Training in progress, step 550, checkpoint

checkpoint-550/README.md

@@ -0,0 +1,202 @@
+---
+base_model: microsoft/Phi-3-mini-4k-instruct
+library_name: peft
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.14.0

checkpoint-550/adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "microsoft/Phi-3-mini-4k-instruct",
+  "bias": "none",
+  "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": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "down_proj",
+    "qkv_proj",
+    "o_proj",
+    "gate_up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-550/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf562e0725b165446bb7ecc8f792860b14b7cf9fe7fa98020e93ae0e56c7e3f5
+size 25200088

checkpoint-550/global_step550/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:dcd9ce4f5a9ae1729b9405ef0147ad9f4b15ca83ebd6efd9f47ce860608a8953
+size 18881328

checkpoint-550/global_step550/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:361a416d09e569c10fa91b97cf53b64f97d56c812ef7a230cc21073f2d936881
+size 18881328

checkpoint-550/global_step550/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:320926f69b2a1995be0413a65d66108018ae67a787ab73b69813ddfa2b20a7ba
+size 18881328

checkpoint-550/global_step550/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:438db5b1b571fa09c60a04e63cf8badcac0666d3d773b6925bdf38f930fa10b1
+size 18881392

checkpoint-550/global_step550/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e4ff47c429aed1ff62f8ced0ece50a932051acc38f0b0889de60b52479b53cf
+size 18881392

checkpoint-550/global_step550/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:659265ff9d11f69a5e0f6d55f9cdd392f191fa4cba3fa71046bddd2daa651e20
+size 18881392

checkpoint-550/global_step550/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:864b8fd7c3a6a0679d8a0a1a210e4befd39f5b974c171f154a2b111dc3e1824b
+size 18881392

checkpoint-550/global_step550/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b03ee8f85592a21559bd91bd62779c88096586400291dbf59a2aa920a8d7d997
+size 18881392

checkpoint-550/global_step550/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:847138aec4bc01b5dcf65880c7a6f248d80a4daea884a8a23f4043877d21ef92
+size 25379244

checkpoint-550/latest

@@ -0,0 +1 @@
+global_step550

checkpoint-550/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e5cd4ad571350abe2eb98424dd5c5dd650f79de5be8de2b9ff4da9d030d723b
+size 15984

checkpoint-550/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcff98006be86afc3f75b37d6113fdf5b62db51c94b6f68b33f555f4ac346822
+size 15984

checkpoint-550/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f220fd74a6757e167d014f721e96b7e5710e8f5c97f48c9fe6d72e19ebbbd65c
+size 15984

checkpoint-550/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21565575b5db0aa139865ffb0d9df6ceb55078dc7b218f601419cc3d7b873134
+size 15984

checkpoint-550/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:487a03a3b6c36091572b8fbb74add1eb3c753efe5ab0eee791c8d03f495e5c98
+size 15984

checkpoint-550/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98f8c6e22cfd0b3668705becc42fb2c443ef5e4cfe38d4ba5e3dfdc565094143
+size 15984

checkpoint-550/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:379eebc7ccebea3c24281c6604242d09589a64d4774ea37b6d5cf6e7bbece645
+size 15984

checkpoint-550/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3668b553f323a1aa5806c5d8feff7c926f6116dc2b7f961e9746634c8e825c0
+size 15984

checkpoint-550/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0a8efee4112d2dd8949a8275b8c557402eda5213e3f06fa3f59f89e1c62ee85
+size 1064

checkpoint-550/special_tokens_map.json

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

checkpoint-550/tokenizer_config.json

@@ -0,0 +1,133 @@
+{
+  "add_bos_token": false,
+  "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": true,
+      "single_word": false,
+      "special": false
+    },
+    "32000": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32001": {
+      "content": "<|assistant|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32002": {
+      "content": "<|placeholder1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32003": {
+      "content": "<|placeholder2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32004": {
+      "content": "<|placeholder3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32005": {
+      "content": "<|placeholder4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32006": {
+      "content": "<|system|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32007": {
+      "content": "<|end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32008": {
+      "content": "<|placeholder5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32009": {
+      "content": "<|placeholder6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32010": {
+      "content": "<|user|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "chat_template": "{% set system_message = 'You are a helpful AI assistant.' %}{% if messages[0]['role'] == 'system' %}{% set system_message = messages[0]['content'] %}{% endif %}{% if system_message is defined %}{{ '<s>' + '<|system|>\n' + system_message + '<|end|>\n' }}{% endif %}{% for message in messages %}{% set content = message['content'] %}{% if message['role'] == 'user' %}{{ '<|user|>\n' + content + '<|end|>\n<|assistant|>\n' }}{% elif message['role'] == 'assistant' %}{{ content + '<|end|>' + '\n' }}{% endif %}{% endfor %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|end|>",
+  "extra_special_tokens": {},
+  "legacy": false,
+  "model_max_length": 4096,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "split_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

