Upload folder using huggingface_hub

mistral7b_dpo_sft/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
+  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
+
+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/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": [
+    "up_proj",
+    "o_proj",
+    "v_proj",
+    "gate_proj",
+    "k_proj",
+    "down_proj",
+    "q_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

mistral7b_dpo_sft/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d09fe8c44b3b0830544b4fe8506ae9a989c31719ed826d8f03b73123cc2d173d
+size 42002584

mistral7b_dpo_sft/all_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "total_flos": 7.072674331683193e+18,
+    "train_loss": 1.8980152759016777,
+    "train_runtime": 4342.782,
+    "train_samples_per_second": 23.027,
+    "train_steps_per_second": 0.045
+}

mistral7b_dpo_sft/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/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/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": [
+    "up_proj",
+    "o_proj",
+    "v_proj",
+    "gate_proj",
+    "k_proj",
+    "down_proj",
+    "q_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

mistral7b_dpo_sft/checkpoint-196/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d09fe8c44b3b0830544b4fe8506ae9a989c31719ed826d8f03b73123cc2d173d
+size 42002584

mistral7b_dpo_sft/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/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 sha256:7f8f60873a91fc3ff16365d530ac9833977f7b85569ff96bd20bfbc12517668e
+size 31467333

mistral7b_dpo_sft/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 sha256:0c9039b0650a041805d18275cd84d36d5144e4d0d84bd1cdb45bd92a666dcd66
+size 31467333

mistral7b_dpo_sft/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 sha256:9ec759f7f1cdb2e4ab7bd64a176b5df18fe1762fd737ad4b996648f2237047cb
+size 31467333

mistral7b_dpo_sft/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:a1c5254c2e604601cabd106bb070332523d53a7523cf8a205d789ca6baf74b90
+size 31467397

mistral7b_dpo_sft/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 sha256:6fccdfa618adda9d36838ff0b341e2c121c724a74083f3a87143397101e9d551
+size 31467397

mistral7b_dpo_sft/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:a8f5c443ecc1a51f2b13957bb036bdf6f8cdbf5131a0f09c846e97a51e85ce54
+size 31467397

mistral7b_dpo_sft/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 sha256:6589c2f7fea3279770bc7727f79e12cc88f94daba18e581cb779a1bb44e83fc4
+size 31467397

mistral7b_dpo_sft/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 sha256:63dca69769f912a6c0c4bfba295ae3e9473f3b0b74aa58f30d5a466e5cfa65d8
+size 31467397

mistral7b_dpo_sft/checkpoint-196/global_step196/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7c10fe43c1eaf82ff428dd724a5f1a295e19fa934c83ee4eaee8e1bf6aa6374
+size 42288409

mistral7b_dpo_sft/checkpoint-196/latest

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

mistral7b_dpo_sft/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/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/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/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/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/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/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/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/checkpoint-196/scheduler.pt

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

mistral7b_dpo_sft/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/checkpoint-196/tokenizer.model

