "step": 110
+    },
+    {
+      "epoch": 0.30962343096234307,
+      "grad_norm": 1.5555627133101146,
+      "learning_rate": 1.9475507846883377e-05,
+      "loss": 1.4626,
+      "step": 111
+    },
+    {
+      "epoch": 0.3124128312412831,
+      "grad_norm": 1.264890715784298,
+      "learning_rate": 1.9460552477414972e-05,
+      "loss": 1.2826,
+      "step": 112
+    },
+    {
+      "epoch": 0.31520223152022314,
+      "grad_norm": 1.2628748380778831,
+      "learning_rate": 1.9445392782681523e-05,
+      "loss": 1.232,
+      "step": 113
+    },
+    {
+      "epoch": 0.3179916317991632,
+      "grad_norm": 1.3550404859387195,
+      "learning_rate": 1.9430029090096118e-05,
+      "loss": 1.3034,
+      "step": 114
+    },
+    {
+      "epoch": 0.3207810320781032,
+      "grad_norm": 1.1496127251687267,
+      "learning_rate": 1.94144617314777e-05,
+      "loss": 1.3215,
+      "step": 115
+    },
+    {
+      "epoch": 0.32357043235704325,
+      "grad_norm": 1.3116085375581363,
+      "learning_rate": 1.939869104304392e-05,
+      "loss": 1.4154,
+      "step": 116
+    },
+    {
+      "epoch": 0.3263598326359833,
+      "grad_norm": 1.2589366985022925,
+      "learning_rate": 1.9382717365403854e-05,
+      "loss": 1.517,
+      "step": 117
+    },
+    {
+      "epoch": 0.3291492329149233,
+      "grad_norm": 1.3051575397158521,
+      "learning_rate": 1.9366541043550667e-05,
+      "loss": 1.3795,
+      "step": 118
+    },
+    {
+      "epoch": 0.3319386331938633,
+      "grad_norm": 1.514184766493585,
+      "learning_rate": 1.9350162426854152e-05,
+      "loss": 1.5974,
+      "step": 119
+    },
+    {
+      "epoch": 0.33472803347280333,
+      "grad_norm": 1.3773144705212048,
+      "learning_rate": 1.933358186905318e-05,
+      "loss": 1.3696,
+      "step": 120
+    },
+    {
+      "epoch": 0.33751743375174337,
+      "grad_norm": 1.070700275666401,
+      "learning_rate": 1.9316799728248074e-05,
+      "loss": 1.1661,
+      "step": 121
+    },
+    {
+      "epoch": 0.3403068340306834,
+      "grad_norm": 1.4181760004495874,
+      "learning_rate": 1.9299816366892865e-05,
+      "loss": 1.4376,
+      "step": 122
+    },
+    {
+      "epoch": 0.34309623430962344,
+      "grad_norm": 1.6034791315185046,
+      "learning_rate": 1.9282632151787462e-05,
+      "loss": 1.5891,
+      "step": 123
+    },
+    {
+      "epoch": 0.3458856345885635,
+      "grad_norm": 1.1884570183288778,
+      "learning_rate": 1.9265247454069736e-05,
+      "loss": 1.3432,
+      "step": 124
+    },
+    {
+      "epoch": 0.3486750348675035,
+      "grad_norm": 1.477555714731358,
+      "learning_rate": 1.924766264920751e-05,
+      "loss": 1.3683,
+      "step": 125
+    },
+    {
+      "epoch": 0.3514644351464435,
+      "grad_norm": 1.2504857458807423,
+      "learning_rate": 1.922987811699042e-05,
+      "loss": 1.2643,
+      "step": 126
+    },
+    {
+      "epoch": 0.3542538354253835,
+      "grad_norm": 1.3679090790056112,
+      "learning_rate": 1.9211894241521757e-05,
+      "loss": 1.5962,
+      "step": 127
+    },
+    {
+      "epoch": 0.35704323570432356,
+      "grad_norm": 1.2309836816544464,
+      "learning_rate": 1.9193711411210138e-05,
+      "loss": 1.3361,
+      "step": 128
+    },
+    {
+      "epoch": 0.3598326359832636,
+      "grad_norm": 1.1714321663378802,