checkpoint-550/trainer_state.json

@@ -0,0 +1,1034 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.8632528938591328,
+  "eval_steps": 50,
+  "global_step": 550,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.015695507161075144,
+      "grad_norm": 0.04355761408805847,
+      "learning_rate": 4.999451708687114e-06,
+      "logits/chosen": 14.845781326293945,
+      "logits/rejected": 14.576438903808594,
+      "logps/chosen": -0.31864267587661743,
+      "logps/rejected": -0.24545662105083466,
+      "loss": 1.0492,
+      "rewards/accuracies": 0.25,
+      "rewards/chosen": -0.47796401381492615,
+      "rewards/margins": -0.10977902263402939,
+      "rewards/rejected": -0.3681849539279938,
+      "step": 10
+    },
+    {
+      "epoch": 0.03139101432215029,
+      "grad_norm": 0.04919258877635002,
+      "learning_rate": 4.997807075247147e-06,
+      "logits/chosen": 15.27595043182373,
+      "logits/rejected": 14.872761726379395,
+      "logps/chosen": -0.3344747722148895,
+      "logps/rejected": -0.24258682131767273,
+      "loss": 1.0487,
+      "rewards/accuracies": 0.16249999403953552,
+      "rewards/chosen": -0.5017121434211731,
+      "rewards/margins": -0.1378319263458252,
+      "rewards/rejected": -0.3638802468776703,
+      "step": 20
+    },
+    {
+      "epoch": 0.047086521483225424,
+      "grad_norm": 0.049933061003685,
+      "learning_rate": 4.9950668210706795e-06,
+      "logits/chosen": 15.913165092468262,
+      "logits/rejected": 15.607622146606445,
+      "logps/chosen": -0.3440183997154236,
+      "logps/rejected": -0.2831566333770752,
+      "loss": 1.0405,
+      "rewards/accuracies": 0.1875,
+      "rewards/chosen": -0.516027569770813,
+      "rewards/margins": -0.09129264950752258,
+      "rewards/rejected": -0.4247349202632904,
+      "step": 30
+    },
+    {
+      "epoch": 0.06278202864430057,
+      "grad_norm": 0.05693503096699715,
+      "learning_rate": 4.9912321481237616e-06,
+      "logits/chosen": 15.402900695800781,
+      "logits/rejected": 14.99272632598877,
+      "logps/chosen": -0.3297731578350067,
+      "logps/rejected": -0.2746916711330414,
+      "loss": 1.0369,
+      "rewards/accuracies": 0.21250000596046448,
+      "rewards/chosen": -0.49465981125831604,
+      "rewards/margins": -0.08262218534946442,
+      "rewards/rejected": -0.41203755140304565,
+      "step": 40
+    },
+    {
+      "epoch": 0.07847753580537571,
+      "grad_norm": 0.05467928573489189,
+      "learning_rate": 4.986304738420684e-06,
+      "logits/chosen": 15.64543342590332,
+      "logits/rejected": 15.632547378540039,
+      "logps/chosen": -0.30952686071395874,
+      "logps/rejected": -0.24847058951854706,
+      "loss": 1.0367,
+      "rewards/accuracies": 0.21250000596046448,
+      "rewards/chosen": -0.46429023146629333,
+      "rewards/margins": -0.09158438444137573,
+      "rewards/rejected": -0.37270587682724,
+      "step": 50
+    },
+    {
+      "epoch": 0.07847753580537571,
+      "eval_logits/chosen": 15.850138664245605,
+      "eval_logits/rejected": 15.368529319763184,
+      "eval_logps/chosen": -0.3222965598106384,
+      "eval_logps/rejected": -0.26877468824386597,
+      "eval_loss": 1.0326261520385742,
+      "eval_rewards/accuracies": 0.26923078298568726,
+      "eval_rewards/chosen": -0.4834447503089905,
+      "eval_rewards/margins": -0.08028276264667511,
+      "eval_rewards/rejected": -0.40316200256347656,
+      "eval_runtime": 14.5044,
+      "eval_samples_per_second": 28.405,
+      "eval_steps_per_second": 3.585,
+      "step": 50
+    },
+    {
+      "epoch": 0.09417304296645085,
+      "grad_norm": 0.06174452602863312,
+      "learning_rate": 4.980286753286196e-06,
+      "logits/chosen": 15.443066596984863,
+      "logits/rejected": 15.192205429077148,
+      "logps/chosen": -0.31090402603149414,
+      "logps/rejected": -0.26281923055648804,
+      "loss": 1.04,
+      "rewards/accuracies": 0.25,
+      "rewards/chosen": -0.466356098651886,
+      "rewards/margins": -0.07212716341018677,
+      "rewards/rejected": -0.39422887563705444,
+      "step": 60
+    },
+    {
+      "epoch": 0.109868550127526,
+      "grad_norm": 0.06952528655529022,
+      "learning_rate": 4.973180832407471e-06,
+      "logits/chosen": 16.024200439453125,
+      "logits/rejected": 15.82934284210205,
+      "logps/chosen": -0.348227322101593,
+      "logps/rejected": -0.26220566034317017,
+      "loss": 1.043,
+      "rewards/accuracies": 0.21250000596046448,
+      "rewards/chosen": -0.5223408937454224,
+      "rewards/margins": -0.1290324479341507,
+      "rewards/rejected": -0.39330852031707764,
+      "step": 70
+    },
+    {
+      "epoch": 0.12556405728860115,
+      "grad_norm": 0.07572082430124283,
+      "learning_rate": 4.964990092676263e-06,
+      "logits/chosen": 15.884915351867676,
+      "logits/rejected": 15.603845596313477,
+      "logps/chosen": -0.34849274158477783,
+      "logps/rejected": -0.26585355401039124,
+      "loss": 1.0285,
+      "rewards/accuracies": 0.1875,
+      "rewards/chosen": -0.5227391719818115,
+      "rewards/margins": -0.12395882606506348,
+      "rewards/rejected": -0.39878037571907043,
+      "step": 80
+    },
+    {
+      "epoch": 0.14125956444967627,