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

mistral7b_dpo_sft/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/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": 1.914654016494751,
+      "learning_rate": 4e-08,
+      "logits/chosen": -2.272939682006836,
+      "logits/rejected": -2.1233606338500977,
+      "logps/chosen": -375.065185546875,
+      "logps/rejected": -210.8690948486328,
+      "loss": 1.9254,
+      "metrics/advantage_var": 0.04025906324386597,
+      "rewards/accuracies": 0.29804688692092896,
+      "rewards/chosen": 0.0010315654799342155,
+      "rewards/margins": 0.0005036827642470598,
+      "rewards/rejected": 0.0005278826574794948,
+      "step": 5
+    },
+    {
+      "epoch": 0.05118362124120281,
+      "grad_norm": 1.9914286136627197,
+      "learning_rate": 9e-08,
+      "logits/chosen": -2.2681591510772705,
+      "logits/rejected": -2.1167702674865723,
+      "logps/chosen": -382.0304870605469,
+      "logps/rejected": -202.4506072998047,
+      "loss": 1.9158,
+      "metrics/advantage_var": 0.07169340550899506,
+      "rewards/accuracies": 0.5109375715255737,
+      "rewards/chosen": 0.0012701328378170729,
+      "rewards/margins": 0.0005846476415172219,
+      "rewards/rejected": 0.0006854852545075119,
+      "stability/repetition_rate_mean": 0.918988049030304,
+      "stability/response_length_mean": 602.7999877929688,
+      "stability/response_length_std": 295.1842346191406,
+      "stability/response_length_var": 87133.734375,
+      "stability/token_entropy_mean": 6.987500190734863,
+      "step": 10
+    },
+    {
+      "epoch": 0.07677543186180422,
+      "grad_norm": 2.024305582046509,
+      "learning_rate": 1.4e-07,
+      "logits/chosen": -2.2606420516967773,
+      "logits/rejected": -2.0991973876953125,
+      "logps/chosen": -389.207763671875,
+      "logps/rejected": -207.05108642578125,
+      "loss": 1.9201,
+      "metrics/advantage_var": 0.07511603087186813,
+      "rewards/accuracies": 0.501171886920929,
+      "rewards/chosen": 0.0017871337477117777,
+      "rewards/margins": 0.0005336535396054387,
+      "rewards/rejected": 0.0012534803245216608,
+      "stability/repetition_rate_mean": 0.852084755897522,
+      "stability/response_length_mean": 345.70001220703125,
+      "stability/response_length_std": 212.67974853515625,
+      "stability/response_length_var": 45232.6796875,
+      "stability/token_entropy_mean": 6.317187309265137,
+      "step": 15
+    },
+    {
+      "epoch": 0.10236724248240563,
+      "grad_norm": 1.9839270114898682,
+      "learning_rate": 1.8999999999999998e-07,
+      "logits/chosen": -2.2611918449401855,
+      "logits/rejected": -2.114309549331665,
+      "logps/chosen": -368.709716796875,
+      "logps/rejected": -198.56033325195312,
+      "loss": 1.9381,
+      "metrics/advantage_var": 0.06882387399673462,
+      "rewards/accuracies": 0.49726563692092896,
+      "rewards/chosen": 0.001441984553821385,
+      "rewards/margins": -3.1705712899565697e-07,
+      "rewards/rejected": 0.0014423017855733633,
+      "stability/repetition_rate_mean": 0.934249997138977,
+      "stability/response_length_mean": 739.4000244140625,
+      "stability/response_length_std": 281.282470703125,
+      "stability/response_length_var": 79119.8203125,
+      "stability/token_entropy_mean": 6.940625190734863,
+      "step": 20
+    },
+    {
+      "epoch": 0.12795905310300704,
+      "grad_norm": 2.0085482597351074,
+      "learning_rate": 2.4e-07,
+      "logits/chosen": -2.2772860527038574,
+      "logits/rejected": -2.125190019607544,
+      "logps/chosen": -378.1678161621094,
+      "logps/rejected": -205.85311889648438,
+      "loss": 1.9146,
+      "metrics/advantage_var": 0.06963051110506058,
+      "rewards/accuracies": 0.49296876788139343,
+      "rewards/chosen": 0.0017121940618380904,
+      "rewards/margins": -0.0007601291872560978,
+      "rewards/rejected": 0.00247232336550951,
+      "stability/repetition_rate_mean": 0.9178038835525513,
+      "stability/response_length_mean": 658.5999755859375,
+      "stability/response_length_std": 288.0155029296875,
+      "stability/response_length_var": 82952.9296875,
+      "stability/token_entropy_mean": 7.03125,
+      "step": 25
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 2.009305477142334,
+      "learning_rate": 2.9e-07,
+      "logits/chosen": -2.270566940307617,
+      "logits/rejected": -2.123317241668701,
+      "logps/chosen": -373.8929443359375,
+      "logps/rejected": -208.92208862304688,
+      "loss": 1.9331,
+      "metrics/advantage_var": 0.07247471809387207,
+      "rewards/accuracies": 0.516406238079071,
+      "rewards/chosen": 0.0037445530761033297,
+      "rewards/margins": 0.0013395380228757858,