+      "learning_rate": 1.917533001876113e-05,
+      "loss": 1.3019,
+      "step": 129
+    },
+    {
+      "epoch": 0.36262203626220363,
+      "grad_norm": 1.2813447334304535,
+      "learning_rate": 1.9156750461168768e-05,
+      "loss": 1.4319,
+      "step": 130
+    },
+    {
+      "epoch": 0.36541143654114366,
+      "grad_norm": 1.427767209588848,
+      "learning_rate": 1.9137973139706973e-05,
+      "loss": 1.4712,
+      "step": 131
+    },
+    {
+      "epoch": 0.3682008368200837,
+      "grad_norm": 1.3426640935974619,
+      "learning_rate": 1.91189984599209e-05,
+      "loss": 1.4397,
+      "step": 132
+    },
+    {
+      "epoch": 0.37099023709902373,
+      "grad_norm": 1.3173830250699687,
+      "learning_rate": 1.9099826831618168e-05,
+      "loss": 1.4799,
+      "step": 133
+    },
+    {
+      "epoch": 0.3737796373779637,
+      "grad_norm": 1.5260053920559482,
+      "learning_rate": 1.908045866886001e-05,
+      "loss": 1.5167,
+      "step": 134
+    },
+    {
+      "epoch": 0.37656903765690375,
+      "grad_norm": 1.2838799058228236,
+      "learning_rate": 1.9060894389952328e-05,
+      "loss": 1.3957,
+      "step": 135
+    },
+    {
+      "epoch": 0.37656903765690375,
+      "eval_loss": 1.629968523979187,
+      "eval_runtime": 155.8796,
+      "eval_samples_per_second": 2.04,
+      "eval_steps_per_second": 0.128,
+      "step": 135
+    },
+    {
+      "epoch": 0.3793584379358438,
+      "grad_norm": 1.2595723963522658,
+      "learning_rate": 1.9041134417436674e-05,
+      "loss": 1.3773,
+      "step": 136
+    },
+    {
+      "epoch": 0.3821478382147838,
+      "grad_norm": 1.380977078216302,
+      "learning_rate": 1.9021179178081107e-05,
+      "loss": 1.5358,
+      "step": 137
+    },
+    {
+      "epoch": 0.38493723849372385,
+      "grad_norm": 1.3245015694120374,
+      "learning_rate": 1.9001029102870982e-05,
+      "loss": 1.506,
+      "step": 138
+    },
+    {
+      "epoch": 0.3877266387726639,
+      "grad_norm": 1.2602456347612,
+      "learning_rate": 1.898068462699964e-05,
+      "loss": 1.4246,
+      "step": 139
+    },
+    {
+      "epoch": 0.3905160390516039,
+      "grad_norm": 1.3989105118625398,
+      "learning_rate": 1.8960146189859014e-05,
+      "loss": 1.4692,
+      "step": 140
+    },
+    {
+      "epoch": 0.39330543933054396,
+      "grad_norm": 1.2616111800783727,
+      "learning_rate": 1.8939414235030137e-05,
+      "loss": 1.3755,
+      "step": 141
+    },
+    {
+      "epoch": 0.39609483960948394,
+      "grad_norm": 1.2686523296470786,
+      "learning_rate": 1.891848921027355e-05,
+      "loss": 1.3442,
+      "step": 142
+    },
+    {
+      "epoch": 0.398884239888424,
+      "grad_norm": 1.37417212405681,
+      "learning_rate": 1.889737156751965e-05,
+      "loss": 1.3651,
+      "step": 143
+    },
+    {
+      "epoch": 0.401673640167364,
+      "grad_norm": 1.373244156828418,
+      "learning_rate": 1.887606176285893e-05,
+      "loss": 1.4961,
+      "step": 144
+    },
+    {
+      "epoch": 0.40446304044630405,
+      "grad_norm": 1.1872961109156253,
+      "learning_rate": 1.8854560256532098e-05,
+      "loss": 1.2587,
+      "step": 145
+    },
+    {
+      "epoch": 0.4072524407252441,
+      "grad_norm": 1.313036403643509,
+      "learning_rate": 1.883286751292018e-05,
+      "loss": 1.4575,
+      "step": 146
+    },
+    {
