Upload folder using huggingface_hub

mistral7b_dpo_sft_ftx0.001/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_sft_ftx0.001
+  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_sft_ftx0.001
+
+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 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_sft_ftx0.001/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",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "up_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_sft_ftx0.001/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adba71763a0b978112916236cdb4683d914754109a626707cbd40efa6ea76b6c
+size 42002584

mistral7b_dpo_sft_ftx0.001/all_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "total_flos": 7.072674331683193e+18,
+    "train_loss": 0.6755896362723136,
+    "train_runtime": 4339.5435,
+    "train_samples_per_second": 23.044,
+    "train_steps_per_second": 0.045
+}

mistral7b_dpo_sft_ftx0.001/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_sft_ftx0.001/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_sft_ftx0.001/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",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "up_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_sft_ftx0.001/checkpoint-196/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adba71763a0b978112916236cdb4683d914754109a626707cbd40efa6ea76b6c
+size 42002584

mistral7b_dpo_sft_ftx0.001/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_sft_ftx0.001/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 sha256:6bd4e93946e86264d5dd6b28fee5c35f4770e24a5d633cd386865f70be8df73b
+size 31467333

mistral7b_dpo_sft_ftx0.001/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 sha256:98334fadad056bb02ac22a100a47a8d7c38d5e40c749559c43f192a7688bab62
+size 31467333

mistral7b_dpo_sft_ftx0.001/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 sha256:47075113ef694ca288b7a6d7b192acdc98eebc2a365a31ff721557ec5ff0484c
+size 31467333

mistral7b_dpo_sft_ftx0.001/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:e59c077d72e3458f724f775cf352a941e3354c86613de58ab6a570c5d5f8a1fb
+size 31467397

mistral7b_dpo_sft_ftx0.001/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 sha256:8e744c5e8e117d0e7b76a98e4a36cf42da1ac03fdb0536b51b308d9f80d0480c
+size 31467397

mistral7b_dpo_sft_ftx0.001/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 sha256:0bec1daf4f7a90ef1490b065b1eddae61cbc2e372688fa6318e6969710b2b36e
+size 31467397

mistral7b_dpo_sft_ftx0.001/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 sha256:9abd02ecc58fe63e7d053da362f8e3c489c836160c21fb2639cb9c1cf3904bbf
+size 31467397

mistral7b_dpo_sft_ftx0.001/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:d8851beb40a892501459f1be00515dcf008a4a1cf64970bb37c8b339aa06ff84
+size 31467397

mistral7b_dpo_sft_ftx0.001/checkpoint-196/global_step196/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07c75626d6e6b7277fb0431ab9f4b58fccc018374397464a8c7df636f2966258
+size 42288409

mistral7b_dpo_sft_ftx0.001/checkpoint-196/latest

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

mistral7b_dpo_sft_ftx0.001/checkpoint-196/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7efbc3b39e557ed047a255851058c3d62216b6bc5794c19289879cf099f52035
+size 16389

mistral7b_dpo_sft_ftx0.001/checkpoint-196/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce29a8767a7d907dd24987aa2c3e654d4317f3042fbc13b5b72cadb46d43311a
+size 16389

mistral7b_dpo_sft_ftx0.001/checkpoint-196/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61a48db011646b4e9a867bf12f4a233cad5dfbfe309686f8996c250196d3783a
+size 16389

mistral7b_dpo_sft_ftx0.001/checkpoint-196/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9562ee822472a4f01dcd6349ab3d1ef42a48915fe3b92e843a0c37db53c8421
+size 16389

mistral7b_dpo_sft_ftx0.001/checkpoint-196/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7d2767d83c3bf27f12db022b0632e2c4f8c164274ba75e380cf18f9d5f21819
+size 16389

mistral7b_dpo_sft_ftx0.001/checkpoint-196/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76816358d4e5db8149d60d55234db658d67a13c0c1ce05d7404cf7125a676a5c
+size 16389

mistral7b_dpo_sft_ftx0.001/checkpoint-196/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1562e7520c977d178183d641f70abcf3f57da2489938756cfbebf9b6e6c1a9fd
+size 16389

mistral7b_dpo_sft_ftx0.001/checkpoint-196/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6b6cabaed045c5398cd1b732f7ec48bd363f3b43cd24e0e70e641a42bd00c28
+size 16389

mistral7b_dpo_sft_ftx0.001/checkpoint-196/scheduler.pt

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

mistral7b_dpo_sft_ftx0.001/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_sft_ftx0.001/checkpoint-196/tokenizer.model

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

mistral7b_dpo_sft_ftx0.001/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_sft_ftx0.001/checkpoint-196/trainer_state.json

