Training in progress, step 250, checkpoint

checkpoint-250/README.md

@@ -0,0 +1,202 @@
+---
+base_model: meta-llama/CodeLlama-7b-Instruct-hf
+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-250/adapter_config.json

@@ -0,0 +1,37 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "meta-llama/CodeLlama-7b-Instruct-hf",
+  "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": [
+    "k_proj",
+    "gate_proj",
+    "o_proj",
+    "v_proj",
+    "down_proj",
+    "q_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

checkpoint-250/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b266be409db7d879882292a4790ace968b4aae610e8e251fee9bd6a8c6f685cf
+size 40036488

checkpoint-250/global_step250/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:fc5c4da5e35a00fe9afcaaf0e929b043d44f001df9e654c66a7766f451798f20
+size 29992112

checkpoint-250/global_step250/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:a46de46647fedb4954f944c91555663389bff03b721cbd547f9c72c8689aa563
+size 29992112

checkpoint-250/global_step250/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:d203055acf472677528eb0d4d18e6369cf7b71a91bee518fe7591e7fca30e06b
+size 29992176

checkpoint-250/global_step250/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e41eabfc1e815c11c2b48b34520cc47ed39cde456b69e524b40f64cab8430d0
+size 29992176

checkpoint-250/global_step250/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76db4053a3c5b78cb784a68632784461dbd775cba09dc19d7aa14ab96faf5897
+size 29992176

checkpoint-250/global_step250/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:ef117051da661594970aa56af41480fdc0239df6debec88b0b83d46b3c7794af
+size 29992176

checkpoint-250/global_step250/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20257efce253d3668d71b187f351eb20f7478b85cdfdf57a9254540bf27643a7
+size 29992176

checkpoint-250/global_step250/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1adaa1cf3a98ff4ffa9ea44907082d0b052e5b25d76a59489eed497da8a0fcba
+size 29992176

checkpoint-250/global_step250/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e352f5c83dae1ef761ea89ad342b9fde2c1a669be6c8bcac8814033db3ca7db
+size 40324204

checkpoint-250/latest

@@ -0,0 +1 @@
+global_step250

checkpoint-250/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69ec6e3926fa071bede113523efa3dc6e630c3c7958c54a9ca321cf4d62ed145
+size 15984

checkpoint-250/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6127ee4f0c13500ec5038fce65af8f7beec63c137c7d4b7c157aa6303cf5879
+size 15984

checkpoint-250/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da01d1c5eb2cc3a323f97c1f590d13ccfac2a4c5b1479bd378b4e643304f5a4f
+size 15984

checkpoint-250/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49a3f04d76c0d3acc7d3dd95a04215f368f35a451ae8cba8a2fdba38cda9ca0a
+size 15984

checkpoint-250/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df7d2c9825dba80cb544920f8cc0c72122f96514e6cd259052a8765b034393e2
+size 15984

checkpoint-250/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a20a42d44ff48cc162224010190e898fe28598ddad8cd1896d330a3bb1d8ec3
+size 15984

checkpoint-250/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18ac0dc4f09f25179860561fcea7c5c8f997aabdc46a170665f9dc5a72bc27c6
+size 15984

checkpoint-250/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a16fcb5411ff961b47eff7378d85105fe9837e0492d19ea5ce3b7c4b77aa3b6
+size 15984

checkpoint-250/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75885307ef1cc180826720de67db9b95b3068a98db8ed9ec5c12cb5fecf9dc4f
+size 1064

checkpoint-250/special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "additional_special_tokens": [
+    "▁<PRE>",
+    "▁<MID>",
+    "▁<SUF>",
+    "▁<EOT>"
+  ],
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "</s>",
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-250/tokenizer_config.json

