Upload folder using huggingface_hub

mistral7b_dpo_baseline/README.md

@@ -0,0 +1,64 @@
+---
+library_name: peft
+license: other
+base_model: mistralai/Mistral-7B-v0.1
+tags:
+- base_model:adapter:mistralai/Mistral-7B-v0.1
+- llama-factory
+- lora
+- transformers
+pipeline_tag: text-generation
+model-index:
+- name: mistral7b_dpo_baseline
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# mistral7b_dpo_baseline
+
+This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the shp_dpo_train dataset.
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 1e-06
+- train_batch_size: 8
+- eval_batch_size: 8
+- seed: 42
+- distributed_type: multi-GPU
+- num_devices: 8
+- gradient_accumulation_steps: 8
+- total_train_batch_size: 512
+- total_eval_batch_size: 64
+- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_steps: 100
+- num_epochs: 1.0
+
+### Training results
+
+
+
+### Framework versions
+
+- PEFT 0.17.1
+- Transformers 4.57.1
+- Pytorch 2.10.0+cu128
+- Datasets 4.0.0
+- Tokenizers 0.22.2

mistral7b_dpo_baseline/adapter_config.json

@@ -0,0 +1,42 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "mistralai/Mistral-7B-v0.1",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "down_proj",
+    "v_proj",
+    "k_proj",
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "gate_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

mistral7b_dpo_baseline/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb08bf36b7c4c22c2a271042abc4374e96b79bb94961046f1199a0fd2b826a92
+size 42002584

mistral7b_dpo_baseline/all_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "total_flos": 7.072674331683193e+18,
+    "train_loss": 0.6730708850889789,
+    "train_runtime": 4328.1955,
+    "train_samples_per_second": 23.104,
+    "train_steps_per_second": 0.045
+}

mistral7b_dpo_baseline/chat_template.jinja

@@ -0,0 +1,3 @@
+{{ '<s>' }}{% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% else %}{% set loop_messages = messages %}{% endif %}{% for message in loop_messages %}{% if loop.index0 == 0 and system_message is defined %}{% set content = system_message + '
+
+' + message['content'] %}{% else %}{% set content = message['content'] %}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + content + '[/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + content + '</s>' }}{% endif %}{% endfor %}

mistral7b_dpo_baseline/checkpoint-196/README.md

@@ -0,0 +1,208 @@
+---
+base_model: mistralai/Mistral-7B-v0.1
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:mistralai/Mistral-7B-v0.1
+- llama-factory
+- lora
+- transformers
+---
+
+# 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 0.17.1

mistral7b_dpo_baseline/checkpoint-196/adapter_config.json

@@ -0,0 +1,42 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "mistralai/Mistral-7B-v0.1",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "down_proj",
+    "v_proj",
+    "k_proj",
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "gate_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

mistral7b_dpo_baseline/checkpoint-196/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb08bf36b7c4c22c2a271042abc4374e96b79bb94961046f1199a0fd2b826a92
+size 42002584

mistral7b_dpo_baseline/checkpoint-196/chat_template.jinja

@@ -0,0 +1,3 @@
+{{ '<s>' }}{% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% else %}{% set loop_messages = messages %}{% endif %}{% for message in loop_messages %}{% if loop.index0 == 0 and system_message is defined %}{% set content = system_message + '
+
+' + message['content'] %}{% else %}{% set content = message['content'] %}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + content + '[/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + content + '</s>' }}{% endif %}{% endfor %}

mistral7b_dpo_baseline/checkpoint-196/global_step196/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:b3412bfdde07e8ccf6ee2c7a6e1013a77897179bd8d87fd1c0151d672fbe6f51
+size 31467333

mistral7b_dpo_baseline/checkpoint-196/global_step196/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:ef0b9c0fb09709cbb0aa2a2bce03029535ba57f2cbad89db5c10bcbe612b87e5
+size 31467333

mistral7b_dpo_baseline/checkpoint-196/global_step196/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95d9fb085f3cd871d003937efad16763bdce2391e4fd3f73bd041d3edd03cbef
+size 31467333

mistral7b_dpo_baseline/checkpoint-196/global_step196/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fc05b69463faeeff9a036941984b5294c80fcdf446c0d7a15e278131f3d764e
+size 31467397

mistral7b_dpo_baseline/checkpoint-196/global_step196/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:adffb6efc08ea80b66ec41b88bb00c41bc3c709ef1b413c847a0f24976d12e91
+size 31467397

mistral7b_dpo_baseline/checkpoint-196/global_step196/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48f16a737fc4da5c39205c1ac838aaf36ce81483ba38bddbbdc55d7ee028aef3
+size 31467397

mistral7b_dpo_baseline/checkpoint-196/global_step196/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90ec802fe879331461fd52e833a64dd8f8b4be0931e5ae4ce081cc8bcfbe0dc0
+size 31467397

mistral7b_dpo_baseline/checkpoint-196/global_step196/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53a28de969dcc56420659a48fc2236e12a3a6b1097741c78a3d5ca6d598a7328
+size 31467397

mistral7b_dpo_baseline/checkpoint-196/global_step196/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e127229070dfd66bdda8267978c980385dd0d845e366ced2838eb74062130772
+size 42288409

mistral7b_dpo_baseline/checkpoint-196/latest

@@ -0,0 +1 @@
+global_step196

mistral7b_dpo_baseline/checkpoint-196/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21d297146ab677fd643661f63c3b273b50343b968f3faa199a7fa80366501310
+size 16389

mistral7b_dpo_baseline/checkpoint-196/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ed40a0a4e9f365d2c6cc004d97e6705894eba46c8be4c160c1455bc3062dee1
+size 16389

mistral7b_dpo_baseline/checkpoint-196/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d688b304d19c260b5cfa471535ed51d7e1d60b3a0d0159dfd1a04b87904a9f42
+size 16389

mistral7b_dpo_baseline/checkpoint-196/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9967425ebcaee80d9b518fa0244d52f739b1b983d87cda71d5fede0c073e9d3b
+size 16389

mistral7b_dpo_baseline/checkpoint-196/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:469900fd39c667ffbd49c3c407c0ba317a1e9f5f9339a99b5d38423b7d0ce6d4
+size 16389

mistral7b_dpo_baseline/checkpoint-196/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:124688471ff2a6e80f2fcefedbf741fb18d08dd539d5bd07a52e81be545142a5
+size 16389

mistral7b_dpo_baseline/checkpoint-196/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e69f1ced9f992a72c948698e5eb06088610788988cdb2fdbdd624e064319d60
+size 16389

mistral7b_dpo_baseline/checkpoint-196/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a376268a55d6ee10c371c06aa952334c4c6a1af9ea2d71b1951a57367a0c6722
+size 16389

mistral7b_dpo_baseline/checkpoint-196/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83820952e8eb59075696a2d6e1f7ef27d70c5576cb2781806606f38f86685d3e
+size 1401

mistral7b_dpo_baseline/checkpoint-196/special_tokens_map.json

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

mistral7b_dpo_baseline/checkpoint-196/tokenizer.model

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

mistral7b_dpo_baseline/checkpoint-196/tokenizer_config.json

@@ -0,0 +1,46 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": null,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [],
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "extra_special_tokens": {},
+  "legacy": false,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "</s>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "split_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