@@ -0,0 +1,848 @@
+{
+  "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": 1.70222008228302,
+      "learning_rate": 4e-08,
+      "logits/chosen": -2.273049831390381,
+      "logits/rejected": -2.1233928203582764,
+      "logps/chosen": -375.07659912109375,
+      "logps/rejected": -210.87295532226562,
+      "loss": 0.6959,
+      "metrics/advantage_var": 0.04196161404252052,
+      "rewards/accuracies": 0.29609373211860657,
+      "rewards/chosen": -0.00011618183634709567,
+      "rewards/margins": -0.00025905348593369126,
+      "rewards/rejected": 0.00014287166413851082,
+      "step": 5
+    },
+    {
+      "epoch": 0.05118362124120281,
+      "grad_norm": 1.7635924816131592,
+      "learning_rate": 9e-08,
+      "logits/chosen": -2.267812967300415,
+      "logits/rejected": -2.1164932250976562,
+      "logps/chosen": -382.0225830078125,
+      "logps/rejected": -202.45101928710938,
+      "loss": 0.6951,
+      "metrics/advantage_var": 0.07391166687011719,
+      "rewards/accuracies": 0.5093749761581421,
+      "rewards/chosen": 0.0020605388563126326,
+      "rewards/margins": 0.001415093895047903,
+      "rewards/rejected": 0.0006454450776800513,
+      "stability/repetition_rate_mean": 0.9129446148872375,
+      "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": 1.8181178569793701,
+      "learning_rate": 1.4e-07,
+      "logits/chosen": -2.2609243392944336,
+      "logits/rejected": -2.099472761154175,
+      "logps/chosen": -389.2162170410156,
+      "logps/rejected": -207.0449981689453,
+      "loss": 0.6963,
+      "metrics/advantage_var": 0.07241873443126678,
+      "rewards/accuracies": 0.4898437559604645,
+      "rewards/chosen": 0.000939845631364733,
+      "rewards/margins": -0.0009244689717888832,
+      "rewards/rejected": 0.0018643148941919208,
+      "stability/repetition_rate_mean": 0.8466346859931946,
+      "stability/response_length_mean": 345.70001220703125,
+      "stability/response_length_std": 212.67974853515625,
+      "stability/response_length_var": 45232.6796875,
+      "stability/token_entropy_mean": 6.3515625,
+      "step": 15
+    },
+    {
+      "epoch": 0.10236724248240563,
+      "grad_norm": 1.7609531879425049,
+      "learning_rate": 1.8999999999999998e-07,
+      "logits/chosen": -2.2611491680145264,
+      "logits/rejected": -2.1143441200256348,
+      "logps/chosen": -368.7252502441406,
+      "logps/rejected": -198.56080627441406,
+      "loss": 0.6966,
+      "metrics/advantage_var": 0.06753375381231308,
+      "rewards/accuracies": 0.49375006556510925,
+      "rewards/chosen": -0.00011375403846614063,
+      "rewards/margins": -0.0015103723853826523,
+      "rewards/rejected": 0.001396618434228003,
+      "stability/repetition_rate_mean": 0.9300774335861206,
+      "stability/response_length_mean": 739.4000244140625,
+      "stability/response_length_std": 281.282470703125,
+      "stability/response_length_var": 79119.8203125,
+      "stability/token_entropy_mean": 6.934374809265137,
+      "step": 20
+    },
+    {
+      "epoch": 0.12795905310300704,
+      "grad_norm": 1.733801245689392,
+      "learning_rate": 2.4e-07,
+      "logits/chosen": -2.277261972427368,
+      "logits/rejected": -2.125120162963867,
+      "logps/chosen": -378.173828125,
+      "logps/rejected": -205.8581085205078,
+      "loss": 0.6962,
+      "metrics/advantage_var": 0.07016502320766449,
+      "rewards/accuracies": 0.49179691076278687,
+      "rewards/chosen": 0.0011066627921536565,
+      "rewards/margins": -0.0008675385033711791,
+      "rewards/rejected": 0.0019742015283554792,
+      "stability/repetition_rate_mean": 0.9150856137275696,
+      "stability/response_length_mean": 658.5999755859375,
+      "stability/response_length_std": 288.0155029296875,
+      "stability/response_length_var": 82952.9296875,
+      "stability/token_entropy_mean": 7.012499809265137,
+      "step": 25
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 1.8615390062332153,
+      "learning_rate": 2.9e-07,
+      "logits/chosen": -2.2703747749328613,
+      "logits/rejected": -2.123135805130005,
+      "logps/chosen": -373.9203796386719,
+      "logps/rejected": -208.94192504882812,
+      "loss": 0.6955,
+      "metrics/advantage_var": 0.07473797351121902,
+      "rewards/accuracies": 0.4957031309604645,
+      "rewards/chosen": 0.0009970426326617599,
+      "rewards/margins": 0.0005779300117865205,
+      "rewards/rejected": 0.00041911250445991755,
+      "stability/repetition_rate_mean": 0.8125225901603699,
+      "stability/response_length_mean": 356.0,
+      "stability/response_length_std": 249.6766815185547,
+      "stability/response_length_var": 62338.4453125,
+      "stability/token_entropy_mean": 6.229687690734863,
+      "step": 30
+    },
+    {
+      "epoch": 0.17914267434420986,
+      "grad_norm": 1.722904920578003,
+      "learning_rate": 3.4000000000000003e-07,
+      "logits/chosen": -2.2611172199249268,
+      "logits/rejected": -2.1200122833251953,
+      "logps/chosen": -366.32049560546875,
+      "logps/rejected": -203.98924255371094,
+      "loss": 0.6955,
+      "metrics/advantage_var": 0.06772881746292114,
+      "rewards/accuracies": 0.5160156488418579,
+      "rewards/chosen": 0.002536697080358863,
+      "rewards/margins": 0.0006712052272632718,
+      "rewards/rejected": 0.0018654920859262347,
+      "stability/repetition_rate_mean": 0.9350665807723999,
+      "stability/response_length_mean": 606.2000122070312,
+      "stability/response_length_std": 394.1366882324219,
+      "stability/response_length_var": 155343.734375,
+      "stability/token_entropy_mean": 7.150000095367432,
+      "step": 35
+    },
+    {
+      "epoch": 0.20473448496481125,
+      "grad_norm": 1.8076204061508179,
+      "learning_rate": 3.8999999999999997e-07,
+      "logits/chosen": -2.2763009071350098,
+      "logits/rejected": -2.122572422027588,
+      "logps/chosen": -369.1820373535156,
+      "logps/rejected": -197.3726043701172,
+      "loss": 0.6956,
+      "metrics/advantage_var": 0.06771942228078842,
+      "rewards/accuracies": 0.516796886920929,
+      "rewards/chosen": 0.0037511326372623444,
+      "rewards/margins": 0.0003795797238126397,
+      "rewards/rejected": 0.003371552564203739,
+      "stability/repetition_rate_mean": 0.9299213290214539,
+      "stability/response_length_mean": 547.2000122070312,
+      "stability/response_length_std": 290.8897399902344,
+      "stability/response_length_var": 84616.84375,
+      "stability/token_entropy_mean": 6.993750095367432,
+      "step": 40
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 1.8154666423797607,
+      "learning_rate": 4.3999999999999997e-07,
+      "logits/chosen": -2.2571003437042236,
+      "logits/rejected": -2.107235908508301,
+      "logps/chosen": -385.1767578125,
+      "logps/rejected": -212.5702667236328,
+      "loss": 0.6956,
+      "metrics/advantage_var": 0.07769347727298737,
+      "rewards/accuracies": 0.5101562142372131,
+      "rewards/chosen": 0.00422213040292263,
+      "rewards/margins": 0.00046065542846918106,
+      "rewards/rejected": 0.0037614754401147366,
+      "stability/repetition_rate_mean": 0.8824867010116577,
+      "stability/response_length_mean": 366.5,
+      "stability/response_length_std": 118.86243438720703,
+      "stability/response_length_var": 14128.27734375,
+      "stability/token_entropy_mean": 6.659375190734863,
+      "step": 45
+    },
+    {
+      "epoch": 0.2559181062060141,
+      "grad_norm": 1.8462119102478027,
+      "learning_rate": 4.9e-07,
+      "logits/chosen": -2.264322280883789,
+      "logits/rejected": -2.109330415725708,
+      "logps/chosen": -378.2926940917969,
+      "logps/rejected": -207.4957275390625,
+      "loss": 0.6943,
+      "metrics/advantage_var": 0.07516211271286011,
+      "rewards/accuracies": 0.53125,
+      "rewards/chosen": 0.006509767379611731,
+      "rewards/margins": 0.00300016300752759,
+      "rewards/rejected": 0.0035096043720841408,
+      "stability/repetition_rate_mean": 0.8690369725227356,
+      "stability/response_length_mean": 384.1000061035156,
+      "stability/response_length_std": 178.4809112548828,
+      "stability/response_length_var": 31855.43359375,
+      "stability/token_entropy_mean": 6.496874809265137,
+      "step": 50
+    },
+    {
+      "epoch": 0.28150991682661547,
+      "grad_norm": 1.9448425769805908,
+      "learning_rate": 5.4e-07,
+      "logits/chosen": -2.2531988620758057,
+      "logits/rejected": -2.094651937484741,
+      "logps/chosen": -393.789306640625,
+      "logps/rejected": -210.3695068359375,
+      "loss": 0.6933,
+      "metrics/advantage_var": 0.07928746938705444,
+      "rewards/accuracies": 0.5472656488418579,
+      "rewards/chosen": 0.009074963629245758,
+      "rewards/margins": 0.005103425122797489,
+      "rewards/rejected": 0.003971538506448269,
+      "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": 1.8275038003921509,
+      "learning_rate": 5.9e-07,
+      "logits/chosen": -2.273451805114746,
+      "logits/rejected": -2.1119141578674316,
+      "logps/chosen": -387.7010803222656,
+      "logps/rejected": -206.8184051513672,
+      "loss": 0.6935,
+      "metrics/advantage_var": 0.07989054918289185,
+      "rewards/accuracies": 0.54296875,
+      "rewards/chosen": 0.0105532705783844,
+      "rewards/margins": 0.004686878528445959,
+      "rewards/rejected": 0.005866392515599728,
+      "stability/repetition_rate_mean": 0.8754798173904419,
+      "stability/response_length_mean": 655.2999877929688,
+      "stability/response_length_std": 282.005126953125,
+      "stability/response_length_var": 79526.8984375,
+      "stability/token_entropy_mean": 6.606249809265137,
+      "step": 60
+    },
+    {
+      "epoch": 0.3326935380678183,
+      "grad_norm": 1.697847843170166,
+      "learning_rate": 6.4e-07,
+      "logits/chosen": -2.2514572143554688,
+      "logits/rejected": -2.0982723236083984,
+      "logps/chosen": -377.4744567871094,
+      "logps/rejected": -204.39608764648438,
+      "loss": 0.6929,
+      "metrics/advantage_var": 0.08167043328285217,
+      "rewards/accuracies": 0.561718761920929,
+      "rewards/chosen": 0.012515384703874588,
+      "rewards/margins": 0.005867896601557732,
+      "rewards/rejected": 0.006647488102316856,
+      "stability/repetition_rate_mean": 0.916172981262207,
+      "stability/response_length_mean": 306.3999938964844,
+      "stability/response_length_std": 141.7236785888672,
+      "stability/response_length_var": 20085.599609375,
+      "stability/token_entropy_mean": 6.837500095367432,
+      "step": 65
+    },
+    {
+      "epoch": 0.3582853486884197,
+      "grad_norm": 1.9988712072372437,
+      "learning_rate": 6.9e-07,
+      "logits/chosen": -2.271535873413086,
+      "logits/rejected": -2.1066555976867676,
+      "logps/chosen": -393.74591064453125,
+      "logps/rejected": -206.18272399902344,
+      "loss": 0.692,
+      "metrics/advantage_var": 0.09693735092878342,
+      "rewards/accuracies": 0.5687500238418579,
+      "rewards/chosen": 0.016287218779325485,
+      "rewards/margins": 0.00758810481056571,
+      "rewards/rejected": 0.008699115365743637,
+      "stability/repetition_rate_mean": 0.9707034230232239,
+      "stability/response_length_mean": 472.70001220703125,
+      "stability/response_length_std": 159.89791870117188,
+      "stability/response_length_var": 25567.34375,
+      "stability/token_entropy_mean": 7.21875,
+      "step": 70
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 1.882232666015625,
+      "learning_rate": 7.4e-07,
+      "logits/chosen": -2.2751991748809814,
+      "logits/rejected": -2.116259813308716,
+      "logps/chosen": -381.05670166015625,
+      "logps/rejected": -207.9344482421875,
+      "loss": 0.6915,
+      "metrics/advantage_var": 0.10063213855028152,
+      "rewards/accuracies": 0.5695313215255737,
+      "rewards/chosen": 0.019038334488868713,
+      "rewards/margins": 0.008797663263976574,
+      "rewards/rejected": 0.010240670293569565,
+      "stability/repetition_rate_mean": 0.9109193086624146,
+      "stability/response_length_mean": 527.0,
+      "stability/response_length_std": 293.7500915527344,
+      "stability/response_length_var": 86289.109375,
+      "stability/token_entropy_mean": 6.035937309265137,
+      "step": 75
+    },
+    {
+      "epoch": 0.4094689699296225,
+      "grad_norm": 1.8204421997070312,
+      "learning_rate": 7.9e-07,
+      "logits/chosen": -2.261033058166504,
+      "logits/rejected": -2.103379726409912,
+      "logps/chosen": -382.4222106933594,
+      "logps/rejected": -208.96932983398438,
+      "loss": 0.6916,
+      "metrics/advantage_var": 0.11322233080863953,
+      "rewards/accuracies": 0.5718750357627869,
+      "rewards/chosen": 0.021087391301989555,
+      "rewards/margins": 0.008481135591864586,
+      "rewards/rejected": 0.012606254778802395,
+      "stability/repetition_rate_mean": 0.9449335336685181,
+      "stability/response_length_mean": 635.5,
+      "stability/response_length_std": 262.53900146484375,
+      "stability/response_length_var": 68926.71875,
+      "stability/token_entropy_mean": 7.178124904632568,
+      "step": 80
+    },
+    {
+      "epoch": 0.4350607805502239,
+      "grad_norm": 1.8702808618545532,
+      "learning_rate": 8.399999999999999e-07,
+      "logits/chosen": -2.258106231689453,
+      "logits/rejected": -2.1010942459106445,
+      "logps/chosen": -387.320068359375,
+      "logps/rejected": -208.63490295410156,
+      "loss": 0.6896,
+      "metrics/advantage_var": 0.13779063522815704,
+      "rewards/accuracies": 0.592578113079071,
+      "rewards/chosen": 0.02714032307267189,
+      "rewards/margins": 0.012744016945362091,
+      "rewards/rejected": 0.014396305195987225,
+      "stability/repetition_rate_mean": 0.9021533727645874,
+      "stability/response_length_mean": 600.0999755859375,
+      "stability/response_length_std": 349.2030334472656,
+      "stability/response_length_var": 121942.765625,
+      "stability/token_entropy_mean": 6.732031345367432,
+      "step": 85
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 1.8974494934082031,
+      "learning_rate": 8.9e-07,
+      "logits/chosen": -2.2613511085510254,
+      "logits/rejected": -2.113387107849121,
+      "logps/chosen": -375.201904296875,
+      "logps/rejected": -206.21514892578125,
+      "loss": 0.6886,
+      "metrics/advantage_var": 0.15796051919460297,
+      "rewards/accuracies": 0.6050781607627869,
+      "rewards/chosen": 0.03241811320185661,
+      "rewards/margins": 0.014857406727969646,
+      "rewards/rejected": 0.017560705542564392,
+      "stability/repetition_rate_mean": 0.9763043522834778,
+      "stability/response_length_mean": 692.9000244140625,
+      "stability/response_length_std": 127.31452941894531,
+      "stability/response_length_var": 16208.9892578125,
+      "stability/token_entropy_mean": 7.112500190734863,
+      "step": 90
+    },
+    {
+      "epoch": 0.48624440179142675,
+      "grad_norm": 1.8197991847991943,
+      "learning_rate": 9.399999999999999e-07,
+      "logits/chosen": -2.2584359645843506,
+      "logits/rejected": -2.1117799282073975,
+      "logps/chosen": -366.8218078613281,
+      "logps/rejected": -203.42640686035156,
+      "loss": 0.686,
+      "metrics/advantage_var": 0.20491762459278107,
+      "rewards/accuracies": 0.6281249523162842,
+      "rewards/chosen": 0.03878289833664894,
+      "rewards/margins": 0.02026660367846489,
+      "rewards/rejected": 0.0185162965208292,
+      "stability/repetition_rate_mean": 0.9486184120178223,
+      "stability/response_length_mean": 727.0,
+      "stability/response_length_std": 335.0764465332031,
+      "stability/response_length_var": 112276.21875,
+      "stability/token_entropy_mean": 7.303124904632568,
+      "step": 95
+    },
+    {
+      "epoch": 0.5118362124120281,
+      "grad_norm": 1.816606879234314,
+      "learning_rate": 9.9e-07,
+      "logits/chosen": -2.2744193077087402,
+      "logits/rejected": -2.1189730167388916,
+      "logps/chosen": -384.5008239746094,
+      "logps/rejected": -210.61135864257812,
+      "loss": 0.6849,
+      "metrics/advantage_var": 0.26820605993270874,
+      "rewards/accuracies": 0.6503906846046448,
+      "rewards/chosen": 0.04569867625832558,
+      "rewards/margins": 0.022690609097480774,
+      "rewards/rejected": 0.023008069023489952,
+      "stability/repetition_rate_mean": 0.9514176249504089,
+      "stability/response_length_mean": 419.0,
+      "stability/response_length_std": 232.0225067138672,
+      "stability/response_length_var": 53834.4453125,
