Upload 16 files

Multitask/README.md

@@ -0,0 +1,202 @@
+---
+base_model: /home/bingxing2/ailab/gaoben/models/Mistral-7B-Instruct/Mistral-7B-Instruct-v0.2/AI-ModelScope/Mistral-7B-Instruct-v0___2
+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

Multitask/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "/home/bingxing2/ailab/gaoben/models/Mistral-7B-Instruct/Mistral-7B-Instruct-v0.2/AI-ModelScope/Mistral-7B-Instruct-v0___2",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj",
+    "v_proj",
+    "up_proj",
+    "k_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

Multitask/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e66c1a723e86dc9e67dd728d08922e3d823c3d0d9887376028c7d0b8597435cd
+size 335605144

Multitask/latest

@@ -0,0 +1 @@
+global_step1748

Multitask/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7d74de51245105e1fbf57a6707ef3538b353952485508f6e2f8f74dc5d479d4
+size 15024

Multitask/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0617c9eb6cf7df57b2e0bb53cfe17c05f0910de56fe5b14427fe39ab54a44782
+size 15024

Multitask/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed68a365057022897d9645ee60902a77102f43215dcdf2ddd5d3842b6a8446d8
+size 15024

Multitask/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63ebaa0c302cadbdfcd9f8ee2289e35ecf9c9fc8c9968fc0c05f100dac20c6b9
+size 15024

Multitask/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d84bce0dc28d54d9c075e29d38cb4d4e03938d7b6ffa3b8d32674f5dbaa337a
+size 1064

Multitask/special_tokens_map.json

@@ -0,0 +1,24 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "</s>",
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

Multitask/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055
+size 493443

Multitask/tokenizer_config.json

@@ -0,0 +1,47 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": null,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [],
+  "bos_token": "<s>",
+  "chat_template": "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token}}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "extra_special_tokens": {},
+  "legacy": true,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "</s>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "split_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