@@ -0,0 +1,87 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32007": {
+      "content": "▁<PRE>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32008": {
+      "content": "▁<SUF>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32009": {
+      "content": "▁<MID>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32010": {
+      "content": "▁<EOT>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "▁<PRE>",
+    "▁<MID>",
+    "▁<SUF>",
+    "▁<EOT>"
+  ],
+  "bos_token": "<s>",
+  "chat_template": "{% if messages[0]['role'] == 'system' %}{% set system_message = messages[0]['content'] %}{% endif %}{% for message in messages %}{% set content = message['content'] %}{% if loop.index0 == 0 and system_message is defined %}{% set content = '<<SYS>>\n' + system_message + '\n<</SYS>>\n\n' + message['content'] %}{% endif %}{% if message['role'] == 'user' %}{{ '<s>' + '[INST] ' + content + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + content + ' ' + '</s>' }}{% endif %}{% endfor %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "eot_token": "▁<EOT>",
+  "extra_special_tokens": {},
+  "fill_token": "<FILL_ME>",
+  "legacy": null,
+  "middle_token": "▁<MID>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "</s>",
+  "padding_side": "right",
+  "prefix_token": "▁<PRE>",
+  "sp_model_kwargs": {},
+  "split_special_tokens": false,
+  "suffix_token": "▁<SUF>",
+  "tokenizer_class": "CodeLlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

