Training in progress, step 450, checkpoint

checkpoint-450/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-450/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": [
+    "qkv_proj",
+    "down_proj",
+    "gate_up_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-450/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf8d88445e3673c5be168c15627c10287c9887a96a6312a9010cd25543711cbe
+size 25200088

checkpoint-450/global_step450/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:f60bde29d4addfc3f61d628afea4e4dd66c156700180b402585fe6165326586f
+size 18881328

checkpoint-450/global_step450/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:c3a86f8ca13f948dcdaf9e7c736c41f71fefbe9ce9a6780f4cb53e852f0e5996
+size 18881328

checkpoint-450/global_step450/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:629af85b82cff8912e88f36133ce91eecb32026b41f83d78c8d2c545ab51582f
+size 18881328

checkpoint-450/global_step450/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7711900796ada0c8d213efd3d48948a2642f8b6fae46ba67c5b1669df934437a
+size 18881392

checkpoint-450/global_step450/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:d8654b2a3ad6f48e39471cb3a6b94a59903f8cd192eebe1ea05d6066531293e4
+size 18881392

checkpoint-450/global_step450/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:fa52140c027703546ba9b1d7b352aed429cd6af2c99bc61e718736f1b4c98e8d
+size 18881392

checkpoint-450/global_step450/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68a91d0b49a222f0f4dd0758f9f3fe1895ae4b5cf06f5af855579c0651406752
+size 18881392

checkpoint-450/global_step450/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:284d98794537780312f039233040496cb1364be24feb76a061ba57cdb40c3aa9
+size 18881392

checkpoint-450/global_step450/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6813e08d90d4907cf7692dadd344215f6c4e4bdd911a7d8b663b99d474bec341
+size 25379244

checkpoint-450/latest

@@ -0,0 +1 @@
+global_step450

checkpoint-450/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ca51d4b33edcedf9568d5202767b896d828b5aeca18f2cdd82617688464b784
+size 15984

checkpoint-450/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10e6ea705ea5a1704cd5773090c827a2013c8caab967a116ff24a5f57ce3ce90
+size 15984

checkpoint-450/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6ff0b4da12dd0cdcb6e90b04160e41685d9ccc1fa1cc74bb7949edf700200d4
+size 15984

checkpoint-450/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:322470b09ac4f5d9443d55c37c8b8e7d0e8a1702208c81e52e3a58a8de515b5b
+size 15984

checkpoint-450/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ababe8505205ca2bcb959a2abbe2fbc8b6ad677bd43b1f2ee9055b3cb400b061
+size 15984

checkpoint-450/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed1c3ba656fdb40a72824e366a08e148a30e1089a6ecd019eaf28fa4a17859fa
+size 15984

checkpoint-450/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:042c62210b9e1c9ed394e0a3362b1c773c07591d94f2716a8e928676134742b7
+size 15984

checkpoint-450/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb54f0278c663494261026658652f845bae43245e75ccc213c6897de179f542a
+size 15984

checkpoint-450/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d19884021ab25eea895d76397965f6ac5add520c43a6de4e0967f0b97a74efaa
+size 1064

checkpoint-450/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-450/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-450/trainer_state.json