+      "stability/token_entropy_mean": 7.068749904632568,
+      "step": 100
+    },
+    {
+      "epoch": 0.5374280230326296,
+      "grad_norm": 1.7418373823165894,
+      "learning_rate": 9.999815554351224e-07,
+      "logits/chosen": -2.2584712505340576,
+      "logits/rejected": -2.1078848838806152,
+      "logps/chosen": -377.9892578125,
+      "logps/rejected": -209.6846160888672,
+      "loss": 0.6829,
+      "metrics/advantage_var": 0.34556183218955994,
+      "rewards/accuracies": 0.6472656726837158,
+      "rewards/chosen": 0.053732603788375854,
+      "rewards/margins": 0.026961183175444603,
+      "rewards/rejected": 0.02677142061293125,
+      "stability/repetition_rate_mean": 0.9020894169807434,
+      "stability/response_length_mean": 669.5,
+      "stability/response_length_std": 271.60888671875,
+      "stability/response_length_var": 73771.390625,
+      "stability/token_entropy_mean": 6.571093559265137,
+      "step": 105
+    },
+    {
+      "epoch": 0.5630198336532309,
+      "grad_norm": 1.6666159629821777,
+      "learning_rate": 9.999066267225444e-07,
+      "logits/chosen": -2.274813413619995,
+      "logits/rejected": -2.1170949935913086,
+      "logps/chosen": -385.2557678222656,
+      "logps/rejected": -206.71839904785156,
+      "loss": 0.6818,
+      "metrics/advantage_var": 0.43270087242126465,
+      "rewards/accuracies": 0.6460937857627869,
+      "rewards/chosen": 0.06057017669081688,
+      "rewards/margins": 0.0296179112046957,
+      "rewards/rejected": 0.03095226362347603,
+      "stability/repetition_rate_mean": 0.937423050403595,
+      "stability/response_length_mean": 255.60000610351562,
+      "stability/response_length_std": 114.30874633789062,
+      "stability/response_length_var": 13066.4892578125,
+      "stability/token_entropy_mean": 6.990624904632568,
+      "step": 110
+    },
+    {
+      "epoch": 0.5886116442738324,
+      "grad_norm": 1.798413872718811,
+      "learning_rate": 9.997740697079592e-07,
+      "logits/chosen": -2.2516493797302246,
+      "logits/rejected": -2.1109843254089355,
+      "logps/chosen": -372.72344970703125,
+      "logps/rejected": -213.9393310546875,
+      "loss": 0.6786,
+      "metrics/advantage_var": 0.6110946536064148,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": 0.07168412208557129,
+      "rewards/margins": 0.03667932376265526,
+      "rewards/rejected": 0.03500479832291603,
+      "stability/repetition_rate_mean": 0.8393692970275879,
+      "stability/response_length_mean": 499.29998779296875,
+      "stability/response_length_std": 282.31072998046875,
+      "stability/response_length_var": 79699.34375,
+      "stability/token_entropy_mean": 6.525000095367432,
+      "step": 115
+    },
+    {
+      "epoch": 0.6142034548944337,
+      "grad_norm": 1.7711139917373657,
+      "learning_rate": 9.995838996722914e-07,
+      "logits/chosen": -2.248577117919922,
+      "logits/rejected": -2.102853298187256,
+      "logps/chosen": -367.78424072265625,
+      "logps/rejected": -207.5933837890625,
+      "loss": 0.6757,
+      "metrics/advantage_var": 0.744879961013794,
+      "rewards/accuracies": 0.6742187738418579,
+      "rewards/chosen": 0.08194158971309662,
+      "rewards/margins": 0.04303481802344322,
+      "rewards/rejected": 0.0389067716896534,
+      "stability/repetition_rate_mean": 0.9263461828231812,
+      "stability/response_length_mean": 363.6000061035156,
+      "stability/response_length_std": 146.0701446533203,
+      "stability/response_length_var": 21336.48828125,
+      "stability/token_entropy_mean": 6.887499809265137,
+      "step": 120
+    },
+    {
+      "epoch": 0.6397952655150352,
+      "grad_norm": 1.7923411130905151,
+      "learning_rate": 9.993361385379874e-07,
+      "logits/chosen": -2.2532808780670166,
+      "logits/rejected": -2.0951650142669678,
+      "logps/chosen": -385.71148681640625,
+      "logps/rejected": -205.10792541503906,
+      "loss": 0.6732,
+      "metrics/advantage_var": 1.0200822353363037,
+      "rewards/accuracies": 0.6714844107627869,
+      "rewards/chosen": 0.09237537533044815,
+      "rewards/margins": 0.049354877322912216,
+      "rewards/rejected": 0.04302050173282623,
+      "stability/repetition_rate_mean": 0.8936500549316406,
+      "stability/response_length_mean": 433.79998779296875,
+      "stability/response_length_std": 126.92412567138672,
+      "stability/response_length_var": 16109.7333984375,
+      "stability/token_entropy_mean": 6.706250190734863,
+      "step": 125
+    },
+    {
+      "epoch": 0.6653870761356366,
+      "grad_norm": 1.8727540969848633,
+      "learning_rate": 9.99030814866488e-07,
+      "logits/chosen": -2.244988203048706,
+      "logits/rejected": -2.096203565597534,
+      "logps/chosen": -375.5982360839844,
+      "logps/rejected": -203.75221252441406,
+      "loss": 0.6698,
+      "metrics/advantage_var": 1.1310168504714966,
+      "rewards/accuracies": 0.682812511920929,
+      "rewards/chosen": 0.09967245161533356,
+      "rewards/margins": 0.056794874370098114,
+      "rewards/rejected": 0.042877569794654846,
+      "stability/repetition_rate_mean": 0.8148199319839478,
+      "stability/response_length_mean": 296.1000061035156,
+      "stability/response_length_std": 167.77593994140625,
+      "stability/response_length_var": 28148.767578125,
+      "stability/token_entropy_mean": 6.009375095367432,
+      "step": 130
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 1.7467596530914307,
+      "learning_rate": 9.986679638549358e-07,
+      "logits/chosen": -2.2671823501586914,
+      "logits/rejected": -2.107846736907959,
+      "logps/chosen": -387.3779602050781,
+      "logps/rejected": -213.74169921875,
+      "loss": 0.6672,
+      "metrics/advantage_var": 1.38654625415802,
+      "rewards/accuracies": 0.6800780892372131,
+      "rewards/chosen": 0.11093107610940933,
+      "rewards/margins": 0.06296703219413757,
+      "rewards/rejected": 0.04796404391527176,
+      "stability/repetition_rate_mean": 0.8634861707687378,
+      "stability/response_length_mean": 596.4000244140625,
+      "stability/response_length_std": 291.9129333496094,
+      "stability/response_length_var": 85213.15625,
+      "stability/token_entropy_mean": 6.521874904632568,
+      "step": 135
+    },
+    {
+      "epoch": 0.7165706973768394,
+      "grad_norm": 1.827934741973877,
+      "learning_rate": 9.982476273321175e-07,
+      "logits/chosen": -2.2552778720855713,
+      "logits/rejected": -2.0999696254730225,
+      "logps/chosen": -389.03033447265625,
+      "logps/rejected": -211.16603088378906,
+      "loss": 0.6637,
+      "metrics/advantage_var": 1.8156306743621826,
+      "rewards/accuracies": 0.6910156607627869,
+      "rewards/chosen": 0.11952749639749527,
+      "rewards/margins": 0.07165504992008209,
+      "rewards/rejected": 0.04787244647741318,
+      "stability/repetition_rate_mean": 0.8619750142097473,
+      "stability/response_length_mean": 173.89999389648438,
+      "stability/response_length_std": 70.74908447265625,
+      "stability/response_length_var": 5005.43310546875,
+      "stability/token_entropy_mean": 6.634375095367432,
+      "step": 140
+    },
+    {
+      "epoch": 0.7421625079974408,
+      "grad_norm": 1.7593910694122314,
+      "learning_rate": 9.977698537536417e-07,
+      "logits/chosen": -2.2668190002441406,
+      "logits/rejected": -2.0994153022766113,
+      "logps/chosen": -387.7881164550781,
+      "logps/rejected": -203.63453674316406,
+      "loss": 0.6611,
+      "metrics/advantage_var": 2.1796822547912598,
+      "rewards/accuracies": 0.6863281726837158,
+      "rewards/chosen": 0.1324867308139801,
+      "rewards/margins": 0.07788518816232681,
+      "rewards/rejected": 0.05460155010223389,
+      "stability/repetition_rate_mean": 0.9642826914787292,
+      "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": 1.8130406141281128,
+      "learning_rate": 9.972346981963546e-07,
+      "logits/chosen": -2.2341678142547607,
+      "logits/rejected": -2.091710090637207,
+      "logps/chosen": -373.8421630859375,
+      "logps/rejected": -211.25048828125,
+      "loss": 0.6593,
+      "metrics/advantage_var": 3.361509084701538,
+      "rewards/accuracies": 0.6628906726837158,
+      "rewards/chosen": 0.14231964945793152,
+      "rewards/margins": 0.08618739247322083,
+      "rewards/rejected": 0.0561322346329689,
+      "stability/repetition_rate_mean": 0.8522264361381531,
+      "stability/response_length_mean": 654.2000122070312,
+      "stability/response_length_std": 372.5482177734375,
+      "stability/response_length_var": 138792.171875,
+      "stability/token_entropy_mean": 6.565625190734863,
+      "step": 150
+    },
+    {
+      "epoch": 0.7933461292386437,
+      "grad_norm": 1.8375900983810425,
+      "learning_rate": 9.966422223519885e-07,
+      "logits/chosen": -2.259756565093994,
+      "logits/rejected": -2.108833074569702,
+      "logps/chosen": -377.3822326660156,
+      "logps/rejected": -206.42324829101562,
+      "loss": 0.6549,
+      "metrics/advantage_var": 2.9679672718048096,
+      "rewards/accuracies": 0.6792968511581421,
+      "rewards/chosen": 0.15822532773017883,
+      "rewards/margins": 0.09387710690498352,
+      "rewards/rejected": 0.06434821337461472,
+      "stability/repetition_rate_mean": 0.8794426918029785,
+      "stability/response_length_mean": 502.3999938964844,
+      "stability/response_length_std": 200.0795440673828,
+      "stability/response_length_var": 40031.8203125,
+      "stability/token_entropy_mean": 6.564062595367432,
+      "step": 155
+    },
+    {
+      "epoch": 0.818937939859245,
+      "grad_norm": 1.7692912817001343,
+      "learning_rate": 9.959924945200523e-07,
+      "logits/chosen": -2.253265619277954,
+      "logits/rejected": -2.0869619846343994,
+      "logps/chosen": -383.5110778808594,
+      "logps/rejected": -207.28126525878906,
+      "loss": 0.6519,
+      "metrics/advantage_var": 4.039853096008301,
+      "rewards/accuracies": 0.6902344226837158,
+      "rewards/chosen": 0.16477353870868683,
+      "rewards/margins": 0.10416299849748611,
+      "rewards/rejected": 0.060610540211200714,
+      "stability/repetition_rate_mean": 0.9666600227355957,
+      "stability/response_length_mean": 870.4000244140625,
+      "stability/response_length_std": 147.6581573486328,
+      "stability/response_length_var": 21802.93359375,
+      "stability/token_entropy_mean": 7.25,
+      "step": 160
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.6458988189697266,
+      "learning_rate": 9.952855895999568e-07,
+      "logits/chosen": -2.238316535949707,
+      "logits/rejected": -2.090843915939331,
+      "logps/chosen": -362.42230224609375,
+      "logps/rejected": -203.6447296142578,
+      "loss": 0.6502,
+      "metrics/advantage_var": 3.8923609256744385,
+      "rewards/accuracies": 0.6910156607627869,
+      "rewards/chosen": 0.16530370712280273,
+      "rewards/margins": 0.10715127736330032,
+      "rewards/rejected": 0.05815241485834122,
+      "stability/repetition_rate_mean": 0.8666970133781433,
+      "stability/response_length_mean": 547.7000122070312,
+      "stability/response_length_std": 249.60325622558594,
+      "stability/response_length_var": 62301.7890625,
+      "stability/token_entropy_mean": 6.614062309265137,
+      "step": 165
+    },
+    {
+      "epoch": 0.8701215611004478,
+      "grad_norm": 1.7035630941390991,
+      "learning_rate": 9.9452158908238e-07,
+      "logits/chosen": -2.263615131378174,
+      "logits/rejected": -2.0976061820983887,
+      "logps/chosen": -376.9706115722656,
+      "logps/rejected": -203.80648803710938,
+      "loss": 0.6447,
+      "metrics/advantage_var": 4.90874719619751,
+      "rewards/accuracies": 0.7003906965255737,
+      "rewards/chosen": 0.18038088083267212,
+      "rewards/margins": 0.1220773383975029,
+      "rewards/rejected": 0.05830354988574982,
+      "stability/repetition_rate_mean": 0.9128576517105103,
+      "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": 1.8031121492385864,
+      "learning_rate": 9.937005810398745e-07,
+      "logits/chosen": -2.248379945755005,
+      "logits/rejected": -2.092766046524048,
+      "logps/chosen": -380.500732421875,
+      "logps/rejected": -212.28701782226562,
+      "loss": 0.6424,
+      "metrics/advantage_var": 6.325986385345459,
+      "rewards/accuracies": 0.693359375,
+      "rewards/chosen": 0.18881221115589142,
+      "rewards/margins": 0.13121359050273895,
+      "rewards/rejected": 0.05759861320257187,
+      "stability/repetition_rate_mean": 0.9426466226577759,
+      "stability/response_length_mean": 460.6000061035156,
+      "stability/response_length_std": 165.9894256591797,
+      "stability/response_length_var": 27552.48828125,
+      "stability/token_entropy_mean": 6.956250190734863,
+      "step": 175
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 1.771286129951477,
+      "learning_rate": 9.928226601167138e-07,
+      "logits/chosen": -2.247504472732544,
+      "logits/rejected": -2.0860483646392822,
+      "logps/chosen": -387.04644775390625,
+      "logps/rejected": -212.97767639160156,
+      "loss": 0.6404,
+      "metrics/advantage_var": 5.818198204040527,
+      "rewards/accuracies": 0.6871094107627869,
+      "rewards/chosen": 0.20346775650978088,
+      "rewards/margins": 0.13489243388175964,
+      "rewards/rejected": 0.06857535243034363,
+      "stability/repetition_rate_mean": 0.8947866559028625,
+      "stability/response_length_mean": 716.2000122070312,
+      "stability/response_length_std": 329.3815002441406,
+      "stability/response_length_var": 108492.1796875,
+      "stability/token_entropy_mean": 6.510937690734863,
+      "step": 180
+    },
+    {
+      "epoch": 0.946896992962252,
+      "grad_norm": 1.6673380136489868,
+      "learning_rate": 9.918879275179817e-07,
+      "logits/chosen": -2.2497382164001465,
+      "logits/rejected": -2.098439931869507,
+      "logps/chosen": -374.5255126953125,
+      "logps/rejected": -210.45721435546875,
+      "loss": 0.6393,
+      "metrics/advantage_var": 7.5517191886901855,
+      "rewards/accuracies": 0.6832031607627869,
+      "rewards/chosen": 0.21010692417621613,
+      "rewards/margins": 0.1418810933828354,
+      "rewards/rejected": 0.06822583824396133,
+      "stability/repetition_rate_mean": 0.881879985332489,
+      "stability/response_length_mean": 486.3999938964844,
+      "stability/response_length_std": 213.8624725341797,
+      "stability/response_length_var": 45737.15625,
+      "stability/token_entropy_mean": 6.560937404632568,
+      "step": 185
+    },
+    {
+      "epoch": 0.9724888035828535,
+      "grad_norm": 1.5536139011383057,
+      "learning_rate": 9.90896490997906e-07,
+      "logits/chosen": -2.233537435531616,
+      "logits/rejected": -2.077605724334717,
+      "logps/chosen": -377.6812744140625,
+      "logps/rejected": -206.32400512695312,
+      "loss": 0.6319,
+      "metrics/advantage_var": 8.11306095123291,
+      "rewards/accuracies": 0.7000000476837158,
+      "rewards/chosen": 0.2151639461517334,
+      "rewards/margins": 0.16041463613510132,
+      "rewards/rejected": 0.05474931001663208,
+      "stability/repetition_rate_mean": 0.9174295663833618,
+      "stability/response_length_mean": 736.7999877929688,
+      "stability/response_length_std": 246.57106018066406,
+      "stability/response_length_var": 60797.2890625,
+      "stability/token_entropy_mean": 6.928124904632568,
+      "step": 190
+    },
+    {
+      "epoch": 0.9980806142034548,
+      "grad_norm": 1.6391992568969727,
+      "learning_rate": 9.898484648474362e-07,
+      "logits/chosen": -2.2502989768981934,
+      "logits/rejected": -2.0969607830047607,
+      "logps/chosen": -361.8087158203125,
+      "logps/rejected": -208.4525604248047,
+      "loss": 0.6318,
+      "metrics/advantage_var": 8.544903755187988,
+      "rewards/accuracies": 0.7019531726837158,
+      "rewards/chosen": 0.21875369548797607,
+      "rewards/margins": 0.16276590526103973,
+      "rewards/rejected": 0.05598779395222664,
+      "stability/repetition_rate_mean": 0.7733734846115112,
+      "stability/response_length_mean": 356.3999938964844,
+      "stability/response_length_std": 207.94882202148438,
+      "stability/response_length_var": 43242.7109375,
+      "stability/token_entropy_mean": 5.948437690734863,
+      "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_sft_ftx0.001/checkpoint-196/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4512cc16205733853f6f345ca6097937972575afbbdedf6e03f7b46d17f21d8
+size 8081