+      "grad_norm": 0.2423778474330902,
+      "learning_rate": 4.9557181268217225e-06,
+      "logits/chosen": 15.978216171264648,
+      "logits/rejected": 15.76471996307373,
+      "logps/chosen": -0.327436238527298,
+      "logps/rejected": -0.25457051396369934,
+      "loss": 1.03,
+      "rewards/accuracies": 0.22499999403953552,
+      "rewards/chosen": -0.49115434288978577,
+      "rewards/margins": -0.10929858684539795,
+      "rewards/rejected": -0.3818557560443878,
+      "step": 90
+    },
+    {
+      "epoch": 0.15695507161075142,
+      "grad_norm": 0.1594536453485489,
+      "learning_rate": 4.9453690018345144e-06,
+      "logits/chosen": 16.307537078857422,
+      "logits/rejected": 16.138330459594727,
+      "logps/chosen": -0.3342314660549164,
+      "logps/rejected": -0.27582648396492004,
+      "loss": 1.0309,
+      "rewards/accuracies": 0.26249998807907104,
+      "rewards/chosen": -0.5013472437858582,
+      "rewards/margins": -0.0876075029373169,
+      "rewards/rejected": -0.41373974084854126,
+      "step": 100
+    },
+    {
+      "epoch": 0.15695507161075142,
+      "eval_logits/chosen": 16.4310245513916,
+      "eval_logits/rejected": 15.98912525177002,
+      "eval_logps/chosen": -0.3239763677120209,
+      "eval_logps/rejected": -0.28784558176994324,
+      "eval_loss": 1.020836353302002,
+      "eval_rewards/accuracies": 0.3076923191547394,
+      "eval_rewards/chosen": -0.4859645664691925,
+      "eval_rewards/margins": -0.054196178913116455,
+      "eval_rewards/rejected": -0.43176835775375366,
+      "eval_runtime": 14.5049,
+      "eval_samples_per_second": 28.404,
+      "eval_steps_per_second": 3.585,
+      "step": 100
+    },
+    {
+      "epoch": 0.17265057877182657,
+      "grad_norm": 0.07431349903345108,
+      "learning_rate": 4.933947257182901e-06,
+      "logits/chosen": 16.56686782836914,
+      "logits/rejected": 16.093189239501953,
+      "logps/chosen": -0.34455060958862305,
+      "logps/rejected": -0.2834388315677643,
+      "loss": 1.0388,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.5168259739875793,
+      "rewards/margins": -0.09166768193244934,
+      "rewards/rejected": -0.4251582622528076,
+      "step": 110
+    },
+    {
+      "epoch": 0.1883460859329017,
+      "grad_norm": 0.08802352845668793,
+      "learning_rate": 4.921457902821578e-06,
+      "logits/chosen": 16.50200843811035,
+      "logits/rejected": 16.286388397216797,
+      "logps/chosen": -0.30845317244529724,
+      "logps/rejected": -0.2677682936191559,
+      "loss": 1.0247,
+      "rewards/accuracies": 0.25,
+      "rewards/chosen": -0.4626797139644623,
+      "rewards/margins": -0.06102731078863144,
+      "rewards/rejected": -0.40165242552757263,
+      "step": 120
+    },
+    {
+      "epoch": 0.20404159309397685,
+      "grad_norm": 0.10464702546596527,
+      "learning_rate": 4.907906416994146e-06,
+      "logits/chosen": 16.163082122802734,
+      "logits/rejected": 16.158031463623047,
+      "logps/chosen": -0.3138599991798401,
+      "logps/rejected": -0.28097471594810486,
+      "loss": 1.0169,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.47078999876976013,
+      "rewards/margins": -0.04932791367173195,
+      "rewards/rejected": -0.4214620590209961,
+      "step": 130
+    },
+    {
+      "epoch": 0.219737100255052,
+      "grad_norm": 0.16971275210380554,
+      "learning_rate": 4.893298743830168e-06,
+      "logits/chosen": 16.28864860534668,
+      "logits/rejected": 16.151805877685547,
+      "logps/chosen": -0.3283368945121765,
+      "logps/rejected": -0.2850198745727539,
+      "loss": 0.9964,
+      "rewards/accuracies": 0.3499999940395355,
+      "rewards/chosen": -0.49250537157058716,
+      "rewards/margins": -0.06497551500797272,
+      "rewards/rejected": -0.42752987146377563,
+      "step": 140
+    },
+    {
+      "epoch": 0.23543260741612712,
+      "grad_norm": 0.18377964198589325,
+      "learning_rate": 4.8776412907378845e-06,
+      "logits/chosen": 16.890087127685547,
+      "logits/rejected": 16.42388153076172,
+      "logps/chosen": -0.33256903290748596,
+      "logps/rejected": -0.2939595878124237,
+      "loss": 1.0073,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": -0.49885353446006775,
+      "rewards/margins": -0.057914119213819504,
+      "rewards/rejected": -0.44093936681747437,
+      "step": 150
+    },
+    {
+      "epoch": 0.23543260741612712,
+      "eval_logits/chosen": 16.833438873291016,
+      "eval_logits/rejected": 16.328977584838867,
+      "eval_logps/chosen": -0.32567569613456726,
+      "eval_logps/rejected": -0.35700783133506775,
+      "eval_loss": 0.9802881479263306,
+      "eval_rewards/accuracies": 0.42307692766189575,
+      "eval_rewards/chosen": -0.4885135293006897,
+      "eval_rewards/margins": 0.04699822515249252,
+      "eval_rewards/rejected": -0.5355117321014404,
+      "eval_runtime": 14.5005,
+      "eval_samples_per_second": 28.413,
+      "eval_steps_per_second": 3.586,
+      "step": 150
+    },
+    {
+      "epoch": 0.2511281145772023,
+      "grad_norm": 0.12049826234579086,
+      "learning_rate": 4.860940925593703e-06,
+      "logits/chosen": 16.505878448486328,
+      "logits/rejected": 16.178979873657227,
+      "logps/chosen": -0.3397011458873749,