@@ -0,0 +1,852 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.7062978222483814,
+  "eval_steps": 50,
+  "global_step": 450,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.015695507161075144,
+      "grad_norm": 0.07061820477247238,
+      "learning_rate": 4.999451708687114e-06,
+      "logits/chosen": 15.001790046691895,
+      "logits/rejected": 14.624488830566406,
+      "logps/chosen": -0.33085882663726807,
+      "logps/rejected": -0.24924471974372864,
+      "loss": 1.0519,
+      "rewards/accuracies": 0.23749999701976776,
+      "rewards/chosen": -0.4962882399559021,
+      "rewards/margins": -0.12242116779088974,
+      "rewards/rejected": -0.37386709451675415,
+      "step": 10
+    },
+    {
+      "epoch": 0.03139101432215029,
+      "grad_norm": 0.060370542109012604,
+      "learning_rate": 4.997807075247147e-06,
+      "logits/chosen": 15.272351264953613,
+      "logits/rejected": 14.801017761230469,
+      "logps/chosen": -0.3379867672920227,
+      "logps/rejected": -0.24759705364704132,
+      "loss": 1.0494,
+      "rewards/accuracies": 0.15000000596046448,
+      "rewards/chosen": -0.5069801211357117,
+      "rewards/margins": -0.13558456301689148,
+      "rewards/rejected": -0.3713955581188202,
+      "step": 20
+    },
+    {
+      "epoch": 0.047086521483225424,
+      "grad_norm": 0.050332240760326385,
+      "learning_rate": 4.9950668210706795e-06,
+      "logits/chosen": 15.569372177124023,
+      "logits/rejected": 15.249380111694336,
+      "logps/chosen": -0.3347483277320862,
+      "logps/rejected": -0.275604784488678,
+      "loss": 1.0375,
+      "rewards/accuracies": 0.21250000596046448,
+      "rewards/chosen": -0.5021225214004517,
+      "rewards/margins": -0.08871528506278992,
+      "rewards/rejected": -0.41340717673301697,
+      "step": 30
+    },
+    {
+      "epoch": 0.06278202864430057,
+      "grad_norm": 0.054660990834236145,
+      "learning_rate": 4.9912321481237616e-06,
+      "logits/chosen": 15.344067573547363,
+      "logits/rejected": 14.881669998168945,
+      "logps/chosen": -0.3336474299430847,
+      "logps/rejected": -0.27252259850502014,
+      "loss": 1.0375,
+      "rewards/accuracies": 0.20000000298023224,
+      "rewards/chosen": -0.5004712343215942,
+      "rewards/margins": -0.09168727695941925,
+      "rewards/rejected": -0.40878385305404663,
+      "step": 40
+    },
+    {
+      "epoch": 0.07847753580537571,
+      "grad_norm": 0.060405269265174866,
+      "learning_rate": 4.986304738420684e-06,
+      "logits/chosen": 15.5701904296875,
+      "logits/rejected": 15.60998821258545,
+      "logps/chosen": -0.3075069785118103,
+      "logps/rejected": -0.24850216507911682,
+      "loss": 1.0341,
+      "rewards/accuracies": 0.20000000298023224,
+      "rewards/chosen": -0.46126049757003784,
+      "rewards/margins": -0.08850721269845963,
+      "rewards/rejected": -0.3727532923221588,
+      "step": 50
+    },
+    {
+      "epoch": 0.07847753580537571,
+      "eval_logits/chosen": 15.719181060791016,
+      "eval_logits/rejected": 15.20205307006836,
+      "eval_logps/chosen": -0.32490479946136475,
+      "eval_logps/rejected": -0.26058924198150635,
+      "eval_loss": 1.0295902490615845,
+      "eval_rewards/accuracies": 0.26923078298568726,
+      "eval_rewards/chosen": -0.48735716938972473,
+      "eval_rewards/margins": -0.09647335112094879,
+      "eval_rewards/rejected": -0.39088380336761475,
+      "eval_runtime": 14.6102,
+      "eval_samples_per_second": 28.199,
+      "eval_steps_per_second": 3.559,
+      "step": 50
+    },
+    {
+      "epoch": 0.09417304296645085,
+      "grad_norm": 0.06511708348989487,
+      "learning_rate": 4.980286753286196e-06,
+      "logits/chosen": 15.571348190307617,
+      "logits/rejected": 15.418828010559082,
+      "logps/chosen": -0.3217025101184845,
+      "logps/rejected": -0.2813836932182312,
+      "loss": 1.0385,
+      "rewards/accuracies": 0.23749999701976776,
+      "rewards/chosen": -0.48255378007888794,
+      "rewards/margins": -0.060478221625089645,
+      "rewards/rejected": -0.4220755696296692,
+      "step": 60
+    },
+    {
+      "epoch": 0.109868550127526,
+      "grad_norm": 0.06922140717506409,
+      "learning_rate": 4.973180832407471e-06,
+      "logits/chosen": 15.926614761352539,
+      "logits/rejected": 15.90565299987793,