+      "rewards/rejected": 0.002405015053227544,
+      "stability/repetition_rate_mean": 0.8267838358879089,
+      "stability/response_length_mean": 356.0,
+      "stability/response_length_std": 249.6766815185547,
+      "stability/response_length_var": 62338.4453125,
+      "stability/token_entropy_mean": 6.321875095367432,
+      "step": 30
+    },
+    {
+      "epoch": 0.17914267434420986,
+      "grad_norm": 1.9789156913757324,
+      "learning_rate": 3.4000000000000003e-07,
+      "logits/chosen": -2.261219024658203,
+      "logits/rejected": -2.1199522018432617,
+      "logps/chosen": -366.2962341308594,
+      "logps/rejected": -203.96969604492188,
+      "loss": 1.9336,
+      "metrics/advantage_var": 0.07087455689907074,
+      "rewards/accuracies": 0.510937511920929,
+      "rewards/chosen": 0.004964467138051987,
+      "rewards/margins": 0.0011428671423345804,
+      "rewards/rejected": 0.0038215997628867626,
+      "stability/repetition_rate_mean": 0.9363252520561218,
+      "stability/response_length_mean": 606.2000122070312,
+      "stability/response_length_std": 394.1366882324219,
+      "stability/response_length_var": 155343.734375,
+      "stability/token_entropy_mean": 7.153124809265137,
+      "step": 35
+    },
+    {
+      "epoch": 0.20473448496481125,
+      "grad_norm": 2.0971174240112305,
+      "learning_rate": 3.8999999999999997e-07,
+      "logits/chosen": -2.276848554611206,
+      "logits/rejected": -2.123274087905884,
+      "logps/chosen": -369.1612243652344,
+      "logps/rejected": -197.34510803222656,
+      "loss": 1.9288,
+      "metrics/advantage_var": 0.07022758573293686,
+      "rewards/accuracies": 0.4937499761581421,
+      "rewards/chosen": 0.005836769472807646,
+      "rewards/margins": -0.0002842957037501037,
+      "rewards/rejected": 0.006121065001934767,
+      "stability/repetition_rate_mean": 0.9261367917060852,
+      "stability/response_length_mean": 547.2000122070312,
+      "stability/response_length_std": 290.8897399902344,
+      "stability/response_length_var": 84616.84375,
+      "stability/token_entropy_mean": 6.974999904632568,
+      "step": 40
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 2.0916647911071777,
+      "learning_rate": 4.3999999999999997e-07,
+      "logits/chosen": -2.2577290534973145,
+      "logits/rejected": -2.1079652309417725,
+      "logps/chosen": -385.138671875,
+      "logps/rejected": -212.5291290283203,
+      "loss": 1.9343,
+      "metrics/advantage_var": 0.07318010926246643,
+      "rewards/accuracies": 0.49335938692092896,
+      "rewards/chosen": 0.008032082580029964,
+      "rewards/margins": 0.00015603320207446814,
+      "rewards/rejected": 0.007876048795878887,
+      "stability/repetition_rate_mean": 0.8869589567184448,
+      "stability/response_length_mean": 366.5,
+      "stability/response_length_std": 118.86243438720703,
+      "stability/response_length_var": 14128.27734375,
+      "stability/token_entropy_mean": 6.696875095367432,
+      "step": 45
+    },
+    {
+      "epoch": 0.2559181062060141,
+      "grad_norm": 2.067622423171997,
+      "learning_rate": 4.9e-07,
+      "logits/chosen": -2.265531539916992,
+      "logits/rejected": -2.1107473373413086,
+      "logps/chosen": -378.2447509765625,
+      "logps/rejected": -207.44895935058594,
+      "loss": 1.9417,
+      "metrics/advantage_var": 0.07468117028474808,
+      "rewards/accuracies": 0.524609386920929,
+      "rewards/chosen": 0.011302120983600616,
+      "rewards/margins": 0.00311595038510859,
+      "rewards/rejected": 0.00818617083132267,
+      "stability/repetition_rate_mean": 0.8679786920547485,
+      "stability/response_length_mean": 384.1000061035156,
+      "stability/response_length_std": 178.4809112548828,
+      "stability/response_length_var": 31855.43359375,
+      "stability/token_entropy_mean": 6.493750095367432,
+      "step": 50
+    },
+    {
+      "epoch": 0.28150991682661547,
+      "grad_norm": 2.2453255653381348,
+      "learning_rate": 5.4e-07,
+      "logits/chosen": -2.2538602352142334,
+      "logits/rejected": -2.095405101776123,
+      "logps/chosen": -393.7201232910156,
+      "logps/rejected": -210.2903289794922,
+      "loss": 1.9151,
+      "metrics/advantage_var": 0.08342276513576508,
+      "rewards/accuracies": 0.5433593988418579,
+      "rewards/chosen": 0.015990909188985825,
+      "rewards/margins": 0.00410159258171916,
+      "rewards/rejected": 0.011889317072927952,
+      "stability/repetition_rate_mean": 0.9476622343063354,
+      "stability/response_length_mean": 332.8999938964844,
+      "stability/response_length_std": 137.39193725585938,
+      "stability/response_length_var": 18876.544921875,