mistral7b_dpo_sft_ftx0.001/checkpoint-196/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, 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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

mistral7b_dpo_sft_ftx0.001/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_sft_ftx0.001/stability_metrics.jsonl

@@ -0,0 +1,38 @@
+{"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": 0.9129446148872375, "step": 10}
+{"stability/response_length_mean": 345.70001220703125, "stability/response_length_std": 212.67974853515625, "stability/response_length_var": 45232.6796875, "stability/token_entropy_mean": 6.3515625, "stability/repetition_rate_mean": 0.8466346859931946, "step": 15}
+{"stability/response_length_mean": 739.4000244140625, "stability/response_length_std": 281.282470703125, "stability/response_length_var": 79119.8203125, "stability/token_entropy_mean": 6.934374809265137, "stability/repetition_rate_mean": 0.9300774335861206, "step": 20}
+{"stability/response_length_mean": 658.5999755859375, "stability/response_length_std": 288.0155029296875, "stability/response_length_var": 82952.9296875, "stability/token_entropy_mean": 7.012499809265137, "stability/repetition_rate_mean": 0.9150856137275696, "step": 25}
+{"stability/response_length_mean": 356.0, "stability/response_length_std": 249.6766815185547, "stability/response_length_var": 62338.4453125, "stability/token_entropy_mean": 6.229687690734863, "stability/repetition_rate_mean": 0.8125225901603699, "step": 30}
+{"stability/response_length_mean": 606.2000122070312, "stability/response_length_std": 394.1366882324219, "stability/response_length_var": 155343.734375, "stability/token_entropy_mean": 7.150000095367432, "stability/repetition_rate_mean": 0.9350665807723999, "step": 35}
+{"stability/response_length_mean": 547.2000122070312, "stability/response_length_std": 290.8897399902344, "stability/response_length_var": 84616.84375, "stability/token_entropy_mean": 6.993750095367432, "stability/repetition_rate_mean": 0.9299213290214539, "step": 40}
+{"stability/response_length_mean": 366.5, "stability/response_length_std": 118.86243438720703, "stability/response_length_var": 14128.27734375, "stability/token_entropy_mean": 6.659375190734863, "stability/repetition_rate_mean": 0.8824867010116577, "step": 45}
+{"stability/response_length_mean": 384.1000061035156, "stability/response_length_std": 178.4809112548828, "stability/response_length_var": 31855.43359375, "stability/token_entropy_mean": 6.496874809265137, "stability/repetition_rate_mean": 0.8690369725227356, "step": 50}
+{"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, "step": 55}
+{"stability/response_length_mean": 655.2999877929688, "stability/response_length_std": 282.005126953125, "stability/response_length_var": 79526.8984375, "stability/token_entropy_mean": 6.606249809265137, "stability/repetition_rate_mean": 0.8754798173904419, "step": 60}
+{"stability/response_length_mean": 306.3999938964844, "stability/response_length_std": 141.7236785888672, "stability/response_length_var": 20085.599609375, "stability/token_entropy_mean": 6.837500095367432, "stability/repetition_rate_mean": 0.916172981262207, "step": 65}
+{"stability/response_length_mean": 472.70001220703125, "stability/response_length_std": 159.89791870117188, "stability/response_length_var": 25567.34375, "stability/token_entropy_mean": 7.21875, "stability/repetition_rate_mean": 0.9707034230232239, "step": 70}
+{"stability/response_length_mean": 527.0, "stability/response_length_std": 293.7500915527344, "stability/response_length_var": 86289.109375, "stability/token_entropy_mean": 6.035937309265137, "stability/repetition_rate_mean": 0.9109193086624146, "step": 75}
+{"stability/response_length_mean": 635.5, "stability/response_length_std": 262.53900146484375, "stability/response_length_var": 68926.71875, "stability/token_entropy_mean": 7.178124904632568, "stability/repetition_rate_mean": 0.9449335336685181, "step": 80}
+{"stability/response_length_mean": 600.0999755859375, "stability/response_length_std": 349.2030334472656, "stability/response_length_var": 121942.765625, "stability/token_entropy_mean": 6.732031345367432, "stability/repetition_rate_mean": 0.9021533727645874, "step": 85}
+{"stability/response_length_mean": 692.9000244140625, "stability/response_length_std": 127.31452941894531, "stability/response_length_var": 16208.9892578125, "stability/token_entropy_mean": 7.112500190734863, "stability/repetition_rate_mean": 0.9763043522834778, "step": 90}
+{"stability/response_length_mean": 727.0, "stability/response_length_std": 335.0764465332031, "stability/response_length_var": 112276.21875, "stability/token_entropy_mean": 7.303124904632568, "stability/repetition_rate_mean": 0.9486184120178223, "step": 95}
+{"stability/response_length_mean": 419.0, "stability/response_length_std": 232.0225067138672, "stability/response_length_var": 53834.4453125, "stability/token_entropy_mean": 7.068749904632568, "stability/repetition_rate_mean": 0.9514176249504089, "step": 100}
+{"stability/response_length_mean": 669.5, "stability/response_length_std": 271.60888671875, "stability/response_length_var": 73771.390625, "stability/token_entropy_mean": 6.571093559265137, "stability/repetition_rate_mean": 0.9020894169807434, "step": 105}
+{"stability/response_length_mean": 255.60000610351562, "stability/response_length_std": 114.30874633789062, "stability/response_length_var": 13066.4892578125, "stability/token_entropy_mean": 6.990624904632568, "stability/repetition_rate_mean": 0.937423050403595, "step": 110}
+{"stability/response_length_mean": 499.29998779296875, "stability/response_length_std": 282.31072998046875, "stability/response_length_var": 79699.34375, "stability/token_entropy_mean": 6.525000095367432, "stability/repetition_rate_mean": 0.8393692970275879, "step": 115}
+{"stability/response_length_mean": 363.6000061035156, "stability/response_length_std": 146.0701446533203, "stability/response_length_var": 21336.48828125, "stability/token_entropy_mean": 6.887499809265137, "stability/repetition_rate_mean": 0.9263461828231812, "step": 120}
+{"stability/response_length_mean": 433.79998779296875, "stability/response_length_std": 126.92412567138672, "stability/response_length_var": 16109.7333984375, "stability/token_entropy_mean": 6.706250190734863, "stability/repetition_rate_mean": 0.8936500549316406, "step": 125}
+{"stability/response_length_mean": 296.1000061035156, "stability/response_length_std": 167.77593994140625, "stability/response_length_var": 28148.767578125, "stability/token_entropy_mean": 6.009375095367432, "stability/repetition_rate_mean": 0.8148199319839478, "step": 130}
+{"stability/response_length_mean": 596.4000244140625, "stability/response_length_std": 291.9129333496094, "stability/response_length_var": 85213.15625, "stability/token_entropy_mean": 6.521874904632568, "stability/repetition_rate_mean": 0.8634861707687378, "step": 135}
+{"stability/response_length_mean": 173.89999389648438, "stability/response_length_std": 70.74908447265625, "stability/response_length_var": 5005.43310546875, "stability/token_entropy_mean": 6.634375095367432, "stability/repetition_rate_mean": 0.8619750142097473, "step": 140}
+{"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": 0.9642826914787292, "step": 145}
+{"stability/response_length_mean": 654.2000122070312, "stability/response_length_std": 372.5482177734375, "stability/response_length_var": 138792.171875, "stability/token_entropy_mean": 6.565625190734863, "stability/repetition_rate_mean": 0.8522264361381531, "step": 150}
+{"stability/response_length_mean": 502.3999938964844, "stability/response_length_std": 200.0795440673828, "stability/response_length_var": 40031.8203125, "stability/token_entropy_mean": 6.564062595367432, "stability/repetition_rate_mean": 0.8794426918029785, "step": 155}
+{"stability/response_length_mean": 870.4000244140625, "stability/response_length_std": 147.6581573486328, "stability/response_length_var": 21802.93359375, "stability/token_entropy_mean": 7.25, "stability/repetition_rate_mean": 0.9666600227355957, "step": 160}
+{"stability/response_length_mean": 547.7000122070312, "stability/response_length_std": 249.60325622558594, "stability/response_length_var": 62301.7890625, "stability/token_entropy_mean": 6.614062309265137, "stability/repetition_rate_mean": 0.8666970133781433, "step": 165}
+{"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": 0.9128576517105103, "step": 170}
+{"stability/response_length_mean": 460.6000061035156, "stability/response_length_std": 165.9894256591797, "stability/response_length_var": 27552.48828125, "stability/token_entropy_mean": 6.956250190734863, "stability/repetition_rate_mean": 0.9426466226577759, "step": 175}
+{"stability/response_length_mean": 716.2000122070312, "stability/response_length_std": 329.3815002441406, "stability/response_length_var": 108492.1796875, "stability/token_entropy_mean": 6.510937690734863, "stability/repetition_rate_mean": 0.8947866559028625, "step": 180}
+{"stability/response_length_mean": 486.3999938964844, "stability/response_length_std": 213.8624725341797, "stability/response_length_var": 45737.15625, "stability/token_entropy_mean": 6.560937404632568, "stability/repetition_rate_mean": 0.881879985332489, "step": 185}
+{"stability/response_length_mean": 736.7999877929688, "stability/response_length_std": 246.57106018066406, "stability/response_length_var": 60797.2890625, "stability/token_entropy_mean": 6.928124904632568, "stability/repetition_rate_mean": 0.9174295663833618, "step": 190}
+{"stability/response_length_mean": 356.3999938964844, "stability/response_length_std": 207.94882202148438, "stability/response_length_var": 43242.7109375, "stability/token_entropy_mean": 5.948437690734863, "stability/repetition_rate_mean": 0.7733734846115112, "step": 195}

mistral7b_dpo_sft_ftx0.001/tokenizer.model

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

mistral7b_dpo_sft_ftx0.001/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_sft_ftx0.001/train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "total_flos": 7.072674331683193e+18,
+    "train_loss": 0.6755896362723136,
+    "train_runtime": 4339.5435,
+    "train_samples_per_second": 23.044,
+    "train_steps_per_second": 0.045
+}