mistral7b_dpo_baseline/checkpoint-196/trainer_state.json

@@ -0,0 +1,809 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 196,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.025591810620601407,
+      "grad_norm": 1.705037236213684,
+      "learning_rate": 4e-08,
+      "logits/chosen": -2.273033380508423,
+      "logits/rejected": -2.1234283447265625,
+      "logps/chosen": -375.063232421875,
+      "logps/rejected": -210.8657684326172,
+      "loss": 0.6931,
+      "rewards/accuracies": 0.3062500059604645,
+      "rewards/chosen": 0.0012258010683581233,
+      "rewards/margins": 0.0003632208681665361,
+      "rewards/rejected": 0.0008625800255686045,
+      "step": 5
+    },
+    {
+      "epoch": 0.05118362124120281,
+      "grad_norm": 1.766357421875,
+      "learning_rate": 9e-08,
+      "logits/chosen": -2.2679283618927,
+      "logits/rejected": -2.116572141647339,
+      "logps/chosen": -382.0296325683594,
+      "logps/rejected": -202.4493408203125,
+      "loss": 0.6931,
+      "rewards/accuracies": 0.5082031488418579,
+      "rewards/chosen": 0.0013576988130807877,
+      "rewards/margins": 0.0005467134760692716,
+      "rewards/rejected": 0.0008109854534268379,
+      "stability/repetition_rate_mean": 0.9135147929191589,
+      "stability/response_length_mean": 602.7999877929688,
+      "stability/response_length_std": 295.1842346191406,
+      "stability/response_length_var": 87133.734375,
+      "stability/token_entropy_mean": 6.959374904632568,
+      "step": 10
+    },
+    {
+      "epoch": 0.07677543186180422,
+      "grad_norm": 1.8122901916503906,
+      "learning_rate": 1.4e-07,
+      "logits/chosen": -2.260676860809326,
+      "logits/rejected": -2.0991530418395996,
+      "logps/chosen": -389.2082214355469,
+      "logps/rejected": -207.0491180419922,
+      "loss": 0.6932,
+      "rewards/accuracies": 0.5082031488418579,
+      "rewards/chosen": 0.0017401383956894279,
+      "rewards/margins": 0.00028905909857712686,
+      "rewards/rejected": 0.0014510790351778269,
+      "stability/repetition_rate_mean": 0.8363126516342163,
+      "stability/response_length_mean": 345.70001220703125,
+      "stability/response_length_std": 212.67974853515625,
+      "stability/response_length_var": 45232.6796875,
+      "stability/token_entropy_mean": 6.342187404632568,
+      "step": 15
+    },
+    {
+      "epoch": 0.10236724248240563,
+      "grad_norm": 1.7545065879821777,
+      "learning_rate": 1.8999999999999998e-07,
+      "logits/chosen": -2.2613608837127686,
+      "logits/rejected": -2.1145310401916504,
+      "logps/chosen": -368.71954345703125,
+      "logps/rejected": -198.5596160888672,
+      "loss": 0.6939,
+      "rewards/accuracies": 0.49140626192092896,
+      "rewards/chosen": 0.0004601169202942401,
+      "rewards/margins": -0.0010562256211414933,
+      "rewards/rejected": 0.0015163427451625466,
+      "stability/repetition_rate_mean": 0.9317001104354858,
+      "stability/response_length_mean": 739.4000244140625,
+      "stability/response_length_std": 281.282470703125,
+      "stability/response_length_var": 79119.8203125,
+      "stability/token_entropy_mean": 6.931250095367432,
+      "step": 20
+    },
+    {
+      "epoch": 0.12795905310300704,
+      "grad_norm": 1.731654167175293,
+      "learning_rate": 2.4e-07,
+      "logits/chosen": -2.277377128601074,
+      "logits/rejected": -2.125220775604248,
+      "logps/chosen": -378.1647033691406,
+      "logps/rejected": -205.86363220214844,
+      "loss": 0.693,
+      "rewards/accuracies": 0.5082031488418579,
+      "rewards/chosen": 0.002021136460825801,
+      "rewards/margins": 0.0006010310025885701,
+      "rewards/rejected": 0.0014201054582372308,
+      "stability/repetition_rate_mean": 0.8950179815292358,
+      "stability/response_length_mean": 658.5999755859375,
+      "stability/response_length_std": 288.0155029296875,
+      "stability/response_length_var": 82952.9296875,
+      "stability/token_entropy_mean": 6.821875095367432,
+      "step": 25
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 1.8716338872909546,
+      "learning_rate": 2.9e-07,
+      "logits/chosen": -2.2701575756073,
+      "logits/rejected": -2.1229705810546875,
+      "logps/chosen": -373.9202575683594,
+      "logps/rejected": -208.93287658691406,
+      "loss": 0.6935,
+      "rewards/accuracies": 0.490234375,
+      "rewards/chosen": 0.0010083412053063512,
+      "rewards/margins": -0.0003178715705871582,
+      "rewards/rejected": 0.0013262125430628657,
+      "stability/repetition_rate_mean": 0.8135284185409546,
+      "stability/response_length_mean": 356.0,
+      "stability/response_length_std": 249.6766815185547,
+      "stability/response_length_var": 62338.4453125,
+      "stability/token_entropy_mean": 6.199999809265137,
+      "step": 30
+    },
+    {
+      "epoch": 0.17914267434420986,
+      "grad_norm": 1.733734369277954,
+      "learning_rate": 3.4000000000000003e-07,
+      "logits/chosen": -2.2612318992614746,
+      "logits/rejected": -2.120094060897827,
+      "logps/chosen": -366.3136901855469,
+      "logps/rejected": -203.98719787597656,
+      "loss": 0.6928,
+      "rewards/accuracies": 0.5265624523162842,
+      "rewards/chosen": 0.0032140661496669054,
+      "rewards/margins": 0.0011429906589910388,
+      "rewards/rejected": 0.0020710753742605448,
+      "stability/repetition_rate_mean": 0.9289250373840332,
+      "stability/response_length_mean": 606.2000122070312,
+      "stability/response_length_std": 394.1366882324219,
+      "stability/response_length_var": 155343.734375,
+      "stability/token_entropy_mean": 7.112500190734863,
+      "step": 35
+    },
+    {
+      "epoch": 0.20473448496481125,
+      "grad_norm": 1.8034971952438354,
+      "learning_rate": 3.8999999999999997e-07,
+      "logits/chosen": -2.2764394283294678,
+      "logits/rejected": -2.122685432434082,
+      "logps/chosen": -369.1784973144531,
+      "logps/rejected": -197.37686157226562,
+      "loss": 0.6928,
+      "rewards/accuracies": 0.5144531726837158,
+      "rewards/chosen": 0.004107568878680468,
+      "rewards/margins": 0.0011615146650001407,
+      "rewards/rejected": 0.0029460543300956488,
+      "stability/repetition_rate_mean": 0.8629743456840515,
+      "stability/response_length_mean": 547.2000122070312,
+      "stability/response_length_std": 290.8897399902344,
+      "stability/response_length_var": 84616.84375,
+      "stability/token_entropy_mean": 6.557812690734863,
+      "step": 40
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 1.8129643201828003,
+      "learning_rate": 4.3999999999999997e-07,
+      "logits/chosen": -2.257107734680176,
+      "logits/rejected": -2.1072099208831787,
+      "logps/chosen": -385.1796569824219,
+      "logps/rejected": -212.56887817382812,
+      "loss": 0.6933,
+      "rewards/accuracies": 0.512890636920929,
+      "rewards/chosen": 0.0039331079460680485,
+      "rewards/margins": 3.291945904493332e-05,
+      "rewards/rejected": 0.003900188487023115,
+      "stability/repetition_rate_mean": 0.8764154314994812,
+      "stability/response_length_mean": 366.5,
+      "stability/response_length_std": 118.86243438720703,
+      "stability/response_length_var": 14128.27734375,
+      "stability/token_entropy_mean": 6.640625,
+      "step": 45
+    },
+    {
+      "epoch": 0.2559181062060141,
+      "grad_norm": 1.8471362590789795,
+      "learning_rate": 4.9e-07,
+      "logits/chosen": -2.264720916748047,
+      "logits/rejected": -2.109816312789917,
+      "logps/chosen": -378.2891540527344,
+      "logps/rejected": -207.4978790283203,
+      "loss": 0.6916,
+      "rewards/accuracies": 0.5316406488418579,
+      "rewards/chosen": 0.0068646748550236225,
+      "rewards/margins": 0.0035711233504116535,
+      "rewards/rejected": 0.003293551504611969,
+      "stability/repetition_rate_mean": 0.8785834312438965,
+      "stability/response_length_mean": 384.1000061035156,
+      "stability/response_length_std": 178.4809112548828,
+      "stability/response_length_var": 31855.43359375,
+      "stability/token_entropy_mean": 6.543749809265137,
+      "step": 50
+    },
+    {
+      "epoch": 0.28150991682661547,
+      "grad_norm": 1.940763235092163,
+      "learning_rate": 5.4e-07,
+      "logits/chosen": -2.2527079582214355,
+      "logits/rejected": -2.094048261642456,
+      "logps/chosen": -393.7845764160156,
+      "logps/rejected": -210.35861206054688,
+      "loss": 0.6911,
+      "rewards/accuracies": 0.536328136920929,
+      "rewards/chosen": 0.009546511806547642,
+      "rewards/margins": 0.004484717734158039,
+      "rewards/rejected": 0.005061793606728315,
+      "stability/repetition_rate_mean": 0.9481571912765503,
+      "stability/response_length_mean": 332.8999938964844,
+      "stability/response_length_std": 137.39193725585938,
+      "stability/response_length_var": 18876.544921875,
+      "stability/token_entropy_mean": 7.009375095367432,
+      "step": 55
+    },
+    {
+      "epoch": 0.30710172744721687,
+      "grad_norm": 1.8226360082626343,
+      "learning_rate": 5.9e-07,
+      "logits/chosen": -2.272801160812378,
+      "logits/rejected": -2.111156463623047,
+      "logps/chosen": -387.7115173339844,
+      "logps/rejected": -206.819580078125,
+      "loss": 0.6915,
+      "rewards/accuracies": 0.5328124761581421,
+      "rewards/chosen": 0.009509808383882046,
+      "rewards/margins": 0.00375912943854928,
+      "rewards/rejected": 0.005750678479671478,
+      "stability/repetition_rate_mean": 0.8699707984924316,
+      "stability/response_length_mean": 655.2999877929688,
+      "stability/response_length_std": 282.005126953125,
+      "stability/response_length_var": 79526.8984375,
+      "stability/token_entropy_mean": 6.559374809265137,
+      "step": 60
+    },
+    {
+      "epoch": 0.3326935380678183,
+      "grad_norm": 1.6994986534118652,
+      "learning_rate": 6.4e-07,
+      "logits/chosen": -2.2521181106567383,
+      "logits/rejected": -2.098989963531494,
+      "logps/chosen": -377.4696960449219,
+      "logps/rejected": -204.3833465576172,
+      "loss": 0.6908,
+      "rewards/accuracies": 0.5625000596046448,
+      "rewards/chosen": 0.012991372495889664,
+      "rewards/margins": 0.00506935128942132,
+      "rewards/rejected": 0.007922019809484482,
+      "stability/repetition_rate_mean": 0.9148252606391907,
+      "stability/response_length_mean": 306.3999938964844,
+      "stability/response_length_std": 141.7236785888672,
+      "stability/response_length_var": 20085.599609375,
+      "stability/token_entropy_mean": 6.840624809265137,
+      "step": 65
+    },
+    {
+      "epoch": 0.3582853486884197,
+      "grad_norm": 2.000992774963379,
+      "learning_rate": 6.9e-07,
+      "logits/chosen": -2.2713451385498047,
+      "logits/rejected": -2.106651782989502,
+      "logps/chosen": -393.73406982421875,
+      "logps/rejected": -206.1870574951172,
+      "loss": 0.6888,
+      "rewards/accuracies": 0.5777344107627869,
+      "rewards/chosen": 0.017473159357905388,
+      "rewards/margins": 0.009205814450979233,
+      "rewards/rejected": 0.008267344906926155,
+      "stability/repetition_rate_mean": 0.9625760316848755,
+      "stability/response_length_mean": 472.70001220703125,
+      "stability/response_length_std": 159.89791870117188,
+      "stability/response_length_var": 25567.34375,
+      "stability/token_entropy_mean": 7.162499904632568,
+      "step": 70
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 1.8905284404754639,
+      "learning_rate": 7.4e-07,
+      "logits/chosen": -2.275454521179199,
+      "logits/rejected": -2.116396903991699,
+      "logps/chosen": -381.0571594238281,
+      "logps/rejected": -207.9349365234375,
+      "loss": 0.689,
+      "rewards/accuracies": 0.577343761920929,
+      "rewards/chosen": 0.018993159756064415,
+      "rewards/margins": 0.008799022994935513,
+      "rewards/rejected": 0.010194134898483753,
+      "stability/repetition_rate_mean": 0.9068478345870972,
+      "stability/response_length_mean": 527.0,
+      "stability/response_length_std": 293.7500915527344,
+      "stability/response_length_var": 86289.109375,
+      "stability/token_entropy_mean": 6.139062404632568,
+      "step": 75
+    },
+    {
+      "epoch": 0.4094689699296225,
+      "grad_norm": 1.8269047737121582,
+      "learning_rate": 7.9e-07,
+      "logits/chosen": -2.2604687213897705,
+      "logits/rejected": -2.1025872230529785,
+      "logps/chosen": -382.4193420410156,
+      "logps/rejected": -208.97756958007812,
+      "loss": 0.6886,
+      "rewards/accuracies": 0.5726562738418579,
+      "rewards/chosen": 0.021377386525273323,
+      "rewards/margins": 0.009593295864760876,
+      "rewards/rejected": 0.011784089729189873,
+      "stability/repetition_rate_mean": 0.9464001655578613,
+      "stability/response_length_mean": 635.5,
+      "stability/response_length_std": 262.53900146484375,
+      "stability/response_length_var": 68926.71875,
+      "stability/token_entropy_mean": 7.199999809265137,
+      "step": 80
+    },
+    {
+      "epoch": 0.4350607805502239,
+      "grad_norm": 1.8734734058380127,
+      "learning_rate": 8.399999999999999e-07,
+      "logits/chosen": -2.258051872253418,
+      "logits/rejected": -2.100944757461548,
+      "logps/chosen": -387.3235778808594,
+      "logps/rejected": -208.6320343017578,
+      "loss": 0.6875,
+      "rewards/accuracies": 0.5863280892372131,
+      "rewards/chosen": 0.026788339018821716,
+      "rewards/margins": 0.012103421613574028,
+      "rewards/rejected": 0.014684918336570263,
+      "stability/repetition_rate_mean": 0.9034587144851685,
+      "stability/response_length_mean": 600.0999755859375,
+      "stability/response_length_std": 349.2030334472656,
+      "stability/response_length_var": 121942.765625,
+      "stability/token_entropy_mean": 6.737500190734863,
+      "step": 85
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 1.8927897214889526,
+      "learning_rate": 8.9e-07,
+      "logits/chosen": -2.261427164077759,
+      "logits/rejected": -2.1134793758392334,
+      "logps/chosen": -375.204345703125,
+      "logps/rejected": -206.2209930419922,
+      "loss": 0.6859,
+      "rewards/accuracies": 0.608593761920929,
+      "rewards/chosen": 0.032174669206142426,
+      "rewards/margins": 0.015198366716504097,
+      "rewards/rejected": 0.01697630062699318,
+      "stability/repetition_rate_mean": 0.9765706062316895,
+      "stability/response_length_mean": 692.9000244140625,
+      "stability/response_length_std": 127.31452941894531,
+      "stability/response_length_var": 16208.9892578125,
+      "stability/token_entropy_mean": 7.109375,
+      "step": 90
+    },
+    {
+      "epoch": 0.48624440179142675,
+      "grad_norm": 1.8181637525558472,
+      "learning_rate": 9.399999999999999e-07,
+      "logits/chosen": -2.2588284015655518,
+      "logits/rejected": -2.112243175506592,
+      "logps/chosen": -366.8333740234375,
+      "logps/rejected": -203.4322967529297,
+      "loss": 0.6838,
+      "rewards/accuracies": 0.6300780773162842,
+      "rewards/chosen": 0.03762410953640938,
+      "rewards/margins": 0.01969875767827034,
+      "rewards/rejected": 0.01792534813284874,
+      "stability/repetition_rate_mean": 0.9483324289321899,
+      "stability/response_length_mean": 727.0,
+      "stability/response_length_std": 335.0764465332031,
+      "stability/response_length_var": 112276.21875,
+      "stability/token_entropy_mean": 7.3125,
+      "step": 95
+    },
+    {
+      "epoch": 0.5118362124120281,
+      "grad_norm": 1.8130452632904053,
+      "learning_rate": 9.9e-07,
+      "logits/chosen": -2.27445387840271,
+      "logits/rejected": -2.11898136138916,
+      "logps/chosen": -384.503662109375,
+      "logps/rejected": -210.62109375,
+      "loss": 0.6821,
+      "rewards/accuracies": 0.6527343988418579,
+      "rewards/chosen": 0.045418523252010345,
+      "rewards/margins": 0.02338458225131035,
+      "rewards/rejected": 0.0220339372754097,
+      "stability/repetition_rate_mean": 0.9498335123062134,
+      "stability/response_length_mean": 419.0,
+      "stability/response_length_std": 232.0225067138672,