Multitask/trainer_state.json

@@ -0,0 +1,1426 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 14.993576017130621,
+  "eval_steps": 500,
+  "global_step": 1740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.08565310492505353,
+      "grad_norm": 5.411635875701904,
+      "learning_rate": 9.999339889379647e-06,
+      "loss": 1.5609,
+      "num_input_tokens_seen": 255328,
+      "step": 10
+    },
+    {
+      "epoch": 0.17130620985010706,
+      "grad_norm": 0.7214002013206482,
+      "learning_rate": 9.997058249278764e-06,
+      "loss": 0.2055,
+      "num_input_tokens_seen": 510640,
+      "step": 20
+    },
+    {
+      "epoch": 0.2569593147751606,
+      "grad_norm": 0.47501006722450256,
+      "learning_rate": 9.993147673772869e-06,
+      "loss": 0.184,
+      "num_input_tokens_seen": 763808,
+      "step": 30
+    },
+    {
+      "epoch": 0.3426124197002141,
+      "grad_norm": 0.20641829073429108,
+      "learning_rate": 9.987609437626955e-06,
+      "loss": 0.1767,
+      "num_input_tokens_seen": 1017472,
+      "step": 40
+    },
+    {
+      "epoch": 0.4282655246252677,
+      "grad_norm": 0.49966660141944885,
+      "learning_rate": 9.98044534618898e-06,
+      "loss": 0.1821,
+      "num_input_tokens_seen": 1273488,
+      "step": 50
+    },
+    {
+      "epoch": 0.5139186295503212,
+      "grad_norm": 0.8994255065917969,
+      "learning_rate": 9.971657734801385e-06,
+      "loss": 0.1819,
+      "num_input_tokens_seen": 1527744,
+      "step": 60
+    },
+    {
+      "epoch": 0.5995717344753747,
+      "grad_norm": 1.7688418626785278,
+      "learning_rate": 9.961249468039806e-06,
+      "loss": 0.1797,
+      "num_input_tokens_seen": 1785520,
+      "step": 70
+    },
+    {
+      "epoch": 0.6852248394004282,
+      "grad_norm": 0.7211973071098328,
+      "learning_rate": 9.949223938779286e-06,
+      "loss": 0.1765,
+      "num_input_tokens_seen": 2037648,
+      "step": 80
+    },
+    {
+      "epoch": 0.7708779443254818,
+      "grad_norm": 0.94338458776474,
+      "learning_rate": 9.935585067088276e-06,
+      "loss": 0.1766,
+      "num_input_tokens_seen": 2292464,
+      "step": 90
+    },
+    {
+      "epoch": 0.8565310492505354,
+      "grad_norm": 0.6227909326553345,
+      "learning_rate": 9.920337298950767e-06,
+      "loss": 0.1714,
+      "num_input_tokens_seen": 2547872,
+      "step": 100
+    },
+    {
+      "epoch": 0.9421841541755889,
+      "grad_norm": 0.5941164493560791,
+      "learning_rate": 9.903485604816993e-06,
+      "loss": 0.1728,
+      "num_input_tokens_seen": 2801536,
+      "step": 110
+    },
+    {
+      "epoch": 1.0342612419700214,
+      "grad_norm": 0.4057783782482147,
+      "learning_rate": 9.885035477983184e-06,
+      "loss": 0.1868,
+      "num_input_tokens_seen": 3074416,
+      "step": 120
+    },
+    {
+      "epoch": 1.119914346895075,
+      "grad_norm": 0.5549430251121521,
+      "learning_rate": 9.864992932800845e-06,
+      "loss": 0.1679,
+      "num_input_tokens_seen": 3327568,
+      "step": 130
+    },
+    {
+      "epoch": 1.2055674518201285,
+      "grad_norm": 0.41472136974334717,
+      "learning_rate": 9.843364502716225e-06,
+      "loss": 0.1671,
+      "num_input_tokens_seen": 3582240,
+      "step": 140
+    },
+    {
+      "epoch": 1.291220556745182,
+      "grad_norm": 0.5156757235527039,
+      "learning_rate": 9.820157238140535e-06,
+      "loss": 0.1682,
+      "num_input_tokens_seen": 3838160,
+      "step": 150
+    },
+    {
+      "epoch": 1.3768736616702355,
+      "grad_norm": 0.5046593546867371,
+      "learning_rate": 9.795378704151675e-06,
+      "loss": 0.1651,
+      "num_input_tokens_seen": 4092304,
+      "step": 160
+    },
+    {
+      "epoch": 1.462526766595289,
+      "grad_norm": 0.5588434338569641,
+      "learning_rate": 9.76903697802817e-06,
+      "loss": 0.1649,
+      "num_input_tokens_seen": 4346640,
+      "step": 170
+    },
+    {
+      "epoch": 1.5481798715203428,
+      "grad_norm": 0.46262454986572266,
+      "learning_rate": 9.741140646616161e-06,
+      "loss": 0.1669,
+      "num_input_tokens_seen": 4602192,
+      "step": 180
+    },
+    {
+      "epoch": 1.633832976445396,
+      "grad_norm": 0.45427972078323364,
+      "learning_rate": 9.711698803530253e-06,
+      "loss": 0.1674,
+      "num_input_tokens_seen": 4858240,
+      "step": 190
+    },
+    {
+      "epoch": 1.7194860813704498,
+      "grad_norm": 0.4514879882335663,
+      "learning_rate": 9.68072104618921e-06,
+      "loss": 0.1641,
+      "num_input_tokens_seen": 5113408,
+      "step": 200
+    },
+    {
+      "epoch": 1.805139186295503,
+      "grad_norm": 0.7933849692344666,
+      "learning_rate": 9.648217472687385e-06,
+      "loss": 0.1614,
+      "num_input_tokens_seen": 5368352,
+      "step": 210
+    },
+    {
+      "epoch": 1.8907922912205568,
+      "grad_norm": 0.6207934021949768,
+      "learning_rate": 9.614198678502965e-06,
+      "loss": 0.163,
+      "num_input_tokens_seen": 5622128,
+      "step": 220
+    },
+    {
+      "epoch": 1.9764453961456103,
+      "grad_norm": 0.8193040490150452,
+      "learning_rate": 9.57867575304406e-06,
+      "loss": 0.1589,
+      "num_input_tokens_seen": 5876816,
+      "step": 230
+    },
+    {
+      "epoch": 2.068522483940043,
+      "grad_norm": 1.0469202995300293,
+      "learning_rate": 9.541660276033795e-06,
+      "loss": 0.1755,
+      "num_input_tokens_seen": 6145392,
+      "step": 240
+    },
+    {
+      "epoch": 2.154175588865096,
+      "grad_norm": 0.9274189472198486,
+      "learning_rate": 9.503164313735566e-06,
+      "loss": 0.1595,
+      "num_input_tokens_seen": 6399504,
+      "step": 250
+    },
+    {
+      "epoch": 2.23982869379015,
+      "grad_norm": 0.6875982880592346,
+      "learning_rate": 9.46320041501969e-06,