+      "logps/rejected": -0.35640352964401245,
+      "loss": 0.9795,
+      "rewards/accuracies": 0.4124999940395355,
+      "rewards/chosen": -0.5095517039299011,
+      "rewards/margins": 0.0250535998493433,
+      "rewards/rejected": -0.5346053242683411,
+      "step": 160
+    },
+    {
+      "epoch": 0.2668236217382774,
+      "grad_norm": 0.09485407918691635,
+      "learning_rate": 4.84320497372973e-06,
+      "logits/chosen": 16.245588302612305,
+      "logits/rejected": 15.922958374023438,
+      "logps/chosen": -0.29733315110206604,
+      "logps/rejected": -0.3461209237575531,
+      "loss": 0.9694,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": -0.44599977135658264,
+      "rewards/margins": 0.0731816366314888,
+      "rewards/rejected": -0.5191814303398132,
+      "step": 170
+    },
+    {
+      "epoch": 0.28251912889935255,
+      "grad_norm": 0.155483216047287,
+      "learning_rate": 4.824441214720629e-06,
+      "logits/chosen": 16.339645385742188,
+      "logits/rejected": 16.115110397338867,
+      "logps/chosen": -0.3076801002025604,
+      "logps/rejected": -0.3655286729335785,
+      "loss": 0.9488,
+      "rewards/accuracies": 0.4625000059604645,
+      "rewards/chosen": -0.4615201950073242,
+      "rewards/margins": 0.0867728441953659,
+      "rewards/rejected": -0.5482929944992065,
+      "step": 180
+    },
+    {
+      "epoch": 0.2982146360604277,
+      "grad_norm": 0.21345795691013336,
+      "learning_rate": 4.804657878971252e-06,
+      "logits/chosen": 16.491886138916016,
+      "logits/rejected": 16.376684188842773,
+      "logps/chosen": -0.32437095046043396,
+      "logps/rejected": -0.39715105295181274,
+      "loss": 0.9423,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": -0.4865564703941345,
+      "rewards/margins": 0.10917013883590698,
+      "rewards/rejected": -0.5957266092300415,
+      "step": 190
+    },
+    {
+      "epoch": 0.31391014322150285,
+      "grad_norm": 0.17633090913295746,
+      "learning_rate": 4.783863644106502e-06,
+      "logits/chosen": 16.6339168548584,
+      "logits/rejected": 16.79404640197754,
+      "logps/chosen": -0.33018192648887634,
+      "logps/rejected": -0.384196400642395,
+      "loss": 0.939,
+      "rewards/accuracies": 0.42500001192092896,
+      "rewards/chosen": -0.4952728748321533,
+      "rewards/margins": 0.0810217633843422,
+      "rewards/rejected": -0.5762946009635925,
+      "step": 200
+    },
+    {
+      "epoch": 0.31391014322150285,
+      "eval_logits/chosen": 17.17803192138672,
+      "eval_logits/rejected": 16.59328269958496,
+      "eval_logps/chosen": -0.33600664138793945,
+      "eval_logps/rejected": -0.47861453890800476,
+      "eval_loss": 0.9303967356681824,
+      "eval_rewards/accuracies": 0.4615384638309479,
+      "eval_rewards/chosen": -0.5040098428726196,
+      "eval_rewards/margins": 0.21391186118125916,
+      "eval_rewards/rejected": -0.717921793460846,
+      "eval_runtime": 14.5,
+      "eval_samples_per_second": 28.414,
+      "eval_steps_per_second": 3.586,
+      "step": 200
+    },
+    {
+      "epoch": 0.329605650382578,
+      "grad_norm": 0.1562221795320511,
+      "learning_rate": 4.762067631165049e-06,
+      "logits/chosen": 16.597665786743164,
+      "logits/rejected": 16.3507022857666,
+      "logps/chosen": -0.33709320425987244,
+      "logps/rejected": -0.43131333589553833,
+      "loss": 0.9107,
+      "rewards/accuracies": 0.4749999940395355,
+      "rewards/chosen": -0.5056397914886475,
+      "rewards/margins": 0.14133022725582123,
+      "rewards/rejected": -0.6469700932502747,
+      "step": 210
+    },
+    {
+      "epoch": 0.34530115754365315,
+      "grad_norm": 0.17680124938488007,
+      "learning_rate": 4.7392794005985324e-06,
+      "logits/chosen": 16.873676300048828,
+      "logits/rejected": 16.61945152282715,
+      "logps/chosen": -0.35083168745040894,
+      "logps/rejected": -0.537697434425354,
+      "loss": 0.9094,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": -0.526247501373291,
+      "rewards/margins": 0.2802986204624176,
+      "rewards/rejected": -0.8065462112426758,
+      "step": 220
+    },
+    {
+      "epoch": 0.3609966647047283,
+      "grad_norm": 0.2662070393562317,
+      "learning_rate": 4.715508948078037e-06,
+      "logits/chosen": 16.803974151611328,
+      "logits/rejected": 16.698320388793945,
+      "logps/chosen": -0.3500753343105316,
+      "logps/rejected": -0.5399882793426514,
+      "loss": 0.8871,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": -0.5251129865646362,
+      "rewards/margins": 0.2848694324493408,
+      "rewards/rejected": -0.8099824786186218,
+      "step": 230
+    },
+    {
+      "epoch": 0.3766921718658034,
+      "grad_norm": 0.17161667346954346,
+      "learning_rate": 4.690766700109659e-06,
+      "logits/chosen": 16.67904281616211,
+      "logits/rejected": 16.410579681396484,
+      "logps/chosen": -0.33549198508262634,
+      "logps/rejected": -0.4875665605068207,
+      "loss": 0.9037,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.5032380223274231,
+      "rewards/margins": 0.22811183333396912,
+      "rewards/rejected": -0.7313498258590698,
+      "step": 240
+    },
+    {
+      "epoch": 0.39238767902687854,