checkpoint-250/trainer_state.json

@@ -0,0 +1,488 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.1812086617740328,
+  "eval_steps": 50,
+  "global_step": 250,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.007248346470961312,
+      "grad_norm": 0.07201674580574036,
+      "learning_rate": 4.999451708687114e-06,
+      "logits/chosen": -1.9793262481689453,
+      "logits/rejected": -2.5381760597229004,
+      "logps/chosen": -0.28126341104507446,
+      "logps/rejected": -0.3779803514480591,
+      "loss": 7.3904,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.4218950867652893,
+      "rewards/margins": 0.14507544040679932,
+      "rewards/rejected": -0.5669704675674438,
+      "step": 10
+    },
+    {
+      "epoch": 0.014496692941922623,
+      "grad_norm": 0.07562297582626343,
+      "learning_rate": 4.997807075247147e-06,
+      "logits/chosen": -2.0567643642425537,
+      "logits/rejected": -2.4989147186279297,
+      "logps/chosen": -0.27690139412879944,
+      "logps/rejected": -0.33544114232063293,
+      "loss": 7.3756,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": -0.41535210609436035,
+      "rewards/margins": 0.08780960738658905,
+      "rewards/rejected": -0.5031617283821106,
+      "step": 20
+    },
+    {
+      "epoch": 0.021745039412883936,
+      "grad_norm": 0.09685570746660233,
+      "learning_rate": 4.9950668210706795e-06,
+      "logits/chosen": -2.10174298286438,
+      "logits/rejected": -2.378197431564331,
+      "logps/chosen": -0.26717427372932434,
+      "logps/rejected": -0.30565372109413147,
+      "loss": 7.451,
+      "rewards/accuracies": 0.4375,
+      "rewards/chosen": -0.4007614254951477,
+      "rewards/margins": 0.05771917849779129,
+      "rewards/rejected": -0.4584805369377136,
+      "step": 30
+    },
+    {
+      "epoch": 0.028993385883845247,
+      "grad_norm": 0.08213861286640167,
+      "learning_rate": 4.9912321481237616e-06,
+      "logits/chosen": -2.1633317470550537,
+      "logits/rejected": -2.387866497039795,
+      "logps/chosen": -0.27634260058403015,
+      "logps/rejected": -0.37035584449768066,
+      "loss": 7.3892,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.4145139157772064,
+      "rewards/margins": 0.14101983606815338,
+      "rewards/rejected": -0.555533766746521,
+      "step": 40
+    },
+    {
+      "epoch": 0.03624173235480656,
+      "grad_norm": 0.08846044540405273,
+      "learning_rate": 4.986304738420684e-06,
+      "logits/chosen": -2.1402599811553955,
+      "logits/rejected": -2.4459526538848877,
+      "logps/chosen": -0.2535383999347687,
+      "logps/rejected": -0.3090876042842865,
+      "loss": 7.5171,
+      "rewards/accuracies": 0.512499988079071,
+      "rewards/chosen": -0.3803076148033142,
+      "rewards/margins": 0.08332376182079315,
+      "rewards/rejected": -0.4636313319206238,
+      "step": 50
+    },
+    {
+      "epoch": 0.03624173235480656,
+      "eval_logits/chosen": -2.1165692806243896,
+      "eval_logits/rejected": -2.476428747177124,
+      "eval_logps/chosen": -0.2828062176704407,
+      "eval_logps/rejected": -0.3432886600494385,
+      "eval_loss": 0.9120001792907715,
+      "eval_rewards/accuracies": 0.5089285969734192,
+      "eval_rewards/chosen": -0.4242093861103058,
+      "eval_rewards/margins": 0.09072363376617432,
+      "eval_rewards/rejected": -0.5149329900741577,
+      "eval_runtime": 30.971,
+      "eval_samples_per_second": 28.801,
+      "eval_steps_per_second": 3.616,
+      "step": 50
+    },
+    {
+      "epoch": 0.04349007882576787,
+      "grad_norm": 0.11753705143928528,
+      "learning_rate": 4.980286753286196e-06,
+      "logits/chosen": -2.0575368404388428,
+      "logits/rejected": -2.5456700325012207,
+      "logps/chosen": -0.24622221291065216,
+      "logps/rejected": -0.32402220368385315,
+      "loss": 7.3926,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.36933332681655884,
+      "rewards/margins": 0.11669999361038208,
+      "rewards/rejected": -0.48603329062461853,
+      "step": 60
+    },
+    {
+      "epoch": 0.05073842529672919,
+      "grad_norm": 0.09996571391820908,
+      "learning_rate": 4.973180832407471e-06,
+      "logits/chosen": -1.9278606176376343,
+      "logits/rejected": -2.4620182514190674,
+      "logps/chosen": -0.2596542239189148,