+      "loss": 0.1563,
+      "num_input_tokens_seen": 6654160,
+      "step": 260
+    },
+    {
+      "epoch": 2.325481798715203,
+      "grad_norm": 0.5835751295089722,
+      "learning_rate": 9.421781607272741e-06,
+      "loss": 0.1554,
+      "num_input_tokens_seen": 6910752,
+      "step": 270
+    },
+    {
+      "epoch": 2.411134903640257,
+      "grad_norm": 0.6475698351860046,
+      "learning_rate": 9.378921392150893e-06,
+      "loss": 0.1579,
+      "num_input_tokens_seen": 7166960,
+      "step": 280
+    },
+    {
+      "epoch": 2.4967880085653107,
+      "grad_norm": 0.6029316782951355,
+      "learning_rate": 9.33463374117867e-06,
+      "loss": 0.1577,
+      "num_input_tokens_seen": 7420288,
+      "step": 290
+    },
+    {
+      "epoch": 2.582441113490364,
+      "grad_norm": 0.6444355845451355,
+      "learning_rate": 9.288933091194524e-06,
+      "loss": 0.1564,
+      "num_input_tokens_seen": 7675184,
+      "step": 300
+    },
+    {
+      "epoch": 2.6680942184154177,
+      "grad_norm": 0.5541071891784668,
+      "learning_rate": 9.241834339644726e-06,
+      "loss": 0.1528,
+      "num_input_tokens_seen": 7926976,
+      "step": 310
+    },
+    {
+      "epoch": 2.753747323340471,
+      "grad_norm": 0.6703725457191467,
+      "learning_rate": 9.193352839727122e-06,
+      "loss": 0.1549,
+      "num_input_tokens_seen": 8184992,
+      "step": 320
+    },
+    {
+      "epoch": 2.8394004282655247,
+      "grad_norm": 0.6515584588050842,
+      "learning_rate": 9.143504395386302e-06,
+      "loss": 0.157,
+      "num_input_tokens_seen": 8439712,
+      "step": 330
+    },
+    {
+      "epoch": 2.925053533190578,
+      "grad_norm": 0.5527693629264832,
+      "learning_rate": 9.09230525616186e-06,
+      "loss": 0.157,
+      "num_input_tokens_seen": 8694080,
+      "step": 340
+    },
+    {
+      "epoch": 3.017130620985011,
+      "grad_norm": 0.6593677401542664,
+      "learning_rate": 9.039772111891383e-06,
+      "loss": 0.1672,
+      "num_input_tokens_seen": 8965488,
+      "step": 350
+    },
+    {
+      "epoch": 3.102783725910064,
+      "grad_norm": 0.5042828917503357,
+      "learning_rate": 8.985922087269916e-06,
+      "loss": 0.1483,
+      "num_input_tokens_seen": 9220480,
+      "step": 360
+    },
+    {
+      "epoch": 3.188436830835118,
+      "grad_norm": 0.4123888611793518,
+      "learning_rate": 8.930772736267675e-06,
+      "loss": 0.1532,
+      "num_input_tokens_seen": 9477024,
+      "step": 370
+    },
+    {
+      "epoch": 3.274089935760171,
+      "grad_norm": 0.7851901054382324,
+      "learning_rate": 8.874342036407815e-06,
+      "loss": 0.1508,
+      "num_input_tokens_seen": 9731840,
+      "step": 380
+    },
+    {
+      "epoch": 3.359743040685225,
+      "grad_norm": 0.7545840740203857,
+      "learning_rate": 8.816648382906154e-06,
+      "loss": 0.1516,
+      "num_input_tokens_seen": 9986704,
+      "step": 390
+    },
+    {
+      "epoch": 3.445396145610278,
+      "grad_norm": 0.7439327239990234,
+      "learning_rate": 8.757710582674708e-06,
+      "loss": 0.1506,
+      "num_input_tokens_seen": 10238720,
+      "step": 400
+    },
+    {
+      "epoch": 3.531049250535332,
+      "grad_norm": 0.8343164920806885,
+      "learning_rate": 8.697547848191037e-06,
+      "loss": 0.1516,
+      "num_input_tokens_seen": 10491856,
+      "step": 410
+    },
+    {
+      "epoch": 3.6167023554603857,
+      "grad_norm": 0.817565381526947,
+      "learning_rate": 8.63617979123539e-06,
+      "loss": 0.1542,
+      "num_input_tokens_seen": 10744240,
+      "step": 420
+    },
+    {
+      "epoch": 3.702355460385439,
+      "grad_norm": 0.5334470272064209,
+      "learning_rate": 8.573626416497669e-06,
+      "loss": 0.1446,
+      "num_input_tokens_seen": 10996768,
+      "step": 430
+    },
+    {
+      "epoch": 3.7880085653104922,
+      "grad_norm": 0.9441611766815186,
+      "learning_rate": 8.509908115056334e-06,
+      "loss": 0.1515,
+      "num_input_tokens_seen": 11254560,
+      "step": 440
+    },
+    {
+      "epoch": 3.873661670235546,
+      "grad_norm": 0.6177489757537842,
+      "learning_rate": 8.445045657731329e-06,
+      "loss": 0.1513,
+      "num_input_tokens_seen": 11512992,
+      "step": 450
+    },
+    {
+      "epoch": 3.9593147751605997,
+      "grad_norm": 0.5743350982666016,
+      "learning_rate": 8.379060188313244e-06,
+      "loss": 0.1458,
+      "num_input_tokens_seen": 11765808,
+      "step": 460
+    },
+    {
+      "epoch": 4.0513918629550325,
+      "grad_norm": 0.8525713086128235,
+      "learning_rate": 8.311973216670888e-06,
+      "loss": 0.1598,
+      "num_input_tokens_seen": 12036784,
+      "step": 470
+    },
+    {
+      "epoch": 4.137044967880086,
+      "grad_norm": 0.6399952173233032,
+      "learning_rate": 8.243806611739516e-06,
+      "loss": 0.1448,
+      "num_input_tokens_seen": 12290592,
+      "step": 480
+    },
+    {
+      "epoch": 4.222698072805139,
+      "grad_norm": 0.657546877861023,
+      "learning_rate": 8.17458259439202e-06,
+      "loss": 0.144,
+      "num_input_tokens_seen": 12542464,
+      "step": 490
+    },
+    {
+      "epoch": 4.308351177730192,
+      "grad_norm": 0.6414650678634644,
+      "learning_rate": 8.104323730195407e-06,
+      "loss": 0.1406,
+      "num_input_tokens_seen": 12796848,
+      "step": 500
+    },
+    {
+      "epoch": 4.394004282655247,
+      "grad_norm": 0.7480872869491577,
+      "learning_rate": 8.033052922054882e-06,
+      "loss": 0.1436,
+      "num_input_tokens_seen": 13051760,
+      "step": 510
+    },
+    {
+      "epoch": 4.4796573875803,
+      "grad_norm": 0.7025752067565918,
+      "learning_rate": 7.960793402748001e-06,
+      "loss": 0.147,
+      "num_input_tokens_seen": 13305808,
+      "step": 520
+    },
+    {
+      "epoch": 4.565310492505353,