+      "logps/chosen": -0.34742942452430725,
+      "logps/rejected": -0.26853513717651367,
+      "loss": 1.0417,
+      "rewards/accuracies": 0.21250000596046448,
+      "rewards/chosen": -0.5211440324783325,
+      "rewards/margins": -0.11834144592285156,
+      "rewards/rejected": -0.40280264616012573,
+      "step": 70
+    },
+    {
+      "epoch": 0.12556405728860115,
+      "grad_norm": 0.07853339612483978,
+      "learning_rate": 4.964990092676263e-06,
+      "logits/chosen": 15.876957893371582,
+      "logits/rejected": 15.636863708496094,
+      "logps/chosen": -0.34999752044677734,
+      "logps/rejected": -0.2652502655982971,
+      "loss": 1.0255,
+      "rewards/accuracies": 0.1875,
+      "rewards/chosen": -0.524996280670166,
+      "rewards/margins": -0.12712089717388153,
+      "rewards/rejected": -0.3978753685951233,
+      "step": 80
+    },
+    {
+      "epoch": 0.14125956444967627,
+      "grad_norm": 0.07262148708105087,
+      "learning_rate": 4.9557181268217225e-06,
+      "logits/chosen": 16.037456512451172,
+      "logits/rejected": 15.742408752441406,
+      "logps/chosen": -0.32482510805130005,
+      "logps/rejected": -0.2520079016685486,
+      "loss": 1.0297,
+      "rewards/accuracies": 0.21250000596046448,
+      "rewards/chosen": -0.48723769187927246,
+      "rewards/margins": -0.1092257872223854,
+      "rewards/rejected": -0.3780118525028229,
+      "step": 90
+    },
+    {
+      "epoch": 0.15695507161075142,
+      "grad_norm": 0.08407289534807205,
+      "learning_rate": 4.9453690018345144e-06,
+      "logits/chosen": 16.34904670715332,
+      "logits/rejected": 16.30767059326172,
+      "logps/chosen": -0.3302846848964691,
+      "logps/rejected": -0.2799247205257416,
+      "loss": 1.0285,
+      "rewards/accuracies": 0.2750000059604645,
+      "rewards/chosen": -0.4954269826412201,
+      "rewards/margins": -0.07553993165493011,
+      "rewards/rejected": -0.41988706588745117,
+      "step": 100
+    },
+    {
+      "epoch": 0.15695507161075142,
+      "eval_logits/chosen": 16.43848419189453,
+      "eval_logits/rejected": 15.978095054626465,
+      "eval_logps/chosen": -0.3256901502609253,
+      "eval_logps/rejected": -0.2794351279735565,
+      "eval_loss": 1.0174708366394043,
+      "eval_rewards/accuracies": 0.2884615361690521,
+      "eval_rewards/chosen": -0.4885352551937103,
+      "eval_rewards/margins": -0.06938254088163376,
+      "eval_rewards/rejected": -0.41915270686149597,
+      "eval_runtime": 14.607,
+      "eval_samples_per_second": 28.206,
+      "eval_steps_per_second": 3.56,
+      "step": 100
+    },
+    {
+      "epoch": 0.17265057877182657,
+      "grad_norm": 0.07448805868625641,
+      "learning_rate": 4.933947257182901e-06,
+      "logits/chosen": 16.681476593017578,
+      "logits/rejected": 16.208908081054688,
+      "logps/chosen": -0.3410753309726715,
+      "logps/rejected": -0.28177398443222046,
+      "loss": 1.0364,
+      "rewards/accuracies": 0.2874999940395355,
+      "rewards/chosen": -0.5116130113601685,
+      "rewards/margins": -0.08895199000835419,
+      "rewards/rejected": -0.4226610064506531,
+      "step": 110
+    },
+    {
+      "epoch": 0.1883460859329017,
+      "grad_norm": 0.17266640067100525,
+      "learning_rate": 4.921457902821578e-06,
+      "logits/chosen": 16.73276710510254,
+      "logits/rejected": 16.46255111694336,
+      "logps/chosen": -0.30808666348457336,
+      "logps/rejected": -0.2663067877292633,
+      "loss": 1.0242,
+      "rewards/accuracies": 0.23749999701976776,
+      "rewards/chosen": -0.46213001012802124,
+      "rewards/margins": -0.0626697838306427,
+      "rewards/rejected": -0.39946022629737854,
+      "step": 120
+    },
+    {
+      "epoch": 0.20404159309397685,
+      "grad_norm": 0.1035998985171318,
+      "learning_rate": 4.907906416994146e-06,
+      "logits/chosen": 16.380239486694336,
+      "logits/rejected": 16.233612060546875,
+      "logps/chosen": -0.32250356674194336,
+      "logps/rejected": -0.27460020780563354,
+      "loss": 1.0162,
+      "rewards/accuracies": 0.2874999940395355,
+      "rewards/chosen": -0.48375529050827026,
+      "rewards/margins": -0.0718550831079483,
+      "rewards/rejected": -0.41190028190612793,
+      "step": 130
+    },
+    {
+      "epoch": 0.219737100255052,
+      "grad_norm": 0.0961030125617981,