+      "logps/rejected": -0.3665880560874939,
+      "loss": 7.185,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.3894812762737274,
+      "rewards/margins": 0.160400852560997,
+      "rewards/rejected": -0.5498821139335632,
+      "step": 70
+    },
+    {
+      "epoch": 0.057986771767690494,
+      "grad_norm": 0.07287321239709854,
+      "learning_rate": 4.964990092676263e-06,
+      "logits/chosen": -2.078031063079834,
+      "logits/rejected": -2.461479663848877,
+      "logps/chosen": -0.24513795971870422,
+      "logps/rejected": -0.3448730707168579,
+      "loss": 7.365,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.36770695447921753,
+      "rewards/margins": 0.14960262179374695,
+      "rewards/rejected": -0.5173095464706421,
+      "step": 80
+    },
+    {
+      "epoch": 0.0652351182386518,
+      "grad_norm": 0.09656044095754623,
+      "learning_rate": 4.9557181268217225e-06,
+      "logits/chosen": -1.9979238510131836,
+      "logits/rejected": -2.4381277561187744,
+      "logps/chosen": -0.24047240614891052,
+      "logps/rejected": -0.3277527987957001,
+      "loss": 7.2664,
+      "rewards/accuracies": 0.6000000238418579,
+      "rewards/chosen": -0.36070865392684937,
+      "rewards/margins": 0.13092057406902313,
+      "rewards/rejected": -0.4916292130947113,
+      "step": 90
+    },
+    {
+      "epoch": 0.07248346470961312,
+      "grad_norm": 0.08125138282775879,
+      "learning_rate": 4.9453690018345144e-06,
+      "logits/chosen": -1.8948112726211548,
+      "logits/rejected": -2.4755520820617676,
+      "logps/chosen": -0.20189261436462402,
+      "logps/rejected": -0.29732149839401245,
+      "loss": 7.2542,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.30283889174461365,
+      "rewards/margins": 0.14314329624176025,
+      "rewards/rejected": -0.44598227739334106,
+      "step": 100
+    },
+    {
+      "epoch": 0.07248346470961312,
+      "eval_logits/chosen": -2.12322735786438,
+      "eval_logits/rejected": -2.481174945831299,
+      "eval_logps/chosen": -0.2438412606716156,
+      "eval_logps/rejected": -0.3260033428668976,
+      "eval_loss": 0.891861081123352,
+      "eval_rewards/accuracies": 0.5625,
+      "eval_rewards/chosen": -0.3657619059085846,
+      "eval_rewards/margins": 0.12324309349060059,
+      "eval_rewards/rejected": -0.48900502920150757,
+      "eval_runtime": 30.3299,
+      "eval_samples_per_second": 29.41,
+      "eval_steps_per_second": 3.693,
+      "step": 100
+    },
+    {
+      "epoch": 0.07973181118057443,
+      "grad_norm": 0.05962231010198593,
+      "learning_rate": 4.933947257182901e-06,
+      "logits/chosen": -2.1134355068206787,
+      "logits/rejected": -2.516538381576538,
+      "logps/chosen": -0.237023264169693,
+      "logps/rejected": -0.31476154923439026,
+      "loss": 7.1749,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": -0.3555349111557007,
+      "rewards/margins": 0.11660744249820709,
+      "rewards/rejected": -0.4721423089504242,
+      "step": 110
+    },
+    {
+      "epoch": 0.08698015765153574,
+      "grad_norm": 0.06015922501683235,
+      "learning_rate": 4.921457902821578e-06,
+      "logits/chosen": -2.0215041637420654,
+      "logits/rejected": -2.4902031421661377,
+      "logps/chosen": -0.1890055537223816,
+      "logps/rejected": -0.3192065358161926,
+      "loss": 7.142,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.2835083603858948,
+      "rewards/margins": 0.19530144333839417,
+      "rewards/rejected": -0.47880974411964417,
+      "step": 120
+    },
+    {
+      "epoch": 0.09422850412249706,
+      "grad_norm": 0.06430571526288986,
+      "learning_rate": 4.907906416994146e-06,
+      "logits/chosen": -2.0684752464294434,
+      "logits/rejected": -2.510018825531006,
+      "logps/chosen": -0.2073744535446167,
+      "logps/rejected": -0.3121300935745239,
+      "loss": 7.1438,
+      "rewards/accuracies": 0.612500011920929,
+      "rewards/chosen": -0.31106168031692505,
+      "rewards/margins": 0.15713343024253845,
+      "rewards/rejected": -0.4681951403617859,
+      "step": 130
+    },
+    {
+      "epoch": 0.10147685059345837,
+      "grad_norm": 0.08829955011606216,
+      "learning_rate": 4.893298743830168e-06,
+      "logits/chosen": -2.026458263397217,