+      "grad_norm": 0.5708986520767212,
+      "learning_rate": 7.887568727351262e-06,
+      "loss": 0.1456,
+      "num_input_tokens_seen": 13563056,
+      "step": 530
+    },
+    {
+      "epoch": 4.650963597430406,
+      "grad_norm": 0.6903087496757507,
+      "learning_rate": 7.813402765561664e-06,
+      "loss": 0.143,
+      "num_input_tokens_seen": 13816992,
+      "step": 540
+    },
+    {
+      "epoch": 4.736616702355461,
+      "grad_norm": 0.6083903908729553,
+      "learning_rate": 7.738319693915673e-06,
+      "loss": 0.1439,
+      "num_input_tokens_seen": 14071936,
+      "step": 550
+    },
+    {
+      "epoch": 4.822269807280514,
+      "grad_norm": 0.6583831906318665,
+      "learning_rate": 7.662343987908195e-06,
+      "loss": 0.147,
+      "num_input_tokens_seen": 14327440,
+      "step": 560
+    },
+    {
+      "epoch": 4.907922912205567,
+      "grad_norm": 0.8827478885650635,
+      "learning_rate": 7.585500414014077e-06,
+      "loss": 0.1467,
+      "num_input_tokens_seen": 14582832,
+      "step": 570
+    },
+    {
+      "epoch": 4.993576017130621,
+      "grad_norm": 0.8274891972541809,
+      "learning_rate": 7.507814021614761e-06,
+      "loss": 0.1478,
+      "num_input_tokens_seen": 14839136,
+      "step": 580
+    },
+    {
+      "epoch": 5.085653104925053,
+      "grad_norm": 1.3195112943649292,
+      "learning_rate": 7.429310134832709e-06,
+      "loss": 0.1517,
+      "num_input_tokens_seen": 15109264,
+      "step": 590
+    },
+    {
+      "epoch": 5.1713062098501075,
+      "grad_norm": 0.7981224656105042,
+      "learning_rate": 7.35001434427628e-06,
+      "loss": 0.1396,
+      "num_input_tokens_seen": 15363824,
+      "step": 600
+    },
+    {
+      "epoch": 5.256959314775161,
+      "grad_norm": 0.6522560715675354,
+      "learning_rate": 7.269952498697734e-06,
+      "loss": 0.142,
+      "num_input_tokens_seen": 15618576,
+      "step": 610
+    },
+    {
+      "epoch": 5.342612419700214,
+      "grad_norm": 0.7629905343055725,
+      "learning_rate": 7.189150696567081e-06,
+      "loss": 0.1384,
+      "num_input_tokens_seen": 15871056,
+      "step": 620
+    },
+    {
+      "epoch": 5.428265524625267,
+      "grad_norm": 0.9554848670959473,
+      "learning_rate": 7.10763527756453e-06,
+      "loss": 0.1405,
+      "num_input_tokens_seen": 16124976,
+      "step": 630
+    },
+    {
+      "epoch": 5.5139186295503215,
+      "grad_norm": 0.8175866603851318,
+      "learning_rate": 7.025432813994315e-06,
+      "loss": 0.1357,
+      "num_input_tokens_seen": 16381680,
+      "step": 640
+    },
+    {
+      "epoch": 5.599571734475375,
+      "grad_norm": 0.7990790009498596,
+      "learning_rate": 6.942570102122679e-06,
+      "loss": 0.1387,
+      "num_input_tokens_seen": 16638048,
+      "step": 650
+    },
+    {
+      "epoch": 5.685224839400428,
+      "grad_norm": 0.9116854667663574,
+      "learning_rate": 6.859074153442864e-06,
+      "loss": 0.1414,
+      "num_input_tokens_seen": 16894688,
+      "step": 660
+    },
+    {
+      "epoch": 5.770877944325482,
+      "grad_norm": 0.7633938789367676,
+      "learning_rate": 6.774972185869928e-06,
+      "loss": 0.1389,
+      "num_input_tokens_seen": 17147808,
+      "step": 670
+    },
+    {
+      "epoch": 5.856531049250536,
+      "grad_norm": 0.8924551606178284,
+      "learning_rate": 6.690291614868287e-06,
+      "loss": 0.1361,
+      "num_input_tokens_seen": 17403280,
+      "step": 680
+    },
+    {
+      "epoch": 5.942184154175589,
+      "grad_norm": 0.8566009998321533,
+      "learning_rate": 6.60506004451485e-06,
+      "loss": 0.1356,
+      "num_input_tokens_seen": 17657888,
+      "step": 690
+    },
+    {
+      "epoch": 6.034261241970022,
+      "grad_norm": 0.9057173132896423,
+      "learning_rate": 6.5193052585006666e-06,
+      "loss": 0.1483,
+      "num_input_tokens_seen": 17927520,
+      "step": 700
+    },
+    {
+      "epoch": 6.119914346895075,
+      "grad_norm": 0.9895085692405701,
+      "learning_rate": 6.433055211074042e-06,
+      "loss": 0.1308,
+      "num_input_tokens_seen": 18184352,
+      "step": 710
+    },
+    {
+      "epoch": 6.205567451820128,
+      "grad_norm": 1.0845868587493896,
+      "learning_rate": 6.346338017928036e-06,
+      "loss": 0.1269,
+      "num_input_tokens_seen": 18437792,
+      "step": 720
+    },
+    {
+      "epoch": 6.291220556745182,
+      "grad_norm": 1.021283745765686,
+      "learning_rate": 6.2591819470353424e-06,
+      "loss": 0.1301,
+      "num_input_tokens_seen": 18690144,
+      "step": 730
+    },
+    {
+      "epoch": 6.376873661670236,
+      "grad_norm": 1.1350120306015015,
+      "learning_rate": 6.171615409433525e-06,
+      "loss": 0.1275,
+      "num_input_tokens_seen": 18944688,
+      "step": 740
+    },
+    {
+      "epoch": 6.462526766595289,
+      "grad_norm": 1.0572874546051025,
+      "learning_rate": 6.0836669499636255e-06,
+      "loss": 0.1264,
+      "num_input_tokens_seen": 19199984,
+      "step": 750
+    },
+    {
+      "epoch": 6.548179871520342,
+      "grad_norm": 1.1884225606918335,
+      "learning_rate": 5.995365237965144e-06,
+      "loss": 0.1294,
+      "num_input_tokens_seen": 19452032,
+      "step": 760
+    },
+    {
+      "epoch": 6.6338329764453965,
+      "grad_norm": 0.9745492339134216,
+      "learning_rate": 5.906739057930439e-06,
+      "loss": 0.1262,
+      "num_input_tokens_seen": 19707040,
+      "step": 770
+    },
+    {
+      "epoch": 6.71948608137045,
+      "grad_norm": 1.090391755104065,
+      "learning_rate": 5.817817300121592e-06,
+      "loss": 0.1266,
+      "num_input_tokens_seen": 19962960,
+      "step": 780
+    },
+    {
+      "epoch": 6.805139186295503,
+      "grad_norm": 1.1640676259994507,
+      "learning_rate": 5.728628951152799e-06,