+      "stability/token_entropy_mean": 6.981249809265137,
+      "step": 55
+    },
+    {
+      "epoch": 0.30710172744721687,
+      "grad_norm": 2.1323249340057373,
+      "learning_rate": 5.9e-07,
+      "logits/chosen": -2.27402400970459,
+      "logits/rejected": -2.1127142906188965,
+      "logps/chosen": -387.6095275878906,
+      "logps/rejected": -206.73020935058594,
+      "loss": 1.9258,
+      "metrics/advantage_var": 0.08840229362249374,
+      "rewards/accuracies": 0.5367187857627869,
+      "rewards/chosen": 0.01970873773097992,
+      "rewards/margins": 0.005020757671445608,
+      "rewards/rejected": 0.014687979593873024,
+      "stability/repetition_rate_mean": 0.8755375146865845,
+      "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": 1.914339542388916,
+      "learning_rate": 6.4e-07,
+      "logits/chosen": -2.2534067630767822,
+      "logits/rejected": -2.1004819869995117,
+      "logps/chosen": -377.36798095703125,
+      "logps/rejected": -204.2776336669922,
+      "loss": 1.9516,
+      "metrics/advantage_var": 0.0927530974149704,
+      "rewards/accuracies": 0.5406250357627869,
+      "rewards/chosen": 0.023161547258496284,
+      "rewards/margins": 0.004670025315135717,
+      "rewards/rejected": 0.01849152334034443,
+      "stability/repetition_rate_mean": 0.9074932336807251,
+      "stability/response_length_mean": 306.3999938964844,
+      "stability/response_length_std": 141.7236785888672,
+      "stability/response_length_var": 20085.599609375,
+      "stability/token_entropy_mean": 6.787499904632568,
+      "step": 65
+    },
+    {
+      "epoch": 0.3582853486884197,
+      "grad_norm": 2.256079912185669,
+      "learning_rate": 6.9e-07,
+      "logits/chosen": -2.2728006839752197,
+      "logits/rejected": -2.1082444190979004,
+      "logps/chosen": -393.5992736816406,
+      "logps/rejected": -206.0418701171875,
+      "loss": 1.907,
+      "metrics/advantage_var": 0.10806500911712646,
+      "rewards/accuracies": 0.571093738079071,
+      "rewards/chosen": 0.030948856845498085,
+      "rewards/margins": 0.008165466599166393,
+      "rewards/rejected": 0.022783391177654266,
+      "stability/repetition_rate_mean": 0.970089316368103,
+      "stability/response_length_mean": 472.70001220703125,
+      "stability/response_length_std": 159.89791870117188,
+      "stability/response_length_var": 25567.34375,
+      "stability/token_entropy_mean": 7.224999904632568,
+      "step": 70
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 2.153575897216797,
+      "learning_rate": 7.4e-07,
+      "logits/chosen": -2.27734637260437,
+      "logits/rejected": -2.1185312271118164,
+      "logps/chosen": -380.871337890625,
+      "logps/rejected": -207.7386474609375,
+      "loss": 1.9254,
+      "metrics/advantage_var": 0.12323446571826935,
+      "rewards/accuracies": 0.5691406726837158,
+      "rewards/chosen": 0.0375763475894928,
+      "rewards/margins": 0.007752560544759035,
+      "rewards/rejected": 0.0298237893730402,
+      "stability/repetition_rate_mean": 0.9063584208488464,
+      "stability/response_length_mean": 527.0,
+      "stability/response_length_std": 293.7500915527344,
+      "stability/response_length_var": 86289.109375,
+      "stability/token_entropy_mean": 6.067187309265137,
+      "step": 75
+    },
+    {
+      "epoch": 0.4094689699296225,
+      "grad_norm": 2.0587069988250732,
+      "learning_rate": 7.9e-07,
+      "logits/chosen": -2.262965440750122,
+      "logits/rejected": -2.1055397987365723,
+      "logps/chosen": -382.1772766113281,
+      "logps/rejected": -208.73495483398438,
+      "loss": 1.9115,
+      "metrics/advantage_var": 0.14562615752220154,
+      "rewards/accuracies": 0.5769531726837158,
+      "rewards/chosen": 0.04558176547288895,
+      "rewards/margins": 0.009537179954349995,
+      "rewards/rejected": 0.036044586449861526,
+      "stability/repetition_rate_mean": 0.945895791053772,
+      "stability/response_length_mean": 635.5,
+      "stability/response_length_std": 262.53900146484375,
+      "stability/response_length_var": 68926.71875,
+      "stability/token_entropy_mean": 7.193749904632568,
+      "step": 80
+    },
+    {
+      "epoch": 0.4350607805502239,
+      "grad_norm": 2.1112332344055176,
+      "learning_rate": 8.399999999999999e-07,
+      "logits/chosen": -2.2605109214782715,
+      "logits/rejected": -2.1037368774414062,
+      "logps/chosen": -387.02532958984375,
+      "logps/rejected": -208.32882690429688,
+      "loss": 1.9267,
+      "metrics/advantage_var": 0.19675803184509277,
+      "rewards/accuracies": 0.5796875357627869,
+      "rewards/chosen": 0.056618086993694305,
+      "rewards/margins": 0.011614463292062283,
+      "rewards/rejected": 0.045003630220890045,