mistral7b_dpo_sft_ftx0.001/trainer_log.jsonl

@@ -0,0 +1,40 @@
+{"current_steps": 5, "total_steps": 196, "loss": 0.6959, "accuracy": 0.29609373211860657, "lr": 4e-08, "epoch": 0.025591810620601407, "percentage": 2.55, "elapsed_time": "0:01:48", "remaining_time": "1:08:49", "rewards/chosen": -0.00011618183634709567, "rewards/rejected": 0.00014287166413851082, "rewards/accuracies": 0.29609373211860657, "rewards/margins": -0.00025905348593369126, "logps/chosen": -375.07659912109375, "logps/rejected": -210.87295532226562, "logits/chosen": -2.273049831390381, "logits/rejected": -2.1233928203582764}
+{"current_steps": 10, "total_steps": 196, "loss": 0.6951, "accuracy": 0.5093749761581421, "lr": 9e-08, "epoch": 0.05118362124120281, "percentage": 5.1, "elapsed_time": "0:03:39", "remaining_time": "1:08:01", "rewards/chosen": 0.0020605388563126326, "rewards/rejected": 0.0006454450776800513, "rewards/accuracies": 0.5093749761581421, "rewards/margins": 0.001415093895047903, "logps/chosen": -382.0225830078125, "logps/rejected": -202.45101928710938, "logits/chosen": -2.267812967300415, "logits/rejected": -2.1164932250976562, "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": 0.9129446148872375}
+{"current_steps": 15, "total_steps": 196, "loss": 0.6963, "accuracy": 0.4898437559604645, "lr": 1.4e-07, "epoch": 0.07677543186180422, "percentage": 7.65, "elapsed_time": "0:05:29", "remaining_time": "1:06:21", "rewards/chosen": 0.000939845631364733, "rewards/rejected": 0.0018643148941919208, "rewards/accuracies": 0.4898437559604645, "rewards/margins": -0.0009244689717888832, "logps/chosen": -389.2162170410156, "logps/rejected": -207.0449981689453, "logits/chosen": -2.2609243392944336, "logits/rejected": -2.099472761154175, "stability/response_length_mean": 345.70001220703125, "stability/response_length_std": 212.67974853515625, "stability/response_length_var": 45232.6796875, "stability/token_entropy_mean": 6.3515625, "stability/repetition_rate_mean": 0.8466346859931946}
+{"current_steps": 20, "total_steps": 196, "loss": 0.6966, "accuracy": 0.49375006556510925, "lr": 1.8999999999999998e-07, "epoch": 0.10236724248240563, "percentage": 10.2, "elapsed_time": "0:07:20", "remaining_time": "1:04:39", "rewards/chosen": -0.00011375403846614063, "rewards/rejected": 0.001396618434228003, "rewards/accuracies": 0.49375006556510925, "rewards/margins": -0.0015103723853826523, "logps/chosen": -368.7252502441406, "logps/rejected": -198.56080627441406, "logits/chosen": -2.2611491680145264, "logits/rejected": -2.1143441200256348, "stability/response_length_mean": 739.4000244140625, "stability/response_length_std": 281.282470703125, "stability/response_length_var": 79119.8203125, "stability/token_entropy_mean": 6.934374809265137, "stability/repetition_rate_mean": 0.9300774335861206}
+{"current_steps": 25, "total_steps": 196, "loss": 0.6962, "accuracy": 0.49179691076278687, "lr": 2.4e-07, "epoch": 0.12795905310300704, "percentage": 12.76, "elapsed_time": "0:09:12", "remaining_time": "1:02:57", "rewards/chosen": 0.0011066627921536565, "rewards/rejected": 0.0019742015283554792, "rewards/accuracies": 0.49179691076278687, "rewards/margins": -0.0008675385033711791, "logps/chosen": -378.173828125, "logps/rejected": -205.8581085205078, "logits/chosen": -2.277261972427368, "logits/rejected": -2.125120162963867, "stability/response_length_mean": 658.5999755859375, "stability/response_length_std": 288.0155029296875, "stability/response_length_var": 82952.9296875, "stability/token_entropy_mean": 7.012499809265137, "stability/repetition_rate_mean": 0.9150856137275696}
+{"current_steps": 30, "total_steps": 196, "loss": 0.6955, "accuracy": 0.4957031309604645, "lr": 2.9e-07, "epoch": 0.15355086372360843, "percentage": 15.31, "elapsed_time": "0:11:03", "remaining_time": "1:01:09", "rewards/chosen": 0.0009970426326617599, "rewards/rejected": 0.00041911250445991755, "rewards/accuracies": 0.4957031309604645, "rewards/margins": 0.0005779300117865205, "logps/chosen": -373.9203796386719, "logps/rejected": -208.94192504882812, "logits/chosen": -2.2703747749328613, "logits/rejected": -2.123135805130005, "stability/response_length_mean": 356.0, "stability/response_length_std": 249.6766815185547, "stability/response_length_var": 62338.4453125, "stability/token_entropy_mean": 6.229687690734863, "stability/repetition_rate_mean": 0.8125225901603699}
+{"current_steps": 35, "total_steps": 196, "loss": 0.6955, "accuracy": 0.5160156488418579, "lr": 3.4000000000000003e-07, "epoch": 0.17914267434420986, "percentage": 17.86, "elapsed_time": "0:12:54", "remaining_time": "0:59:23", "rewards/chosen": 0.002536697080358863, "rewards/rejected": 0.0018654920859262347, "rewards/accuracies": 0.5160156488418579, "rewards/margins": 0.0006712052272632718, "logps/chosen": -366.32049560546875, "logps/rejected": -203.98924255371094, "logits/chosen": -2.2611172199249268, "logits/rejected": -2.1200122833251953, "stability/response_length_mean": 606.2000122070312, "stability/response_length_std": 394.1366882324219, "stability/response_length_var": 155343.734375, "stability/token_entropy_mean": 7.150000095367432, "stability/repetition_rate_mean": 0.9350665807723999}
+{"current_steps": 40, "total_steps": 196, "loss": 0.6956, "accuracy": 0.516796886920929, "lr": 3.8999999999999997e-07, "epoch": 0.20473448496481125, "percentage": 20.41, "elapsed_time": "0:14:46", "remaining_time": "0:57:37", "rewards/chosen": 0.0037511326372623444, "rewards/rejected": 0.003371552564203739, "rewards/accuracies": 0.516796886920929, "rewards/margins": 0.0003795797238126397, "logps/chosen": -369.1820373535156, "logps/rejected": -197.3726043701172, "logits/chosen": -2.2763009071350098, "logits/rejected": -2.122572422027588, "stability/response_length_mean": 547.2000122070312, "stability/response_length_std": 290.8897399902344, "stability/response_length_var": 84616.84375, "stability/token_entropy_mean": 6.993750095367432, "stability/repetition_rate_mean": 0.9299213290214539}
+{"current_steps": 45, "total_steps": 196, "loss": 0.6956, "accuracy": 0.5101562142372131, "lr": 4.3999999999999997e-07, "epoch": 0.23032629558541268, "percentage": 22.96, "elapsed_time": "0:16:36", "remaining_time": "0:55:44", "rewards/chosen": 0.00422213040292263, "rewards/rejected": 0.0037614754401147366, "rewards/accuracies": 0.5101562142372131, "rewards/margins": 0.00046065542846918106, "logps/chosen": -385.1767578125, "logps/rejected": -212.5702667236328, "logits/chosen": -2.2571003437042236, "logits/rejected": -2.107235908508301, "stability/response_length_mean": 366.5, "stability/response_length_std": 118.86243438720703, "stability/response_length_var": 14128.27734375, "stability/token_entropy_mean": 6.659375190734863, "stability/repetition_rate_mean": 0.8824867010116577}
+{"current_steps": 50, "total_steps": 196, "loss": 0.6943, "accuracy": 0.53125, "lr": 4.9e-07, "epoch": 0.2559181062060141, "percentage": 25.51, "elapsed_time": "0:18:27", "remaining_time": "0:53:52", "rewards/chosen": 0.006509767379611731, "rewards/rejected": 0.0035096043720841408, "rewards/accuracies": 0.53125, "rewards/margins": 0.00300016300752759, "logps/chosen": -378.2926940917969, "logps/rejected": -207.4957275390625, "logits/chosen": -2.264322280883789, "logits/rejected": -2.109330415725708, "stability/response_length_mean": 384.1000061035156, "stability/response_length_std": 178.4809112548828, "stability/response_length_var": 31855.43359375, "stability/token_entropy_mean": 6.496874809265137, "stability/repetition_rate_mean": 0.8690369725227356}
+{"current_steps": 55, "total_steps": 196, "loss": 0.6933, "accuracy": 0.5472656488418579, "lr": 5.4e-07, "epoch": 0.28150991682661547, "percentage": 28.06, "elapsed_time": "0:20:17", "remaining_time": "0:52:01", "rewards/chosen": 0.009074963629245758, "rewards/rejected": 0.003971538506448269, "rewards/accuracies": 0.5472656488418579, "rewards/margins": 0.005103425122797489, "logps/chosen": -393.789306640625, "logps/rejected": -210.3695068359375, "logits/chosen": -2.2531988620758057, "logits/rejected": -2.094651937484741, "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": 0.6935, "accuracy": 0.54296875, "lr": 5.9e-07, "epoch": 0.30710172744721687, "percentage": 30.61, "elapsed_time": "0:22:09", "remaining_time": "0:50:13", "rewards/chosen": 0.0105532705783844, "rewards/rejected": 0.005866392515599728, "rewards/accuracies": 0.54296875, "rewards/margins": 0.004686878528445959, "logps/chosen": -387.7010803222656, "logps/rejected": -206.8184051513672, "logits/chosen": -2.273451805114746, "logits/rejected": -2.1119141578674316, "stability/response_length_mean": 655.2999877929688, "stability/response_length_std": 282.005126953125, "stability/response_length_var": 79526.8984375, "stability/token_entropy_mean": 6.606249809265137, "stability/repetition_rate_mean": 0.8754798173904419}
+{"current_steps": 65, "total_steps": 196, "loss": 0.6929, "accuracy": 0.561718761920929, "lr": 6.4e-07, "epoch": 0.3326935380678183, "percentage": 33.16, "elapsed_time": "0:23:59", "remaining_time": "0:48:21", "rewards/chosen": 0.012515384703874588, "rewards/rejected": 0.006647488102316856, "rewards/accuracies": 0.561718761920929, "rewards/margins": 0.005867896601557732, "logps/chosen": -377.4744567871094, "logps/rejected": -204.39608764648438, "logits/chosen": -2.2514572143554688, "logits/rejected": -2.0982723236083984, "stability/response_length_mean": 306.3999938964844, "stability/response_length_std": 141.7236785888672, "stability/response_length_var": 20085.599609375, "stability/token_entropy_mean": 6.837500095367432, "stability/repetition_rate_mean": 0.916172981262207}
+{"current_steps": 70, "total_steps": 196, "loss": 0.692, "accuracy": 0.5687500238418579, "lr": 6.9e-07, "epoch": 0.3582853486884197, "percentage": 35.71, "elapsed_time": "0:25:50", "remaining_time": "0:46:31", "rewards/chosen": 0.016287218779325485, "rewards/rejected": 0.008699115365743637, "rewards/accuracies": 0.5687500238418579, "rewards/margins": 0.00758810481056571, "logps/chosen": -393.74591064453125, "logps/rejected": -206.18272399902344, "logits/chosen": -2.271535873413086, "logits/rejected": -2.1066555976867676, "stability/response_length_mean": 472.70001220703125, "stability/response_length_std": 159.89791870117188, "stability/response_length_var": 25567.34375, "stability/token_entropy_mean": 7.21875, "stability/repetition_rate_mean": 0.9707034230232239}
+{"current_steps": 75, "total_steps": 196, "loss": 0.6915, "accuracy": 0.5695313215255737, "lr": 7.4e-07, "epoch": 0.3838771593090211, "percentage": 38.27, "elapsed_time": "0:27:41", "remaining_time": "0:44:40", "rewards/chosen": 0.019038334488868713, "rewards/rejected": 0.010240670293569565, "rewards/accuracies": 0.5695313215255737, "rewards/margins": 0.008797663263976574, "logps/chosen": -381.05670166015625, "logps/rejected": -207.9344482421875, "logits/chosen": -2.2751991748809814, "logits/rejected": -2.116259813308716, "stability/response_length_mean": 527.0, "stability/response_length_std": 293.7500915527344, "stability/response_length_var": 86289.109375, "stability/token_entropy_mean": 6.035937309265137, "stability/repetition_rate_mean": 0.9109193086624146}
+{"current_steps": 80, "total_steps": 196, "loss": 0.6916, "accuracy": 0.5718750357627869, "lr": 7.9e-07, "epoch": 0.4094689699296225, "percentage": 40.82, "elapsed_time": "0:29:32", "remaining_time": "0:42:50", "rewards/chosen": 0.021087391301989555, "rewards/rejected": 0.012606254778802395, "rewards/accuracies": 0.5718750357627869, "rewards/margins": 0.008481135591864586, "logps/chosen": -382.4222106933594, "logps/rejected": -208.96932983398438, "logits/chosen": -2.261033058166504, "logits/rejected": -2.103379726409912, "stability/response_length_mean": 635.5, "stability/response_length_std": 262.53900146484375, "stability/response_length_var": 68926.71875, "stability/token_entropy_mean": 7.178124904632568, "stability/repetition_rate_mean": 0.9449335336685181}
+{"current_steps": 85, "total_steps": 196, "loss": 0.6896, "accuracy": 0.592578113079071, "lr": 8.399999999999999e-07, "epoch": 0.4350607805502239, "percentage": 43.37, "elapsed_time": "0:31:24", "remaining_time": "0:41:01", "rewards/chosen": 0.02714032307267189, "rewards/rejected": 0.014396305195987225, "rewards/accuracies": 0.592578113079071, "rewards/margins": 0.012744016945362091, "logps/chosen": -387.320068359375, "logps/rejected": -208.63490295410156, "logits/chosen": -2.258106231689453, "logits/rejected": -2.1010942459106445, "stability/response_length_mean": 600.0999755859375, "stability/response_length_std": 349.2030334472656, "stability/response_length_var": 121942.765625, "stability/token_entropy_mean": 6.732031345367432, "stability/repetition_rate_mean": 0.9021533727645874}
+{"current_steps": 90, "total_steps": 196, "loss": 0.6886, "accuracy": 0.6050781607627869, "lr": 8.9e-07, "epoch": 0.46065259117082535, "percentage": 45.92, "elapsed_time": "0:33:16", "remaining_time": "0:39:11", "rewards/chosen": 0.03241811320185661, "rewards/rejected": 0.017560705542564392, "rewards/accuracies": 0.6050781607627869, "rewards/margins": 0.014857406727969646, "logps/chosen": -375.201904296875, "logps/rejected": -206.21514892578125, "logits/chosen": -2.2613511085510254, "logits/rejected": -2.113387107849121, "stability/response_length_mean": 692.9000244140625, "stability/response_length_std": 127.31452941894531, "stability/response_length_var": 16208.9892578125, "stability/token_entropy_mean": 7.112500190734863, "stability/repetition_rate_mean": 0.9763043522834778}
+{"current_steps": 95, "total_steps": 196, "loss": 0.686, "accuracy": 0.6281249523162842, "lr": 9.399999999999999e-07, "epoch": 0.48624440179142675, "percentage": 48.47, "elapsed_time": "0:35:07", "remaining_time": "0:37:20", "rewards/chosen": 0.03878289833664894, "rewards/rejected": 0.0185162965208292, "rewards/accuracies": 0.6281249523162842, "rewards/margins": 0.02026660367846489, "logps/chosen": -366.8218078613281, "logps/rejected": -203.42640686035156, "logits/chosen": -2.2584359645843506, "logits/rejected": -2.1117799282073975, "stability/response_length_mean": 727.0, "stability/response_length_std": 335.0764465332031, "stability/response_length_var": 112276.21875, "stability/token_entropy_mean": 7.303124904632568, "stability/repetition_rate_mean": 0.9486184120178223}
+{"current_steps": 100, "total_steps": 196, "loss": 0.6849, "accuracy": 0.6503906846046448, "lr": 9.9e-07, "epoch": 0.5118362124120281, "percentage": 51.02, "elapsed_time": "0:36:58", "remaining_time": "0:35:30", "rewards/chosen": 0.04569867625832558, "rewards/rejected": 0.023008069023489952, "rewards/accuracies": 0.6503906846046448, "rewards/margins": 0.022690609097480774, "logps/chosen": -384.5008239746094, "logps/rejected": -210.61135864257812, "logits/chosen": -2.2744193077087402, "logits/rejected": -2.1189730167388916, "stability/response_length_mean": 419.0, "stability/response_length_std": 232.0225067138672, "stability/response_length_var": 53834.4453125, "stability/token_entropy_mean": 7.068749904632568, "stability/repetition_rate_mean": 0.9514176249504089}