+      "stability/response_length_var": 53834.4453125,
+      "stability/token_entropy_mean": 7.074999809265137,
+      "step": 100
+    },
+    {
+      "epoch": 0.5374280230326296,
+      "grad_norm": 1.743505597114563,
+      "learning_rate": 9.999815554351224e-07,
+      "logits/chosen": -2.258531093597412,
+      "logits/rejected": -2.108144998550415,
+      "logps/chosen": -378.0056457519531,
+      "logps/rejected": -209.6881103515625,
+      "loss": 0.6811,
+      "rewards/accuracies": 0.626953125,
+      "rewards/chosen": 0.052092913538217545,
+      "rewards/margins": 0.025670701637864113,
+      "rewards/rejected": 0.02642221562564373,
+      "stability/repetition_rate_mean": 0.901488184928894,
+      "stability/response_length_mean": 669.5,
+      "stability/response_length_std": 271.60888671875,
+      "stability/response_length_var": 73771.390625,
+      "stability/token_entropy_mean": 6.564843654632568,
+      "step": 105
+    },
+    {
+      "epoch": 0.5630198336532309,
+      "grad_norm": 1.6861857175827026,
+      "learning_rate": 9.999066267225444e-07,
+      "logits/chosen": -2.2750279903411865,
+      "logits/rejected": -2.117339611053467,
+      "logps/chosen": -385.2655029296875,
+      "logps/rejected": -206.72357177734375,
+      "loss": 0.6795,
+      "rewards/accuracies": 0.6519531607627869,
+      "rewards/chosen": 0.059595055878162384,
+      "rewards/margins": 0.029159152880311012,
+      "rewards/rejected": 0.03043590858578682,
+      "stability/repetition_rate_mean": 0.9347068071365356,
+      "stability/response_length_mean": 255.60000610351562,
+      "stability/response_length_std": 114.30874633789062,
+      "stability/response_length_var": 13066.4892578125,
+      "stability/token_entropy_mean": 7.015625,
+      "step": 110
+    },
+    {
+      "epoch": 0.5886116442738324,
+      "grad_norm": 1.7960835695266724,
+      "learning_rate": 9.997740697079592e-07,
+      "logits/chosen": -2.251587390899658,
+      "logits/rejected": -2.110828399658203,
+      "logps/chosen": -372.7311096191406,
+      "logps/rejected": -213.94320678710938,
+      "loss": 0.6763,
+      "rewards/accuracies": 0.6519531607627869,
+      "rewards/chosen": 0.07092021405696869,
+      "rewards/margins": 0.036303263157606125,
+      "rewards/rejected": 0.034616947174072266,
+      "stability/repetition_rate_mean": 0.8279488682746887,
+      "stability/response_length_mean": 499.29998779296875,
+      "stability/response_length_std": 282.31072998046875,
+      "stability/response_length_var": 79699.34375,
+      "stability/token_entropy_mean": 6.553124904632568,
+      "step": 115
+    },
+    {
+      "epoch": 0.6142034548944337,
+      "grad_norm": 1.773605227470398,
+      "learning_rate": 9.995838996722914e-07,
+      "logits/chosen": -2.2482666969299316,
+      "logits/rejected": -2.102483034133911,
+      "logps/chosen": -367.7881164550781,
+      "logps/rejected": -207.59573364257812,
+      "loss": 0.6733,
+      "rewards/accuracies": 0.6703125238418579,
+      "rewards/chosen": 0.08155383914709091,
+      "rewards/margins": 0.04288302734494209,
+      "rewards/rejected": 0.03867081552743912,
+      "stability/repetition_rate_mean": 0.9149526357650757,
+      "stability/response_length_mean": 363.6000061035156,
+      "stability/response_length_std": 146.0701446533203,
+      "stability/response_length_var": 21336.48828125,
+      "stability/token_entropy_mean": 6.8125,
+      "step": 120
+    },
+    {
+      "epoch": 0.6397952655150352,
+      "grad_norm": 1.7966293096542358,
+      "learning_rate": 9.993361385379874e-07,
+      "logits/chosen": -2.253580093383789,
+      "logits/rejected": -2.0954601764678955,
+      "logps/chosen": -385.7173767089844,
+      "logps/rejected": -205.1151123046875,
+      "loss": 0.6707,
+      "rewards/accuracies": 0.6781249642372131,
+      "rewards/chosen": 0.09179403632879257,
+      "rewards/margins": 0.04949168488383293,
+      "rewards/rejected": 0.04230235144495964,
+      "stability/repetition_rate_mean": 0.8946477770805359,
+      "stability/response_length_mean": 433.79998779296875,
+      "stability/response_length_std": 126.92412567138672,
+      "stability/response_length_var": 16109.7333984375,
+      "stability/token_entropy_mean": 6.532812595367432,
+      "step": 125
+    },
+    {
+      "epoch": 0.6653870761356366,
+      "grad_norm": 1.873657464981079,
+      "learning_rate": 9.99030814866488e-07,
+      "logits/chosen": -2.2453463077545166,
+      "logits/rejected": -2.096496105194092,
+      "logps/chosen": -375.60565185546875,
+      "logps/rejected": -203.76229858398438,
+      "loss": 0.6672,
+      "rewards/accuracies": 0.682421863079071,
+      "rewards/chosen": 0.09893052279949188,
+      "rewards/margins": 0.057060565799474716,
+      "rewards/rejected": 0.041869960725307465,
+      "stability/repetition_rate_mean": 0.8153165578842163,
+      "stability/response_length_mean": 296.1000061035156,
+      "stability/response_length_std": 167.77593994140625,
+      "stability/response_length_var": 28148.767578125,
+      "stability/token_entropy_mean": 6.018750190734863,
+      "step": 130
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 1.7402511835098267,
+      "learning_rate": 9.986679638549358e-07,
+      "logits/chosen": -2.2670061588287354,
+      "logits/rejected": -2.1075525283813477,
+      "logps/chosen": -387.3724060058594,
+      "logps/rejected": -213.73989868164062,
+      "loss": 0.6645,
+      "rewards/accuracies": 0.676953136920929,
+      "rewards/chosen": 0.11148682981729507,
+      "rewards/margins": 0.06334061175584793,
+      "rewards/rejected": 0.048146218061447144,
+      "stability/repetition_rate_mean": 0.8604398965835571,
+      "stability/response_length_mean": 596.4000244140625,
+      "stability/response_length_std": 291.9129333496094,
+      "stability/response_length_var": 85213.15625,
+      "stability/token_entropy_mean": 6.510937690734863,
+      "step": 135
+    },
+    {
+      "epoch": 0.7165706973768394,
+      "grad_norm": 1.8281584978103638,
+      "learning_rate": 9.982476273321175e-07,
+      "logits/chosen": -2.25555157661438,
+      "logits/rejected": -2.100205898284912,
+      "logps/chosen": -389.03497314453125,
+      "logps/rejected": -211.17449951171875,
+      "loss": 0.6611,
+      "rewards/accuracies": 0.6976562738418579,
+      "rewards/chosen": 0.11906532198190689,
+      "rewards/margins": 0.07203800976276398,
+      "rewards/rejected": 0.04702730476856232,
+      "stability/repetition_rate_mean": 0.8607560396194458,
+      "stability/response_length_mean": 173.89999389648438,
+      "stability/response_length_std": 70.74908447265625,
+      "stability/response_length_var": 5005.43310546875,
+      "stability/token_entropy_mean": 6.612500190734863,
+      "step": 140
+    },
+    {
+      "epoch": 0.7421625079974408,
+      "grad_norm": 1.7499624490737915,
+      "learning_rate": 9.977698537536417e-07,
+      "logits/chosen": -2.2664294242858887,
+      "logits/rejected": -2.0990548133850098,
+      "logps/chosen": -387.792724609375,
+      "logps/rejected": -203.63430786132812,
+      "loss": 0.6589,
+      "rewards/accuracies": 0.678906261920929,
+      "rewards/chosen": 0.13202722370624542,
+      "rewards/margins": 0.07740288227796555,
+      "rewards/rejected": 0.054624345153570175,
+      "stability/repetition_rate_mean": 0.9638816714286804,
+      "stability/response_length_mean": 720.2000122070312,
+      "stability/response_length_std": 284.2842102050781,
+      "stability/response_length_var": 80817.5078125,
+      "stability/token_entropy_mean": 7.106249809265137,
+      "step": 145
+    },
+    {
+      "epoch": 0.7677543186180422,
+      "grad_norm": 1.8286625146865845,
+      "learning_rate": 9.972346981963546e-07,
+      "logits/chosen": -2.2344398498535156,
+      "logits/rejected": -2.0921382904052734,
+      "logps/chosen": -373.849853515625,
+      "logps/rejected": -211.25674438476562,
+      "loss": 0.6568,
+      "rewards/accuracies": 0.6703124642372131,
+      "rewards/chosen": 0.14155012369155884,
+      "rewards/margins": 0.08604539185762405,
+      "rewards/rejected": 0.05550474300980568,
+      "stability/repetition_rate_mean": 0.8590916395187378,
+      "stability/response_length_mean": 654.2000122070312,
+      "stability/response_length_std": 372.5482177734375,
+      "stability/response_length_var": 138792.171875,
+      "stability/token_entropy_mean": 6.5859375,
+      "step": 150
+    },
+    {
+      "epoch": 0.7933461292386437,
+      "grad_norm": 1.8375192880630493,
+      "learning_rate": 9.966422223519885e-07,
+      "logits/chosen": -2.259735107421875,
+      "logits/rejected": -2.108813762664795,
+      "logps/chosen": -377.39886474609375,
+      "logps/rejected": -206.43038940429688,
+      "loss": 0.6528,
+      "rewards/accuracies": 0.682421863079071,
+      "rewards/chosen": 0.1565590798854828,
+      "rewards/margins": 0.09292585402727127,
+      "rewards/rejected": 0.06363321840763092,
+      "stability/repetition_rate_mean": 0.8765794038772583,
+      "stability/response_length_mean": 502.3999938964844,
+      "stability/response_length_std": 200.0795440673828,
+      "stability/response_length_var": 40031.8203125,
+      "stability/token_entropy_mean": 6.557812690734863,
+      "step": 155
+    },
+    {
+      "epoch": 0.818937939859245,
+      "grad_norm": 1.7681745290756226,
+      "learning_rate": 9.959924945200523e-07,
+      "logits/chosen": -2.2527999877929688,
+      "logits/rejected": -2.086219549179077,
+      "logps/chosen": -383.5099182128906,
+      "logps/rejected": -207.28614807128906,
+      "loss": 0.6491,
+      "rewards/accuracies": 0.6898438334465027,
+      "rewards/chosen": 0.16489063203334808,
+      "rewards/margins": 0.10477059334516525,
+      "rewards/rejected": 0.06012001633644104,
+      "stability/repetition_rate_mean": 0.9663667678833008,
+      "stability/response_length_mean": 870.4000244140625,
+      "stability/response_length_std": 147.6581573486328,
+      "stability/response_length_var": 21802.93359375,
+      "stability/token_entropy_mean": 7.243750095367432,
+      "step": 160
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.6438355445861816,
+      "learning_rate": 9.952855895999568e-07,
+      "logits/chosen": -2.23742413520813,
+      "logits/rejected": -2.089754104614258,
+      "logps/chosen": -362.43914794921875,
+      "logps/rejected": -203.66177368164062,
+      "loss": 0.6477,
+      "rewards/accuracies": 0.684374988079071,
+      "rewards/chosen": 0.1636190414428711,
+      "rewards/margins": 0.10717189311981201,
+      "rewards/rejected": 0.05644712969660759,
+      "stability/repetition_rate_mean": 0.8620624542236328,
+      "stability/response_length_mean": 547.7000122070312,
+      "stability/response_length_std": 249.60325622558594,
+      "stability/response_length_var": 62301.7890625,
+      "stability/token_entropy_mean": 6.599999904632568,
+      "step": 165
+    },
+    {
+      "epoch": 0.8701215611004478,
+      "grad_norm": 1.7001264095306396,
+      "learning_rate": 9.9452158908238e-07,
+      "logits/chosen": -2.2632334232330322,
+      "logits/rejected": -2.097275733947754,
+      "logps/chosen": -376.97039794921875,
+      "logps/rejected": -203.81222534179688,
+      "loss": 0.6419,
+      "rewards/accuracies": 0.705078125,
+      "rewards/chosen": 0.18040212988853455,
+      "rewards/margins": 0.12267370522022247,
+      "rewards/rejected": 0.057728420943021774,
+      "stability/repetition_rate_mean": 0.9189292192459106,
+      "stability/response_length_mean": 385.70001220703125,
+      "stability/response_length_std": 143.96685791015625,
+      "stability/response_length_var": 20726.455078125,
+      "stability/token_entropy_mean": 6.831250190734863,
+      "step": 170
+    },
+    {
+      "epoch": 0.8957133717210493,
+      "grad_norm": 1.804345726966858,
+      "learning_rate": 9.937005810398745e-07,
+      "logits/chosen": -2.2485477924346924,
+      "logits/rejected": -2.0931143760681152,
+      "logps/chosen": -380.5006408691406,
+      "logps/rejected": -212.2974853515625,
+      "loss": 0.6396,
+      "rewards/accuracies": 0.6957032084465027,
+      "rewards/chosen": 0.18881921470165253,
+      "rewards/margins": 0.13226722180843353,
+      "rewards/rejected": 0.0565519854426384,
+      "stability/repetition_rate_mean": 0.9413634538650513,
+      "stability/response_length_mean": 460.6000061035156,
+      "stability/response_length_std": 165.9894256591797,
+      "stability/response_length_var": 27552.48828125,
+      "stability/token_entropy_mean": 6.959374904632568,
+      "step": 175
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 1.76821768283844,
+      "learning_rate": 9.928226601167138e-07,
+      "logits/chosen": -2.2474441528320312,
+      "logits/rejected": -2.0858802795410156,
+      "logps/chosen": -387.0555419921875,
+      "logps/rejected": -212.9912109375,
+      "loss": 0.6377,
+      "rewards/accuracies": 0.690234363079071,
+      "rewards/chosen": 0.20255513489246368,
+      "rewards/margins": 0.13533423840999603,
+      "rewards/rejected": 0.06722089648246765,
+      "stability/repetition_rate_mean": 0.8933365941047668,
+      "stability/response_length_mean": 716.2000122070312,
+      "stability/response_length_std": 329.3815002441406,
+      "stability/response_length_var": 108492.1796875,
+      "stability/token_entropy_mean": 6.526562690734863,
+      "step": 180
+    },
+    {
+      "epoch": 0.946896992962252,
+      "grad_norm": 1.6571131944656372,
+      "learning_rate": 9.918879275179817e-07,
+      "logits/chosen": -2.249807596206665,
+      "logits/rejected": -2.0986175537109375,
+      "logps/chosen": -374.5240173339844,
+      "logps/rejected": -210.4625244140625,
+      "loss": 0.6365,
+      "rewards/accuracies": 0.6898437738418579,
+      "rewards/chosen": 0.21025450527668,
+      "rewards/margins": 0.14256080985069275,
+      "rewards/rejected": 0.06769368052482605,
+      "stability/repetition_rate_mean": 0.8755763173103333,
+      "stability/response_length_mean": 486.3999938964844,
+      "stability/response_length_std": 213.8624725341797,
+      "stability/response_length_var": 45737.15625,
+      "stability/token_entropy_mean": 6.559374809265137,
+      "step": 185
+    },
+    {
+      "epoch": 0.9724888035828535,
+      "grad_norm": 1.5521091222763062,
+      "learning_rate": 9.90896490997906e-07,
+      "logits/chosen": -2.233643054962158,
+      "logits/rejected": -2.0779027938842773,
+      "logps/chosen": -377.6806640625,
+      "logps/rejected": -206.34112548828125,
+      "loss": 0.6287,
+      "rewards/accuracies": 0.701953113079071,
+      "rewards/chosen": 0.2152232825756073,
+      "rewards/margins": 0.16218551993370056,
+      "rewards/rejected": 0.053037747740745544,
+      "stability/repetition_rate_mean": 0.8903291821479797,
+      "stability/response_length_mean": 736.7999877929688,
+      "stability/response_length_std": 246.57106018066406,
+      "stability/response_length_var": 60797.2890625,
+      "stability/token_entropy_mean": 6.7109375,
+      "step": 190
+    },
+    {
+      "epoch": 0.9980806142034548,
+      "grad_norm": 1.6728918552398682,
+      "learning_rate": 9.898484648474362e-07,
+      "logits/chosen": -2.250211477279663,
+      "logits/rejected": -2.096865653991699,
+      "logps/chosen": -361.8232116699219,
+      "logps/rejected": -208.4629364013672,
+      "loss": 0.6296,
+      "rewards/accuracies": 0.7027343511581421,
+      "rewards/chosen": 0.2173018455505371,
+      "rewards/margins": 0.16235104203224182,
+      "rewards/rejected": 0.054950814694166183,
+      "stability/repetition_rate_mean": 0.7688959836959839,
+      "stability/response_length_mean": 356.3999938964844,
+      "stability/response_length_std": 207.94882202148438,
+      "stability/response_length_var": 43242.7109375,
+      "stability/token_entropy_mean": 5.928124904632568,
+      "step": 195
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 196,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 200,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.072674331683193e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