+      "epoch": 0.4100418410041841,
+      "grad_norm": 1.390974479025736,
+      "learning_rate": 1.8810984000534457e-05,
+      "loss": 1.3891,
+      "step": 147
+    },
+    {
+      "epoch": 0.41283124128312415,
+      "grad_norm": 1.3368020257438076,
+      "learning_rate": 1.8788910192006363e-05,
+      "loss": 1.3415,
+      "step": 148
+    },
+    {
+      "epoch": 0.41562064156206413,
+      "grad_norm": 1.2329515511987332,
+      "learning_rate": 1.8766646564077265e-05,
+      "loss": 1.3339,
+      "step": 149
+    },
+    {
+      "epoch": 0.41841004184100417,
+      "grad_norm": 1.276208090352136,
+      "learning_rate": 1.8744193597588185e-05,
+      "loss": 1.4495,
+      "step": 150
+    },
+    {
+      "epoch": 0.4211994421199442,
+      "grad_norm": 1.1548092334953926,
+      "learning_rate": 1.8721551777469397e-05,
+      "loss": 1.2099,
+      "step": 151
+    },
+    {
+      "epoch": 0.42398884239888424,
+      "grad_norm": 1.3184986062317001,
+      "learning_rate": 1.869872159272997e-05,
+      "loss": 1.2136,
+      "step": 152
+    },
+    {
+      "epoch": 0.42677824267782427,
+      "grad_norm": 1.1962726760049636,
+      "learning_rate": 1.8675703536447178e-05,
+      "loss": 1.2598,
+      "step": 153
+    },
+    {
+      "epoch": 0.4295676429567643,
+      "grad_norm": 1.3302312487017707,
+      "learning_rate": 1.8652498105755898e-05,
+      "loss": 1.3719,
+      "step": 154
+    },
+    {
+      "epoch": 0.43235704323570434,
+      "grad_norm": 1.55872330252289,
+      "learning_rate": 1.862910580183782e-05,
+      "loss": 1.4867,
+      "step": 155
+    },
+    {
+      "epoch": 0.4351464435146444,
+      "grad_norm": 1.3004017172912832,
+      "learning_rate": 1.8605527129910663e-05,
+      "loss": 1.3089,
+      "step": 156
+    },
+    {
+      "epoch": 0.43793584379358436,
+      "grad_norm": 1.2650294480346138,
+      "learning_rate": 1.858176259921724e-05,
+      "loss": 1.2856,
+      "step": 157
+    },
+    {
+      "epoch": 0.4407252440725244,
+      "grad_norm": 1.414189467122214,
+      "learning_rate": 1.8557812723014476e-05,
+      "loss": 1.4827,
+      "step": 158
+    },
+    {
+      "epoch": 0.4435146443514644,
+      "grad_norm": 1.2634959665686347,
+      "learning_rate": 1.853367801856231e-05,
+      "loss": 1.3568,
+      "step": 159
+    },
+    {
+      "epoch": 0.44630404463040446,
+      "grad_norm": 1.3164047095667806,
+      "learning_rate": 1.8509359007112523e-05,
+      "loss": 1.5306,
+      "step": 160
+    },
+    {
+      "epoch": 0.4490934449093445,
+      "grad_norm": 1.3232600329575266,
+      "learning_rate": 1.8484856213897496e-05,
+      "loss": 1.4279,
+      "step": 161
+    },
+    {
+      "epoch": 0.45188284518828453,
+      "grad_norm": 1.286194256752095,
+      "learning_rate": 1.8460170168118857e-05,
+      "loss": 1.2656,
+      "step": 162
+    },
+    {
+      "epoch": 0.45467224546722457,
+      "grad_norm": 1.393553888811773,
+      "learning_rate": 1.843530140293603e-05,
+      "loss": 1.4003,
+      "step": 163
+    },
+    {
+      "epoch": 0.45746164574616455,
+      "grad_norm": 1.2952418440030533,
+      "learning_rate": 1.841025045545477e-05,
+      "loss": 1.365,
+      "step": 164
+    },
+    {
+      "epoch": 0.4602510460251046,
+      "grad_norm": 1.2556167978994803,
+      "learning_rate": 1.8385017866715507e-05,