+{"current_steps": 105, "total_steps": 196, "loss": 0.6829, "accuracy": 0.6472656726837158, "lr": 9.999815554351224e-07, "epoch": 0.5374280230326296, "percentage": 53.57, "elapsed_time": "0:38:50", "remaining_time": "0:33:39", "rewards/chosen": 0.053732603788375854, "rewards/rejected": 0.02677142061293125, "rewards/accuracies": 0.6472656726837158, "rewards/margins": 0.026961183175444603, "logps/chosen": -377.9892578125, "logps/rejected": -209.6846160888672, "logits/chosen": -2.2584712505340576, "logits/rejected": -2.1078848838806152, "stability/response_length_mean": 669.5, "stability/response_length_std": 271.60888671875, "stability/response_length_var": 73771.390625, "stability/token_entropy_mean": 6.571093559265137, "stability/repetition_rate_mean": 0.9020894169807434}
+{"current_steps": 110, "total_steps": 196, "loss": 0.6818, "accuracy": 0.6460937857627869, "lr": 9.999066267225444e-07, "epoch": 0.5630198336532309, "percentage": 56.12, "elapsed_time": "0:40:40", "remaining_time": "0:31:47", "rewards/chosen": 0.06057017669081688, "rewards/rejected": 0.03095226362347603, "rewards/accuracies": 0.6460937857627869, "rewards/margins": 0.0296179112046957, "logps/chosen": -385.2557678222656, "logps/rejected": -206.71839904785156, "logits/chosen": -2.274813413619995, "logits/rejected": -2.1170949935913086, "stability/response_length_mean": 255.60000610351562, "stability/response_length_std": 114.30874633789062, "stability/response_length_var": 13066.4892578125, "stability/token_entropy_mean": 6.990624904632568, "stability/repetition_rate_mean": 0.937423050403595}
+{"current_steps": 115, "total_steps": 196, "loss": 0.6786, "accuracy": 0.65625, "lr": 9.997740697079592e-07, "epoch": 0.5886116442738324, "percentage": 58.67, "elapsed_time": "0:42:31", "remaining_time": "0:29:57", "rewards/chosen": 0.07168412208557129, "rewards/rejected": 0.03500479832291603, "rewards/accuracies": 0.65625, "rewards/margins": 0.03667932376265526, "logps/chosen": -372.72344970703125, "logps/rejected": -213.9393310546875, "logits/chosen": -2.2516493797302246, "logits/rejected": -2.1109843254089355, "stability/response_length_mean": 499.29998779296875, "stability/response_length_std": 282.31072998046875, "stability/response_length_var": 79699.34375, "stability/token_entropy_mean": 6.525000095367432, "stability/repetition_rate_mean": 0.8393692970275879}
+{"current_steps": 120, "total_steps": 196, "loss": 0.6757, "accuracy": 0.6742187738418579, "lr": 9.995838996722914e-07, "epoch": 0.6142034548944337, "percentage": 61.22, "elapsed_time": "0:44:21", "remaining_time": "0:28:05", "rewards/chosen": 0.08194158971309662, "rewards/rejected": 0.0389067716896534, "rewards/accuracies": 0.6742187738418579, "rewards/margins": 0.04303481802344322, "logps/chosen": -367.78424072265625, "logps/rejected": -207.5933837890625, "logits/chosen": -2.248577117919922, "logits/rejected": -2.102853298187256, "stability/response_length_mean": 363.6000061035156, "stability/response_length_std": 146.0701446533203, "stability/response_length_var": 21336.48828125, "stability/token_entropy_mean": 6.887499809265137, "stability/repetition_rate_mean": 0.9263461828231812}
+{"current_steps": 125, "total_steps": 196, "loss": 0.6732, "accuracy": 0.6714844107627869, "lr": 9.993361385379874e-07, "epoch": 0.6397952655150352, "percentage": 63.78, "elapsed_time": "0:46:11", "remaining_time": "0:26:14", "rewards/chosen": 0.09237537533044815, "rewards/rejected": 0.04302050173282623, "rewards/accuracies": 0.6714844107627869, "rewards/margins": 0.049354877322912216, "logps/chosen": -385.71148681640625, "logps/rejected": -205.10792541503906, "logits/chosen": -2.2532808780670166, "logits/rejected": -2.0951650142669678, "stability/response_length_mean": 433.79998779296875, "stability/response_length_std": 126.92412567138672, "stability/response_length_var": 16109.7333984375, "stability/token_entropy_mean": 6.706250190734863, "stability/repetition_rate_mean": 0.8936500549316406}
+{"current_steps": 130, "total_steps": 196, "loss": 0.6698, "accuracy": 0.682812511920929, "lr": 9.99030814866488e-07, "epoch": 0.6653870761356366, "percentage": 66.33, "elapsed_time": "0:48:01", "remaining_time": "0:24:23", "rewards/chosen": 0.09967245161533356, "rewards/rejected": 0.042877569794654846, "rewards/accuracies": 0.682812511920929, "rewards/margins": 0.056794874370098114, "logps/chosen": -375.5982360839844, "logps/rejected": -203.75221252441406, "logits/chosen": -2.244988203048706, "logits/rejected": -2.096203565597534, "stability/response_length_mean": 296.1000061035156, "stability/response_length_std": 167.77593994140625, "stability/response_length_var": 28148.767578125, "stability/token_entropy_mean": 6.009375095367432, "stability/repetition_rate_mean": 0.8148199319839478}
+{"current_steps": 135, "total_steps": 196, "loss": 0.6672, "accuracy": 0.6800780892372131, "lr": 9.986679638549358e-07, "epoch": 0.690978886756238, "percentage": 68.88, "elapsed_time": "0:49:53", "remaining_time": "0:22:32", "rewards/chosen": 0.11093107610940933, "rewards/rejected": 0.04796404391527176, "rewards/accuracies": 0.6800780892372131, "rewards/margins": 0.06296703219413757, "logps/chosen": -387.3779602050781, "logps/rejected": -213.74169921875, "logits/chosen": -2.2671823501586914, "logits/rejected": -2.107846736907959, "stability/response_length_mean": 596.4000244140625, "stability/response_length_std": 291.9129333496094, "stability/response_length_var": 85213.15625, "stability/token_entropy_mean": 6.521874904632568, "stability/repetition_rate_mean": 0.8634861707687378}
+{"current_steps": 140, "total_steps": 196, "loss": 0.6637, "accuracy": 0.6910156607627869, "lr": 9.982476273321175e-07, "epoch": 0.7165706973768394, "percentage": 71.43, "elapsed_time": "0:51:42", "remaining_time": "0:20:41", "rewards/chosen": 0.11952749639749527, "rewards/rejected": 0.04787244647741318, "rewards/accuracies": 0.6910156607627869, "rewards/margins": 0.07165504992008209, "logps/chosen": -389.03033447265625, "logps/rejected": -211.16603088378906, "logits/chosen": -2.2552778720855713, "logits/rejected": -2.0999696254730225, "stability/response_length_mean": 173.89999389648438, "stability/response_length_std": 70.74908447265625, "stability/response_length_var": 5005.43310546875, "stability/token_entropy_mean": 6.634375095367432, "stability/repetition_rate_mean": 0.8619750142097473}
+{"current_steps": 145, "total_steps": 196, "loss": 0.6611, "accuracy": 0.6863281726837158, "lr": 9.977698537536417e-07, "epoch": 0.7421625079974408, "percentage": 73.98, "elapsed_time": "0:53:34", "remaining_time": "0:18:50", "rewards/chosen": 0.1324867308139801, "rewards/rejected": 0.05460155010223389, "rewards/accuracies": 0.6863281726837158, "rewards/margins": 0.07788518816232681, "logps/chosen": -387.7881164550781, "logps/rejected": -203.63453674316406, "logits/chosen": -2.2668190002441406, "logits/rejected": -2.0994153022766113, "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": 0.9642826914787292}
+{"current_steps": 150, "total_steps": 196, "loss": 0.6593, "accuracy": 0.6628906726837158, "lr": 9.972346981963546e-07, "epoch": 0.7677543186180422, "percentage": 76.53, "elapsed_time": "0:55:26", "remaining_time": "0:17:00", "rewards/chosen": 0.14231964945793152, "rewards/rejected": 0.0561322346329689, "rewards/accuracies": 0.6628906726837158, "rewards/margins": 0.08618739247322083, "logps/chosen": -373.8421630859375, "logps/rejected": -211.25048828125, "logits/chosen": -2.2341678142547607, "logits/rejected": -2.091710090637207, "stability/response_length_mean": 654.2000122070312, "stability/response_length_std": 372.5482177734375, "stability/response_length_var": 138792.171875, "stability/token_entropy_mean": 6.565625190734863, "stability/repetition_rate_mean": 0.8522264361381531}
+{"current_steps": 155, "total_steps": 196, "loss": 0.6549, "accuracy": 0.6792968511581421, "lr": 9.966422223519885e-07, "epoch": 0.7933461292386437, "percentage": 79.08, "elapsed_time": "0:57:17", "remaining_time": "0:15:09", "rewards/chosen": 0.15822532773017883, "rewards/rejected": 0.06434821337461472, "rewards/accuracies": 0.6792968511581421, "rewards/margins": 0.09387710690498352, "logps/chosen": -377.3822326660156, "logps/rejected": -206.42324829101562, "logits/chosen": -2.259756565093994, "logits/rejected": -2.108833074569702, "stability/response_length_mean": 502.3999938964844, "stability/response_length_std": 200.0795440673828, "stability/response_length_var": 40031.8203125, "stability/token_entropy_mean": 6.564062595367432, "stability/repetition_rate_mean": 0.8794426918029785}
+{"current_steps": 160, "total_steps": 196, "loss": 0.6519, "accuracy": 0.6902344226837158, "lr": 9.959924945200523e-07, "epoch": 0.818937939859245, "percentage": 81.63, "elapsed_time": "0:59:08", "remaining_time": "0:13:18", "rewards/chosen": 0.16477353870868683, "rewards/rejected": 0.060610540211200714, "rewards/accuracies": 0.6902344226837158, "rewards/margins": 0.10416299849748611, "logps/chosen": -383.5110778808594, "logps/rejected": -207.28126525878906, "logits/chosen": -2.253265619277954, "logits/rejected": -2.0869619846343994, "stability/response_length_mean": 870.4000244140625, "stability/response_length_std": 147.6581573486328, "stability/response_length_var": 21802.93359375, "stability/token_entropy_mean": 7.25, "stability/repetition_rate_mean": 0.9666600227355957}
+{"current_steps": 165, "total_steps": 196, "loss": 0.6502, "accuracy": 0.6910156607627869, "lr": 9.952855895999568e-07, "epoch": 0.8445297504798465, "percentage": 84.18, "elapsed_time": "1:00:59", "remaining_time": "0:11:27", "rewards/chosen": 0.16530370712280273, "rewards/rejected": 0.05815241485834122, "rewards/accuracies": 0.6910156607627869, "rewards/margins": 0.10715127736330032, "logps/chosen": -362.42230224609375, "logps/rejected": -203.6447296142578, "logits/chosen": -2.238316535949707, "logits/rejected": -2.090843915939331, "stability/response_length_mean": 547.7000122070312, "stability/response_length_std": 249.60325622558594, "stability/response_length_var": 62301.7890625, "stability/token_entropy_mean": 6.614062309265137, "stability/repetition_rate_mean": 0.8666970133781433}
+{"current_steps": 170, "total_steps": 196, "loss": 0.6447, "accuracy": 0.7003906965255737, "lr": 9.9452158908238e-07, "epoch": 0.8701215611004478, "percentage": 86.73, "elapsed_time": "1:02:50", "remaining_time": "0:09:36", "rewards/chosen": 0.18038088083267212, "rewards/rejected": 0.05830354988574982, "rewards/accuracies": 0.7003906965255737, "rewards/margins": 0.1220773383975029, "logps/chosen": -376.9706115722656, "logps/rejected": -203.80648803710938, "logits/chosen": -2.263615131378174, "logits/rejected": -2.0976061820983887, "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": 0.9128576517105103}
+{"current_steps": 175, "total_steps": 196, "loss": 0.6424, "accuracy": 0.693359375, "lr": 9.937005810398745e-07, "epoch": 0.8957133717210493, "percentage": 89.29, "elapsed_time": "1:04:40", "remaining_time": "0:07:45", "rewards/chosen": 0.18881221115589142, "rewards/rejected": 0.05759861320257187, "rewards/accuracies": 0.693359375, "rewards/margins": 0.13121359050273895, "logps/chosen": -380.500732421875, "logps/rejected": -212.28701782226562, "logits/chosen": -2.248379945755005, "logits/rejected": -2.092766046524048, "stability/response_length_mean": 460.6000061035156, "stability/response_length_std": 165.9894256591797, "stability/response_length_var": 27552.48828125, "stability/token_entropy_mean": 6.956250190734863, "stability/repetition_rate_mean": 0.9426466226577759}
+{"current_steps": 180, "total_steps": 196, "loss": 0.6404, "accuracy": 0.6871094107627869, "lr": 9.928226601167138e-07, "epoch": 0.9213051823416507, "percentage": 91.84, "elapsed_time": "1:06:32", "remaining_time": "0:05:54", "rewards/chosen": 0.20346775650978088, "rewards/rejected": 0.06857535243034363, "rewards/accuracies": 0.6871094107627869, "rewards/margins": 0.13489243388175964, "logps/chosen": -387.04644775390625, "logps/rejected": -212.97767639160156, "logits/chosen": -2.247504472732544, "logits/rejected": -2.0860483646392822, "stability/response_length_mean": 716.2000122070312, "stability/response_length_std": 329.3815002441406, "stability/response_length_var": 108492.1796875, "stability/token_entropy_mean": 6.510937690734863, "stability/repetition_rate_mean": 0.8947866559028625}
+{"current_steps": 185, "total_steps": 196, "loss": 0.6393, "accuracy": 0.6832031607627869, "lr": 9.918879275179817e-07, "epoch": 0.946896992962252, "percentage": 94.39, "elapsed_time": "1:08:23", "remaining_time": "0:04:03", "rewards/chosen": 0.21010692417621613, "rewards/rejected": 0.06822583824396133, "rewards/accuracies": 0.6832031607627869, "rewards/margins": 0.1418810933828354, "logps/chosen": -374.5255126953125, "logps/rejected": -210.45721435546875, "logits/chosen": -2.2497382164001465, "logits/rejected": -2.098439931869507, "stability/response_length_mean": 486.3999938964844, "stability/response_length_std": 213.8624725341797, "stability/response_length_var": 45737.15625, "stability/token_entropy_mean": 6.560937404632568, "stability/repetition_rate_mean": 0.881879985332489}
+{"current_steps": 190, "total_steps": 196, "loss": 0.6319, "accuracy": 0.7000000476837158, "lr": 9.90896490997906e-07, "epoch": 0.9724888035828535, "percentage": 96.94, "elapsed_time": "1:10:14", "remaining_time": "0:02:13", "rewards/chosen": 0.2151639461517334, "rewards/rejected": 0.05474931001663208, "rewards/accuracies": 0.7000000476837158, "rewards/margins": 0.16041463613510132, "logps/chosen": -377.6812744140625, "logps/rejected": -206.32400512695312, "logits/chosen": -2.233537435531616, "logits/rejected": -2.077605724334717, "stability/response_length_mean": 736.7999877929688, "stability/response_length_std": 246.57106018066406, "stability/response_length_var": 60797.2890625, "stability/token_entropy_mean": 6.928124904632568, "stability/repetition_rate_mean": 0.9174295663833618}
+{"current_steps": 195, "total_steps": 196, "loss": 0.6318, "accuracy": 0.7019531726837158, "lr": 9.898484648474362e-07, "epoch": 0.9980806142034548, "percentage": 99.49, "elapsed_time": "1:12:03", "remaining_time": "0:00:22", "rewards/chosen": 0.21875369548797607, "rewards/rejected": 0.05598779395222664, "rewards/accuracies": 0.7019531726837158, "rewards/margins": 0.16276590526103973, "logps/chosen": -361.8087158203125, "logps/rejected": -208.4525604248047, "logits/chosen": -2.2502989768981934, "logits/rejected": -2.0969607830047607, "stability/response_length_mean": 356.3999938964844, "stability/response_length_std": 207.94882202148438, "stability/response_length_var": 43242.7109375, "stability/token_entropy_mean": 5.948437690734863, "stability/repetition_rate_mean": 0.7733734846115112}
+{"current_steps": 196, "total_steps": 196, "epoch": 1.0, "percentage": 100.0, "elapsed_time": "1:12:19", "remaining_time": "0:00:00"}