+      "stability/repetition_rate_mean": 0.9032164812088013,
+      "stability/response_length_mean": 600.0999755859375,
+      "stability/response_length_std": 349.2030334472656,
+      "stability/response_length_var": 121942.765625,
+      "stability/token_entropy_mean": 6.728906154632568,
+      "step": 85
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 2.1287057399749756,
+      "learning_rate": 8.9e-07,
+      "logits/chosen": -2.263827323913574,
+      "logits/rejected": -2.1160788536071777,
+      "logps/chosen": -374.8363342285156,
+      "logps/rejected": -205.86585998535156,
+      "loss": 1.9156,
+      "metrics/advantage_var": 0.23069456219673157,
+      "rewards/accuracies": 0.608593761920929,
+      "rewards/chosen": 0.06897309422492981,
+      "rewards/margins": 0.016483532264828682,
+      "rewards/rejected": 0.05248956382274628,
+      "stability/repetition_rate_mean": 0.9750186204910278,
+      "stability/response_length_mean": 692.9000244140625,
+      "stability/response_length_std": 127.31452941894531,
+      "stability/response_length_var": 16208.9892578125,
+      "stability/token_entropy_mean": 7.096875190734863,
+      "step": 90
+    },
+    {
+      "epoch": 0.48624440179142675,
+      "grad_norm": 2.0764377117156982,
+      "learning_rate": 9.399999999999999e-07,
+      "logits/chosen": -2.2615463733673096,
+      "logits/rejected": -2.115648031234741,
+      "logps/chosen": -366.39752197265625,
+      "logps/rejected": -203.00827026367188,
+      "loss": 1.91,
+      "metrics/advantage_var": 0.3129298686981201,
+      "rewards/accuracies": 0.6187500357627869,
+      "rewards/chosen": 0.08120815455913544,
+      "rewards/margins": 0.020878203213214874,
+      "rewards/rejected": 0.060329943895339966,
+      "stability/repetition_rate_mean": 0.9472225904464722,
+      "stability/response_length_mean": 727.0,
+      "stability/response_length_std": 335.0764465332031,
+      "stability/response_length_var": 112276.21875,
+      "stability/token_entropy_mean": 7.3125,
+      "step": 95
+    },
+    {
+      "epoch": 0.5118362124120281,
+      "grad_norm": 2.0339155197143555,
+      "learning_rate": 9.9e-07,
+      "logits/chosen": -2.2777926921844482,
+      "logits/rejected": -2.123100757598877,
+      "logps/chosen": -383.9757995605469,
+      "logps/rejected": -210.1046905517578,
+      "loss": 1.9049,
+      "metrics/advantage_var": 0.4217294156551361,
+      "rewards/accuracies": 0.6289063096046448,
+      "rewards/chosen": 0.09820398688316345,
+      "rewards/margins": 0.024528376758098602,
+      "rewards/rejected": 0.07367560267448425,
+      "stability/repetition_rate_mean": 0.9518136978149414,
+      "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": 1.9553194046020508,
+      "learning_rate": 9.999815554351224e-07,
+      "logits/chosen": -2.2625222206115723,
+      "logits/rejected": -2.112849473953247,
+      "logps/chosen": -377.36602783203125,
+      "logps/rejected": -209.0551300048828,
+      "loss": 1.9191,
+      "metrics/advantage_var": 0.5498611330986023,
+      "rewards/accuracies": 0.6265625357627869,
+      "rewards/chosen": 0.116055428981781,
+      "rewards/margins": 0.02633403241634369,
+      "rewards/rejected": 0.08972139656543732,
+      "stability/repetition_rate_mean": 0.8962067365646362,
+      "stability/response_length_mean": 669.5,
+      "stability/response_length_std": 271.60888671875,
+      "stability/response_length_var": 73771.390625,
+      "stability/token_entropy_mean": 6.586718559265137,
+      "step": 105
+    },
+    {
+      "epoch": 0.5630198336532309,
+      "grad_norm": 1.9055349826812744,
+      "learning_rate": 9.999066267225444e-07,
+      "logits/chosen": -2.279107093811035,
+      "logits/rejected": -2.122042655944824,
+      "logps/chosen": -384.4992370605469,
+      "logps/rejected": -205.9490203857422,
+      "loss": 1.9087,
+      "metrics/advantage_var": 0.6977776885032654,
+      "rewards/accuracies": 0.6167968511581421,
+      "rewards/chosen": 0.13622210919857025,
+      "rewards/margins": 0.0283319354057312,
+      "rewards/rejected": 0.10789017379283905,
+      "stability/repetition_rate_mean": 0.9286434054374695,
+      "stability/response_length_mean": 255.60000610351562,
+      "stability/response_length_std": 114.30874633789062,
+      "stability/response_length_var": 13066.4892578125,
+      "stability/token_entropy_mean": 6.981249809265137,
+      "step": 110
+    },
+    {
+      "epoch": 0.5886116442738324,
+      "grad_norm": 1.9901360273361206,
+      "learning_rate": 9.997740697079592e-07,
+      "logits/chosen": -2.2564876079559326,
+      "logits/rejected": -2.1166574954986572,
+      "logps/chosen": -371.8219299316406,
+      "logps/rejected": -213.05772399902344,