+      "loss": 0.1324,
+      "num_input_tokens_seen": 20219008,
+      "step": 790
+    },
+    {
+      "epoch": 6.890792291220556,
+      "grad_norm": 0.9813507199287415,
+      "learning_rate": 5.639203084541338e-06,
+      "loss": 0.1338,
+      "num_input_tokens_seen": 20473664,
+      "step": 800
+    },
+    {
+      "epoch": 6.9764453961456105,
+      "grad_norm": 1.11289644241333,
+      "learning_rate": 5.549568851230219e-06,
+      "loss": 0.1273,
+      "num_input_tokens_seen": 20727296,
+      "step": 810
+    },
+    {
+      "epoch": 7.0685224839400425,
+      "grad_norm": 1.5624918937683105,
+      "learning_rate": 5.459755470085595e-06,
+      "loss": 0.1332,
+      "num_input_tokens_seen": 20996432,
+      "step": 820
+    },
+    {
+      "epoch": 7.154175588865097,
+      "grad_norm": 1.3339862823486328,
+      "learning_rate": 5.369792218372026e-06,
+      "loss": 0.1104,
+      "num_input_tokens_seen": 21252272,
+      "step": 830
+    },
+    {
+      "epoch": 7.23982869379015,
+      "grad_norm": 1.5236716270446777,
+      "learning_rate": 5.2797084222087105e-06,
+      "loss": 0.1114,
+      "num_input_tokens_seen": 21508208,
+      "step": 840
+    },
+    {
+      "epoch": 7.325481798715203,
+      "grad_norm": 1.4154669046401978,
+      "learning_rate": 5.189533447009795e-06,
+      "loss": 0.1134,
+      "num_input_tokens_seen": 21765536,
+      "step": 850
+    },
+    {
+      "epoch": 7.4111349036402565,
+      "grad_norm": 1.5260732173919678,
+      "learning_rate": 5.099296687911858e-06,
+      "loss": 0.1102,
+      "num_input_tokens_seen": 22020160,
+      "step": 860
+    },
+    {
+      "epoch": 7.496788008565311,
+      "grad_norm": 1.2989623546600342,
+      "learning_rate": 5.009027560191732e-06,
+      "loss": 0.1122,
+      "num_input_tokens_seen": 22274400,
+      "step": 870
+    },
+    {
+      "epoch": 7.582441113490364,
+      "grad_norm": 1.4925442934036255,
+      "learning_rate": 4.918755489677729e-06,
+      "loss": 0.1094,
+      "num_input_tokens_seen": 22526464,
+      "step": 880
+    },
+    {
+      "epoch": 7.668094218415417,
+      "grad_norm": 1.3059921264648438,
+      "learning_rate": 4.828509903157451e-06,
+      "loss": 0.1128,
+      "num_input_tokens_seen": 22779664,
+      "step": 890
+    },
+    {
+      "epoch": 7.7537473233404715,
+      "grad_norm": 1.6819276809692383,
+      "learning_rate": 4.738320218785281e-06,
+      "loss": 0.1146,
+      "num_input_tokens_seen": 23036160,
+      "step": 900
+    },
+    {
+      "epoch": 7.839400428265525,
+      "grad_norm": 1.3909580707550049,
+      "learning_rate": 4.648215836492682e-06,
+      "loss": 0.1145,
+      "num_input_tokens_seen": 23292016,
+      "step": 910
+    },
+    {
+      "epoch": 7.925053533190578,
+      "grad_norm": 1.7210851907730103,
+      "learning_rate": 4.5582261284044385e-06,
+      "loss": 0.1156,
+      "num_input_tokens_seen": 23544800,
+      "step": 920
+    },
+    {
+      "epoch": 8.01713062098501,
+      "grad_norm": 1.2723944187164307,
+      "learning_rate": 4.468380429263973e-06,
+      "loss": 0.1197,
+      "num_input_tokens_seen": 23816288,
+      "step": 930
+    },
+    {
+      "epoch": 8.102783725910065,
+      "grad_norm": 1.9091925621032715,
+      "learning_rate": 4.378708026870825e-06,
+      "loss": 0.0916,
+      "num_input_tokens_seen": 24071488,
+      "step": 940
+    },
+    {
+      "epoch": 8.188436830835117,
+      "grad_norm": 1.7839370965957642,
+      "learning_rate": 4.289238152533465e-06,
+      "loss": 0.0893,
+      "num_input_tokens_seen": 24324720,
+      "step": 950
+    },
+    {
+      "epoch": 8.274089935760172,
+      "grad_norm": 2.00311541557312,
+      "learning_rate": 4.199999971540489e-06,
+      "loss": 0.0889,
+      "num_input_tokens_seen": 24579648,
+      "step": 960
+    },
+    {
+      "epoch": 8.359743040685224,
+      "grad_norm": 2.047337293624878,
+      "learning_rate": 4.111022573653366e-06,
+      "loss": 0.0873,
+      "num_input_tokens_seen": 24833840,
+      "step": 970
+    },
+    {
+      "epoch": 8.445396145610278,
+      "grad_norm": 1.9115785360336304,
+      "learning_rate": 4.0223349636237766e-06,
+      "loss": 0.0904,
+      "num_input_tokens_seen": 25089776,
+      "step": 980
+    },
+    {
+      "epoch": 8.531049250535332,
+      "grad_norm": 1.8445810079574585,
+      "learning_rate": 3.933966051738684e-06,
+      "loss": 0.088,
+      "num_input_tokens_seen": 25345264,
+      "step": 990
+    },
+    {
+      "epoch": 8.616702355460385,
+      "grad_norm": 1.6529115438461304,
+      "learning_rate": 3.845944644396194e-06,
+      "loss": 0.0919,
+      "num_input_tokens_seen": 25598112,
+      "step": 1000
+    },
+    {
+      "epoch": 8.702355460385439,
+      "grad_norm": 2.129995346069336,
+      "learning_rate": 3.758299434715268e-06,
+      "loss": 0.0906,
+      "num_input_tokens_seen": 25851728,
+      "step": 1010
+    },
+    {
+      "epoch": 8.788008565310493,
+      "grad_norm": 2.1039373874664307,
+      "learning_rate": 3.6710589931823837e-06,
+      "loss": 0.0895,
+      "num_input_tokens_seen": 26104704,
+      "step": 1020
+    },
+    {
+      "epoch": 8.873661670235546,
+      "grad_norm": 2.058598518371582,
+      "learning_rate": 3.584251758338151e-06,
+      "loss": 0.0923,
+      "num_input_tokens_seen": 26361680,
+      "step": 1030
+    },
+    {
+      "epoch": 8.9593147751606,
+      "grad_norm": 1.8930065631866455,
+      "learning_rate": 3.4979060275069576e-06,
+      "loss": 0.0908,
+      "num_input_tokens_seen": 26617536,
+      "step": 1040
+    },
+    {
+      "epoch": 9.051391862955033,
+      "grad_norm": 1.8233646154403687,
+      "learning_rate": 3.4120499475726266e-06,
+      "loss": 0.0847,
+      "num_input_tokens_seen": 26888160,
+      "step": 1050