mistral7b_dpo_baseline/checkpoint-196/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94e6d5df42ddf678581bc105236846c8bb9630da14f49c0e71f40b3a546b04ca
+size 8017

mistral7b_dpo_baseline/checkpoint-196/zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device, weights_only=False)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
+
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

mistral7b_dpo_baseline/special_tokens_map.json

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

mistral7b_dpo_baseline/tokenizer.model

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

mistral7b_dpo_baseline/tokenizer_config.json

@@ -0,0 +1,46 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": null,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [],
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "extra_special_tokens": {},
+  "legacy": false,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "</s>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "split_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

mistral7b_dpo_baseline/train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "total_flos": 7.072674331683193e+18,
+    "train_loss": 0.6730708850889789,
+    "train_runtime": 4328.1955,
+    "train_samples_per_second": 23.104,
+    "train_steps_per_second": 0.045
+}

mistral7b_dpo_baseline/trainer_log.jsonl

@@ -0,0 +1,40 @@
+{"current_steps": 5, "total_steps": 196, "loss": 0.6931, "accuracy": 0.3062500059604645, "lr": 4e-08, "epoch": 0.025591810620601407, "percentage": 2.55, "elapsed_time": "0:01:47", "remaining_time": "1:08:36", "rewards/chosen": 0.0012258010683581233, "rewards/rejected": 0.0008625800255686045, "rewards/accuracies": 0.3062500059604645, "rewards/margins": 0.0003632208681665361, "logps/chosen": -375.063232421875, "logps/rejected": -210.8657684326172, "logits/chosen": -2.273033380508423, "logits/rejected": -2.1234283447265625}
+{"current_steps": 10, "total_steps": 196, "loss": 0.6931, "accuracy": 0.5082031488418579, "lr": 9e-08, "epoch": 0.05118362124120281, "percentage": 5.1, "elapsed_time": "0:03:38", "remaining_time": "1:07:50", "rewards/chosen": 0.0013576988130807877, "rewards/rejected": 0.0008109854534268379, "rewards/accuracies": 0.5082031488418579, "rewards/margins": 0.0005467134760692716, "logps/chosen": -382.0296325683594, "logps/rejected": -202.4493408203125, "logits/chosen": -2.2679283618927, "logits/rejected": -2.116572141647339, "stability/response_length_mean": 602.7999877929688, "stability/response_length_std": 295.1842346191406, "stability/response_length_var": 87133.734375, "stability/token_entropy_mean": 6.959374904632568, "stability/repetition_rate_mean": 0.9135147929191589}
+{"current_steps": 15, "total_steps": 196, "loss": 0.6932, "accuracy": 0.5082031488418579, "lr": 1.4e-07, "epoch": 0.07677543186180422, "percentage": 7.65, "elapsed_time": "0:05:29", "remaining_time": "1:06:10", "rewards/chosen": 0.0017401383956894279, "rewards/rejected": 0.0014510790351778269, "rewards/accuracies": 0.5082031488418579, "rewards/margins": 0.00028905909857712686, "logps/chosen": -389.2082214355469, "logps/rejected": -207.0491180419922, "logits/chosen": -2.260676860809326, "logits/rejected": -2.0991530418395996, "stability/response_length_mean": 345.70001220703125, "stability/response_length_std": 212.67974853515625, "stability/response_length_var": 45232.6796875, "stability/token_entropy_mean": 6.342187404632568, "stability/repetition_rate_mean": 0.8363126516342163}
+{"current_steps": 20, "total_steps": 196, "loss": 0.6939, "accuracy": 0.49140626192092896, "lr": 1.8999999999999998e-07, "epoch": 0.10236724248240563, "percentage": 10.2, "elapsed_time": "0:07:19", "remaining_time": "1:04:29", "rewards/chosen": 0.0004601169202942401, "rewards/rejected": 0.0015163427451625466, "rewards/accuracies": 0.49140626192092896, "rewards/margins": -0.0010562256211414933, "logps/chosen": -368.71954345703125, "logps/rejected": -198.5596160888672, "logits/chosen": -2.2613608837127686, "logits/rejected": -2.1145310401916504, "stability/response_length_mean": 739.4000244140625, "stability/response_length_std": 281.282470703125, "stability/response_length_var": 79119.8203125, "stability/token_entropy_mean": 6.931250095367432, "stability/repetition_rate_mean": 0.9317001104354858}
+{"current_steps": 25, "total_steps": 196, "loss": 0.693, "accuracy": 0.5082031488418579, "lr": 2.4e-07, "epoch": 0.12795905310300704, "percentage": 12.76, "elapsed_time": "0:09:10", "remaining_time": "1:02:48", "rewards/chosen": 0.002021136460825801, "rewards/rejected": 0.0014201054582372308, "rewards/accuracies": 0.5082031488418579, "rewards/margins": 0.0006010310025885701, "logps/chosen": -378.1647033691406, "logps/rejected": -205.86363220214844, "logits/chosen": -2.277377128601074, "logits/rejected": -2.125220775604248, "stability/response_length_mean": 658.5999755859375, "stability/response_length_std": 288.0155029296875, "stability/response_length_var": 82952.9296875, "stability/token_entropy_mean": 6.821875095367432, "stability/repetition_rate_mean": 0.8950179815292358}
+{"current_steps": 30, "total_steps": 196, "loss": 0.6935, "accuracy": 0.490234375, "lr": 2.9e-07, "epoch": 0.15355086372360843, "percentage": 15.31, "elapsed_time": "0:11:01", "remaining_time": "1:01:00", "rewards/chosen": 0.0010083412053063512, "rewards/rejected": 0.0013262125430628657, "rewards/accuracies": 0.490234375, "rewards/margins": -0.0003178715705871582, "logps/chosen": -373.9202575683594, "logps/rejected": -208.93287658691406, "logits/chosen": -2.2701575756073, "logits/rejected": -2.1229705810546875, "stability/response_length_mean": 356.0, "stability/response_length_std": 249.6766815185547, "stability/response_length_var": 62338.4453125, "stability/token_entropy_mean": 6.199999809265137, "stability/repetition_rate_mean": 0.8135284185409546}
+{"current_steps": 35, "total_steps": 196, "loss": 0.6928, "accuracy": 0.5265624523162842, "lr": 3.4000000000000003e-07, "epoch": 0.17914267434420986, "percentage": 17.86, "elapsed_time": "0:12:52", "remaining_time": "0:59:14", "rewards/chosen": 0.0032140661496669054, "rewards/rejected": 0.0020710753742605448, "rewards/accuracies": 0.5265624523162842, "rewards/margins": 0.0011429906589910388, "logps/chosen": -366.3136901855469, "logps/rejected": -203.98719787597656, "logits/chosen": -2.2612318992614746, "logits/rejected": -2.120094060897827, "stability/response_length_mean": 606.2000122070312, "stability/response_length_std": 394.1366882324219, "stability/response_length_var": 155343.734375, "stability/token_entropy_mean": 7.112500190734863, "stability/repetition_rate_mean": 0.9289250373840332}
+{"current_steps": 40, "total_steps": 196, "loss": 0.6928, "accuracy": 0.5144531726837158, "lr": 3.8999999999999997e-07, "epoch": 0.20473448496481125, "percentage": 20.41, "elapsed_time": "0:14:44", "remaining_time": "0:57:28", "rewards/chosen": 0.004107568878680468, "rewards/rejected": 0.0029460543300956488, "rewards/accuracies": 0.5144531726837158, "rewards/margins": 0.0011615146650001407, "logps/chosen": -369.1784973144531, "logps/rejected": -197.37686157226562, "logits/chosen": -2.2764394283294678, "logits/rejected": -2.122685432434082, "stability/response_length_mean": 547.2000122070312, "stability/response_length_std": 290.8897399902344, "stability/response_length_var": 84616.84375, "stability/token_entropy_mean": 6.557812690734863, "stability/repetition_rate_mean": 0.8629743456840515}
+{"current_steps": 45, "total_steps": 196, "loss": 0.6933, "accuracy": 0.512890636920929, "lr": 4.3999999999999997e-07, "epoch": 0.23032629558541268, "percentage": 22.96, "elapsed_time": "0:16:34", "remaining_time": "0:55:36", "rewards/chosen": 0.0039331079460680485, "rewards/rejected": 0.003900188487023115, "rewards/accuracies": 0.512890636920929, "rewards/margins": 3.291945904493332e-05, "logps/chosen": -385.1796569824219, "logps/rejected": -212.56887817382812, "logits/chosen": -2.257107734680176, "logits/rejected": -2.1072099208831787, "stability/response_length_mean": 366.5, "stability/response_length_std": 118.86243438720703, "stability/response_length_var": 14128.27734375, "stability/token_entropy_mean": 6.640625, "stability/repetition_rate_mean": 0.8764154314994812}
+{"current_steps": 50, "total_steps": 196, "loss": 0.6916, "accuracy": 0.5316406488418579, "lr": 4.9e-07, "epoch": 0.2559181062060141, "percentage": 25.51, "elapsed_time": "0:18:24", "remaining_time": "0:53:45", "rewards/chosen": 0.0068646748550236225, "rewards/rejected": 0.003293551504611969, "rewards/accuracies": 0.5316406488418579, "rewards/margins": 0.0035711233504116535, "logps/chosen": -378.2891540527344, "logps/rejected": -207.4978790283203, "logits/chosen": -2.264720916748047, "logits/rejected": -2.109816312789917, "stability/response_length_mean": 384.1000061035156, "stability/response_length_std": 178.4809112548828, "stability/response_length_var": 31855.43359375, "stability/token_entropy_mean": 6.543749809265137, "stability/repetition_rate_mean": 0.8785834312438965}
+{"current_steps": 55, "total_steps": 196, "loss": 0.6911, "accuracy": 0.536328136920929, "lr": 5.4e-07, "epoch": 0.28150991682661547, "percentage": 28.06, "elapsed_time": "0:20:14", "remaining_time": "0:51:54", "rewards/chosen": 0.009546511806547642, "rewards/rejected": 0.005061793606728315, "rewards/accuracies": 0.536328136920929, "rewards/margins": 0.004484717734158039, "logps/chosen": -393.7845764160156, "logps/rejected": -210.35861206054688, "logits/chosen": -2.2527079582214355, "logits/rejected": -2.094048261642456, "stability/response_length_mean": 332.8999938964844, "stability/response_length_std": 137.39193725585938, "stability/response_length_var": 18876.544921875, "stability/token_entropy_mean": 7.009375095367432, "stability/repetition_rate_mean": 0.9481571912765503}
+{"current_steps": 60, "total_steps": 196, "loss": 0.6915, "accuracy": 0.5328124761581421, "lr": 5.9e-07, "epoch": 0.30710172744721687, "percentage": 30.61, "elapsed_time": "0:22:06", "remaining_time": "0:50:06", "rewards/chosen": 0.009509808383882046, "rewards/rejected": 0.005750678479671478, "rewards/accuracies": 0.5328124761581421, "rewards/margins": 0.00375912943854928, "logps/chosen": -387.7115173339844, "logps/rejected": -206.819580078125, "logits/chosen": -2.272801160812378, "logits/rejected": -2.111156463623047, "stability/response_length_mean": 655.2999877929688, "stability/response_length_std": 282.005126953125, "stability/response_length_var": 79526.8984375, "stability/token_entropy_mean": 6.559374809265137, "stability/repetition_rate_mean": 0.8699707984924316}