+      "loss": 1.9089,
+      "metrics/advantage_var": 0.9774861931800842,
+      "rewards/accuracies": 0.6238281726837158,
+      "rewards/chosen": 0.16183851659297943,
+      "rewards/margins": 0.03867293521761894,
+      "rewards/rejected": 0.12316557765007019,
+      "stability/repetition_rate_mean": 0.8308541178703308,
+      "stability/response_length_mean": 499.29998779296875,
+      "stability/response_length_std": 282.31072998046875,
+      "stability/response_length_var": 79699.34375,
+      "stability/token_entropy_mean": 6.537499904632568,
+      "step": 115
+    },
+    {
+      "epoch": 0.6142034548944337,
+      "grad_norm": 1.9900635480880737,
+      "learning_rate": 9.995838996722914e-07,
+      "logits/chosen": -2.253622055053711,
+      "logits/rejected": -2.108788251876831,
+      "logps/chosen": -366.7509460449219,
+      "logps/rejected": -206.54931640625,
+      "loss": 1.8995,
+      "metrics/advantage_var": 1.1828315258026123,
+      "rewards/accuracies": 0.6390625238418579,
+      "rewards/chosen": 0.18527083098888397,
+      "rewards/margins": 0.04195715859532356,
+      "rewards/rejected": 0.1433136910200119,
+      "stability/repetition_rate_mean": 0.9338710904121399,
+      "stability/response_length_mean": 363.6000061035156,
+      "stability/response_length_std": 146.0701446533203,
+      "stability/response_length_var": 21336.48828125,
+      "stability/token_entropy_mean": 6.934374809265137,
+      "step": 120
+    },
+    {
+      "epoch": 0.6397952655150352,
+      "grad_norm": 1.9649677276611328,
+      "learning_rate": 9.993361385379874e-07,
+      "logits/chosen": -2.2593374252319336,
+      "logits/rejected": -2.1024961471557617,
+      "logps/chosen": -384.51739501953125,
+      "logps/rejected": -203.9079132080078,
+      "loss": 1.8987,
+      "metrics/advantage_var": 1.6206504106521606,
+      "rewards/accuracies": 0.6304687857627869,
+      "rewards/chosen": 0.21179263293743134,
+      "rewards/margins": 0.04877074807882309,
+      "rewards/rejected": 0.16302187740802765,
+      "stability/repetition_rate_mean": 0.8942571878433228,
+      "stability/response_length_mean": 433.79998779296875,
+      "stability/response_length_std": 126.92412567138672,
+      "stability/response_length_var": 16109.7333984375,
+      "stability/token_entropy_mean": 6.699999809265137,
+      "step": 125
+    },
+    {
+      "epoch": 0.6653870761356366,
+      "grad_norm": 2.029193162918091,
+      "learning_rate": 9.99030814866488e-07,
+      "logits/chosen": -2.251539468765259,
+      "logits/rejected": -2.103938579559326,
+      "logps/chosen": -374.2728271484375,
+      "logps/rejected": -202.3927764892578,
+      "loss": 1.8884,
+      "metrics/advantage_var": 1.7878669500350952,
+      "rewards/accuracies": 0.639843761920929,
+      "rewards/chosen": 0.23221388459205627,
+      "rewards/margins": 0.0533922053873539,
+      "rewards/rejected": 0.17882166802883148,
+      "stability/repetition_rate_mean": 0.8264604806900024,
+      "stability/response_length_mean": 296.1000061035156,
+      "stability/response_length_std": 167.77593994140625,
+      "stability/response_length_var": 28148.767578125,
+      "stability/token_entropy_mean": 6.131249904632568,
+      "step": 130
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 1.9165802001953125,
+      "learning_rate": 9.986679638549358e-07,
+      "logits/chosen": -2.273317813873291,
+      "logits/rejected": -2.1149117946624756,
+      "logps/chosen": -385.8937072753906,
+      "logps/rejected": -212.23419189453125,
+      "loss": 1.8861,
+      "metrics/advantage_var": 2.076320171356201,
+      "rewards/accuracies": 0.647265613079071,
+      "rewards/chosen": 0.25935378670692444,
+      "rewards/margins": 0.06063650920987129,
+      "rewards/rejected": 0.19871726632118225,
+      "stability/repetition_rate_mean": 0.8641007542610168,
+      "stability/response_length_mean": 596.4000244140625,
+      "stability/response_length_std": 291.9129333496094,
+      "stability/response_length_var": 85213.15625,
+      "stability/token_entropy_mean": 6.550000190734863,
+      "step": 135
+    },
+    {
+      "epoch": 0.7165706973768394,
+      "grad_norm": 2.0632386207580566,
+      "learning_rate": 9.982476273321175e-07,
+      "logits/chosen": -2.262234926223755,
+      "logits/rejected": -2.1079959869384766,
+      "logps/chosen": -387.342041015625,
+      "logps/rejected": -209.46954345703125,
+      "loss": 1.8781,
+      "metrics/advantage_var": 2.7123324871063232,
+      "rewards/accuracies": 0.6578124761581421,
+      "rewards/chosen": 0.28835996985435486,
+      "rewards/margins": 0.07083673030138016,
+      "rewards/rejected": 0.2175232321023941,