+      "learning_rate": 4.893298743830168e-06,
+      "logits/chosen": 16.440303802490234,
+      "logits/rejected": 16.233903884887695,
+      "logps/chosen": -0.33421438932418823,
+      "logps/rejected": -0.2857271134853363,
+      "loss": 1.0008,
+      "rewards/accuracies": 0.3125,
+      "rewards/chosen": -0.5013214945793152,
+      "rewards/margins": -0.07273083180189133,
+      "rewards/rejected": -0.42859068512916565,
+      "step": 140
+    },
+    {
+      "epoch": 0.23543260741612712,
+      "grad_norm": 0.14799565076828003,
+      "learning_rate": 4.8776412907378845e-06,
+      "logits/chosen": 16.600738525390625,
+      "logits/rejected": 16.536447525024414,
+      "logps/chosen": -0.3243326246738434,
+      "logps/rejected": -0.3061785101890564,
+      "loss": 1.0085,
+      "rewards/accuracies": 0.3375000059604645,
+      "rewards/chosen": -0.4864989221096039,
+      "rewards/margins": -0.027231160551309586,
+      "rewards/rejected": -0.459267795085907,
+      "step": 150
+    },
+    {
+      "epoch": 0.23543260741612712,
+      "eval_logits/chosen": 16.8078556060791,
+      "eval_logits/rejected": 16.311922073364258,
+      "eval_logps/chosen": -0.326242595911026,
+      "eval_logps/rejected": -0.3398977816104889,
+      "eval_loss": 0.9791463017463684,
+      "eval_rewards/accuracies": 0.4038461446762085,
+      "eval_rewards/chosen": -0.4893638789653778,
+      "eval_rewards/margins": 0.020482787862420082,
+      "eval_rewards/rejected": -0.5098467469215393,
+      "eval_runtime": 14.6072,
+      "eval_samples_per_second": 28.205,
+      "eval_steps_per_second": 3.56,
+      "step": 150
+    },
+    {
+      "epoch": 0.2511281145772023,
+      "grad_norm": 0.11464422941207886,
+      "learning_rate": 4.860940925593703e-06,
+      "logits/chosen": 16.586620330810547,
+      "logits/rejected": 16.253028869628906,
+      "logps/chosen": -0.33331993222236633,
+      "logps/rejected": -0.35580095648765564,
+      "loss": 0.9827,
+      "rewards/accuracies": 0.42500001192092896,
+      "rewards/chosen": -0.49997982382774353,
+      "rewards/margins": 0.03372158855199814,
+      "rewards/rejected": -0.5337014198303223,
+      "step": 160
+    },
+    {
+      "epoch": 0.2668236217382774,
+      "grad_norm": 0.09462621062994003,
+      "learning_rate": 4.84320497372973e-06,
+      "logits/chosen": 16.220312118530273,
+      "logits/rejected": 16.105356216430664,
+      "logps/chosen": -0.2957116663455963,
+      "logps/rejected": -0.3528788089752197,
+      "loss": 0.9716,
+      "rewards/accuracies": 0.4625000059604645,
+      "rewards/chosen": -0.44356757402420044,
+      "rewards/margins": 0.08575066924095154,
+      "rewards/rejected": -0.5293182134628296,
+      "step": 170
+    },
+    {
+      "epoch": 0.28251912889935255,
+      "grad_norm": 0.10995513945817947,
+      "learning_rate": 4.824441214720629e-06,
+      "logits/chosen": 16.811044692993164,
+      "logits/rejected": 16.43205451965332,
+      "logps/chosen": -0.3263992369174957,
+      "logps/rejected": -0.35878634452819824,
+      "loss": 0.9546,
+      "rewards/accuracies": 0.4124999940395355,
+      "rewards/chosen": -0.4895988404750824,
+      "rewards/margins": 0.04858064278960228,
+      "rewards/rejected": -0.5381795167922974,
+      "step": 180
+    },
+    {
+      "epoch": 0.2982146360604277,
+      "grad_norm": 0.1348036825656891,
+      "learning_rate": 4.804657878971252e-06,
+      "logits/chosen": 16.816158294677734,
+      "logits/rejected": 16.55413055419922,
+      "logps/chosen": -0.31842875480651855,
+      "logps/rejected": -0.3791810870170593,
+      "loss": 0.9472,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": -0.47764310240745544,
+      "rewards/margins": 0.09112847596406937,
+      "rewards/rejected": -0.5687715411186218,
+      "step": 190
+    },
+    {
+      "epoch": 0.31391014322150285,
+      "grad_norm": 0.13393694162368774,
+      "learning_rate": 4.783863644106502e-06,
+      "logits/chosen": 17.075885772705078,
+      "logits/rejected": 16.982004165649414,
+      "logps/chosen": -0.3368823528289795,
+      "logps/rejected": -0.3636801540851593,
+      "loss": 0.9455,
+      "rewards/accuracies": 0.375,
+      "rewards/chosen": -0.5053235292434692,
+      "rewards/margins": 0.04019671678543091,
+      "rewards/rejected": -0.5455202460289001,