+      "grad_norm": 1.106021761894226,
+      "learning_rate": 4.665063509461098e-06,
+      "logits/chosen": 17.208127975463867,
+      "logits/rejected": 16.654085159301758,
+      "logps/chosen": -0.3752726912498474,
+      "logps/rejected": -0.4947708249092102,
+      "loss": 0.8606,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.5629090070724487,
+      "rewards/margins": 0.1792471706867218,
+      "rewards/rejected": -0.7421562671661377,
+      "step": 250
+    },
+    {
+      "epoch": 0.39238767902687854,
+      "eval_logits/chosen": 17.32963752746582,
+      "eval_logits/rejected": 16.589412689208984,
+      "eval_logps/chosen": -0.37825876474380493,
+      "eval_logps/rejected": -0.9001243114471436,
+      "eval_loss": 0.8168494701385498,
+      "eval_rewards/accuracies": 0.5192307829856873,
+      "eval_rewards/chosen": -0.5673881769180298,
+      "eval_rewards/margins": 0.7827982306480408,
+      "eval_rewards/rejected": -1.3501865863800049,
+      "eval_runtime": 14.5053,
+      "eval_samples_per_second": 28.403,
+      "eval_steps_per_second": 3.585,
+      "step": 250
+    },
+    {
+      "epoch": 0.4080831861879537,
+      "grad_norm": 0.29772138595581055,
+      "learning_rate": 4.638410650401267e-06,
+      "logits/chosen": 16.713775634765625,
+      "logits/rejected": 16.35211944580078,
+      "logps/chosen": -0.3877524733543396,
+      "logps/rejected": -0.8511163592338562,
+      "loss": 0.8138,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": -0.581628680229187,
+      "rewards/margins": 0.6950457692146301,
+      "rewards/rejected": -1.276674509048462,
+      "step": 260
+    },
+    {
+      "epoch": 0.42377869334902885,
+      "grad_norm": 0.31860050559043884,
+      "learning_rate": 4.610819813755038e-06,
+      "logits/chosen": 17.09469985961914,
+      "logits/rejected": 16.5472412109375,
+      "logps/chosen": -0.47504258155822754,
+      "logps/rejected": -1.3266533613204956,
+      "loss": 0.7318,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": -0.7125638723373413,
+      "rewards/margins": 1.2774161100387573,
+      "rewards/rejected": -1.9899799823760986,
+      "step": 270
+    },
+    {
+      "epoch": 0.439474200510104,
+      "grad_norm": 0.6508163809776306,
+      "learning_rate": 4.582303101775249e-06,
+      "logits/chosen": 17.11003303527832,
+      "logits/rejected": 16.6564998626709,
+      "logps/chosen": -0.49357643723487854,
+      "logps/rejected": -1.4481580257415771,
+      "loss": 0.7599,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.7403645515441895,
+      "rewards/margins": 1.4318726062774658,
+      "rewards/rejected": -2.1722371578216553,
+      "step": 280
+    },
+    {
+      "epoch": 0.45516970767117915,
+      "grad_norm": 0.32430580258369446,
+      "learning_rate": 4.55287302283426e-06,
+      "logits/chosen": 16.830989837646484,
+      "logits/rejected": 16.384944915771484,
+      "logps/chosen": -0.593712329864502,
+      "logps/rejected": -1.7630856037139893,
+      "loss": 0.7336,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -0.8905684351921082,
+      "rewards/margins": 1.7540600299835205,
+      "rewards/rejected": -2.6446282863616943,
+      "step": 290
+    },
+    {
+      "epoch": 0.47086521483225424,
+      "grad_norm": 0.8555932641029358,
+      "learning_rate": 4.522542485937369e-06,
+      "logits/chosen": 16.72231674194336,
+      "logits/rejected": 16.28726577758789,
+      "logps/chosen": -0.5670709609985352,
+      "logps/rejected": -2.0420775413513184,
+      "loss": 0.6861,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.8506065607070923,
+      "rewards/margins": 2.2125096321105957,
+      "rewards/rejected": -3.0631160736083984,
+      "step": 300
+    },
+    {
+      "epoch": 0.47086521483225424,
+      "eval_logits/chosen": 17.13121795654297,
+      "eval_logits/rejected": 16.268341064453125,
+      "eval_logps/chosen": -0.6842947602272034,
+      "eval_logps/rejected": -2.119321584701538,
+      "eval_loss": 0.7583853602409363,
+      "eval_rewards/accuracies": 0.75,
+      "eval_rewards/chosen": -1.026442289352417,
+      "eval_rewards/margins": 2.1525399684906006,
+      "eval_rewards/rejected": -3.1789822578430176,
+      "eval_runtime": 14.5007,
+      "eval_samples_per_second": 28.412,
+      "eval_steps_per_second": 3.586,
+      "step": 300
+    },
+    {
+      "epoch": 0.4865607219933294,
+      "grad_norm": 0.6059070825576782,
+      "learning_rate": 4.491324795060491e-06,
+      "logits/chosen": 17.147808074951172,
+      "logits/rejected": 16.194652557373047,
+      "logps/chosen": -0.8036400079727173,
+      "logps/rejected": -2.289825201034546,
+      "loss": 0.7163,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -1.2054599523544312,
+      "rewards/margins": 2.229278087615967,
+      "rewards/rejected": -3.4347376823425293,
+      "step": 310
+    },
+    {
+      "epoch": 0.5022562291544046,
+      "grad_norm": 1.8073927164077759,
+      "learning_rate": 4.4592336433146e-06,
+      "logits/chosen": 16.94902992248535,
+      "logits/rejected": 16.066068649291992,
+      "logps/chosen": -1.198162317276001,
+      "logps/rejected": -2.2922632694244385,
+      "loss": 0.6894,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": -1.7972434759140015,