+{"current_steps": 65, "total_steps": 196, "loss": 0.6908, "accuracy": 0.5625000596046448, "lr": 6.4e-07, "epoch": 0.3326935380678183, "percentage": 33.16, "elapsed_time": "0:23:56", "remaining_time": "0:48:14", "rewards/chosen": 0.012991372495889664, "rewards/rejected": 0.007922019809484482, "rewards/accuracies": 0.5625000596046448, "rewards/margins": 0.00506935128942132, "logps/chosen": -377.4696960449219, "logps/rejected": -204.3833465576172, "logits/chosen": -2.2521181106567383, "logits/rejected": -2.098989963531494, "stability/response_length_mean": 306.3999938964844, "stability/response_length_std": 141.7236785888672, "stability/response_length_var": 20085.599609375, "stability/token_entropy_mean": 6.840624809265137, "stability/repetition_rate_mean": 0.9148252606391907}
+{"current_steps": 70, "total_steps": 196, "loss": 0.6888, "accuracy": 0.5777344107627869, "lr": 6.9e-07, "epoch": 0.3582853486884197, "percentage": 35.71, "elapsed_time": "0:25:46", "remaining_time": "0:46:24", "rewards/chosen": 0.017473159357905388, "rewards/rejected": 0.008267344906926155, "rewards/accuracies": 0.5777344107627869, "rewards/margins": 0.009205814450979233, "logps/chosen": -393.73406982421875, "logps/rejected": -206.1870574951172, "logits/chosen": -2.2713451385498047, "logits/rejected": -2.106651782989502, "stability/response_length_mean": 472.70001220703125, "stability/response_length_std": 159.89791870117188, "stability/response_length_var": 25567.34375, "stability/token_entropy_mean": 7.162499904632568, "stability/repetition_rate_mean": 0.9625760316848755}
+{"current_steps": 75, "total_steps": 196, "loss": 0.689, "accuracy": 0.577343761920929, "lr": 7.4e-07, "epoch": 0.3838771593090211, "percentage": 38.27, "elapsed_time": "0:27:37", "remaining_time": "0:44:33", "rewards/chosen": 0.018993159756064415, "rewards/rejected": 0.010194134898483753, "rewards/accuracies": 0.577343761920929, "rewards/margins": 0.008799022994935513, "logps/chosen": -381.0571594238281, "logps/rejected": -207.9349365234375, "logits/chosen": -2.275454521179199, "logits/rejected": -2.116396903991699, "stability/response_length_mean": 527.0, "stability/response_length_std": 293.7500915527344, "stability/response_length_var": 86289.109375, "stability/token_entropy_mean": 6.139062404632568, "stability/repetition_rate_mean": 0.9068478345870972}
+{"current_steps": 80, "total_steps": 196, "loss": 0.6886, "accuracy": 0.5726562738418579, "lr": 7.9e-07, "epoch": 0.4094689699296225, "percentage": 40.82, "elapsed_time": "0:29:28", "remaining_time": "0:42:43", "rewards/chosen": 0.021377386525273323, "rewards/rejected": 0.011784089729189873, "rewards/accuracies": 0.5726562738418579, "rewards/margins": 0.009593295864760876, "logps/chosen": -382.4193420410156, "logps/rejected": -208.97756958007812, "logits/chosen": -2.2604687213897705, "logits/rejected": -2.1025872230529785, "stability/response_length_mean": 635.5, "stability/response_length_std": 262.53900146484375, "stability/response_length_var": 68926.71875, "stability/token_entropy_mean": 7.199999809265137, "stability/repetition_rate_mean": 0.9464001655578613}
+{"current_steps": 85, "total_steps": 196, "loss": 0.6875, "accuracy": 0.5863280892372131, "lr": 8.399999999999999e-07, "epoch": 0.4350607805502239, "percentage": 43.37, "elapsed_time": "0:31:19", "remaining_time": "0:40:54", "rewards/chosen": 0.026788339018821716, "rewards/rejected": 0.014684918336570263, "rewards/accuracies": 0.5863280892372131, "rewards/margins": 0.012103421613574028, "logps/chosen": -387.3235778808594, "logps/rejected": -208.6320343017578, "logits/chosen": -2.258051872253418, "logits/rejected": -2.100944757461548, "stability/response_length_mean": 600.0999755859375, "stability/response_length_std": 349.2030334472656, "stability/response_length_var": 121942.765625, "stability/token_entropy_mean": 6.737500190734863, "stability/repetition_rate_mean": 0.9034587144851685}
+{"current_steps": 90, "total_steps": 196, "loss": 0.6859, "accuracy": 0.608593761920929, "lr": 8.9e-07, "epoch": 0.46065259117082535, "percentage": 45.92, "elapsed_time": "0:33:10", "remaining_time": "0:39:04", "rewards/chosen": 0.032174669206142426, "rewards/rejected": 0.01697630062699318, "rewards/accuracies": 0.608593761920929, "rewards/margins": 0.015198366716504097, "logps/chosen": -375.204345703125, "logps/rejected": -206.2209930419922, "logits/chosen": -2.261427164077759, "logits/rejected": -2.1134793758392334, "stability/response_length_mean": 692.9000244140625, "stability/response_length_std": 127.31452941894531, "stability/response_length_var": 16208.9892578125, "stability/token_entropy_mean": 7.109375, "stability/repetition_rate_mean": 0.9765706062316895}
+{"current_steps": 95, "total_steps": 196, "loss": 0.6838, "accuracy": 0.6300780773162842, "lr": 9.399999999999999e-07, "epoch": 0.48624440179142675, "percentage": 48.47, "elapsed_time": "0:35:01", "remaining_time": "0:37:14", "rewards/chosen": 0.03762410953640938, "rewards/rejected": 0.01792534813284874, "rewards/accuracies": 0.6300780773162842, "rewards/margins": 0.01969875767827034, "logps/chosen": -366.8333740234375, "logps/rejected": -203.4322967529297, "logits/chosen": -2.2588284015655518, "logits/rejected": -2.112243175506592, "stability/response_length_mean": 727.0, "stability/response_length_std": 335.0764465332031, "stability/response_length_var": 112276.21875, "stability/token_entropy_mean": 7.3125, "stability/repetition_rate_mean": 0.9483324289321899}
+{"current_steps": 100, "total_steps": 196, "loss": 0.6821, "accuracy": 0.6527343988418579, "lr": 9.9e-07, "epoch": 0.5118362124120281, "percentage": 51.02, "elapsed_time": "0:36:52", "remaining_time": "0:35:24", "rewards/chosen": 0.045418523252010345, "rewards/rejected": 0.0220339372754097, "rewards/accuracies": 0.6527343988418579, "rewards/margins": 0.02338458225131035, "logps/chosen": -384.503662109375, "logps/rejected": -210.62109375, "logits/chosen": -2.27445387840271, "logits/rejected": -2.11898136138916, "stability/response_length_mean": 419.0, "stability/response_length_std": 232.0225067138672, "stability/response_length_var": 53834.4453125, "stability/token_entropy_mean": 7.074999809265137, "stability/repetition_rate_mean": 0.9498335123062134}
+{"current_steps": 105, "total_steps": 196, "loss": 0.6811, "accuracy": 0.626953125, "lr": 9.999815554351224e-07, "epoch": 0.5374280230326296, "percentage": 53.57, "elapsed_time": "0:38:43", "remaining_time": "0:33:34", "rewards/chosen": 0.052092913538217545, "rewards/rejected": 0.02642221562564373, "rewards/accuracies": 0.626953125, "rewards/margins": 0.025670701637864113, "logps/chosen": -378.0056457519531, "logps/rejected": -209.6881103515625, "logits/chosen": -2.258531093597412, "logits/rejected": -2.108144998550415, "stability/response_length_mean": 669.5, "stability/response_length_std": 271.60888671875, "stability/response_length_var": 73771.390625, "stability/token_entropy_mean": 6.564843654632568, "stability/repetition_rate_mean": 0.901488184928894}
+{"current_steps": 110, "total_steps": 196, "loss": 0.6795, "accuracy": 0.6519531607627869, "lr": 9.999066267225444e-07, "epoch": 0.5630198336532309, "percentage": 56.12, "elapsed_time": "0:40:33", "remaining_time": "0:31:42", "rewards/chosen": 0.059595055878162384, "rewards/rejected": 0.03043590858578682, "rewards/accuracies": 0.6519531607627869, "rewards/margins": 0.029159152880311012, "logps/chosen": -385.2655029296875, "logps/rejected": -206.72357177734375, "logits/chosen": -2.2750279903411865, "logits/rejected": -2.117339611053467, "stability/response_length_mean": 255.60000610351562, "stability/response_length_std": 114.30874633789062, "stability/response_length_var": 13066.4892578125, "stability/token_entropy_mean": 7.015625, "stability/repetition_rate_mean": 0.9347068071365356}
+{"current_steps": 115, "total_steps": 196, "loss": 0.6763, "accuracy": 0.6519531607627869, "lr": 9.997740697079592e-07, "epoch": 0.5886116442738324, "percentage": 58.67, "elapsed_time": "0:42:24", "remaining_time": "0:29:52", "rewards/chosen": 0.07092021405696869, "rewards/rejected": 0.034616947174072266, "rewards/accuracies": 0.6519531607627869, "rewards/margins": 0.036303263157606125, "logps/chosen": -372.7311096191406, "logps/rejected": -213.94320678710938, "logits/chosen": -2.251587390899658, "logits/rejected": -2.110828399658203, "stability/response_length_mean": 499.29998779296875, "stability/response_length_std": 282.31072998046875, "stability/response_length_var": 79699.34375, "stability/token_entropy_mean": 6.553124904632568, "stability/repetition_rate_mean": 0.8279488682746887}
+{"current_steps": 120, "total_steps": 196, "loss": 0.6733, "accuracy": 0.6703125238418579, "lr": 9.995838996722914e-07, "epoch": 0.6142034548944337, "percentage": 61.22, "elapsed_time": "0:44:14", "remaining_time": "0:28:00", "rewards/chosen": 0.08155383914709091, "rewards/rejected": 0.03867081552743912, "rewards/accuracies": 0.6703125238418579, "rewards/margins": 0.04288302734494209, "logps/chosen": -367.7881164550781, "logps/rejected": -207.59573364257812, "logits/chosen": -2.2482666969299316, "logits/rejected": -2.102483034133911, "stability/response_length_mean": 363.6000061035156, "stability/response_length_std": 146.0701446533203, "stability/response_length_var": 21336.48828125, "stability/token_entropy_mean": 6.8125, "stability/repetition_rate_mean": 0.9149526357650757}
+{"current_steps": 125, "total_steps": 196, "loss": 0.6707, "accuracy": 0.6781249642372131, "lr": 9.993361385379874e-07, "epoch": 0.6397952655150352, "percentage": 63.78, "elapsed_time": "0:46:03", "remaining_time": "0:26:09", "rewards/chosen": 0.09179403632879257, "rewards/rejected": 0.04230235144495964, "rewards/accuracies": 0.6781249642372131, "rewards/margins": 0.04949168488383293, "logps/chosen": -385.7173767089844, "logps/rejected": -205.1151123046875, "logits/chosen": -2.253580093383789, "logits/rejected": -2.0954601764678955, "stability/response_length_mean": 433.79998779296875, "stability/response_length_std": 126.92412567138672, "stability/response_length_var": 16109.7333984375, "stability/token_entropy_mean": 6.532812595367432, "stability/repetition_rate_mean": 0.8946477770805359}
+{"current_steps": 130, "total_steps": 196, "loss": 0.6672, "accuracy": 0.682421863079071, "lr": 9.99030814866488e-07, "epoch": 0.6653870761356366, "percentage": 66.33, "elapsed_time": "0:47:53", "remaining_time": "0:24:18", "rewards/chosen": 0.09893052279949188, "rewards/rejected": 0.041869960725307465, "rewards/accuracies": 0.682421863079071, "rewards/margins": 0.057060565799474716, "logps/chosen": -375.60565185546875, "logps/rejected": -203.76229858398438, "logits/chosen": -2.2453463077545166, "logits/rejected": -2.096496105194092, "stability/response_length_mean": 296.1000061035156, "stability/response_length_std": 167.77593994140625, "stability/response_length_var": 28148.767578125, "stability/token_entropy_mean": 6.018750190734863, "stability/repetition_rate_mean": 0.8153165578842163}