+      "logits/rejected": -2.496157646179199,
+      "logps/chosen": -0.19946983456611633,
+      "logps/rejected": -0.32050156593322754,
+      "loss": 7.1196,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": -0.2992047667503357,
+      "rewards/margins": 0.18154758214950562,
+      "rewards/rejected": -0.4807523787021637,
+      "step": 140
+    },
+    {
+      "epoch": 0.10872519706441967,
+      "grad_norm": 0.09773921221494675,
+      "learning_rate": 4.8776412907378845e-06,
+      "logits/chosen": -2.101029872894287,
+      "logits/rejected": -2.5849032402038574,
+      "logps/chosen": -0.18889756500720978,
+      "logps/rejected": -0.36427801847457886,
+      "loss": 7.1227,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.28334635496139526,
+      "rewards/margins": 0.26307064294815063,
+      "rewards/rejected": -0.5464169979095459,
+      "step": 150
+    },
+    {
+      "epoch": 0.10872519706441967,
+      "eval_logits/chosen": -2.1945455074310303,
+      "eval_logits/rejected": -2.559415578842163,
+      "eval_logps/chosen": -0.22209034860134125,
+      "eval_logps/rejected": -0.32476040720939636,
+      "eval_loss": 0.8760393261909485,
+      "eval_rewards/accuracies": 0.5803571343421936,
+      "eval_rewards/chosen": -0.33313554525375366,
+      "eval_rewards/margins": 0.15400508046150208,
+      "eval_rewards/rejected": -0.48714062571525574,
+      "eval_runtime": 30.3484,
+      "eval_samples_per_second": 29.392,
+      "eval_steps_per_second": 3.69,
+      "step": 150
+    },
+    {
+      "epoch": 0.11597354353538099,
+      "grad_norm": 0.09142427891492844,
+      "learning_rate": 4.860940925593703e-06,
+      "logits/chosen": -2.235069751739502,
+      "logits/rejected": -2.5513346195220947,
+      "logps/chosen": -0.19598741829395294,
+      "logps/rejected": -0.2772120535373688,
+      "loss": 7.0272,
+      "rewards/accuracies": 0.5375000238418579,
+      "rewards/chosen": -0.2939811050891876,
+      "rewards/margins": 0.12183700501918793,
+      "rewards/rejected": -0.41581812500953674,
+      "step": 160
+    },
+    {
+      "epoch": 0.1232218900063423,
+      "grad_norm": 0.11735275387763977,
+      "learning_rate": 4.84320497372973e-06,
+      "logits/chosen": -2.154263496398926,
+      "logits/rejected": -2.512010335922241,
+      "logps/chosen": -0.1798369437456131,
+      "logps/rejected": -0.30910637974739075,
+      "loss": 7.1603,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.26975542306900024,
+      "rewards/margins": 0.19390416145324707,
+      "rewards/rejected": -0.4636595845222473,
+      "step": 170
+    },
+    {
+      "epoch": 0.1304702364773036,
+      "grad_norm": 0.09398588538169861,
+      "learning_rate": 4.824441214720629e-06,
+      "logits/chosen": -2.292147159576416,
+      "logits/rejected": -2.581425189971924,
+      "logps/chosen": -0.18498703837394714,
+      "logps/rejected": -0.31237050890922546,
+      "loss": 6.995,
+      "rewards/accuracies": 0.637499988079071,
+      "rewards/chosen": -0.2774805426597595,
+      "rewards/margins": 0.19107523560523987,
+      "rewards/rejected": -0.4685557782649994,
+      "step": 180
+    },
+    {
+      "epoch": 0.13771858294826492,
+      "grad_norm": 0.13551996648311615,
+      "learning_rate": 4.804657878971252e-06,
+      "logits/chosen": -2.274120330810547,
+      "logits/rejected": -2.5797386169433594,
+      "logps/chosen": -0.21168990433216095,
+      "logps/rejected": -0.34927254915237427,
+      "loss": 6.9941,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.317534863948822,
+      "rewards/margins": 0.20637397468090057,
+      "rewards/rejected": -0.5239088535308838,
+      "step": 190
+    },
+    {
+      "epoch": 0.14496692941922623,
+      "grad_norm": 0.15515944361686707,
+      "learning_rate": 4.783863644106502e-06,
+      "logits/chosen": -2.2116379737854004,
+      "logits/rejected": -2.6693203449249268,
+      "logps/chosen": -0.1971816122531891,
+      "logps/rejected": -0.3498842120170593,
+      "loss": 7.0413,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.2957724332809448,
+      "rewards/margins": 0.22905388474464417,
+      "rewards/rejected": -0.5248263478279114,
+      "step": 200
+    },
+    {
+      "epoch": 0.14496692941922623,
+      "eval_logits/chosen": -2.3369696140289307,