+      "rewards/margins": 1.6411516666412354,
+      "rewards/rejected": -3.4383950233459473,
+      "step": 320
+    },
+    {
+      "epoch": 0.5179517363154797,
+      "grad_norm": 3.746042490005493,
+      "learning_rate": 4.426283106939474e-06,
+      "logits/chosen": 16.31036376953125,
+      "logits/rejected": 15.991762161254883,
+      "logps/chosen": -1.6245781183242798,
+      "logps/rejected": -2.553597927093506,
+      "loss": 0.6607,
+      "rewards/accuracies": 0.7875000238418579,
+      "rewards/chosen": -2.4368672370910645,
+      "rewards/margins": 1.3935294151306152,
+      "rewards/rejected": -3.830397129058838,
+      "step": 330
+    },
+    {
+      "epoch": 0.5336472434765548,
+      "grad_norm": 2.098111867904663,
+      "learning_rate": 4.3924876391293915e-06,
+      "logits/chosen": 16.59554672241211,
+      "logits/rejected": 15.915553092956543,
+      "logps/chosen": -2.183227062225342,
+      "logps/rejected": -3.4911434650421143,
+      "loss": 0.6381,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -3.274840831756592,
+      "rewards/margins": 1.9618743658065796,
+      "rewards/rejected": -5.236715316772461,
+      "step": 340
+    },
+    {
+      "epoch": 0.5493427506376299,
+      "grad_norm": 2.153958320617676,
+      "learning_rate": 4.357862063693486e-06,
+      "logits/chosen": 16.332544326782227,
+      "logits/rejected": 15.691922187805176,
+      "logps/chosen": -2.673710346221924,
+      "logps/rejected": -3.8687057495117188,
+      "loss": 0.5752,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -4.010565757751465,
+      "rewards/margins": 1.792493224143982,
+      "rewards/rejected": -5.8030595779418945,
+      "step": 350
+    },
+    {
+      "epoch": 0.5493427506376299,
+      "eval_logits/chosen": 16.272428512573242,
+      "eval_logits/rejected": 15.381678581237793,
+      "eval_logps/chosen": -3.0390572547912598,
+      "eval_logps/rejected": -4.695068836212158,
+      "eval_loss": 0.5928590893745422,
+      "eval_rewards/accuracies": 0.9230769276618958,
+      "eval_rewards/chosen": -4.5585856437683105,
+      "eval_rewards/margins": 2.4840168952941895,
+      "eval_rewards/rejected": -7.0426025390625,
+      "eval_runtime": 14.5008,
+      "eval_samples_per_second": 28.412,
+      "eval_steps_per_second": 3.586,
+      "step": 350
+    },
+    {
+      "epoch": 0.5650382577987051,
+      "grad_norm": 1.5755672454833984,
+      "learning_rate": 4.322421568553529e-06,
+      "logits/chosen": 15.589811325073242,
+      "logits/rejected": 14.780921936035156,
+      "logps/chosen": -3.069565534591675,
+      "logps/rejected": -4.581957817077637,
+      "loss": 0.5275,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -4.604348182678223,
+      "rewards/margins": 2.268587827682495,
+      "rewards/rejected": -6.872936248779297,
+      "step": 360
+    },
+    {
+      "epoch": 0.5807337649597802,
+      "grad_norm": 1.8776415586471558,
+      "learning_rate": 4.286181699082008e-06,
+      "logits/chosen": 15.172673225402832,
+      "logits/rejected": 14.400335311889648,
+      "logps/chosen": -3.9551639556884766,
+      "logps/rejected": -5.95252799987793,
+      "loss": 0.5113,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -5.932745933532715,
+      "rewards/margins": 2.996046543121338,
+      "rewards/rejected": -8.928792953491211,
+      "step": 370
+    },
+    {
+      "epoch": 0.5964292721208554,
+      "grad_norm": 1.5507289171218872,
+      "learning_rate": 4.249158351283414e-06,
+      "logits/chosen": 14.8423433303833,
+      "logits/rejected": 14.592633247375488,
+      "logps/chosen": -3.7946255207061768,
+      "logps/rejected": -5.3750481605529785,
+      "loss": 0.4862,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -5.691938877105713,
+      "rewards/margins": 2.370633125305176,
+      "rewards/rejected": -8.062570571899414,
+      "step": 380
+    },
+    {
+      "epoch": 0.6121247792819305,
+      "grad_norm": 3.4510324001312256,
+      "learning_rate": 4.211367764821722e-06,
+      "logits/chosen": 14.586801528930664,
+      "logits/rejected": 14.031987190246582,
+      "logps/chosen": -4.391470432281494,
+      "logps/rejected": -6.2107672691345215,
+      "loss": 0.5117,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -6.587205410003662,
+      "rewards/margins": 2.72894549369812,
+      "rewards/rejected": -9.316150665283203,
+      "step": 390
+    },
+    {
+      "epoch": 0.6278202864430057,
+      "grad_norm": 3.3816750049591064,
+      "learning_rate": 4.172826515897146e-06,
+      "logits/chosen": 14.523185729980469,
+      "logits/rejected": 14.266815185546875,
+      "logps/chosen": -4.090173244476318,
+      "logps/rejected": -5.582955837249756,
+      "loss": 0.5182,
+      "rewards/accuracies": 0.762499988079071,
+      "rewards/chosen": -6.135260581970215,
+      "rewards/margins": 2.2391738891601562,
+      "rewards/rejected": -8.374434471130371,
+      "step": 400
+    },
+    {
+      "epoch": 0.6278202864430057,
+      "eval_logits/chosen": 15.142684936523438,
+      "eval_logits/rejected": 14.101082801818848,
+      "eval_logps/chosen": -3.866588830947876,
+      "eval_logps/rejected": -6.100707530975342,