+      "step": 200
+    },
+    {
+      "epoch": 0.31391014322150285,
+      "eval_logits/chosen": 17.17239761352539,
+      "eval_logits/rejected": 16.609193801879883,
+      "eval_logps/chosen": -0.33508703112602234,
+      "eval_logps/rejected": -0.4493505656719208,
+      "eval_loss": 0.9283667802810669,
+      "eval_rewards/accuracies": 0.4615384638309479,
+      "eval_rewards/chosen": -0.5026305913925171,
+      "eval_rewards/margins": 0.17139528691768646,
+      "eval_rewards/rejected": -0.67402583360672,
+      "eval_runtime": 14.5913,
+      "eval_samples_per_second": 28.236,
+      "eval_steps_per_second": 3.564,
+      "step": 200
+    },
+    {
+      "epoch": 0.329605650382578,
+      "grad_norm": 0.14985737204551697,
+      "learning_rate": 4.762067631165049e-06,
+      "logits/chosen": 16.92913246154785,
+      "logits/rejected": 16.648632049560547,
+      "logps/chosen": -0.3326043486595154,
+      "logps/rejected": -0.4218205511569977,
+      "loss": 0.9225,
+      "rewards/accuracies": 0.48750001192092896,
+      "rewards/chosen": -0.4989064335823059,
+      "rewards/margins": 0.13382436335086823,
+      "rewards/rejected": -0.6327308416366577,
+      "step": 210
+    },
+    {
+      "epoch": 0.34530115754365315,
+      "grad_norm": 0.16972233355045319,
+      "learning_rate": 4.7392794005985324e-06,
+      "logits/chosen": 17.074562072753906,
+      "logits/rejected": 16.863161087036133,
+      "logps/chosen": -0.3645358383655548,
+      "logps/rejected": -0.5221719145774841,
+      "loss": 0.9264,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": -0.5468038320541382,
+      "rewards/margins": 0.23645417392253876,
+      "rewards/rejected": -0.7832580208778381,
+      "step": 220
+    },
+    {
+      "epoch": 0.3609966647047283,
+      "grad_norm": 0.20796354115009308,
+      "learning_rate": 4.715508948078037e-06,
+      "logits/chosen": 17.4168643951416,
+      "logits/rejected": 17.00813102722168,
+      "logps/chosen": -0.35443297028541565,
+      "logps/rejected": -0.5115998983383179,
+      "loss": 0.9037,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": -0.5316494703292847,
+      "rewards/margins": 0.23575039207935333,
+      "rewards/rejected": -0.767399787902832,
+      "step": 230
+    },
+    {
+      "epoch": 0.3766921718658034,
+      "grad_norm": 0.1932348757982254,
+      "learning_rate": 4.690766700109659e-06,
+      "logits/chosen": 17.034719467163086,
+      "logits/rejected": 16.65166664123535,
+      "logps/chosen": -0.3426091969013214,
+      "logps/rejected": -0.47884297370910645,
+      "loss": 0.9154,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.5139138102531433,
+      "rewards/margins": 0.20435063540935516,
+      "rewards/rejected": -0.7182644605636597,
+      "step": 240
+    },
+    {
+      "epoch": 0.39238767902687854,
+      "grad_norm": 1.016450047492981,
+      "learning_rate": 4.665063509461098e-06,
+      "logits/chosen": 16.935081481933594,
+      "logits/rejected": 16.65024185180664,
+      "logps/chosen": -0.41738080978393555,
+      "logps/rejected": -0.48515433073043823,
+      "loss": 0.8774,
+      "rewards/accuracies": 0.42500001192092896,
+      "rewards/chosen": -0.6260712146759033,
+      "rewards/margins": 0.10166029632091522,
+      "rewards/rejected": -0.7277315258979797,
+      "step": 250
+    },
+    {
+      "epoch": 0.39238767902687854,
+      "eval_logits/chosen": 17.221881866455078,
+      "eval_logits/rejected": 16.536666870117188,
+      "eval_logps/chosen": -0.3716265559196472,
+      "eval_logps/rejected": -0.87992262840271,
+      "eval_loss": 0.8047741651535034,
+      "eval_rewards/accuracies": 0.5,
+      "eval_rewards/chosen": -0.5574398636817932,
+      "eval_rewards/margins": 0.7624441981315613,
+      "eval_rewards/rejected": -1.3198840618133545,
+      "eval_runtime": 14.6008,
+      "eval_samples_per_second": 28.218,
+      "eval_steps_per_second": 3.561,
+      "step": 250
+    },
+    {
+      "epoch": 0.4080831861879537,
+      "grad_norm": 0.2831684648990631,
+      "learning_rate": 4.638410650401267e-06,
+      "logits/chosen": 17.084096908569336,
+      "logits/rejected": 16.564823150634766,
+      "logps/chosen": -0.38100525736808777,
+      "logps/rejected": -0.8173269033432007,
+      "loss": 0.8197,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.5715079307556152,