+    },
+    {
+      "epoch": 9.137044967880085,
+      "grad_norm": 2.1758053302764893,
+      "learning_rate": 3.3267115058031418e-06,
+      "loss": 0.0657,
+      "num_input_tokens_seen": 27142528,
+      "step": 1060
+    },
+    {
+      "epoch": 9.222698072805139,
+      "grad_norm": 2.0327367782592773,
+      "learning_rate": 3.2419185207273816e-06,
+      "loss": 0.0662,
+      "num_input_tokens_seen": 27394144,
+      "step": 1070
+    },
+    {
+      "epoch": 9.308351177730193,
+      "grad_norm": 2.2035434246063232,
+      "learning_rate": 3.157698633066863e-06,
+      "loss": 0.0665,
+      "num_input_tokens_seen": 27649488,
+      "step": 1080
+    },
+    {
+      "epoch": 9.394004282655246,
+      "grad_norm": 1.9066494703292847,
+      "learning_rate": 3.0740792967254606e-06,
+      "loss": 0.0642,
+      "num_input_tokens_seen": 27904992,
+      "step": 1090
+    },
+    {
+      "epoch": 9.4796573875803,
+      "grad_norm": 2.2175674438476562,
+      "learning_rate": 2.991087769840001e-06,
+      "loss": 0.0625,
+      "num_input_tokens_seen": 28160336,
+      "step": 1100
+    },
+    {
+      "epoch": 9.565310492505354,
+      "grad_norm": 2.435115337371826,
+      "learning_rate": 2.9087511058947014e-06,
+      "loss": 0.0643,
+      "num_input_tokens_seen": 28417360,
+      "step": 1110
+    },
+    {
+      "epoch": 9.650963597430406,
+      "grad_norm": 2.237015724182129,
+      "learning_rate": 2.827096144902289e-06,
+      "loss": 0.0645,
+      "num_input_tokens_seen": 28670512,
+      "step": 1120
+    },
+    {
+      "epoch": 9.73661670235546,
+      "grad_norm": 2.473604202270508,
+      "learning_rate": 2.7461495046547436e-06,
+      "loss": 0.068,
+      "num_input_tokens_seen": 28927232,
+      "step": 1130
+    },
+    {
+      "epoch": 9.822269807280513,
+      "grad_norm": 2.220705270767212,
+      "learning_rate": 2.665937572046432e-06,
+      "loss": 0.0647,
+      "num_input_tokens_seen": 29182768,
+      "step": 1140
+    },
+    {
+      "epoch": 9.907922912205567,
+      "grad_norm": 2.652024269104004,
+      "learning_rate": 2.586486494472572e-06,
+      "loss": 0.0644,
+      "num_input_tokens_seen": 29437936,
+      "step": 1150
+    },
+    {
+      "epoch": 9.993576017130621,
+      "grad_norm": 2.180983304977417,
+      "learning_rate": 2.5078221713057048e-06,
+      "loss": 0.0658,
+      "num_input_tokens_seen": 29690944,
+      "step": 1160
+    },
+    {
+      "epoch": 10.085653104925054,
+      "grad_norm": 1.6538355350494385,
+      "learning_rate": 2.4299702454530605e-06,
+      "loss": 0.053,
+      "num_input_tokens_seen": 29964448,
+      "step": 1170
+    },
+    {
+      "epoch": 10.171306209850107,
+      "grad_norm": 2.229673147201538,
+      "learning_rate": 2.3529560949975184e-06,
+      "loss": 0.0446,
+      "num_input_tokens_seen": 30215952,
+      "step": 1180
+    },
+    {
+      "epoch": 10.25695931477516,
+      "grad_norm": 1.8106822967529297,
+      "learning_rate": 2.2768048249248648e-06,
+      "loss": 0.0449,
+      "num_input_tokens_seen": 30471952,
+      "step": 1190
+    },
+    {
+      "epoch": 10.342612419700215,
+      "grad_norm": 2.150508403778076,
+      "learning_rate": 2.201541258940129e-06,
+      "loss": 0.0422,
+      "num_input_tokens_seen": 30727376,
+      "step": 1200
+    },
+    {
+      "epoch": 10.428265524625267,
+      "grad_norm": 2.0471906661987305,
+      "learning_rate": 2.12718993137555e-06,
+      "loss": 0.0461,
+      "num_input_tokens_seen": 30983760,
+      "step": 1210
+    },
+    {
+      "epoch": 10.513918629550322,
+      "grad_norm": 2.299278497695923,
+      "learning_rate": 2.0537750791929296e-06,
+      "loss": 0.0458,
+      "num_input_tokens_seen": 31238720,
+      "step": 1220
+    },
+    {
+      "epoch": 10.599571734475374,
+      "grad_norm": 2.1924257278442383,
+      "learning_rate": 1.981320634082873e-06,
+      "loss": 0.0434,
+      "num_input_tokens_seen": 31494560,
+      "step": 1230
+    },
+    {
+      "epoch": 10.685224839400428,
+      "grad_norm": 2.3524584770202637,
+      "learning_rate": 1.909850214663575e-06,
+      "loss": 0.0452,
+      "num_input_tokens_seen": 31750784,
+      "step": 1240
+    },
+    {
+      "epoch": 10.770877944325482,
+      "grad_norm": 2.2468934059143066,
+      "learning_rate": 1.8393871187816526e-06,
+      "loss": 0.0447,
+      "num_input_tokens_seen": 32005120,
+      "step": 1250
+    },
+    {
+      "epoch": 10.856531049250535,
+      "grad_norm": 2.448117971420288,
+      "learning_rate": 1.7699543159175215e-06,
+      "loss": 0.0449,
+      "num_input_tokens_seen": 32258480,
+      "step": 1260
+    },
+    {
+      "epoch": 10.942184154175589,
+      "grad_norm": 2.0848143100738525,
+      "learning_rate": 1.7015744396978557e-06,
+      "loss": 0.0442,
+      "num_input_tokens_seen": 32510944,
+      "step": 1270
+    },
+    {
+      "epoch": 11.034261241970022,
+      "grad_norm": 1.6036432981491089,
+      "learning_rate": 1.634269780517483e-06,
+      "loss": 0.0435,
+      "num_input_tokens_seen": 32780608,
+      "step": 1280
+    },
+    {
+      "epoch": 11.119914346895074,
+      "grad_norm": 3.015963315963745,
+      "learning_rate": 1.568062278273197e-06,
+      "loss": 0.0286,
+      "num_input_tokens_seen": 33034112,
+      "step": 1290
+    },
+    {
+      "epoch": 11.205567451820128,
+      "grad_norm": 1.6929532289505005,
+      "learning_rate": 1.5029735152118125e-06,
+      "loss": 0.0308,
+      "num_input_tokens_seen": 33290224,
+      "step": 1300
+    },
+    {
+      "epoch": 11.291220556745182,
+      "grad_norm": 1.9741885662078857,
+      "learning_rate": 1.4390247088948073e-06,
+      "loss": 0.0309,
+      "num_input_tokens_seen": 33544448,
+      "step": 1310
+    },
+    {
+      "epoch": 11.376873661670235,