+{"current_steps": 135, "total_steps": 196, "loss": 0.6645, "accuracy": 0.676953136920929, "lr": 9.986679638549358e-07, "epoch": 0.690978886756238, "percentage": 68.88, "elapsed_time": "0:49:44", "remaining_time": "0:22:28", "rewards/chosen": 0.11148682981729507, "rewards/rejected": 0.048146218061447144, "rewards/accuracies": 0.676953136920929, "rewards/margins": 0.06334061175584793, "logps/chosen": -387.3724060058594, "logps/rejected": -213.73989868164062, "logits/chosen": -2.2670061588287354, "logits/rejected": -2.1075525283813477, "stability/response_length_mean": 596.4000244140625, "stability/response_length_std": 291.9129333496094, "stability/response_length_var": 85213.15625, "stability/token_entropy_mean": 6.510937690734863, "stability/repetition_rate_mean": 0.8604398965835571}
+{"current_steps": 140, "total_steps": 196, "loss": 0.6611, "accuracy": 0.6976562738418579, "lr": 9.982476273321175e-07, "epoch": 0.7165706973768394, "percentage": 71.43, "elapsed_time": "0:51:34", "remaining_time": "0:20:37", "rewards/chosen": 0.11906532198190689, "rewards/rejected": 0.04702730476856232, "rewards/accuracies": 0.6976562738418579, "rewards/margins": 0.07203800976276398, "logps/chosen": -389.03497314453125, "logps/rejected": -211.17449951171875, "logits/chosen": -2.25555157661438, "logits/rejected": -2.100205898284912, "stability/response_length_mean": 173.89999389648438, "stability/response_length_std": 70.74908447265625, "stability/response_length_var": 5005.43310546875, "stability/token_entropy_mean": 6.612500190734863, "stability/repetition_rate_mean": 0.8607560396194458}
+{"current_steps": 145, "total_steps": 196, "loss": 0.6589, "accuracy": 0.678906261920929, "lr": 9.977698537536417e-07, "epoch": 0.7421625079974408, "percentage": 73.98, "elapsed_time": "0:53:25", "remaining_time": "0:18:47", "rewards/chosen": 0.13202722370624542, "rewards/rejected": 0.054624345153570175, "rewards/accuracies": 0.678906261920929, "rewards/margins": 0.07740288227796555, "logps/chosen": -387.792724609375, "logps/rejected": -203.63430786132812, "logits/chosen": -2.2664294242858887, "logits/rejected": -2.0990548133850098, "stability/response_length_mean": 720.2000122070312, "stability/response_length_std": 284.2842102050781, "stability/response_length_var": 80817.5078125, "stability/token_entropy_mean": 7.106249809265137, "stability/repetition_rate_mean": 0.9638816714286804}
+{"current_steps": 150, "total_steps": 196, "loss": 0.6568, "accuracy": 0.6703124642372131, "lr": 9.972346981963546e-07, "epoch": 0.7677543186180422, "percentage": 76.53, "elapsed_time": "0:55:17", "remaining_time": "0:16:57", "rewards/chosen": 0.14155012369155884, "rewards/rejected": 0.05550474300980568, "rewards/accuracies": 0.6703124642372131, "rewards/margins": 0.08604539185762405, "logps/chosen": -373.849853515625, "logps/rejected": -211.25674438476562, "logits/chosen": -2.2344398498535156, "logits/rejected": -2.0921382904052734, "stability/response_length_mean": 654.2000122070312, "stability/response_length_std": 372.5482177734375, "stability/response_length_var": 138792.171875, "stability/token_entropy_mean": 6.5859375, "stability/repetition_rate_mean": 0.8590916395187378}
+{"current_steps": 155, "total_steps": 196, "loss": 0.6528, "accuracy": 0.682421863079071, "lr": 9.966422223519885e-07, "epoch": 0.7933461292386437, "percentage": 79.08, "elapsed_time": "0:57:08", "remaining_time": "0:15:06", "rewards/chosen": 0.1565590798854828, "rewards/rejected": 0.06363321840763092, "rewards/accuracies": 0.682421863079071, "rewards/margins": 0.09292585402727127, "logps/chosen": -377.39886474609375, "logps/rejected": -206.43038940429688, "logits/chosen": -2.259735107421875, "logits/rejected": -2.108813762664795, "stability/response_length_mean": 502.3999938964844, "stability/response_length_std": 200.0795440673828, "stability/response_length_var": 40031.8203125, "stability/token_entropy_mean": 6.557812690734863, "stability/repetition_rate_mean": 0.8765794038772583}
+{"current_steps": 160, "total_steps": 196, "loss": 0.6491, "accuracy": 0.6898438334465027, "lr": 9.959924945200523e-07, "epoch": 0.818937939859245, "percentage": 81.63, "elapsed_time": "0:58:59", "remaining_time": "0:13:16", "rewards/chosen": 0.16489063203334808, "rewards/rejected": 0.06012001633644104, "rewards/accuracies": 0.6898438334465027, "rewards/margins": 0.10477059334516525, "logps/chosen": -383.5099182128906, "logps/rejected": -207.28614807128906, "logits/chosen": -2.2527999877929688, "logits/rejected": -2.086219549179077, "stability/response_length_mean": 870.4000244140625, "stability/response_length_std": 147.6581573486328, "stability/response_length_var": 21802.93359375, "stability/token_entropy_mean": 7.243750095367432, "stability/repetition_rate_mean": 0.9663667678833008}
+{"current_steps": 165, "total_steps": 196, "loss": 0.6477, "accuracy": 0.684374988079071, "lr": 9.952855895999568e-07, "epoch": 0.8445297504798465, "percentage": 84.18, "elapsed_time": "1:00:50", "remaining_time": "0:11:25", "rewards/chosen": 0.1636190414428711, "rewards/rejected": 0.05644712969660759, "rewards/accuracies": 0.684374988079071, "rewards/margins": 0.10717189311981201, "logps/chosen": -362.43914794921875, "logps/rejected": -203.66177368164062, "logits/chosen": -2.23742413520813, "logits/rejected": -2.089754104614258, "stability/response_length_mean": 547.7000122070312, "stability/response_length_std": 249.60325622558594, "stability/response_length_var": 62301.7890625, "stability/token_entropy_mean": 6.599999904632568, "stability/repetition_rate_mean": 0.8620624542236328}
+{"current_steps": 170, "total_steps": 196, "loss": 0.6419, "accuracy": 0.705078125, "lr": 9.9452158908238e-07, "epoch": 0.8701215611004478, "percentage": 86.73, "elapsed_time": "1:02:40", "remaining_time": "0:09:35", "rewards/chosen": 0.18040212988853455, "rewards/rejected": 0.057728420943021774, "rewards/accuracies": 0.705078125, "rewards/margins": 0.12267370522022247, "logps/chosen": -376.97039794921875, "logps/rejected": -203.81222534179688, "logits/chosen": -2.2632334232330322, "logits/rejected": -2.097275733947754, "stability/response_length_mean": 385.70001220703125, "stability/response_length_std": 143.96685791015625, "stability/response_length_var": 20726.455078125, "stability/token_entropy_mean": 6.831250190734863, "stability/repetition_rate_mean": 0.9189292192459106}
+{"current_steps": 175, "total_steps": 196, "loss": 0.6396, "accuracy": 0.6957032084465027, "lr": 9.937005810398745e-07, "epoch": 0.8957133717210493, "percentage": 89.29, "elapsed_time": "1:04:30", "remaining_time": "0:07:44", "rewards/chosen": 0.18881921470165253, "rewards/rejected": 0.0565519854426384, "rewards/accuracies": 0.6957032084465027, "rewards/margins": 0.13226722180843353, "logps/chosen": -380.5006408691406, "logps/rejected": -212.2974853515625, "logits/chosen": -2.2485477924346924, "logits/rejected": -2.0931143760681152, "stability/response_length_mean": 460.6000061035156, "stability/response_length_std": 165.9894256591797, "stability/response_length_var": 27552.48828125, "stability/token_entropy_mean": 6.959374904632568, "stability/repetition_rate_mean": 0.9413634538650513}
+{"current_steps": 180, "total_steps": 196, "loss": 0.6377, "accuracy": 0.690234363079071, "lr": 9.928226601167138e-07, "epoch": 0.9213051823416507, "percentage": 91.84, "elapsed_time": "1:06:22", "remaining_time": "0:05:54", "rewards/chosen": 0.20255513489246368, "rewards/rejected": 0.06722089648246765, "rewards/accuracies": 0.690234363079071, "rewards/margins": 0.13533423840999603, "logps/chosen": -387.0555419921875, "logps/rejected": -212.9912109375, "logits/chosen": -2.2474441528320312, "logits/rejected": -2.0858802795410156, "stability/response_length_mean": 716.2000122070312, "stability/response_length_std": 329.3815002441406, "stability/response_length_var": 108492.1796875, "stability/token_entropy_mean": 6.526562690734863, "stability/repetition_rate_mean": 0.8933365941047668}
+{"current_steps": 185, "total_steps": 196, "loss": 0.6365, "accuracy": 0.6898437738418579, "lr": 9.918879275179817e-07, "epoch": 0.946896992962252, "percentage": 94.39, "elapsed_time": "1:08:13", "remaining_time": "0:04:03", "rewards/chosen": 0.21025450527668, "rewards/rejected": 0.06769368052482605, "rewards/accuracies": 0.6898437738418579, "rewards/margins": 0.14256080985069275, "logps/chosen": -374.5240173339844, "logps/rejected": -210.4625244140625, "logits/chosen": -2.249807596206665, "logits/rejected": -2.0986175537109375, "stability/response_length_mean": 486.3999938964844, "stability/response_length_std": 213.8624725341797, "stability/response_length_var": 45737.15625, "stability/token_entropy_mean": 6.559374809265137, "stability/repetition_rate_mean": 0.8755763173103333}
+{"current_steps": 190, "total_steps": 196, "loss": 0.6287, "accuracy": 0.701953113079071, "lr": 9.90896490997906e-07, "epoch": 0.9724888035828535, "percentage": 96.94, "elapsed_time": "1:10:04", "remaining_time": "0:02:12", "rewards/chosen": 0.2152232825756073, "rewards/rejected": 0.053037747740745544, "rewards/accuracies": 0.701953113079071, "rewards/margins": 0.16218551993370056, "logps/chosen": -377.6806640625, "logps/rejected": -206.34112548828125, "logits/chosen": -2.233643054962158, "logits/rejected": -2.0779027938842773, "stability/response_length_mean": 736.7999877929688, "stability/response_length_std": 246.57106018066406, "stability/response_length_var": 60797.2890625, "stability/token_entropy_mean": 6.7109375, "stability/repetition_rate_mean": 0.8903291821479797}
+{"current_steps": 195, "total_steps": 196, "loss": 0.6296, "accuracy": 0.7027343511581421, "lr": 9.898484648474362e-07, "epoch": 0.9980806142034548, "percentage": 99.49, "elapsed_time": "1:11:53", "remaining_time": "0:00:22", "rewards/chosen": 0.2173018455505371, "rewards/rejected": 0.054950814694166183, "rewards/accuracies": 0.7027343511581421, "rewards/margins": 0.16235104203224182, "logps/chosen": -361.8232116699219, "logps/rejected": -208.4629364013672, "logits/chosen": -2.250211477279663, "logits/rejected": -2.096865653991699, "stability/response_length_mean": 356.3999938964844, "stability/response_length_std": 207.94882202148438, "stability/response_length_var": 43242.7109375, "stability/token_entropy_mean": 5.928124904632568, "stability/repetition_rate_mean": 0.7688959836959839}
+{"current_steps": 196, "total_steps": 196, "epoch": 1.0, "percentage": 100.0, "elapsed_time": "1:12:08", "remaining_time": "0:00:00"}