+      "stability/repetition_rate_mean": 0.8578893542289734,
+      "stability/response_length_mean": 173.89999389648438,
+      "stability/response_length_std": 70.74908447265625,
+      "stability/response_length_var": 5005.43310546875,
+      "stability/token_entropy_mean": 6.659375190734863,
+      "step": 140
+    },
+    {
+      "epoch": 0.7421625079974408,
+      "grad_norm": 1.8859580755233765,
+      "learning_rate": 9.977698537536417e-07,
+      "logits/chosen": -2.2734627723693848,
+      "logits/rejected": -2.1078710556030273,
+      "logps/chosen": -385.9046630859375,
+      "logps/rejected": -201.69927978515625,
+      "loss": 1.8752,
+      "metrics/advantage_var": 3.1621975898742676,
+      "rewards/accuracies": 0.6351562738418579,
+      "rewards/chosen": 0.32083240151405334,
+      "rewards/margins": 0.07270360738039017,
+      "rewards/rejected": 0.24812880158424377,
+      "stability/repetition_rate_mean": 0.9656392931938171,
+      "stability/response_length_mean": 720.2000122070312,
+      "stability/response_length_std": 284.2842102050781,
+      "stability/response_length_var": 80817.5078125,
+      "stability/token_entropy_mean": 7.115624904632568,
+      "step": 145
+    },
+    {
+      "epoch": 0.7677543186180422,
+      "grad_norm": 1.9997695684432983,
+      "learning_rate": 9.972346981963546e-07,
+      "logits/chosen": -2.2425670623779297,
+      "logits/rejected": -2.1018757820129395,
+      "logps/chosen": -371.7270812988281,
+      "logps/rejected": -209.07662963867188,
+      "loss": 1.8773,
+      "metrics/advantage_var": 4.346500396728516,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": 0.3538295328617096,
+      "rewards/margins": 0.08031345158815384,
+      "rewards/rejected": 0.27351611852645874,
+      "stability/repetition_rate_mean": 0.8474888801574707,
+      "stability/response_length_mean": 654.2000122070312,
+      "stability/response_length_std": 372.5482177734375,
+      "stability/response_length_var": 138792.171875,
+      "stability/token_entropy_mean": 6.520312309265137,
+      "step": 150
+    },
+    {
+      "epoch": 0.7933461292386437,
+      "grad_norm": 2.04394793510437,
+      "learning_rate": 9.966422223519885e-07,
+      "logits/chosen": -2.2688398361206055,
+      "logits/rejected": -2.1201179027557373,
+      "logps/chosen": -375.017578125,
+      "logps/rejected": -204.01287841796875,
+      "loss": 1.867,
+      "metrics/advantage_var": 4.019099235534668,
+      "rewards/accuracies": 0.641406238079071,
+      "rewards/chosen": 0.39468881487846375,
+      "rewards/margins": 0.08930384367704391,
+      "rewards/rejected": 0.30538496375083923,
+      "stability/repetition_rate_mean": 0.8770626187324524,
+      "stability/response_length_mean": 502.3999938964844,
+      "stability/response_length_std": 200.0795440673828,
+      "stability/response_length_var": 40031.8203125,
+      "stability/token_entropy_mean": 6.5390625,
+      "step": 155
+    },
+    {
+      "epoch": 0.818937939859245,
+      "grad_norm": 1.9942933320999146,
+      "learning_rate": 9.959924945200523e-07,
+      "logits/chosen": -2.2627499103546143,
+      "logits/rejected": -2.098644971847534,
+      "logps/chosen": -380.92755126953125,
+      "logps/rejected": -204.65432739257812,
+      "loss": 1.8557,
+      "metrics/advantage_var": 5.160045146942139,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": 0.42312800884246826,
+      "rewards/margins": 0.09982379525899887,
+      "rewards/rejected": 0.3233042359352112,
+      "stability/repetition_rate_mean": 0.9664923548698425,
+      "stability/response_length_mean": 870.4000244140625,
+      "stability/response_length_std": 147.6581573486328,
+      "stability/response_length_var": 21802.93359375,
+      "stability/token_entropy_mean": 7.256249904632568,
+      "step": 160
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.8497010469436646,
+      "learning_rate": 9.952855895999568e-07,
+      "logits/chosen": -2.249159812927246,
+      "logits/rejected": -2.1040854454040527,
+      "logps/chosen": -359.7124938964844,
+      "logps/rejected": -200.86595153808594,
+      "loss": 1.8541,
+      "metrics/advantage_var": 4.966158866882324,
+      "rewards/accuracies": 0.6617187857627869,
+      "rewards/chosen": 0.43628647923469543,
+      "rewards/margins": 0.10025672614574432,
+      "rewards/rejected": 0.3360297679901123,
+      "stability/repetition_rate_mean": 0.8816134333610535,
+      "stability/response_length_mean": 547.7000122070312,
+      "stability/response_length_std": 249.60325622558594,
+      "stability/response_length_var": 62301.7890625,
+      "stability/token_entropy_mean": 6.637499809265137,
+      "step": 165
+    },
+    {
+      "epoch": 0.8701215611004478,