+      "eval_loss": 0.48699691891670227,
+      "eval_rewards/accuracies": 0.9615384340286255,
+      "eval_rewards/chosen": -5.7998833656311035,
+      "eval_rewards/margins": 3.351177215576172,
+      "eval_rewards/rejected": -9.15106201171875,
+      "eval_runtime": 14.5024,
+      "eval_samples_per_second": 28.409,
+      "eval_steps_per_second": 3.586,
+      "step": 400
+    },
+    {
+      "epoch": 0.6435157936040808,
+      "grad_norm": 1.845595359802246,
+      "learning_rate": 4.133551509975264e-06,
+      "logits/chosen": 14.54790210723877,
+      "logits/rejected": 13.707855224609375,
+      "logps/chosen": -4.112462043762207,
+      "logps/rejected": -6.583975315093994,
+      "loss": 0.5107,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -6.168692111968994,
+      "rewards/margins": 3.707270383834839,
+      "rewards/rejected": -9.875962257385254,
+      "step": 410
+    },
+    {
+      "epoch": 0.659211300765156,
+      "grad_norm": 2.152916193008423,
+      "learning_rate": 4.093559974371725e-06,
+      "logits/chosen": 14.153576850891113,
+      "logits/rejected": 13.553201675415039,
+      "logps/chosen": -4.3408918380737305,
+      "logps/rejected": -6.828585147857666,
+      "loss": 0.41,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -6.511338233947754,
+      "rewards/margins": 3.7315402030944824,
+      "rewards/rejected": -10.242877960205078,
+      "step": 420
+    },
+    {
+      "epoch": 0.6749068079262311,
+      "grad_norm": 2.305433750152588,
+      "learning_rate": 4.052869450695776e-06,
+      "logits/chosen": 13.992170333862305,
+      "logits/rejected": 13.264738082885742,
+      "logps/chosen": -4.259932518005371,
+      "logps/rejected": -6.089646339416504,
+      "loss": 0.4122,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -6.389898777008057,
+      "rewards/margins": 2.7445709705352783,
+      "rewards/rejected": -9.134469985961914,
+      "step": 430
+    },
+    {
+      "epoch": 0.6906023150873063,
+      "grad_norm": 2.397674322128296,
+      "learning_rate": 4.011497787155938e-06,
+      "logits/chosen": 13.975759506225586,
+      "logits/rejected": 13.173799514770508,
+      "logps/chosen": -4.2483391761779785,
+      "logps/rejected": -6.487355709075928,
+      "loss": 0.4397,
+      "rewards/accuracies": 0.8125,
+      "rewards/chosen": -6.3725080490112305,
+      "rewards/margins": 3.3585267066955566,
+      "rewards/rejected": -9.731034278869629,
+      "step": 440
+    },
+    {
+      "epoch": 0.7062978222483814,
+      "grad_norm": 2.3987627029418945,
+      "learning_rate": 3.969463130731183e-06,
+      "logits/chosen": 13.777560234069824,
+      "logits/rejected": 12.936296463012695,
+      "logps/chosen": -3.86059832572937,
+      "logps/rejected": -6.495786190032959,
+      "loss": 0.444,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -5.790897846221924,
+      "rewards/margins": 3.952782392501831,
+      "rewards/rejected": -9.743680000305176,
+      "step": 450
+    },
+    {
+      "epoch": 0.7062978222483814,
+      "eval_logits/chosen": 14.397418975830078,
+      "eval_logits/rejected": 13.271268844604492,
+      "eval_logps/chosen": -3.9544241428375244,
+      "eval_logps/rejected": -6.69989013671875,
+      "eval_loss": 0.435256689786911,
+      "eval_rewards/accuracies": 0.9615384340286255,
+      "eval_rewards/chosen": -5.931635856628418,
+      "eval_rewards/margins": 4.118198871612549,
+      "eval_rewards/rejected": -10.049835205078125,
+      "eval_runtime": 14.5029,
+      "eval_samples_per_second": 28.408,
+      "eval_steps_per_second": 3.585,
+      "step": 450
+    },
+    {
+      "epoch": 0.7219933294094566,
+      "grad_norm": 2.3673465251922607,
+      "learning_rate": 3.92678391921108e-06,
+      "logits/chosen": 13.67004680633545,
+      "logits/rejected": 12.689542770385742,
+      "logps/chosen": -4.037893295288086,
+      "logps/rejected": -6.727287292480469,
+      "loss": 0.3858,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -6.056839942932129,
+      "rewards/margins": 4.034090995788574,
+      "rewards/rejected": -10.09093189239502,
+      "step": 460
+    },
+    {
+      "epoch": 0.7376888365705316,
+      "grad_norm": 2.9868297576904297,
+      "learning_rate": 3.88347887310836e-06,
+      "logits/chosen": 13.124593734741211,
+      "logits/rejected": 12.576837539672852,
+      "logps/chosen": -4.324252605438232,
+      "logps/rejected": -6.761715888977051,
+      "loss": 0.363,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -6.486379146575928,
+      "rewards/margins": 3.6561942100524902,
+      "rewards/rejected": -10.142572402954102,
+      "step": 470
+    },
+    {
+      "epoch": 0.7533843437316068,
+      "grad_norm": 2.613318920135498,
+      "learning_rate": 3.839566987447492e-06,
+      "logits/chosen": 13.514410018920898,
+      "logits/rejected": 12.838116645812988,
+      "logps/chosen": -4.333093166351318,
+      "logps/rejected": -6.704646110534668,
+      "loss": 0.3274,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -6.499639987945557,
+      "rewards/margins": 3.5573298931121826,
+      "rewards/rejected": -10.056970596313477,
+      "step": 480
+    },
+    {
+      "epoch": 0.7690798508926819,
+      "grad_norm": 2.1568610668182373,