+      "grad_norm": 1.5955508947372437,
+      "learning_rate": 1.3762367052818527e-06,
+      "loss": 0.0275,
+      "num_input_tokens_seen": 33799536,
+      "step": 1320
+    },
+    {
+      "epoch": 11.462526766595289,
+      "grad_norm": 2.293123245239258,
+      "learning_rate": 1.3146299719354544e-06,
+      "loss": 0.0304,
+      "num_input_tokens_seen": 34055952,
+      "step": 1330
+    },
+    {
+      "epoch": 11.548179871520343,
+      "grad_norm": 1.8011912107467651,
+      "learning_rate": 1.254224591348983e-06,
+      "loss": 0.0299,
+      "num_input_tokens_seen": 34310000,
+      "step": 1340
+    },
+    {
+      "epoch": 11.633832976445396,
+      "grad_norm": 1.8339879512786865,
+      "learning_rate": 1.1950402544001849e-06,
+      "loss": 0.0311,
+      "num_input_tokens_seen": 34565680,
+      "step": 1350
+    },
+    {
+      "epoch": 11.71948608137045,
+      "grad_norm": 1.6808807849884033,
+      "learning_rate": 1.1370962539323837e-06,
+      "loss": 0.0314,
+      "num_input_tokens_seen": 34820768,
+      "step": 1360
+    },
+    {
+      "epoch": 11.805139186295504,
+      "grad_norm": 1.7647879123687744,
+      "learning_rate": 1.0804114784654158e-06,
+      "loss": 0.0311,
+      "num_input_tokens_seen": 35074016,
+      "step": 1370
+    },
+    {
+      "epoch": 11.890792291220556,
+      "grad_norm": 1.753990650177002,
+      "learning_rate": 1.0250044060383734e-06,
+      "loss": 0.0299,
+      "num_input_tokens_seen": 35328272,
+      "step": 1380
+    },
+    {
+      "epoch": 11.97644539614561,
+      "grad_norm": 2.10841965675354,
+      "learning_rate": 9.708930981861603e-07,
+      "loss": 0.03,
+      "num_input_tokens_seen": 35582880,
+      "step": 1390
+    },
+    {
+      "epoch": 12.068522483940043,
+      "grad_norm": 1.4194451570510864,
+      "learning_rate": 9.180951940518002e-07,
+      "loss": 0.026,
+      "num_input_tokens_seen": 35853280,
+      "step": 1400
+    },
+    {
+      "epoch": 12.154175588865096,
+      "grad_norm": 1.612318515777588,
+      "learning_rate": 8.666279046364595e-07,
+      "loss": 0.0208,
+      "num_input_tokens_seen": 36106816,
+      "step": 1410
+    },
+    {
+      "epoch": 12.23982869379015,
+      "grad_norm": 1.6022765636444092,
+      "learning_rate": 8.165080071890208e-07,
+      "loss": 0.0205,
+      "num_input_tokens_seen": 36359232,
+      "step": 1420
+    },
+    {
+      "epoch": 12.325481798715204,
+      "grad_norm": 1.608430027961731,
+      "learning_rate": 7.677518397370548e-07,
+      "loss": 0.0228,
+      "num_input_tokens_seen": 36614176,
+      "step": 1430
+    },
+    {
+      "epoch": 12.411134903640257,
+      "grad_norm": 1.4423803091049194,
+      "learning_rate": 7.203752957609672e-07,
+      "loss": 0.0207,
+      "num_input_tokens_seen": 36868400,
+      "step": 1440
+    },
+    {
+      "epoch": 12.49678800856531,
+      "grad_norm": 1.6684809923171997,
+      "learning_rate": 6.743938190130616e-07,
+      "loss": 0.0215,
+      "num_input_tokens_seen": 37121536,
+      "step": 1450
+    },
+    {
+      "epoch": 12.582441113490365,
+      "grad_norm": 1.7179003953933716,
+      "learning_rate": 6.298223984832047e-07,
+      "loss": 0.0216,
+      "num_input_tokens_seen": 37377168,
+      "step": 1460
+    },
+    {
+      "epoch": 12.668094218415417,
+      "grad_norm": 1.6454778909683228,
+      "learning_rate": 5.866755635127247e-07,
+      "loss": 0.0207,
+      "num_input_tokens_seen": 37632992,
+      "step": 1470
+    },
+    {
+      "epoch": 12.753747323340471,
+      "grad_norm": 1.8044767379760742,
+      "learning_rate": 5.449673790581611e-07,
+      "loss": 0.0217,
+      "num_input_tokens_seen": 37888640,
+      "step": 1480
+    },
+    {
+      "epoch": 12.839400428265524,
+      "grad_norm": 1.874295711517334,
+      "learning_rate": 5.04711441106382e-07,
+      "loss": 0.0197,
+      "num_input_tokens_seen": 38143760,
+      "step": 1490
+    },
+    {
+      "epoch": 12.925053533190578,
+      "grad_norm": 1.3250926733016968,
+      "learning_rate": 4.659208722425806e-07,
+      "loss": 0.0207,
+      "num_input_tokens_seen": 38398560,
+      "step": 1500
+    },
+    {
+      "epoch": 13.01713062098501,
+      "grad_norm": 1.2411588430404663,
+      "learning_rate": 4.2860831737258857e-07,
+      "loss": 0.0216,
+      "num_input_tokens_seen": 38670912,
+      "step": 1510
+    },
+    {
+      "epoch": 13.102783725910065,
+      "grad_norm": 1.3138427734375,
+      "learning_rate": 3.9278593960090873e-07,
+      "loss": 0.0167,
+      "num_input_tokens_seen": 38925872,
+      "step": 1520
+    },
+    {
+      "epoch": 13.188436830835117,
+      "grad_norm": 1.362457036972046,
+      "learning_rate": 3.5846541626579026e-07,
+      "loss": 0.0159,
+      "num_input_tokens_seen": 39183632,
+      "step": 1530
+    },
+    {
+      "epoch": 13.274089935760172,
+      "grad_norm": 1.515376091003418,
+      "learning_rate": 3.256579351326744e-07,
+      "loss": 0.0156,
+      "num_input_tokens_seen": 39440864,
+      "step": 1540
+    },
+    {
+      "epoch": 13.359743040685224,
+      "grad_norm": 1.4070255756378174,
+      "learning_rate": 2.94374190747212e-07,
+      "loss": 0.0166,
+      "num_input_tokens_seen": 39695712,
+      "step": 1550
+    },
+    {
+      "epoch": 13.445396145610278,
+      "grad_norm": 1.4853448867797852,
+      "learning_rate": 2.64624380949069e-07,
+      "loss": 0.0173,
+      "num_input_tokens_seen": 39950304,
+      "step": 1560
+    },
+    {
+      "epoch": 13.531049250535332,
+      "grad_norm": 1.542286992073059,
+      "learning_rate": 2.3641820354764755e-07,
+      "loss": 0.0165,
+      "num_input_tokens_seen": 40203616,
+      "step": 1570
+    },
+    {
+      "epoch": 13.616702355460385,
+      "grad_norm": 1.565663456916809,