+      "rewards/margins": 0.6544824838638306,
+      "rewards/rejected": -1.2259904146194458,
+      "step": 260
+    },
+    {
+      "epoch": 0.42377869334902885,
+      "grad_norm": 0.9195305109024048,
+      "learning_rate": 4.610819813755038e-06,
+      "logits/chosen": 17.316537857055664,
+      "logits/rejected": 16.623016357421875,
+      "logps/chosen": -0.4713365435600281,
+      "logps/rejected": -1.2184536457061768,
+      "loss": 0.7493,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.7070047855377197,
+      "rewards/margins": 1.120675802230835,
+      "rewards/rejected": -1.8276805877685547,
+      "step": 270
+    },
+    {
+      "epoch": 0.439474200510104,
+      "grad_norm": 0.5542411804199219,
+      "learning_rate": 4.582303101775249e-06,
+      "logits/chosen": 17.173168182373047,
+      "logits/rejected": 16.759702682495117,
+      "logps/chosen": -0.4629506468772888,
+      "logps/rejected": -1.3020483255386353,
+      "loss": 0.7638,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.6944260597229004,
+      "rewards/margins": 1.2586463689804077,
+      "rewards/rejected": -1.9530725479125977,
+      "step": 280
+    },
+    {
+      "epoch": 0.45516970767117915,
+      "grad_norm": 0.2836654782295227,
+      "learning_rate": 4.55287302283426e-06,
+      "logits/chosen": 17.022424697875977,
+      "logits/rejected": 16.43834114074707,
+      "logps/chosen": -0.527222216129303,
+      "logps/rejected": -1.632965087890625,
+      "loss": 0.7466,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.7908332943916321,
+      "rewards/margins": 1.6586145162582397,
+      "rewards/rejected": -2.4494476318359375,
+      "step": 290
+    },
+    {
+      "epoch": 0.47086521483225424,
+      "grad_norm": 0.658358097076416,
+      "learning_rate": 4.522542485937369e-06,
+      "logits/chosen": 16.6541690826416,
+      "logits/rejected": 16.34103012084961,
+      "logps/chosen": -0.50589519739151,
+      "logps/rejected": -1.8910831212997437,
+      "loss": 0.6982,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.7588427662849426,
+      "rewards/margins": 2.077782154083252,
+      "rewards/rejected": -2.8366246223449707,
+      "step": 300
+    },
+    {
+      "epoch": 0.47086521483225424,
+      "eval_logits/chosen": 16.950870513916016,
+      "eval_logits/rejected": 16.154659271240234,
+      "eval_logps/chosen": -0.5594518184661865,
+      "eval_logps/rejected": -1.9617934226989746,
+      "eval_loss": 0.7422243356704712,
+      "eval_rewards/accuracies": 0.7115384340286255,
+      "eval_rewards/chosen": -0.839177668094635,
+      "eval_rewards/margins": 2.103512763977051,
+      "eval_rewards/rejected": -2.942690372467041,
+      "eval_runtime": 14.5969,
+      "eval_samples_per_second": 28.225,
+      "eval_steps_per_second": 3.562,
+      "step": 300
+    },
+    {
+      "epoch": 0.4865607219933294,
+      "grad_norm": 1.1044840812683105,
+      "learning_rate": 4.491324795060491e-06,
+      "logits/chosen": 17.233802795410156,
+      "logits/rejected": 16.292484283447266,
+      "logps/chosen": -0.5940151214599609,
+      "logps/rejected": -2.041588306427002,
+      "loss": 0.7415,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -0.8910226821899414,
+      "rewards/margins": 2.1713600158691406,
+      "rewards/rejected": -3.062382459640503,
+      "step": 310
+    },
+    {
+      "epoch": 0.5022562291544046,
+      "grad_norm": 0.524356484413147,
+      "learning_rate": 4.4592336433146e-06,
+      "logits/chosen": 17.155719757080078,
+      "logits/rejected": 16.328996658325195,
+      "logps/chosen": -0.6095865964889526,
+      "logps/rejected": -1.831730604171753,
+      "loss": 0.7249,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.9143797755241394,
+      "rewards/margins": 1.8332160711288452,
+      "rewards/rejected": -2.747596025466919,
+      "step": 320
+    },
+    {
+      "epoch": 0.5179517363154797,
+      "grad_norm": 0.4383145570755005,
+      "learning_rate": 4.426283106939474e-06,
+      "logits/chosen": 16.774551391601562,
+      "logits/rejected": 16.532238006591797,
+      "logps/chosen": -0.669482946395874,
+      "logps/rejected": -1.7058660984039307,
+      "loss": 0.6868,
+      "rewards/accuracies": 0.6625000238418579,
+      "rewards/chosen": -1.004224419593811,