+      "loss": 1.3313,
+      "step": 165
+    },
+    {
+      "epoch": 0.4630404463040446,
+      "grad_norm": 1.3067564204278406,
+      "learning_rate": 1.8359604181681703e-05,
+      "loss": 1.4114,
+      "step": 166
+    },
+    {
+      "epoch": 0.46582984658298465,
+      "grad_norm": 1.2088590174928828,
+      "learning_rate": 1.833400994922806e-05,
+      "loss": 1.1866,
+      "step": 167
+    },
+    {
+      "epoch": 0.4686192468619247,
+      "grad_norm": 1.334753865529686,
+      "learning_rate": 1.8308235722128674e-05,
+      "loss": 1.2744,
+      "step": 168
+    },
+    {
+      "epoch": 0.4714086471408647,
+      "grad_norm": 1.3612556693462265,
+      "learning_rate": 1.8282282057045087e-05,
+      "loss": 1.3558,
+      "step": 169
+    },
+    {
+      "epoch": 0.47419804741980476,
+      "grad_norm": 1.2660924285869366,
+      "learning_rate": 1.8256149514514284e-05,
+      "loss": 1.2804,
+      "step": 170
+    },
+    {
+      "epoch": 0.4769874476987448,
+      "grad_norm": 1.3192403133250346,
+      "learning_rate": 1.8229838658936566e-05,
+      "loss": 1.3874,
+      "step": 171
+    },
+    {
+      "epoch": 0.4797768479776848,
+      "grad_norm": 1.3924722985838183,
+      "learning_rate": 1.8203350058563366e-05,
+      "loss": 1.5013,
+      "step": 172
+    },
+    {
+      "epoch": 0.4825662482566248,
+      "grad_norm": 1.2874077296032287,
+      "learning_rate": 1.8176684285484985e-05,
+      "loss": 1.3261,
+      "step": 173
+    },
+    {
+      "epoch": 0.48535564853556484,
+      "grad_norm": 1.309597924994118,
+      "learning_rate": 1.814984191561823e-05,
+      "loss": 1.4157,
+      "step": 174
+    },
+    {
+      "epoch": 0.4881450488145049,
+      "grad_norm": 1.1912282212302885,
+      "learning_rate": 1.8122823528693966e-05,
+      "loss": 1.2585,
+      "step": 175
+    },
+    {
+      "epoch": 0.4909344490934449,
+      "grad_norm": 1.2765007106705224,
+      "learning_rate": 1.809562970824462e-05,
+      "loss": 1.3437,
+      "step": 176
+    },
+    {
+      "epoch": 0.49372384937238495,
+      "grad_norm": 1.2571675750543854,
+      "learning_rate": 1.8068261041591548e-05,
+      "loss": 1.4383,
+      "step": 177
+    },
+    {
+      "epoch": 0.496513249651325,
+      "grad_norm": 1.1735759325774886,
+      "learning_rate": 1.8040718119832378e-05,
+      "loss": 1.3234,
+      "step": 178
+    },
+    {
+      "epoch": 0.499302649930265,
+      "grad_norm": 1.1634243266804525,
+      "learning_rate": 1.8013001537828213e-05,
+      "loss": 1.2169,
+      "step": 179
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 716,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 179,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 116435221217280.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-179/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:176779fd314d79db1323a9603622ba6af7ce08e6291ea5cb8066c39938973900
+size 8504

checkpoint-179/zero_to_fp32.py

@@ -0,0 +1,674 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import json
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering Sharded Weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        state_dict_split = split_torch_state_dict_into_shards(state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard, output_path)
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)