mistral7b_dpo_sft_ftx0.001/trainer_state.json

@@ -0,0 +1,857 @@
+{
+  "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": 1.70222008228302,
+      "learning_rate": 4e-08,
+      "logits/chosen": -2.273049831390381,
+      "logits/rejected": -2.1233928203582764,
+      "logps/chosen": -375.07659912109375,
+      "logps/rejected": -210.87295532226562,
+      "loss": 0.6959,
+      "metrics/advantage_var": 0.04196161404252052,
+      "rewards/accuracies": 0.29609373211860657,
+      "rewards/chosen": -0.00011618183634709567,
+      "rewards/margins": -0.00025905348593369126,
+      "rewards/rejected": 0.00014287166413851082,
+      "step": 5
+    },
+    {
+      "epoch": 0.05118362124120281,
+      "grad_norm": 1.7635924816131592,
+      "learning_rate": 9e-08,
+      "logits/chosen": -2.267812967300415,
+      "logits/rejected": -2.1164932250976562,
+      "logps/chosen": -382.0225830078125,
+      "logps/rejected": -202.45101928710938,
+      "loss": 0.6951,
+      "metrics/advantage_var": 0.07391166687011719,
+      "rewards/accuracies": 0.5093749761581421,
+      "rewards/chosen": 0.0020605388563126326,
+      "rewards/margins": 0.001415093895047903,
+      "rewards/rejected": 0.0006454450776800513,
+      "stability/repetition_rate_mean": 0.9129446148872375,
+      "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": 1.8181178569793701,
+      "learning_rate": 1.4e-07,
+      "logits/chosen": -2.2609243392944336,
+      "logits/rejected": -2.099472761154175,
+      "logps/chosen": -389.2162170410156,
+      "logps/rejected": -207.0449981689453,
+      "loss": 0.6963,
+      "metrics/advantage_var": 0.07241873443126678,
+      "rewards/accuracies": 0.4898437559604645,
+      "rewards/chosen": 0.000939845631364733,
+      "rewards/margins": -0.0009244689717888832,
+      "rewards/rejected": 0.0018643148941919208,
+      "stability/repetition_rate_mean": 0.8466346859931946,
+      "stability/response_length_mean": 345.70001220703125,
+      "stability/response_length_std": 212.67974853515625,
+      "stability/response_length_var": 45232.6796875,
+      "stability/token_entropy_mean": 6.3515625,
+      "step": 15
+    },
+    {
+      "epoch": 0.10236724248240563,
+      "grad_norm": 1.7609531879425049,
+      "learning_rate": 1.8999999999999998e-07,
+      "logits/chosen": -2.2611491680145264,
+      "logits/rejected": -2.1143441200256348,
+      "logps/chosen": -368.7252502441406,
+      "logps/rejected": -198.56080627441406,
+      "loss": 0.6966,
+      "metrics/advantage_var": 0.06753375381231308,
+      "rewards/accuracies": 0.49375006556510925,
+      "rewards/chosen": -0.00011375403846614063,
+      "rewards/margins": -0.0015103723853826523,
+      "rewards/rejected": 0.001396618434228003,
+      "stability/repetition_rate_mean": 0.9300774335861206,
+      "stability/response_length_mean": 739.4000244140625,
+      "stability/response_length_std": 281.282470703125,
+      "stability/response_length_var": 79119.8203125,
+      "stability/token_entropy_mean": 6.934374809265137,
+      "step": 20
+    },
+    {
+      "epoch": 0.12795905310300704,
+      "grad_norm": 1.733801245689392,
+      "learning_rate": 2.4e-07,
+      "logits/chosen": -2.277261972427368,
+      "logits/rejected": -2.125120162963867,
+      "logps/chosen": -378.173828125,
+      "logps/rejected": -205.8581085205078,
+      "loss": 0.6962,
+      "metrics/advantage_var": 0.07016502320766449,
+      "rewards/accuracies": 0.49179691076278687,
+      "rewards/chosen": 0.0011066627921536565,
+      "rewards/margins": -0.0008675385033711791,
+      "rewards/rejected": 0.0019742015283554792,
+      "stability/repetition_rate_mean": 0.9150856137275696,
+      "stability/response_length_mean": 658.5999755859375,
+      "stability/response_length_std": 288.0155029296875,
+      "stability/response_length_var": 82952.9296875,
+      "stability/token_entropy_mean": 7.012499809265137,
+      "step": 25
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 1.8615390062332153,
+      "learning_rate": 2.9e-07,
+      "logits/chosen": -2.2703747749328613,
+      "logits/rejected": -2.123135805130005,
+      "logps/chosen": -373.9203796386719,
+      "logps/rejected": -208.94192504882812,
+      "loss": 0.6955,
+      "metrics/advantage_var": 0.07473797351121902,
+      "rewards/accuracies": 0.4957031309604645,
+      "rewards/chosen": 0.0009970426326617599,
+      "rewards/margins": 0.0005779300117865205,
+      "rewards/rejected": 0.00041911250445991755,
+      "stability/repetition_rate_mean": 0.8125225901603699,
+      "stability/response_length_mean": 356.0,
+      "stability/response_length_std": 249.6766815185547,
+      "stability/response_length_var": 62338.4453125,
+      "stability/token_entropy_mean": 6.229687690734863,
+      "step": 30
+    },
+    {
+      "epoch": 0.17914267434420986,
+      "grad_norm": 1.722904920578003,
+      "learning_rate": 3.4000000000000003e-07,
+      "logits/chosen": -2.2611172199249268,
+      "logits/rejected": -2.1200122833251953,
+      "logps/chosen": -366.32049560546875,
+      "logps/rejected": -203.98924255371094,
+      "loss": 0.6955,
+      "metrics/advantage_var": 0.06772881746292114,
+      "rewards/accuracies": 0.5160156488418579,
+      "rewards/chosen": 0.002536697080358863,
+      "rewards/margins": 0.0006712052272632718,
+      "rewards/rejected": 0.0018654920859262347,
+      "stability/repetition_rate_mean": 0.9350665807723999,
+      "stability/response_length_mean": 606.2000122070312,
+      "stability/response_length_std": 394.1366882324219,
+      "stability/response_length_var": 155343.734375,
+      "stability/token_entropy_mean": 7.150000095367432,
+      "step": 35
+    },
+    {
+      "epoch": 0.20473448496481125,
+      "grad_norm": 1.8076204061508179,
+      "learning_rate": 3.8999999999999997e-07,
+      "logits/chosen": -2.2763009071350098,
+      "logits/rejected": -2.122572422027588,
+      "logps/chosen": -369.1820373535156,
+      "logps/rejected": -197.3726043701172,
+      "loss": 0.6956,
+      "metrics/advantage_var": 0.06771942228078842,
+      "rewards/accuracies": 0.516796886920929,
+      "rewards/chosen": 0.0037511326372623444,
+      "rewards/margins": 0.0003795797238126397,
+      "rewards/rejected": 0.003371552564203739,
+      "stability/repetition_rate_mean": 0.9299213290214539,
+      "stability/response_length_mean": 547.2000122070312,
+      "stability/response_length_std": 290.8897399902344,
+      "stability/response_length_var": 84616.84375,
+      "stability/token_entropy_mean": 6.993750095367432,
+      "step": 40
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 1.8154666423797607,
+      "learning_rate": 4.3999999999999997e-07,
+      "logits/chosen": -2.2571003437042236,
+      "logits/rejected": -2.107235908508301,
+      "logps/chosen": -385.1767578125,
+      "logps/rejected": -212.5702667236328,
+      "loss": 0.6956,
+      "metrics/advantage_var": 0.07769347727298737,
+      "rewards/accuracies": 0.5101562142372131,
+      "rewards/chosen": 0.00422213040292263,
+      "rewards/margins": 0.00046065542846918106,
+      "rewards/rejected": 0.0037614754401147366,
+      "stability/repetition_rate_mean": 0.8824867010116577,
+      "stability/response_length_mean": 366.5,
+      "stability/response_length_std": 118.86243438720703,
+      "stability/response_length_var": 14128.27734375,
+      "stability/token_entropy_mean": 6.659375190734863,
+      "step": 45
+    },
+    {
+      "epoch": 0.2559181062060141,
+      "grad_norm": 1.8462119102478027,
+      "learning_rate": 4.9e-07,
+      "logits/chosen": -2.264322280883789,
+      "logits/rejected": -2.109330415725708,
+      "logps/chosen": -378.2926940917969,
+      "logps/rejected": -207.4957275390625,
+      "loss": 0.6943,
+      "metrics/advantage_var": 0.07516211271286011,
+      "rewards/accuracies": 0.53125,
+      "rewards/chosen": 0.006509767379611731,
+      "rewards/margins": 0.00300016300752759,
+      "rewards/rejected": 0.0035096043720841408,
+      "stability/repetition_rate_mean": 0.8690369725227356,
+      "stability/response_length_mean": 384.1000061035156,
+      "stability/response_length_std": 178.4809112548828,
+      "stability/response_length_var": 31855.43359375,
+      "stability/token_entropy_mean": 6.496874809265137,
+      "step": 50
+    },
+    {
+      "epoch": 0.28150991682661547,
+      "grad_norm": 1.9448425769805908,
+      "learning_rate": 5.4e-07,
+      "logits/chosen": -2.2531988620758057,
+      "logits/rejected": -2.094651937484741,
+      "logps/chosen": -393.789306640625,
+      "logps/rejected": -210.3695068359375,
+      "loss": 0.6933,
+      "metrics/advantage_var": 0.07928746938705444,
+      "rewards/accuracies": 0.5472656488418579,
+      "rewards/chosen": 0.009074963629245758,
+      "rewards/margins": 0.005103425122797489,
+      "rewards/rejected": 0.003971538506448269,
+      "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": 1.8275038003921509,
+      "learning_rate": 5.9e-07,
+      "logits/chosen": -2.273451805114746,
+      "logits/rejected": -2.1119141578674316,
+      "logps/chosen": -387.7010803222656,
+      "logps/rejected": -206.8184051513672,
+      "loss": 0.6935,
+      "metrics/advantage_var": 0.07989054918289185,
+      "rewards/accuracies": 0.54296875,
+      "rewards/chosen": 0.0105532705783844,
+      "rewards/margins": 0.004686878528445959,
+      "rewards/rejected": 0.005866392515599728,
+      "stability/repetition_rate_mean": 0.8754798173904419,
+      "stability/response_length_mean": 655.2999877929688,
+      "stability/response_length_std": 282.005126953125,
+      "stability/response_length_var": 79526.8984375,
+      "stability/token_entropy_mean": 6.606249809265137,
+      "step": 60
+    },
+    {
+      "epoch": 0.3326935380678183,
+      "grad_norm": 1.697847843170166,
+      "learning_rate": 6.4e-07,
+      "logits/chosen": -2.2514572143554688,
+      "logits/rejected": -2.0982723236083984,
+      "logps/chosen": -377.4744567871094,
+      "logps/rejected": -204.39608764648438,
+      "loss": 0.6929,
+      "metrics/advantage_var": 0.08167043328285217,
+      "rewards/accuracies": 0.561718761920929,
+      "rewards/chosen": 0.012515384703874588,
+      "rewards/margins": 0.005867896601557732,
+      "rewards/rejected": 0.006647488102316856,
+      "stability/repetition_rate_mean": 0.916172981262207,
+      "stability/response_length_mean": 306.3999938964844,
+      "stability/response_length_std": 141.7236785888672,
+      "stability/response_length_var": 20085.599609375,
+      "stability/token_entropy_mean": 6.837500095367432,
+      "step": 65
+    },
+    {
+      "epoch": 0.3582853486884197,
+      "grad_norm": 1.9988712072372437,
+      "learning_rate": 6.9e-07,
+      "logits/chosen": -2.271535873413086,
+      "logits/rejected": -2.1066555976867676,
+      "logps/chosen": -393.74591064453125,
+      "logps/rejected": -206.18272399902344,
+      "loss": 0.692,
+      "metrics/advantage_var": 0.09693735092878342,
+      "rewards/accuracies": 0.5687500238418579,
+      "rewards/chosen": 0.016287218779325485,
+      "rewards/margins": 0.00758810481056571,
+      "rewards/rejected": 0.008699115365743637,
+      "stability/repetition_rate_mean": 0.9707034230232239,
+      "stability/response_length_mean": 472.70001220703125,
+      "stability/response_length_std": 159.89791870117188,
+      "stability/response_length_var": 25567.34375,
+      "stability/token_entropy_mean": 7.21875,
+      "step": 70
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 1.882232666015625,
+      "learning_rate": 7.4e-07,
+      "logits/chosen": -2.2751991748809814,
+      "logits/rejected": -2.116259813308716,
+      "logps/chosen": -381.05670166015625,
+      "logps/rejected": -207.9344482421875,
+      "loss": 0.6915,
+      "metrics/advantage_var": 0.10063213855028152,
+      "rewards/accuracies": 0.5695313215255737,
+      "rewards/chosen": 0.019038334488868713,
+      "rewards/margins": 0.008797663263976574,
+      "rewards/rejected": 0.010240670293569565,
+      "stability/repetition_rate_mean": 0.9109193086624146,
+      "stability/response_length_mean": 527.0,
+      "stability/response_length_std": 293.7500915527344,
+      "stability/response_length_var": 86289.109375,
+      "stability/token_entropy_mean": 6.035937309265137,
+      "step": 75
+    },
+    {
+      "epoch": 0.4094689699296225,
+      "grad_norm": 1.8204421997070312,
+      "learning_rate": 7.9e-07,
+      "logits/chosen": -2.261033058166504,
+      "logits/rejected": -2.103379726409912,
+      "logps/chosen": -382.4222106933594,
+      "logps/rejected": -208.96932983398438,
+      "loss": 0.6916,
+      "metrics/advantage_var": 0.11322233080863953,
+      "rewards/accuracies": 0.5718750357627869,
+      "rewards/chosen": 0.021087391301989555,
+      "rewards/margins": 0.008481135591864586,
+      "rewards/rejected": 0.012606254778802395,
+      "stability/repetition_rate_mean": 0.9449335336685181,
+      "stability/response_length_mean": 635.5,
+      "stability/response_length_std": 262.53900146484375,
+      "stability/response_length_var": 68926.71875,
+      "stability/token_entropy_mean": 7.178124904632568,
+      "step": 80
+    },
+    {
+      "epoch": 0.4350607805502239,
+      "grad_norm": 1.8702808618545532,
+      "learning_rate": 8.399999999999999e-07,
+      "logits/chosen": -2.258106231689453,
+      "logits/rejected": -2.1010942459106445,
+      "logps/chosen": -387.320068359375,
+      "logps/rejected": -208.63490295410156,
+      "loss": 0.6896,
+      "metrics/advantage_var": 0.13779063522815704,
+      "rewards/accuracies": 0.592578113079071,
+      "rewards/chosen": 0.02714032307267189,
+      "rewards/margins": 0.012744016945362091,
+      "rewards/rejected": 0.014396305195987225,
+      "stability/repetition_rate_mean": 0.9021533727645874,
+      "stability/response_length_mean": 600.0999755859375,
+      "stability/response_length_std": 349.2030334472656,
+      "stability/response_length_var": 121942.765625,
+      "stability/token_entropy_mean": 6.732031345367432,
+      "step": 85
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 1.8974494934082031,
+      "learning_rate": 8.9e-07,
+      "logits/chosen": -2.2613511085510254,
+      "logits/rejected": -2.113387107849121,
+      "logps/chosen": -375.201904296875,
+      "logps/rejected": -206.21514892578125,
+      "loss": 0.6886,
+      "metrics/advantage_var": 0.15796051919460297,
+      "rewards/accuracies": 0.6050781607627869,
+      "rewards/chosen": 0.03241811320185661,
+      "rewards/margins": 0.014857406727969646,
+      "rewards/rejected": 0.017560705542564392,
+      "stability/repetition_rate_mean": 0.9763043522834778,
+      "stability/response_length_mean": 692.9000244140625,
+      "stability/response_length_std": 127.31452941894531,
+      "stability/response_length_var": 16208.9892578125,
+      "stability/token_entropy_mean": 7.112500190734863,
+      "step": 90
+    },
+    {
+      "epoch": 0.48624440179142675,
+      "grad_norm": 1.8197991847991943,
+      "learning_rate": 9.399999999999999e-07,
+      "logits/chosen": -2.2584359645843506,
+      "logits/rejected": -2.1117799282073975,
+      "logps/chosen": -366.8218078613281,
+      "logps/rejected": -203.42640686035156,
+      "loss": 0.686,
+      "metrics/advantage_var": 0.20491762459278107,
+      "rewards/accuracies": 0.6281249523162842,
+      "rewards/chosen": 0.03878289833664894,
+      "rewards/margins": 0.02026660367846489,
+      "rewards/rejected": 0.0185162965208292,
+      "stability/repetition_rate_mean": 0.9486184120178223,
+      "stability/response_length_mean": 727.0,
+      "stability/response_length_std": 335.0764465332031,
+      "stability/response_length_var": 112276.21875,
+      "stability/token_entropy_mean": 7.303124904632568,
+      "step": 95
+    },
+    {
+      "epoch": 0.5118362124120281,
+      "grad_norm": 1.816606879234314,
+      "learning_rate": 9.9e-07,
+      "logits/chosen": -2.2744193077087402,
+      "logits/rejected": -2.1189730167388916,
+      "logps/chosen": -384.5008239746094,
+      "logps/rejected": -210.61135864257812,
+      "loss": 0.6849,
+      "metrics/advantage_var": 0.26820605993270874,
+      "rewards/accuracies": 0.6503906846046448,
+      "rewards/chosen": 0.04569867625832558,
+      "rewards/margins": 0.022690609097480774,
+      "rewards/rejected": 0.023008069023489952,
+      "stability/repetition_rate_mean": 0.9514176249504089,
+      "stability/response_length_mean": 419.0,
+      "stability/response_length_std": 232.0225067138672,
+      "stability/response_length_var": 53834.4453125,
+      "stability/token_entropy_mean": 7.068749904632568,
+      "step": 100