+      "eval_logits/rejected": -2.7298672199249268,
+      "eval_logps/chosen": -0.22749511897563934,
+      "eval_logps/rejected": -0.3613782525062561,
+      "eval_loss": 0.8543878793716431,
+      "eval_rewards/accuracies": 0.5714285969734192,
+      "eval_rewards/chosen": -0.34124264121055603,
+      "eval_rewards/margins": 0.2008247673511505,
+      "eval_rewards/rejected": -0.5420674681663513,
+      "eval_runtime": 30.3397,
+      "eval_samples_per_second": 29.4,
+      "eval_steps_per_second": 3.692,
+      "step": 200
+    },
+    {
+      "epoch": 0.15221527589018755,
+      "grad_norm": 0.13039635121822357,
+      "learning_rate": 4.762067631165049e-06,
+      "logits/chosen": -2.2575619220733643,
+      "logits/rejected": -2.755174160003662,
+      "logps/chosen": -0.199218288064003,
+      "logps/rejected": -0.33260637521743774,
+      "loss": 6.882,
+      "rewards/accuracies": 0.5874999761581421,
+      "rewards/chosen": -0.2988274097442627,
+      "rewards/margins": 0.20008206367492676,
+      "rewards/rejected": -0.49890947341918945,
+      "step": 210
+    },
+    {
+      "epoch": 0.15946362236114886,
+      "grad_norm": 0.15020275115966797,
+      "learning_rate": 4.7392794005985324e-06,
+      "logits/chosen": -2.2949752807617188,
+      "logits/rejected": -2.7288482189178467,
+      "logps/chosen": -0.21987763047218323,
+      "logps/rejected": -0.385539174079895,
+      "loss": 6.9158,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.3298164904117584,
+      "rewards/margins": 0.24849233031272888,
+      "rewards/rejected": -0.5783087611198425,
+      "step": 220
+    },
+    {
+      "epoch": 0.16671196883211017,
+      "grad_norm": 0.16104522347450256,
+      "learning_rate": 4.715508948078037e-06,
+      "logits/chosen": -2.1837964057922363,
+      "logits/rejected": -2.813563585281372,
+      "logps/chosen": -0.2069036215543747,
+      "logps/rejected": -0.3996518552303314,
+      "loss": 6.8503,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.31035539507865906,
+      "rewards/margins": 0.28912240266799927,
+      "rewards/rejected": -0.5994777679443359,
+      "step": 230
+    },
+    {
+      "epoch": 0.1739603153030715,
+      "grad_norm": 0.16533692181110382,
+      "learning_rate": 4.690766700109659e-06,
+      "logits/chosen": -2.314570188522339,
+      "logits/rejected": -2.8867621421813965,
+      "logps/chosen": -0.18964803218841553,
+      "logps/rejected": -0.3997463583946228,
+      "loss": 6.667,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.2844720482826233,
+      "rewards/margins": 0.3151474893093109,
+      "rewards/rejected": -0.5996195077896118,
+      "step": 240
+    },
+    {
+      "epoch": 0.1812086617740328,
+      "grad_norm": 0.24991220235824585,
+      "learning_rate": 4.665063509461098e-06,
+      "logits/chosen": -2.46189546585083,
+      "logits/rejected": -2.847446918487549,
+      "logps/chosen": -0.24189543724060059,
+      "logps/rejected": -0.4183991551399231,
+      "loss": 6.7146,
+      "rewards/accuracies": 0.5625,
+      "rewards/chosen": -0.3628431558609009,
+      "rewards/margins": 0.26475557684898376,
+      "rewards/rejected": -0.6275987029075623,
+      "step": 250
+    },
+    {
+      "epoch": 0.1812086617740328,
+      "eval_logits/chosen": -2.5230772495269775,
+      "eval_logits/rejected": -2.9182300567626953,
+      "eval_logps/chosen": -0.2819940149784088,
+      "eval_logps/rejected": -0.48641347885131836,
+      "eval_loss": 0.8070082664489746,
+      "eval_rewards/accuracies": 0.5803571343421936,
+      "eval_rewards/chosen": -0.4229910671710968,
+      "eval_rewards/margins": 0.3066291809082031,
+      "eval_rewards/rejected": -0.7296201586723328,
+      "eval_runtime": 30.3523,
+      "eval_samples_per_second": 29.388,
+      "eval_steps_per_second": 3.69,
+      "step": 250
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 1500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "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": 9.284468685374751e+17,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-250/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91b9d0e6e8c92ae4a2c435b81fedf969ba2fc02907285b004770a8256166d71d
+size 7224

checkpoint-250/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)