+      "learning_rate": 3.795067523432826e-06,
+      "logits/chosen": 13.456674575805664,
+      "logits/rejected": 12.37132453918457,
+      "logps/chosen": -4.630190849304199,
+      "logps/rejected": -7.871635437011719,
+      "loss": 0.3822,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -6.945285797119141,
+      "rewards/margins": 4.862167835235596,
+      "rewards/rejected": -11.807454109191895,
+      "step": 490
+    },
+    {
+      "epoch": 0.7847753580537571,
+      "grad_norm": 2.516496419906616,
+      "learning_rate": 3.7500000000000005e-06,
+      "logits/chosen": 13.195734024047852,
+      "logits/rejected": 12.557401657104492,
+      "logps/chosen": -3.9327914714813232,
+      "logps/rejected": -6.561183929443359,
+      "loss": 0.352,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -5.899188041687012,
+      "rewards/margins": 3.9425880908966064,
+      "rewards/rejected": -9.841775894165039,
+      "step": 500
+    },
+    {
+      "epoch": 0.7847753580537571,
+      "eval_logits/chosen": 13.891414642333984,
+      "eval_logits/rejected": 12.734375,
+      "eval_logps/chosen": -3.924809455871582,
+      "eval_logps/rejected": -7.003909587860107,
+      "eval_loss": 0.3888731002807617,
+      "eval_rewards/accuracies": 1.0,
+      "eval_rewards/chosen": -5.887214183807373,
+      "eval_rewards/margins": 4.618649482727051,
+      "eval_rewards/rejected": -10.505864143371582,
+      "eval_runtime": 14.5029,
+      "eval_samples_per_second": 28.408,
+      "eval_steps_per_second": 3.585,
+      "step": 500
+    },
+    {
+      "epoch": 0.8004708652148322,
+      "grad_norm": 5.4663896560668945,
+      "learning_rate": 3.7043841852542884e-06,
+      "logits/chosen": 12.861404418945312,
+      "logits/rejected": 11.9797945022583,
+      "logps/chosen": -4.245055198669434,
+      "logps/rejected": -7.063906669616699,
+      "loss": 0.4009,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -6.36758279800415,
+      "rewards/margins": 4.228276252746582,
+      "rewards/rejected": -10.59585952758789,
+      "step": 510
+    },
+    {
+      "epoch": 0.8161663723759074,
+      "grad_norm": 2.4403271675109863,
+      "learning_rate": 3.658240087799655e-06,
+      "logits/chosen": 12.84996223449707,
+      "logits/rejected": 12.256246566772461,
+      "logps/chosen": -4.2785773277282715,
+      "logps/rejected": -7.003395080566406,
+      "loss": 0.3327,
+      "rewards/accuracies": 0.9125000238418579,
+      "rewards/chosen": -6.417865753173828,
+      "rewards/margins": 4.0872273445129395,
+      "rewards/rejected": -10.505093574523926,
+      "step": 520
+    },
+    {
+      "epoch": 0.8318618795369825,
+      "grad_norm": 3.596749782562256,
+      "learning_rate": 3.611587947962319e-06,
+      "logits/chosen": 13.3360595703125,
+      "logits/rejected": 12.449459075927734,
+      "logps/chosen": -4.8148322105407715,
+      "logps/rejected": -7.221930503845215,
+      "loss": 0.3255,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -7.222248077392578,
+      "rewards/margins": 3.610647201538086,
+      "rewards/rejected": -10.832895278930664,
+      "step": 530
+    },
+    {
+      "epoch": 0.8475573866980577,
+      "grad_norm": 4.537969589233398,
+      "learning_rate": 3.564448228912682e-06,
+      "logits/chosen": 12.63983154296875,
+      "logits/rejected": 12.042104721069336,
+      "logps/chosen": -4.933573246002197,
+      "logps/rejected": -7.656388282775879,
+      "loss": 0.2941,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": -7.400360107421875,
+      "rewards/margins": 4.084224224090576,
+      "rewards/rejected": -11.484583854675293,
+      "step": 540
+    },
+    {
+      "epoch": 0.8632528938591328,
+      "grad_norm": 2.4124038219451904,
+      "learning_rate": 3.516841607689501e-06,
+      "logits/chosen": 13.053291320800781,
+      "logits/rejected": 12.351530075073242,
+      "logps/chosen": -5.196786403656006,
+      "logps/rejected": -7.8461012840271,
+      "loss": 0.3906,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -7.795179843902588,
+      "rewards/margins": 3.9739716053009033,
+      "rewards/rejected": -11.76915168762207,
+      "step": 550
+    },
+    {
+      "epoch": 0.8632528938591328,
+      "eval_logits/chosen": 13.596405982971191,
+      "eval_logits/rejected": 12.482269287109375,
+      "eval_logps/chosen": -4.328949928283691,
+      "eval_logps/rejected": -7.602427959442139,
+      "eval_loss": 0.3577499985694885,
+      "eval_rewards/accuracies": 0.9807692170143127,
+      "eval_rewards/chosen": -6.493425369262695,
+      "eval_rewards/margins": 4.910217761993408,
+      "eval_rewards/rejected": -11.403642654418945,
+      "eval_runtime": 14.5032,
+      "eval_samples_per_second": 28.408,
+      "eval_steps_per_second": 3.585,
+      "step": 550
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 1500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 50,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3318133628534784e+18,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-550/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b2203cea5a5bcb79dc529b64d6033ba414f923d9d5e3c378f433027100e50b9
+size 7224

checkpoint-550/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)