+      "rewards/margins": 1.5545748472213745,
+      "rewards/rejected": -2.5587992668151855,
+      "step": 330
+    },
+    {
+      "epoch": 0.5336472434765548,
+      "grad_norm": 3.194026231765747,
+      "learning_rate": 4.3924876391293915e-06,
+      "logits/chosen": 16.963787078857422,
+      "logits/rejected": 16.3092041015625,
+      "logps/chosen": -0.7091449499130249,
+      "logps/rejected": -2.1812453269958496,
+      "loss": 0.706,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": -1.0637174844741821,
+      "rewards/margins": 2.208150863647461,
+      "rewards/rejected": -3.2718684673309326,
+      "step": 340
+    },
+    {
+      "epoch": 0.5493427506376299,
+      "grad_norm": 0.5639687776565552,
+      "learning_rate": 4.357862063693486e-06,
+      "logits/chosen": 17.019062042236328,
+      "logits/rejected": 16.315940856933594,
+      "logps/chosen": -0.8996315002441406,
+      "logps/rejected": -2.0307648181915283,
+      "loss": 0.674,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -1.349447250366211,
+      "rewards/margins": 1.696699857711792,
+      "rewards/rejected": -3.046147108078003,
+      "step": 350
+    },
+    {
+      "epoch": 0.5493427506376299,
+      "eval_logits/chosen": 16.70261001586914,
+      "eval_logits/rejected": 15.845681190490723,
+      "eval_logps/chosen": -0.9138904809951782,
+      "eval_logps/rejected": -2.5003392696380615,
+      "eval_loss": 0.6938430070877075,
+      "eval_rewards/accuracies": 0.8461538553237915,
+      "eval_rewards/chosen": -1.3708356618881226,
+      "eval_rewards/margins": 2.3796732425689697,
+      "eval_rewards/rejected": -3.75050950050354,
+      "eval_runtime": 14.6072,
+      "eval_samples_per_second": 28.205,
+      "eval_steps_per_second": 3.56,
+      "step": 350
+    },
+    {
+      "epoch": 0.5650382577987051,
+      "grad_norm": 0.6917555928230286,
+      "learning_rate": 4.322421568553529e-06,
+      "logits/chosen": 16.269399642944336,
+      "logits/rejected": 15.73193645477295,
+      "logps/chosen": -1.0171478986740112,
+      "logps/rejected": -2.201298236846924,
+      "loss": 0.651,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -1.525721788406372,
+      "rewards/margins": 1.7762253284454346,
+      "rewards/rejected": -3.3019473552703857,
+      "step": 360
+    },
+    {
+      "epoch": 0.5807337649597802,
+      "grad_norm": 1.151584267616272,
+      "learning_rate": 4.286181699082008e-06,
+      "logits/chosen": 15.88188648223877,
+      "logits/rejected": 15.261810302734375,
+      "logps/chosen": -1.2272206544876099,
+      "logps/rejected": -3.0203287601470947,
+      "loss": 0.665,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": -1.8408310413360596,
+      "rewards/margins": 2.689661741256714,
+      "rewards/rejected": -4.530492305755615,
+      "step": 370
+    },
+    {
+      "epoch": 0.5964292721208554,
+      "grad_norm": 1.1879621744155884,
+      "learning_rate": 4.249158351283414e-06,
+      "logits/chosen": 15.906834602355957,
+      "logits/rejected": 15.42895221710205,
+      "logps/chosen": -1.5027008056640625,
+      "logps/rejected": -2.9172558784484863,
+      "loss": 0.5998,
+      "rewards/accuracies": 0.887499988079071,
+      "rewards/chosen": -2.2540509700775146,
+      "rewards/margins": 2.1218326091766357,
+      "rewards/rejected": -4.37588357925415,
+      "step": 380
+    },
+    {
+      "epoch": 0.6121247792819305,
+      "grad_norm": 1.678884744644165,
+      "learning_rate": 4.211367764821722e-06,
+      "logits/chosen": 15.630645751953125,
+      "logits/rejected": 14.888415336608887,
+      "logps/chosen": -2.231367349624634,
+      "logps/rejected": -3.5754780769348145,
+      "loss": 0.5937,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": -3.347050905227661,
+      "rewards/margins": 2.0161664485931396,
+      "rewards/rejected": -5.363216876983643,
+      "step": 390
+    },
+    {
+      "epoch": 0.6278202864430057,
+      "grad_norm": 2.008671998977661,
+      "learning_rate": 4.172826515897146e-06,
+      "logits/chosen": 15.024614334106445,
+      "logits/rejected": 14.981277465820312,
+      "logps/chosen": -2.665980815887451,
+      "logps/rejected": -3.582763195037842,
+      "loss": 0.5722,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": -3.998971462249756,
+      "rewards/margins": 1.3751739263534546,