+    },
+    {
+      "epoch": 0.5374280230326296,
+      "grad_norm": 1.7418373823165894,
+      "learning_rate": 9.999815554351224e-07,
+      "logits/chosen": -2.2584712505340576,
+      "logits/rejected": -2.1078848838806152,
+      "logps/chosen": -377.9892578125,
+      "logps/rejected": -209.6846160888672,
+      "loss": 0.6829,
+      "metrics/advantage_var": 0.34556183218955994,
+      "rewards/accuracies": 0.6472656726837158,
+      "rewards/chosen": 0.053732603788375854,
+      "rewards/margins": 0.026961183175444603,
+      "rewards/rejected": 0.02677142061293125,
+      "stability/repetition_rate_mean": 0.9020894169807434,
+      "stability/response_length_mean": 669.5,
+      "stability/response_length_std": 271.60888671875,
+      "stability/response_length_var": 73771.390625,
+      "stability/token_entropy_mean": 6.571093559265137,
+      "step": 105
+    },
+    {
+      "epoch": 0.5630198336532309,
+      "grad_norm": 1.6666159629821777,
+      "learning_rate": 9.999066267225444e-07,
+      "logits/chosen": -2.274813413619995,
+      "logits/rejected": -2.1170949935913086,
+      "logps/chosen": -385.2557678222656,
+      "logps/rejected": -206.71839904785156,
+      "loss": 0.6818,
+      "metrics/advantage_var": 0.43270087242126465,
+      "rewards/accuracies": 0.6460937857627869,
+      "rewards/chosen": 0.06057017669081688,
+      "rewards/margins": 0.0296179112046957,
+      "rewards/rejected": 0.03095226362347603,
+      "stability/repetition_rate_mean": 0.937423050403595,
+      "stability/response_length_mean": 255.60000610351562,
+      "stability/response_length_std": 114.30874633789062,
+      "stability/response_length_var": 13066.4892578125,
+      "stability/token_entropy_mean": 6.990624904632568,
+      "step": 110
+    },
+    {
+      "epoch": 0.5886116442738324,
+      "grad_norm": 1.798413872718811,
+      "learning_rate": 9.997740697079592e-07,
+      "logits/chosen": -2.2516493797302246,
+      "logits/rejected": -2.1109843254089355,
+      "logps/chosen": -372.72344970703125,
+      "logps/rejected": -213.9393310546875,
+      "loss": 0.6786,
+      "metrics/advantage_var": 0.6110946536064148,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": 0.07168412208557129,
+      "rewards/margins": 0.03667932376265526,
+      "rewards/rejected": 0.03500479832291603,
+      "stability/repetition_rate_mean": 0.8393692970275879,
+      "stability/response_length_mean": 499.29998779296875,
+      "stability/response_length_std": 282.31072998046875,
+      "stability/response_length_var": 79699.34375,
+      "stability/token_entropy_mean": 6.525000095367432,
+      "step": 115
+    },
+    {
+      "epoch": 0.6142034548944337,
+      "grad_norm": 1.7711139917373657,
+      "learning_rate": 9.995838996722914e-07,
+      "logits/chosen": -2.248577117919922,
+      "logits/rejected": -2.102853298187256,
+      "logps/chosen": -367.78424072265625,
+      "logps/rejected": -207.5933837890625,
+      "loss": 0.6757,
+      "metrics/advantage_var": 0.744879961013794,
+      "rewards/accuracies": 0.6742187738418579,
+      "rewards/chosen": 0.08194158971309662,
+      "rewards/margins": 0.04303481802344322,
+      "rewards/rejected": 0.0389067716896534,
+      "stability/repetition_rate_mean": 0.9263461828231812,
+      "stability/response_length_mean": 363.6000061035156,
+      "stability/response_length_std": 146.0701446533203,
+      "stability/response_length_var": 21336.48828125,
+      "stability/token_entropy_mean": 6.887499809265137,
+      "step": 120
+    },
+    {
+      "epoch": 0.6397952655150352,
+      "grad_norm": 1.7923411130905151,
+      "learning_rate": 9.993361385379874e-07,
+      "logits/chosen": -2.2532808780670166,
+      "logits/rejected": -2.0951650142669678,
+      "logps/chosen": -385.71148681640625,
+      "logps/rejected": -205.10792541503906,
+      "loss": 0.6732,
+      "metrics/advantage_var": 1.0200822353363037,
+      "rewards/accuracies": 0.6714844107627869,
+      "rewards/chosen": 0.09237537533044815,
+      "rewards/margins": 0.049354877322912216,
+      "rewards/rejected": 0.04302050173282623,
+      "stability/repetition_rate_mean": 0.8936500549316406,
+      "stability/response_length_mean": 433.79998779296875,
+      "stability/response_length_std": 126.92412567138672,
+      "stability/response_length_var": 16109.7333984375,
+      "stability/token_entropy_mean": 6.706250190734863,
+      "step": 125
+    },
+    {
+      "epoch": 0.6653870761356366,
+      "grad_norm": 1.8727540969848633,
+      "learning_rate": 9.99030814866488e-07,
+      "logits/chosen": -2.244988203048706,
+      "logits/rejected": -2.096203565597534,
+      "logps/chosen": -375.5982360839844,
+      "logps/rejected": -203.75221252441406,
+      "loss": 0.6698,
+      "metrics/advantage_var": 1.1310168504714966,
+      "rewards/accuracies": 0.682812511920929,
+      "rewards/chosen": 0.09967245161533356,
+      "rewards/margins": 0.056794874370098114,
+      "rewards/rejected": 0.042877569794654846,
+      "stability/repetition_rate_mean": 0.8148199319839478,
+      "stability/response_length_mean": 296.1000061035156,
+      "stability/response_length_std": 167.77593994140625,
+      "stability/response_length_var": 28148.767578125,
+      "stability/token_entropy_mean": 6.009375095367432,
+      "step": 130
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 1.7467596530914307,
+      "learning_rate": 9.986679638549358e-07,
+      "logits/chosen": -2.2671823501586914,
+      "logits/rejected": -2.107846736907959,
+      "logps/chosen": -387.3779602050781,
+      "logps/rejected": -213.74169921875,
+      "loss": 0.6672,
+      "metrics/advantage_var": 1.38654625415802,
+      "rewards/accuracies": 0.6800780892372131,
+      "rewards/chosen": 0.11093107610940933,
+      "rewards/margins": 0.06296703219413757,
+      "rewards/rejected": 0.04796404391527176,
+      "stability/repetition_rate_mean": 0.8634861707687378,
+      "stability/response_length_mean": 596.4000244140625,
+      "stability/response_length_std": 291.9129333496094,
+      "stability/response_length_var": 85213.15625,
+      "stability/token_entropy_mean": 6.521874904632568,
+      "step": 135
+    },
+    {
+      "epoch": 0.7165706973768394,
+      "grad_norm": 1.827934741973877,
+      "learning_rate": 9.982476273321175e-07,
+      "logits/chosen": -2.2552778720855713,
+      "logits/rejected": -2.0999696254730225,
+      "logps/chosen": -389.03033447265625,
+      "logps/rejected": -211.16603088378906,
+      "loss": 0.6637,
+      "metrics/advantage_var": 1.8156306743621826,
+      "rewards/accuracies": 0.6910156607627869,
+      "rewards/chosen": 0.11952749639749527,
+      "rewards/margins": 0.07165504992008209,
+      "rewards/rejected": 0.04787244647741318,
+      "stability/repetition_rate_mean": 0.8619750142097473,
+      "stability/response_length_mean": 173.89999389648438,
+      "stability/response_length_std": 70.74908447265625,
+      "stability/response_length_var": 5005.43310546875,
+      "stability/token_entropy_mean": 6.634375095367432,
+      "step": 140
+    },
+    {
+      "epoch": 0.7421625079974408,
+      "grad_norm": 1.7593910694122314,
+      "learning_rate": 9.977698537536417e-07,
+      "logits/chosen": -2.2668190002441406,
+      "logits/rejected": -2.0994153022766113,
+      "logps/chosen": -387.7881164550781,
+      "logps/rejected": -203.63453674316406,
+      "loss": 0.6611,
+      "metrics/advantage_var": 2.1796822547912598,
+      "rewards/accuracies": 0.6863281726837158,
+      "rewards/chosen": 0.1324867308139801,
+      "rewards/margins": 0.07788518816232681,
+      "rewards/rejected": 0.05460155010223389,
+      "stability/repetition_rate_mean": 0.9642826914787292,
+      "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": 1.8130406141281128,
+      "learning_rate": 9.972346981963546e-07,
+      "logits/chosen": -2.2341678142547607,
+      "logits/rejected": -2.091710090637207,
+      "logps/chosen": -373.8421630859375,
+      "logps/rejected": -211.25048828125,
+      "loss": 0.6593,
+      "metrics/advantage_var": 3.361509084701538,
+      "rewards/accuracies": 0.6628906726837158,
+      "rewards/chosen": 0.14231964945793152,
+      "rewards/margins": 0.08618739247322083,
+      "rewards/rejected": 0.0561322346329689,
+      "stability/repetition_rate_mean": 0.8522264361381531,
+      "stability/response_length_mean": 654.2000122070312,
+      "stability/response_length_std": 372.5482177734375,
+      "stability/response_length_var": 138792.171875,
+      "stability/token_entropy_mean": 6.565625190734863,
+      "step": 150
+    },
+    {
+      "epoch": 0.7933461292386437,
+      "grad_norm": 1.8375900983810425,
+      "learning_rate": 9.966422223519885e-07,
+      "logits/chosen": -2.259756565093994,
+      "logits/rejected": -2.108833074569702,
+      "logps/chosen": -377.3822326660156,
+      "logps/rejected": -206.42324829101562,
+      "loss": 0.6549,
+      "metrics/advantage_var": 2.9679672718048096,
+      "rewards/accuracies": 0.6792968511581421,
+      "rewards/chosen": 0.15822532773017883,
+      "rewards/margins": 0.09387710690498352,
+      "rewards/rejected": 0.06434821337461472,
+      "stability/repetition_rate_mean": 0.8794426918029785,
+      "stability/response_length_mean": 502.3999938964844,
+      "stability/response_length_std": 200.0795440673828,
+      "stability/response_length_var": 40031.8203125,
+      "stability/token_entropy_mean": 6.564062595367432,
+      "step": 155
+    },
+    {
+      "epoch": 0.818937939859245,
+      "grad_norm": 1.7692912817001343,
+      "learning_rate": 9.959924945200523e-07,
+      "logits/chosen": -2.253265619277954,
+      "logits/rejected": -2.0869619846343994,
+      "logps/chosen": -383.5110778808594,
+      "logps/rejected": -207.28126525878906,
+      "loss": 0.6519,
+      "metrics/advantage_var": 4.039853096008301,
+      "rewards/accuracies": 0.6902344226837158,
+      "rewards/chosen": 0.16477353870868683,
+      "rewards/margins": 0.10416299849748611,
+      "rewards/rejected": 0.060610540211200714,
+      "stability/repetition_rate_mean": 0.9666600227355957,
+      "stability/response_length_mean": 870.4000244140625,
+      "stability/response_length_std": 147.6581573486328,
+      "stability/response_length_var": 21802.93359375,
+      "stability/token_entropy_mean": 7.25,
+      "step": 160
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.6458988189697266,
+      "learning_rate": 9.952855895999568e-07,
+      "logits/chosen": -2.238316535949707,
+      "logits/rejected": -2.090843915939331,
+      "logps/chosen": -362.42230224609375,
+      "logps/rejected": -203.6447296142578,
+      "loss": 0.6502,
+      "metrics/advantage_var": 3.8923609256744385,
+      "rewards/accuracies": 0.6910156607627869,
+      "rewards/chosen": 0.16530370712280273,
+      "rewards/margins": 0.10715127736330032,
+      "rewards/rejected": 0.05815241485834122,
+      "stability/repetition_rate_mean": 0.8666970133781433,
+      "stability/response_length_mean": 547.7000122070312,
+      "stability/response_length_std": 249.60325622558594,
+      "stability/response_length_var": 62301.7890625,
+      "stability/token_entropy_mean": 6.614062309265137,
+      "step": 165
+    },
+    {
+      "epoch": 0.8701215611004478,
+      "grad_norm": 1.7035630941390991,
+      "learning_rate": 9.9452158908238e-07,
+      "logits/chosen": -2.263615131378174,
+      "logits/rejected": -2.0976061820983887,
+      "logps/chosen": -376.9706115722656,
+      "logps/rejected": -203.80648803710938,
+      "loss": 0.6447,
+      "metrics/advantage_var": 4.90874719619751,
+      "rewards/accuracies": 0.7003906965255737,
+      "rewards/chosen": 0.18038088083267212,
+      "rewards/margins": 0.1220773383975029,
+      "rewards/rejected": 0.05830354988574982,
+      "stability/repetition_rate_mean": 0.9128576517105103,
+      "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": 1.8031121492385864,
+      "learning_rate": 9.937005810398745e-07,
+      "logits/chosen": -2.248379945755005,
+      "logits/rejected": -2.092766046524048,
+      "logps/chosen": -380.500732421875,
+      "logps/rejected": -212.28701782226562,
+      "loss": 0.6424,
+      "metrics/advantage_var": 6.325986385345459,
+      "rewards/accuracies": 0.693359375,
+      "rewards/chosen": 0.18881221115589142,
+      "rewards/margins": 0.13121359050273895,
+      "rewards/rejected": 0.05759861320257187,
+      "stability/repetition_rate_mean": 0.9426466226577759,
+      "stability/response_length_mean": 460.6000061035156,
+      "stability/response_length_std": 165.9894256591797,
+      "stability/response_length_var": 27552.48828125,
+      "stability/token_entropy_mean": 6.956250190734863,
+      "step": 175
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 1.771286129951477,
+      "learning_rate": 9.928226601167138e-07,
+      "logits/chosen": -2.247504472732544,
+      "logits/rejected": -2.0860483646392822,
+      "logps/chosen": -387.04644775390625,
+      "logps/rejected": -212.97767639160156,
+      "loss": 0.6404,
+      "metrics/advantage_var": 5.818198204040527,
+      "rewards/accuracies": 0.6871094107627869,
+      "rewards/chosen": 0.20346775650978088,
+      "rewards/margins": 0.13489243388175964,
+      "rewards/rejected": 0.06857535243034363,
+      "stability/repetition_rate_mean": 0.8947866559028625,
+      "stability/response_length_mean": 716.2000122070312,
+      "stability/response_length_std": 329.3815002441406,
+      "stability/response_length_var": 108492.1796875,
+      "stability/token_entropy_mean": 6.510937690734863,
+      "step": 180
+    },
+    {
+      "epoch": 0.946896992962252,
+      "grad_norm": 1.6673380136489868,
+      "learning_rate": 9.918879275179817e-07,
+      "logits/chosen": -2.2497382164001465,
+      "logits/rejected": -2.098439931869507,
+      "logps/chosen": -374.5255126953125,
+      "logps/rejected": -210.45721435546875,
+      "loss": 0.6393,
+      "metrics/advantage_var": 7.5517191886901855,
+      "rewards/accuracies": 0.6832031607627869,
+      "rewards/chosen": 0.21010692417621613,
+      "rewards/margins": 0.1418810933828354,
+      "rewards/rejected": 0.06822583824396133,
+      "stability/repetition_rate_mean": 0.881879985332489,
+      "stability/response_length_mean": 486.3999938964844,
+      "stability/response_length_std": 213.8624725341797,
+      "stability/response_length_var": 45737.15625,
+      "stability/token_entropy_mean": 6.560937404632568,
+      "step": 185
+    },
+    {
+      "epoch": 0.9724888035828535,
+      "grad_norm": 1.5536139011383057,
+      "learning_rate": 9.90896490997906e-07,
+      "logits/chosen": -2.233537435531616,
+      "logits/rejected": -2.077605724334717,
+      "logps/chosen": -377.6812744140625,
+      "logps/rejected": -206.32400512695312,
+      "loss": 0.6319,
+      "metrics/advantage_var": 8.11306095123291,
+      "rewards/accuracies": 0.7000000476837158,
+      "rewards/chosen": 0.2151639461517334,
+      "rewards/margins": 0.16041463613510132,
+      "rewards/rejected": 0.05474931001663208,
+      "stability/repetition_rate_mean": 0.9174295663833618,
+      "stability/response_length_mean": 736.7999877929688,
+      "stability/response_length_std": 246.57106018066406,
+      "stability/response_length_var": 60797.2890625,
+      "stability/token_entropy_mean": 6.928124904632568,
+      "step": 190
+    },
+    {
+      "epoch": 0.9980806142034548,
+      "grad_norm": 1.6391992568969727,
+      "learning_rate": 9.898484648474362e-07,
+      "logits/chosen": -2.2502989768981934,
+      "logits/rejected": -2.0969607830047607,
+      "logps/chosen": -361.8087158203125,
+      "logps/rejected": -208.4525604248047,
+      "loss": 0.6318,
+      "metrics/advantage_var": 8.544903755187988,
+      "rewards/accuracies": 0.7019531726837158,
+      "rewards/chosen": 0.21875369548797607,
+      "rewards/margins": 0.16276590526103973,
+      "rewards/rejected": 0.05598779395222664,
+      "stability/repetition_rate_mean": 0.7733734846115112,
+      "stability/response_length_mean": 356.3999938964844,
+      "stability/response_length_std": 207.94882202148438,
+      "stability/response_length_var": 43242.7109375,
+      "stability/token_entropy_mean": 5.948437690734863,
+      "step": 195
+    },
+    {
+      "epoch": 1.0,
+      "step": 196,
+      "total_flos": 7.072674331683193e+18,
+      "train_loss": 0.6755896362723136,
+      "train_runtime": 4339.5435,
+      "train_samples_per_second": 23.044,
+      "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_sft_ftx0.001/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4512cc16205733853f6f345ca6097937972575afbbdedf6e03f7b46d17f21d8
+size 8081