mistral7b_dpo_baseline/trainer_state.json

@@ -0,0 +1,818 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 196,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.025591810620601407,
+      "grad_norm": 1.705037236213684,
+      "learning_rate": 4e-08,
+      "logits/chosen": -2.273033380508423,
+      "logits/rejected": -2.1234283447265625,
+      "logps/chosen": -375.063232421875,
+      "logps/rejected": -210.8657684326172,
+      "loss": 0.6931,
+      "rewards/accuracies": 0.3062500059604645,
+      "rewards/chosen": 0.0012258010683581233,
+      "rewards/margins": 0.0003632208681665361,
+      "rewards/rejected": 0.0008625800255686045,
+      "step": 5
+    },
+    {
+      "epoch": 0.05118362124120281,
+      "grad_norm": 1.766357421875,
+      "learning_rate": 9e-08,
+      "logits/chosen": -2.2679283618927,
+      "logits/rejected": -2.116572141647339,
+      "logps/chosen": -382.0296325683594,
+      "logps/rejected": -202.4493408203125,
+      "loss": 0.6931,
+      "rewards/accuracies": 0.5082031488418579,
+      "rewards/chosen": 0.0013576988130807877,
+      "rewards/margins": 0.0005467134760692716,
+      "rewards/rejected": 0.0008109854534268379,
+      "stability/repetition_rate_mean": 0.9135147929191589,
+      "stability/response_length_mean": 602.7999877929688,
+      "stability/response_length_std": 295.1842346191406,
+      "stability/response_length_var": 87133.734375,
+      "stability/token_entropy_mean": 6.959374904632568,
+      "step": 10
+    },
+    {
+      "epoch": 0.07677543186180422,
+      "grad_norm": 1.8122901916503906,
+      "learning_rate": 1.4e-07,
+      "logits/chosen": -2.260676860809326,
+      "logits/rejected": -2.0991530418395996,
+      "logps/chosen": -389.2082214355469,
+      "logps/rejected": -207.0491180419922,
+      "loss": 0.6932,
+      "rewards/accuracies": 0.5082031488418579,
+      "rewards/chosen": 0.0017401383956894279,
+      "rewards/margins": 0.00028905909857712686,
+      "rewards/rejected": 0.0014510790351778269,
+      "stability/repetition_rate_mean": 0.8363126516342163,
+      "stability/response_length_mean": 345.70001220703125,
+      "stability/response_length_std": 212.67974853515625,
+      "stability/response_length_var": 45232.6796875,
+      "stability/token_entropy_mean": 6.342187404632568,
+      "step": 15
+    },
+    {
+      "epoch": 0.10236724248240563,
+      "grad_norm": 1.7545065879821777,
+      "learning_rate": 1.8999999999999998e-07,
+      "logits/chosen": -2.2613608837127686,
+      "logits/rejected": -2.1145310401916504,
+      "logps/chosen": -368.71954345703125,
+      "logps/rejected": -198.5596160888672,
+      "loss": 0.6939,
+      "rewards/accuracies": 0.49140626192092896,
+      "rewards/chosen": 0.0004601169202942401,
+      "rewards/margins": -0.0010562256211414933,
+      "rewards/rejected": 0.0015163427451625466,
+      "stability/repetition_rate_mean": 0.9317001104354858,
+      "stability/response_length_mean": 739.4000244140625,
+      "stability/response_length_std": 281.282470703125,
+      "stability/response_length_var": 79119.8203125,
+      "stability/token_entropy_mean": 6.931250095367432,
+      "step": 20
+    },
+    {
+      "epoch": 0.12795905310300704,
+      "grad_norm": 1.731654167175293,
+      "learning_rate": 2.4e-07,
+      "logits/chosen": -2.277377128601074,
+      "logits/rejected": -2.125220775604248,
+      "logps/chosen": -378.1647033691406,
+      "logps/rejected": -205.86363220214844,
+      "loss": 0.693,
+      "rewards/accuracies": 0.5082031488418579,
+      "rewards/chosen": 0.002021136460825801,
+      "rewards/margins": 0.0006010310025885701,
+      "rewards/rejected": 0.0014201054582372308,
+      "stability/repetition_rate_mean": 0.8950179815292358,
+      "stability/response_length_mean": 658.5999755859375,
+      "stability/response_length_std": 288.0155029296875,
+      "stability/response_length_var": 82952.9296875,
+      "stability/token_entropy_mean": 6.821875095367432,
+      "step": 25
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 1.8716338872909546,
+      "learning_rate": 2.9e-07,
+      "logits/chosen": -2.2701575756073,
+      "logits/rejected": -2.1229705810546875,
+      "logps/chosen": -373.9202575683594,
+      "logps/rejected": -208.93287658691406,
+      "loss": 0.6935,
+      "rewards/accuracies": 0.490234375,
+      "rewards/chosen": 0.0010083412053063512,
+      "rewards/margins": -0.0003178715705871582,
+      "rewards/rejected": 0.0013262125430628657,
+      "stability/repetition_rate_mean": 0.8135284185409546,
+      "stability/response_length_mean": 356.0,
+      "stability/response_length_std": 249.6766815185547,
+      "stability/response_length_var": 62338.4453125,
+      "stability/token_entropy_mean": 6.199999809265137,
+      "step": 30
+    },
+    {
+      "epoch": 0.17914267434420986,
+      "grad_norm": 1.733734369277954,
+      "learning_rate": 3.4000000000000003e-07,
+      "logits/chosen": -2.2612318992614746,
+      "logits/rejected": -2.120094060897827,
+      "logps/chosen": -366.3136901855469,
+      "logps/rejected": -203.98719787597656,
+      "loss": 0.6928,
+      "rewards/accuracies": 0.5265624523162842,
+      "rewards/chosen": 0.0032140661496669054,
+      "rewards/margins": 0.0011429906589910388,
+      "rewards/rejected": 0.0020710753742605448,
+      "stability/repetition_rate_mean": 0.9289250373840332,
+      "stability/response_length_mean": 606.2000122070312,
+      "stability/response_length_std": 394.1366882324219,
+      "stability/response_length_var": 155343.734375,
+      "stability/token_entropy_mean": 7.112500190734863,
+      "step": 35
+    },
+    {
+      "epoch": 0.20473448496481125,
+      "grad_norm": 1.8034971952438354,
+      "learning_rate": 3.8999999999999997e-07,
+      "logits/chosen": -2.2764394283294678,
+      "logits/rejected": -2.122685432434082,
+      "logps/chosen": -369.1784973144531,
+      "logps/rejected": -197.37686157226562,
+      "loss": 0.6928,
+      "rewards/accuracies": 0.5144531726837158,
+      "rewards/chosen": 0.004107568878680468,
+      "rewards/margins": 0.0011615146650001407,
+      "rewards/rejected": 0.0029460543300956488,
+      "stability/repetition_rate_mean": 0.8629743456840515,
+      "stability/response_length_mean": 547.2000122070312,
+      "stability/response_length_std": 290.8897399902344,
+      "stability/response_length_var": 84616.84375,
+      "stability/token_entropy_mean": 6.557812690734863,
+      "step": 40
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 1.8129643201828003,
+      "learning_rate": 4.3999999999999997e-07,
+      "logits/chosen": -2.257107734680176,
+      "logits/rejected": -2.1072099208831787,
+      "logps/chosen": -385.1796569824219,
+      "logps/rejected": -212.56887817382812,
+      "loss": 0.6933,
+      "rewards/accuracies": 0.512890636920929,
+      "rewards/chosen": 0.0039331079460680485,
+      "rewards/margins": 3.291945904493332e-05,
+      "rewards/rejected": 0.003900188487023115,
+      "stability/repetition_rate_mean": 0.8764154314994812,
+      "stability/response_length_mean": 366.5,
+      "stability/response_length_std": 118.86243438720703,
+      "stability/response_length_var": 14128.27734375,
+      "stability/token_entropy_mean": 6.640625,
+      "step": 45
+    },
+    {
+      "epoch": 0.2559181062060141,
+      "grad_norm": 1.8471362590789795,
+      "learning_rate": 4.9e-07,
+      "logits/chosen": -2.264720916748047,
+      "logits/rejected": -2.109816312789917,
+      "logps/chosen": -378.2891540527344,
+      "logps/rejected": -207.4978790283203,
+      "loss": 0.6916,
+      "rewards/accuracies": 0.5316406488418579,
+      "rewards/chosen": 0.0068646748550236225,
+      "rewards/margins": 0.0035711233504116535,
+      "rewards/rejected": 0.003293551504611969,
+      "stability/repetition_rate_mean": 0.8785834312438965,
+      "stability/response_length_mean": 384.1000061035156,
+      "stability/response_length_std": 178.4809112548828,
+      "stability/response_length_var": 31855.43359375,
+      "stability/token_entropy_mean": 6.543749809265137,
+      "step": 50
+    },
+    {
+      "epoch": 0.28150991682661547,
+      "grad_norm": 1.940763235092163,
+      "learning_rate": 5.4e-07,
+      "logits/chosen": -2.2527079582214355,
+      "logits/rejected": -2.094048261642456,
+      "logps/chosen": -393.7845764160156,
+      "logps/rejected": -210.35861206054688,
+      "loss": 0.6911,
+      "rewards/accuracies": 0.536328136920929,
+      "rewards/chosen": 0.009546511806547642,
+      "rewards/margins": 0.004484717734158039,
+      "rewards/rejected": 0.005061793606728315,
+      "stability/repetition_rate_mean": 0.9481571912765503,
+      "stability/response_length_mean": 332.8999938964844,
+      "stability/response_length_std": 137.39193725585938,
+      "stability/response_length_var": 18876.544921875,
+      "stability/token_entropy_mean": 7.009375095367432,
+      "step": 55
+    },
+    {
+      "epoch": 0.30710172744721687,
+      "grad_norm": 1.8226360082626343,
+      "learning_rate": 5.9e-07,
+      "logits/chosen": -2.272801160812378,
+      "logits/rejected": -2.111156463623047,
+      "logps/chosen": -387.7115173339844,
+      "logps/rejected": -206.819580078125,
+      "loss": 0.6915,
+      "rewards/accuracies": 0.5328124761581421,
+      "rewards/chosen": 0.009509808383882046,
+      "rewards/margins": 0.00375912943854928,
+      "rewards/rejected": 0.005750678479671478,
+      "stability/repetition_rate_mean": 0.8699707984924316,
+      "stability/response_length_mean": 655.2999877929688,
+      "stability/response_length_std": 282.005126953125,
+      "stability/response_length_var": 79526.8984375,
+      "stability/token_entropy_mean": 6.559374809265137,
+      "step": 60
+    },
+    {
+      "epoch": 0.3326935380678183,
+      "grad_norm": 1.6994986534118652,
+      "learning_rate": 6.4e-07,
+      "logits/chosen": -2.2521181106567383,
+      "logits/rejected": -2.098989963531494,
+      "logps/chosen": -377.4696960449219,
+      "logps/rejected": -204.3833465576172,
+      "loss": 0.6908,
+      "rewards/accuracies": 0.5625000596046448,
+      "rewards/chosen": 0.012991372495889664,
+      "rewards/margins": 0.00506935128942132,
+      "rewards/rejected": 0.007922019809484482,
+      "stability/repetition_rate_mean": 0.9148252606391907,
+      "stability/response_length_mean": 306.3999938964844,
+      "stability/response_length_std": 141.7236785888672,
+      "stability/response_length_var": 20085.599609375,
+      "stability/token_entropy_mean": 6.840624809265137,
+      "step": 65
+    },
+    {
+      "epoch": 0.3582853486884197,
+      "grad_norm": 2.000992774963379,
+      "learning_rate": 6.9e-07,
+      "logits/chosen": -2.2713451385498047,
+      "logits/rejected": -2.106651782989502,
+      "logps/chosen": -393.73406982421875,
+      "logps/rejected": -206.1870574951172,
+      "loss": 0.6888,
+      "rewards/accuracies": 0.5777344107627869,
+      "rewards/chosen": 0.017473159357905388,
+      "rewards/margins": 0.009205814450979233,
+      "rewards/rejected": 0.008267344906926155,
+      "stability/repetition_rate_mean": 0.9625760316848755,
+      "stability/response_length_mean": 472.70001220703125,
+      "stability/response_length_std": 159.89791870117188,
+      "stability/response_length_var": 25567.34375,
+      "stability/token_entropy_mean": 7.162499904632568,
+      "step": 70
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 1.8905284404754639,
+      "learning_rate": 7.4e-07,
+      "logits/chosen": -2.275454521179199,
+      "logits/rejected": -2.116396903991699,
+      "logps/chosen": -381.0571594238281,
+      "logps/rejected": -207.9349365234375,
+      "loss": 0.689,
+      "rewards/accuracies": 0.577343761920929,
+      "rewards/chosen": 0.018993159756064415,
+      "rewards/margins": 0.008799022994935513,
+      "rewards/rejected": 0.010194134898483753,
+      "stability/repetition_rate_mean": 0.9068478345870972,
+      "stability/response_length_mean": 527.0,
+      "stability/response_length_std": 293.7500915527344,
+      "stability/response_length_var": 86289.109375,
+      "stability/token_entropy_mean": 6.139062404632568,
+      "step": 75
+    },
+    {
+      "epoch": 0.4094689699296225,
+      "grad_norm": 1.8269047737121582,
+      "learning_rate": 7.9e-07,
+      "logits/chosen": -2.2604687213897705,
+      "logits/rejected": -2.1025872230529785,
+      "logps/chosen": -382.4193420410156,
+      "logps/rejected": -208.97756958007812,
+      "loss": 0.6886,
+      "rewards/accuracies": 0.5726562738418579,
+      "rewards/chosen": 0.021377386525273323,
+      "rewards/margins": 0.009593295864760876,
+      "rewards/rejected": 0.011784089729189873,
+      "stability/repetition_rate_mean": 0.9464001655578613,
+      "stability/response_length_mean": 635.5,
+      "stability/response_length_std": 262.53900146484375,
+      "stability/response_length_var": 68926.71875,
+      "stability/token_entropy_mean": 7.199999809265137,
+      "step": 80
+    },
+    {
+      "epoch": 0.4350607805502239,
+      "grad_norm": 1.8734734058380127,
+      "learning_rate": 8.399999999999999e-07,
+      "logits/chosen": -2.258051872253418,
+      "logits/rejected": -2.100944757461548,
+      "logps/chosen": -387.3235778808594,
+      "logps/rejected": -208.6320343017578,
+      "loss": 0.6875,
+      "rewards/accuracies": 0.5863280892372131,
+      "rewards/chosen": 0.026788339018821716,
+      "rewards/margins": 0.012103421613574028,
+      "rewards/rejected": 0.014684918336570263,
+      "stability/repetition_rate_mean": 0.9034587144851685,
+      "stability/response_length_mean": 600.0999755859375,
+      "stability/response_length_std": 349.2030334472656,
+      "stability/response_length_var": 121942.765625,
+      "stability/token_entropy_mean": 6.737500190734863,
+      "step": 85
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 1.8927897214889526,
+      "learning_rate": 8.9e-07,
+      "logits/chosen": -2.261427164077759,
+      "logits/rejected": -2.1134793758392334,
+      "logps/chosen": -375.204345703125,
+      "logps/rejected": -206.2209930419922,
+      "loss": 0.6859,
+      "rewards/accuracies": 0.608593761920929,
+      "rewards/chosen": 0.032174669206142426,
+      "rewards/margins": 0.015198366716504097,
+      "rewards/rejected": 0.01697630062699318,
+      "stability/repetition_rate_mean": 0.9765706062316895,
+      "stability/response_length_mean": 692.9000244140625,
+      "stability/response_length_std": 127.31452941894531,
+      "stability/response_length_var": 16208.9892578125,
+      "stability/token_entropy_mean": 7.109375,
+      "step": 90
+    },
+    {
+      "epoch": 0.48624440179142675,
+      "grad_norm": 1.8181637525558472,
+      "learning_rate": 9.399999999999999e-07,
+      "logits/chosen": -2.2588284015655518,
+      "logits/rejected": -2.112243175506592,
+      "logps/chosen": -366.8333740234375,
+      "logps/rejected": -203.4322967529297,
+      "loss": 0.6838,
+      "rewards/accuracies": 0.6300780773162842,
+      "rewards/chosen": 0.03762410953640938,
+      "rewards/margins": 0.01969875767827034,
+      "rewards/rejected": 0.01792534813284874,
+      "stability/repetition_rate_mean": 0.9483324289321899,
+      "stability/response_length_mean": 727.0,
+      "stability/response_length_std": 335.0764465332031,
+      "stability/response_length_var": 112276.21875,
+      "stability/token_entropy_mean": 7.3125,
+      "step": 95
+    },
+    {
+      "epoch": 0.5118362124120281,
+      "grad_norm": 1.8130452632904053,
+      "learning_rate": 9.9e-07,
+      "logits/chosen": -2.27445387840271,
+      "logits/rejected": -2.11898136138916,
+      "logps/chosen": -384.503662109375,
+      "logps/rejected": -210.62109375,
+      "loss": 0.6821,
+      "rewards/accuracies": 0.6527343988418579,
+      "rewards/chosen": 0.045418523252010345,
+      "rewards/margins": 0.02338458225131035,
+      "rewards/rejected": 0.0220339372754097,
+      "stability/repetition_rate_mean": 0.9498335123062134,
+      "stability/response_length_mean": 419.0,
+      "stability/response_length_std": 232.0225067138672,
+      "stability/response_length_var": 53834.4453125,
+      "stability/token_entropy_mean": 7.074999809265137,
+      "step": 100
+    },
+    {
+      "epoch": 0.5374280230326296,
+      "grad_norm": 1.743505597114563,