+      "learning_rate": 2.0976485316080375e-07,
+      "loss": 0.0167,
+      "num_input_tokens_seen": 40458464,
+      "step": 1580
+    },
+    {
+      "epoch": 13.702355460385439,
+      "grad_norm": 1.3701163530349731,
+      "learning_rate": 1.846730182175993e-07,
+      "loss": 0.017,
+      "num_input_tokens_seen": 40711216,
+      "step": 1590
+    },
+    {
+      "epoch": 13.788008565310493,
+      "grad_norm": 1.4886751174926758,
+      "learning_rate": 1.6115087812605123e-07,
+      "loss": 0.015,
+      "num_input_tokens_seen": 40965856,
+      "step": 1600
+    },
+    {
+      "epoch": 13.873661670235546,
+      "grad_norm": 1.2140471935272217,
+      "learning_rate": 1.392061006068246e-07,
+      "loss": 0.0169,
+      "num_input_tokens_seen": 41220736,
+      "step": 1610
+    },
+    {
+      "epoch": 13.9593147751606,
+      "grad_norm": 1.314063549041748,
+      "learning_rate": 1.1884583919371251e-07,
+      "loss": 0.0164,
+      "num_input_tokens_seen": 41473952,
+      "step": 1620
+    },
+    {
+      "epoch": 14.051391862955033,
+      "grad_norm": 1.2103674411773682,
+      "learning_rate": 1.0007673090173808e-07,
+      "loss": 0.0168,
+      "num_input_tokens_seen": 41742832,
+      "step": 1630
+    },
+    {
+      "epoch": 14.137044967880085,
+      "grad_norm": 1.250216007232666,
+      "learning_rate": 8.29048940636279e-08,
+      "loss": 0.0153,
+      "num_input_tokens_seen": 41998320,
+      "step": 1640
+    },
+    {
+      "epoch": 14.222698072805139,
+      "grad_norm": 1.114964485168457,
+      "learning_rate": 6.733592633536124e-08,
+      "loss": 0.0148,
+      "num_input_tokens_seen": 42253104,
+      "step": 1650
+    },
+    {
+      "epoch": 14.308351177730193,
+      "grad_norm": 1.3133609294891357,
+      "learning_rate": 5.3374902871456965e-08,
+      "loss": 0.0151,
+      "num_input_tokens_seen": 42509584,
+      "step": 1660
+    },
+    {
+      "epoch": 14.394004282655246,
+      "grad_norm": 1.3046901226043701,
+      "learning_rate": 4.102637467057746e-08,
+      "loss": 0.0144,
+      "num_input_tokens_seen": 42764768,
+      "step": 1670
+    },
+    {
+      "epoch": 14.4796573875803,
+      "grad_norm": 1.3270611763000488,
+      "learning_rate": 3.029436709200084e-08,
+      "loss": 0.0142,
+      "num_input_tokens_seen": 43019376,
+      "step": 1680
+    },
+    {
+      "epoch": 14.565310492505354,
+      "grad_norm": 1.1487038135528564,
+      "learning_rate": 2.1182378543438408e-08,
+      "loss": 0.0159,
+      "num_input_tokens_seen": 43273248,
+      "step": 1690
+    },
+    {
+      "epoch": 14.650963597430406,
+      "grad_norm": 1.1392930746078491,
+      "learning_rate": 1.3693379340626867e-08,
+      "loss": 0.0148,
+      "num_input_tokens_seen": 43529200,
+      "step": 1700
+    },
+    {
+      "epoch": 14.73661670235546,
+      "grad_norm": 1.24246084690094,
+      "learning_rate": 7.829810739069521e-09,
+      "loss": 0.0144,
+      "num_input_tokens_seen": 43781760,
+      "step": 1710
+    },
+    {
+      "epoch": 14.822269807280513,
+      "grad_norm": 1.2764571905136108,
+      "learning_rate": 3.593584138237294e-09,
+      "loss": 0.0142,
+      "num_input_tokens_seen": 44036144,
+      "step": 1720
+    },
+    {
+      "epoch": 14.907922912205567,
+      "grad_norm": 1.4254299402236938,
+      "learning_rate": 9.860804584937988e-10,
+      "loss": 0.0144,
+      "num_input_tokens_seen": 44292256,
+      "step": 1730
+    },
+    {
+      "epoch": 14.993576017130621,
+      "grad_norm": 1.1011109352111816,
+      "learning_rate": 8.149690943204391e-12,
+      "loss": 0.014,
+      "num_input_tokens_seen": 44548112,
+      "step": 1740
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 1740,
+  "num_input_tokens_seen": 44548112,
+  "num_train_epochs": 15,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.9454424851110953e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

Multitask/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee78070b6948641e9d45d7eb61fba486bdbaaa146690d84abf77b19e6e3f5333
+size 7480

Multitask/zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/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 gc
+import json
+import numpy as np
+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, weights_only=False)
+
+        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 tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # 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}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] 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")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+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 = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * 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
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[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}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        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 to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=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
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    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.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    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")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+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,
+                                                          lazy_mode=True)
+
+    # 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")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_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 by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # 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)