+      "rewards/rejected": -5.374145030975342,
+      "step": 400
+    },
+    {
+      "epoch": 0.6278202864430057,
+      "eval_logits/chosen": 15.45953369140625,
+      "eval_logits/rejected": 14.663763046264648,
+      "eval_logps/chosen": -2.7269070148468018,
+      "eval_logps/rejected": -4.235719680786133,
+      "eval_loss": 0.5668805837631226,
+      "eval_rewards/accuracies": 0.8846153616905212,
+      "eval_rewards/chosen": -4.090360641479492,
+      "eval_rewards/margins": 2.263219118118286,
+      "eval_rewards/rejected": -6.353579998016357,
+      "eval_runtime": 14.6036,
+      "eval_samples_per_second": 28.212,
+      "eval_steps_per_second": 3.561,
+      "step": 400
+    },
+    {
+      "epoch": 0.6435157936040808,
+      "grad_norm": 1.018362045288086,
+      "learning_rate": 4.133551509975264e-06,
+      "logits/chosen": 15.027656555175781,
+      "logits/rejected": 14.74786376953125,
+      "logps/chosen": -2.76598858833313,
+      "logps/rejected": -4.335003852844238,
+      "loss": 0.5412,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": -4.148982524871826,
+      "rewards/margins": 2.3535237312316895,
+      "rewards/rejected": -6.502506256103516,
+      "step": 410
+    },
+    {
+      "epoch": 0.659211300765156,
+      "grad_norm": 2.2899765968322754,
+      "learning_rate": 4.093559974371725e-06,
+      "logits/chosen": 15.15197467803955,
+      "logits/rejected": 14.681947708129883,
+      "logps/chosen": -2.9831995964050293,
+      "logps/rejected": -4.723050117492676,
+      "loss": 0.4652,
+      "rewards/accuracies": 0.862500011920929,
+      "rewards/chosen": -4.474799633026123,
+      "rewards/margins": 2.6097757816314697,
+      "rewards/rejected": -7.0845746994018555,
+      "step": 420
+    },
+    {
+      "epoch": 0.6749068079262311,
+      "grad_norm": 1.419196605682373,
+      "learning_rate": 4.052869450695776e-06,
+      "logits/chosen": 15.000317573547363,
+      "logits/rejected": 14.538591384887695,
+      "logps/chosen": -3.543004274368286,
+      "logps/rejected": -5.037484169006348,
+      "loss": 0.427,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": -5.314507007598877,
+      "rewards/margins": 2.2417192459106445,
+      "rewards/rejected": -7.556225776672363,
+      "step": 430
+    },
+    {
+      "epoch": 0.6906023150873063,
+      "grad_norm": 1.639215111732483,
+      "learning_rate": 4.011497787155938e-06,
+      "logits/chosen": 14.798017501831055,
+      "logits/rejected": 14.236108779907227,
+      "logps/chosen": -4.154418468475342,
+      "logps/rejected": -6.0606184005737305,
+      "loss": 0.4692,
+      "rewards/accuracies": 0.8374999761581421,
+      "rewards/chosen": -6.231626987457275,
+      "rewards/margins": 2.859299421310425,
+      "rewards/rejected": -9.090926170349121,
+      "step": 440
+    },
+    {
+      "epoch": 0.7062978222483814,
+      "grad_norm": 1.6229957342147827,
+      "learning_rate": 3.969463130731183e-06,
+      "logits/chosen": 14.495908737182617,
+      "logits/rejected": 13.978658676147461,
+      "logps/chosen": -3.4806721210479736,
+      "logps/rejected": -5.746390342712402,
+      "loss": 0.4218,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": -5.221007823944092,
+      "rewards/margins": 3.3985772132873535,
+      "rewards/rejected": -8.619585037231445,
+      "step": 450
+    },
+    {
+      "epoch": 0.7062978222483814,
+      "eval_logits/chosen": 15.033548355102539,
+      "eval_logits/rejected": 14.20669174194336,
+      "eval_logps/chosen": -3.423999547958374,
+      "eval_logps/rejected": -5.855647563934326,
+      "eval_loss": 0.45794492959976196,
+      "eval_rewards/accuracies": 0.942307710647583,
+      "eval_rewards/chosen": -5.135998725891113,
+      "eval_rewards/margins": 3.6474733352661133,
+      "eval_rewards/rejected": -8.783472061157227,
+      "eval_runtime": 14.6088,
+      "eval_samples_per_second": 28.202,
+      "eval_steps_per_second": 3.559,
+      "step": 450
+    }
+  ],
+  "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": 1.0766596336548577e+18,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-450/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5163ac022a6b5f37a143d69c025be3610fb33a65145b769bdd1d9b31cc494442
+size 7224

checkpoint-450/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)