+      "learning_rate": 9.999815554351224e-07,
+      "logits/chosen": -2.258531093597412,
+      "logits/rejected": -2.108144998550415,
+      "logps/chosen": -378.0056457519531,
+      "logps/rejected": -209.6881103515625,
+      "loss": 0.6811,
+      "rewards/accuracies": 0.626953125,
+      "rewards/chosen": 0.052092913538217545,
+      "rewards/margins": 0.025670701637864113,
+      "rewards/rejected": 0.02642221562564373,
+      "stability/repetition_rate_mean": 0.901488184928894,
+      "stability/response_length_mean": 669.5,
+      "stability/response_length_std": 271.60888671875,
+      "stability/response_length_var": 73771.390625,
+      "stability/token_entropy_mean": 6.564843654632568,
+      "step": 105
+    },
+    {
+      "epoch": 0.5630198336532309,
+      "grad_norm": 1.6861857175827026,
+      "learning_rate": 9.999066267225444e-07,
+      "logits/chosen": -2.2750279903411865,
+      "logits/rejected": -2.117339611053467,
+      "logps/chosen": -385.2655029296875,
+      "logps/rejected": -206.72357177734375,
+      "loss": 0.6795,
+      "rewards/accuracies": 0.6519531607627869,
+      "rewards/chosen": 0.059595055878162384,
+      "rewards/margins": 0.029159152880311012,
+      "rewards/rejected": 0.03043590858578682,
+      "stability/repetition_rate_mean": 0.9347068071365356,
+      "stability/response_length_mean": 255.60000610351562,
+      "stability/response_length_std": 114.30874633789062,
+      "stability/response_length_var": 13066.4892578125,
+      "stability/token_entropy_mean": 7.015625,
+      "step": 110
+    },
+    {
+      "epoch": 0.5886116442738324,
+      "grad_norm": 1.7960835695266724,
+      "learning_rate": 9.997740697079592e-07,
+      "logits/chosen": -2.251587390899658,
+      "logits/rejected": -2.110828399658203,
+      "logps/chosen": -372.7311096191406,
+      "logps/rejected": -213.94320678710938,
+      "loss": 0.6763,
+      "rewards/accuracies": 0.6519531607627869,
+      "rewards/chosen": 0.07092021405696869,
+      "rewards/margins": 0.036303263157606125,
+      "rewards/rejected": 0.034616947174072266,
+      "stability/repetition_rate_mean": 0.8279488682746887,
+      "stability/response_length_mean": 499.29998779296875,
+      "stability/response_length_std": 282.31072998046875,
+      "stability/response_length_var": 79699.34375,
+      "stability/token_entropy_mean": 6.553124904632568,
+      "step": 115
+    },
+    {
+      "epoch": 0.6142034548944337,
+      "grad_norm": 1.773605227470398,
+      "learning_rate": 9.995838996722914e-07,
+      "logits/chosen": -2.2482666969299316,
+      "logits/rejected": -2.102483034133911,
+      "logps/chosen": -367.7881164550781,
+      "logps/rejected": -207.59573364257812,
+      "loss": 0.6733,
+      "rewards/accuracies": 0.6703125238418579,
+      "rewards/chosen": 0.08155383914709091,
+      "rewards/margins": 0.04288302734494209,
+      "rewards/rejected": 0.03867081552743912,
+      "stability/repetition_rate_mean": 0.9149526357650757,
+      "stability/response_length_mean": 363.6000061035156,
+      "stability/response_length_std": 146.0701446533203,
+      "stability/response_length_var": 21336.48828125,
+      "stability/token_entropy_mean": 6.8125,
+      "step": 120
+    },
+    {
+      "epoch": 0.6397952655150352,
+      "grad_norm": 1.7966293096542358,
+      "learning_rate": 9.993361385379874e-07,
+      "logits/chosen": -2.253580093383789,
+      "logits/rejected": -2.0954601764678955,
+      "logps/chosen": -385.7173767089844,
+      "logps/rejected": -205.1151123046875,
+      "loss": 0.6707,
+      "rewards/accuracies": 0.6781249642372131,
+      "rewards/chosen": 0.09179403632879257,
+      "rewards/margins": 0.04949168488383293,
+      "rewards/rejected": 0.04230235144495964,
+      "stability/repetition_rate_mean": 0.8946477770805359,
+      "stability/response_length_mean": 433.79998779296875,
+      "stability/response_length_std": 126.92412567138672,
+      "stability/response_length_var": 16109.7333984375,
+      "stability/token_entropy_mean": 6.532812595367432,
+      "step": 125
+    },
+    {
+      "epoch": 0.6653870761356366,
+      "grad_norm": 1.873657464981079,
+      "learning_rate": 9.99030814866488e-07,
+      "logits/chosen": -2.2453463077545166,
+      "logits/rejected": -2.096496105194092,
+      "logps/chosen": -375.60565185546875,
+      "logps/rejected": -203.76229858398438,
+      "loss": 0.6672,
+      "rewards/accuracies": 0.682421863079071,
+      "rewards/chosen": 0.09893052279949188,
+      "rewards/margins": 0.057060565799474716,
+      "rewards/rejected": 0.041869960725307465,
+      "stability/repetition_rate_mean": 0.8153165578842163,
+      "stability/response_length_mean": 296.1000061035156,
+      "stability/response_length_std": 167.77593994140625,
+      "stability/response_length_var": 28148.767578125,
+      "stability/token_entropy_mean": 6.018750190734863,
+      "step": 130
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 1.7402511835098267,
+      "learning_rate": 9.986679638549358e-07,
+      "logits/chosen": -2.2670061588287354,
+      "logits/rejected": -2.1075525283813477,
+      "logps/chosen": -387.3724060058594,
+      "logps/rejected": -213.73989868164062,
+      "loss": 0.6645,
+      "rewards/accuracies": 0.676953136920929,
+      "rewards/chosen": 0.11148682981729507,
+      "rewards/margins": 0.06334061175584793,
+      "rewards/rejected": 0.048146218061447144,
+      "stability/repetition_rate_mean": 0.8604398965835571,
+      "stability/response_length_mean": 596.4000244140625,
+      "stability/response_length_std": 291.9129333496094,
+      "stability/response_length_var": 85213.15625,
+      "stability/token_entropy_mean": 6.510937690734863,
+      "step": 135
+    },
+    {
+      "epoch": 0.7165706973768394,
+      "grad_norm": 1.8281584978103638,
+      "learning_rate": 9.982476273321175e-07,
+      "logits/chosen": -2.25555157661438,
+      "logits/rejected": -2.100205898284912,
+      "logps/chosen": -389.03497314453125,
+      "logps/rejected": -211.17449951171875,
+      "loss": 0.6611,
+      "rewards/accuracies": 0.6976562738418579,
+      "rewards/chosen": 0.11906532198190689,
+      "rewards/margins": 0.07203800976276398,
+      "rewards/rejected": 0.04702730476856232,
+      "stability/repetition_rate_mean": 0.8607560396194458,
+      "stability/response_length_mean": 173.89999389648438,
+      "stability/response_length_std": 70.74908447265625,
+      "stability/response_length_var": 5005.43310546875,
+      "stability/token_entropy_mean": 6.612500190734863,
+      "step": 140
+    },
+    {
+      "epoch": 0.7421625079974408,
+      "grad_norm": 1.7499624490737915,
+      "learning_rate": 9.977698537536417e-07,
+      "logits/chosen": -2.2664294242858887,
+      "logits/rejected": -2.0990548133850098,
+      "logps/chosen": -387.792724609375,
+      "logps/rejected": -203.63430786132812,
+      "loss": 0.6589,
+      "rewards/accuracies": 0.678906261920929,
+      "rewards/chosen": 0.13202722370624542,
+      "rewards/margins": 0.07740288227796555,
+      "rewards/rejected": 0.054624345153570175,
+      "stability/repetition_rate_mean": 0.9638816714286804,
+      "stability/response_length_mean": 720.2000122070312,
+      "stability/response_length_std": 284.2842102050781,
+      "stability/response_length_var": 80817.5078125,
+      "stability/token_entropy_mean": 7.106249809265137,
+      "step": 145
+    },
+    {
+      "epoch": 0.7677543186180422,
+      "grad_norm": 1.8286625146865845,
+      "learning_rate": 9.972346981963546e-07,
+      "logits/chosen": -2.2344398498535156,
+      "logits/rejected": -2.0921382904052734,
+      "logps/chosen": -373.849853515625,
+      "logps/rejected": -211.25674438476562,
+      "loss": 0.6568,
+      "rewards/accuracies": 0.6703124642372131,
+      "rewards/chosen": 0.14155012369155884,
+      "rewards/margins": 0.08604539185762405,
+      "rewards/rejected": 0.05550474300980568,
+      "stability/repetition_rate_mean": 0.8590916395187378,
+      "stability/response_length_mean": 654.2000122070312,
+      "stability/response_length_std": 372.5482177734375,
+      "stability/response_length_var": 138792.171875,
+      "stability/token_entropy_mean": 6.5859375,
+      "step": 150
+    },
+    {
+      "epoch": 0.7933461292386437,
+      "grad_norm": 1.8375192880630493,
+      "learning_rate": 9.966422223519885e-07,
+      "logits/chosen": -2.259735107421875,
+      "logits/rejected": -2.108813762664795,
+      "logps/chosen": -377.39886474609375,
+      "logps/rejected": -206.43038940429688,
+      "loss": 0.6528,
+      "rewards/accuracies": 0.682421863079071,
+      "rewards/chosen": 0.1565590798854828,
+      "rewards/margins": 0.09292585402727127,
+      "rewards/rejected": 0.06363321840763092,
+      "stability/repetition_rate_mean": 0.8765794038772583,
+      "stability/response_length_mean": 502.3999938964844,
+      "stability/response_length_std": 200.0795440673828,
+      "stability/response_length_var": 40031.8203125,
+      "stability/token_entropy_mean": 6.557812690734863,
+      "step": 155
+    },
+    {
+      "epoch": 0.818937939859245,
+      "grad_norm": 1.7681745290756226,
+      "learning_rate": 9.959924945200523e-07,
+      "logits/chosen": -2.2527999877929688,
+      "logits/rejected": -2.086219549179077,
+      "logps/chosen": -383.5099182128906,
+      "logps/rejected": -207.28614807128906,
+      "loss": 0.6491,
+      "rewards/accuracies": 0.6898438334465027,
+      "rewards/chosen": 0.16489063203334808,
+      "rewards/margins": 0.10477059334516525,
+      "rewards/rejected": 0.06012001633644104,
+      "stability/repetition_rate_mean": 0.9663667678833008,
+      "stability/response_length_mean": 870.4000244140625,
+      "stability/response_length_std": 147.6581573486328,
+      "stability/response_length_var": 21802.93359375,
+      "stability/token_entropy_mean": 7.243750095367432,
+      "step": 160
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.6438355445861816,
+      "learning_rate": 9.952855895999568e-07,
+      "logits/chosen": -2.23742413520813,
+      "logits/rejected": -2.089754104614258,
+      "logps/chosen": -362.43914794921875,
+      "logps/rejected": -203.66177368164062,
+      "loss": 0.6477,
+      "rewards/accuracies": 0.684374988079071,
+      "rewards/chosen": 0.1636190414428711,
+      "rewards/margins": 0.10717189311981201,
+      "rewards/rejected": 0.05644712969660759,
+      "stability/repetition_rate_mean": 0.8620624542236328,
+      "stability/response_length_mean": 547.7000122070312,
+      "stability/response_length_std": 249.60325622558594,
+      "stability/response_length_var": 62301.7890625,
+      "stability/token_entropy_mean": 6.599999904632568,
+      "step": 165
+    },
+    {
+      "epoch": 0.8701215611004478,
+      "grad_norm": 1.7001264095306396,
+      "learning_rate": 9.9452158908238e-07,
+      "logits/chosen": -2.2632334232330322,
+      "logits/rejected": -2.097275733947754,
+      "logps/chosen": -376.97039794921875,
+      "logps/rejected": -203.81222534179688,
+      "loss": 0.6419,
+      "rewards/accuracies": 0.705078125,
+      "rewards/chosen": 0.18040212988853455,
+      "rewards/margins": 0.12267370522022247,
+      "rewards/rejected": 0.057728420943021774,
+      "stability/repetition_rate_mean": 0.9189292192459106,
+      "stability/response_length_mean": 385.70001220703125,
+      "stability/response_length_std": 143.96685791015625,
+      "stability/response_length_var": 20726.455078125,
+      "stability/token_entropy_mean": 6.831250190734863,
+      "step": 170
+    },
+    {
+      "epoch": 0.8957133717210493,
+      "grad_norm": 1.804345726966858,
+      "learning_rate": 9.937005810398745e-07,
+      "logits/chosen": -2.2485477924346924,
+      "logits/rejected": -2.0931143760681152,
+      "logps/chosen": -380.5006408691406,
+      "logps/rejected": -212.2974853515625,
+      "loss": 0.6396,
+      "rewards/accuracies": 0.6957032084465027,
+      "rewards/chosen": 0.18881921470165253,
+      "rewards/margins": 0.13226722180843353,
+      "rewards/rejected": 0.0565519854426384,
+      "stability/repetition_rate_mean": 0.9413634538650513,
+      "stability/response_length_mean": 460.6000061035156,
+      "stability/response_length_std": 165.9894256591797,
+      "stability/response_length_var": 27552.48828125,
+      "stability/token_entropy_mean": 6.959374904632568,
+      "step": 175
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 1.76821768283844,
+      "learning_rate": 9.928226601167138e-07,
+      "logits/chosen": -2.2474441528320312,
+      "logits/rejected": -2.0858802795410156,
+      "logps/chosen": -387.0555419921875,
+      "logps/rejected": -212.9912109375,
+      "loss": 0.6377,
+      "rewards/accuracies": 0.690234363079071,
+      "rewards/chosen": 0.20255513489246368,
+      "rewards/margins": 0.13533423840999603,
+      "rewards/rejected": 0.06722089648246765,
+      "stability/repetition_rate_mean": 0.8933365941047668,
+      "stability/response_length_mean": 716.2000122070312,
+      "stability/response_length_std": 329.3815002441406,
+      "stability/response_length_var": 108492.1796875,
+      "stability/token_entropy_mean": 6.526562690734863,
+      "step": 180
+    },
+    {
+      "epoch": 0.946896992962252,
+      "grad_norm": 1.6571131944656372,
+      "learning_rate": 9.918879275179817e-07,
+      "logits/chosen": -2.249807596206665,
+      "logits/rejected": -2.0986175537109375,
+      "logps/chosen": -374.5240173339844,
+      "logps/rejected": -210.4625244140625,
+      "loss": 0.6365,
+      "rewards/accuracies": 0.6898437738418579,
+      "rewards/chosen": 0.21025450527668,
+      "rewards/margins": 0.14256080985069275,
+      "rewards/rejected": 0.06769368052482605,
+      "stability/repetition_rate_mean": 0.8755763173103333,
+      "stability/response_length_mean": 486.3999938964844,
+      "stability/response_length_std": 213.8624725341797,
+      "stability/response_length_var": 45737.15625,
+      "stability/token_entropy_mean": 6.559374809265137,
+      "step": 185
+    },
+    {
+      "epoch": 0.9724888035828535,
+      "grad_norm": 1.5521091222763062,
+      "learning_rate": 9.90896490997906e-07,
+      "logits/chosen": -2.233643054962158,
+      "logits/rejected": -2.0779027938842773,
+      "logps/chosen": -377.6806640625,
+      "logps/rejected": -206.34112548828125,
+      "loss": 0.6287,
+      "rewards/accuracies": 0.701953113079071,
+      "rewards/chosen": 0.2152232825756073,
+      "rewards/margins": 0.16218551993370056,
+      "rewards/rejected": 0.053037747740745544,
+      "stability/repetition_rate_mean": 0.8903291821479797,
+      "stability/response_length_mean": 736.7999877929688,
+      "stability/response_length_std": 246.57106018066406,
+      "stability/response_length_var": 60797.2890625,
+      "stability/token_entropy_mean": 6.7109375,
+      "step": 190
+    },
+    {
+      "epoch": 0.9980806142034548,
+      "grad_norm": 1.6728918552398682,
+      "learning_rate": 9.898484648474362e-07,
+      "logits/chosen": -2.250211477279663,
+      "logits/rejected": -2.096865653991699,
+      "logps/chosen": -361.8232116699219,
+      "logps/rejected": -208.4629364013672,
+      "loss": 0.6296,
+      "rewards/accuracies": 0.7027343511581421,
+      "rewards/chosen": 0.2173018455505371,
+      "rewards/margins": 0.16235104203224182,
+      "rewards/rejected": 0.054950814694166183,
+      "stability/repetition_rate_mean": 0.7688959836959839,
+      "stability/response_length_mean": 356.3999938964844,
+      "stability/response_length_std": 207.94882202148438,
+      "stability/response_length_var": 43242.7109375,
+      "stability/token_entropy_mean": 5.928124904632568,
+      "step": 195
+    },
+    {
+      "epoch": 1.0,
+      "step": 196,
+      "total_flos": 7.072674331683193e+18,
+      "train_loss": 0.6730708850889789,
+      "train_runtime": 4328.1955,
+      "train_samples_per_second": 23.104,
+      "train_steps_per_second": 0.045
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 196,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 200,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.072674331683193e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

mistral7b_dpo_baseline/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94e6d5df42ddf678581bc105236846c8bb9630da14f49c0e71f40b3a546b04ca
+size 8017