+      "grad_norm": 1.8971326351165771,
+      "learning_rate": 9.9452158908238e-07,
+      "logits/chosen": -2.2745604515075684,
+      "logits/rejected": -2.1118056774139404,
+      "logps/chosen": -374.024658203125,
+      "logps/rejected": -200.75364685058594,
+      "loss": 1.8434,
+      "metrics/advantage_var": 6.0166192054748535,
+      "rewards/accuracies": 0.657421886920929,
+      "rewards/chosen": 0.4749736785888672,
+      "rewards/margins": 0.11138726025819778,
+      "rewards/rejected": 0.36358642578125,
+      "stability/repetition_rate_mean": 0.9127447009086609,
+      "stability/response_length_mean": 385.70001220703125,
+      "stability/response_length_std": 143.96685791015625,
+      "stability/response_length_var": 20726.455078125,
+      "stability/token_entropy_mean": 6.84375,
+      "step": 170
+    },
+    {
+      "epoch": 0.8957133717210493,
+      "grad_norm": 1.9797008037567139,
+      "learning_rate": 9.937005810398745e-07,
+      "logits/chosen": -2.2613742351531982,
+      "logits/rejected": -2.1090588569641113,
+      "logps/chosen": -377.3147277832031,
+      "logps/rejected": -209.00302124023438,
+      "loss": 1.8446,
+      "metrics/advantage_var": 7.684011936187744,
+      "rewards/accuracies": 0.6503906846046448,
+      "rewards/chosen": 0.5074092149734497,
+      "rewards/margins": 0.1214134693145752,
+      "rewards/rejected": 0.3859958052635193,
+      "stability/repetition_rate_mean": 0.9286333322525024,
+      "stability/response_length_mean": 460.6000061035156,
+      "stability/response_length_std": 165.9894256591797,
+      "stability/response_length_var": 27552.48828125,
+      "stability/token_entropy_mean": 6.890625,
+      "step": 175
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 1.9558175802230835,
+      "learning_rate": 9.928226601167138e-07,
+      "logits/chosen": -2.261464834213257,
+      "logits/rejected": -2.1034388542175293,
+      "logps/chosen": -383.6528625488281,
+      "logps/rejected": -209.46878051757812,
+      "loss": 1.8445,
+      "metrics/advantage_var": 6.964868068695068,
+      "rewards/accuracies": 0.6527343988418579,
+      "rewards/chosen": 0.5428277254104614,
+      "rewards/margins": 0.12336398661136627,
+      "rewards/rejected": 0.4194636940956116,
+      "stability/repetition_rate_mean": 0.8930152654647827,
+      "stability/response_length_mean": 716.2000122070312,
+      "stability/response_length_std": 329.3815002441406,
+      "stability/response_length_var": 108492.1796875,
+      "stability/token_entropy_mean": 6.537499904632568,
+      "step": 180
+    },
+    {
+      "epoch": 0.946896992962252,
+      "grad_norm": 1.8684709072113037,
+      "learning_rate": 9.918879275179817e-07,
+      "logits/chosen": -2.264568328857422,
+      "logits/rejected": -2.1175734996795654,
+      "logps/chosen": -370.79620361328125,
+      "logps/rejected": -206.6028289794922,
+      "loss": 1.8358,
+      "metrics/advantage_var": 8.66801643371582,
+      "rewards/accuracies": 0.6503906846046448,
+      "rewards/chosen": 0.5830380916595459,
+      "rewards/margins": 0.12937577068805695,
+      "rewards/rejected": 0.45366230607032776,
+      "stability/repetition_rate_mean": 0.8797545433044434,
+      "stability/response_length_mean": 486.3999938964844,
+      "stability/response_length_std": 213.8624725341797,
+      "stability/response_length_var": 45737.15625,
+      "stability/token_entropy_mean": 6.557812690734863,
+      "step": 185
+    },
+    {
+      "epoch": 0.9724888035828535,
+      "grad_norm": 1.716091513633728,
+      "learning_rate": 9.90896490997906e-07,
+      "logits/chosen": -2.2499892711639404,
+      "logits/rejected": -2.098752737045288,
+      "logps/chosen": -373.78045654296875,
+      "logps/rejected": -202.24656677246094,
+      "loss": 1.8332,
+      "metrics/advantage_var": 9.276333808898926,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": 0.6052470803260803,
+      "rewards/margins": 0.1427546888589859,
+      "rewards/rejected": 0.4624924063682556,
+      "stability/repetition_rate_mean": 0.8902595639228821,
+      "stability/response_length_mean": 736.7999877929688,
+      "stability/response_length_std": 246.57106018066406,
+      "stability/response_length_var": 60797.2890625,
+      "stability/token_entropy_mean": 6.739062309265137,
+      "step": 190
+    },
+    {
+      "epoch": 0.9980806142034548,
+      "grad_norm": 1.9551644325256348,
+      "learning_rate": 9.898484648474362e-07,
+      "logits/chosen": -2.267423391342163,
+      "logits/rejected": -2.1186468601226807,
+      "logps/chosen": -357.6914367675781,
+      "logps/rejected": -204.1557159423828,
+      "loss": 1.8325,
+      "metrics/advantage_var": 9.100943565368652,
+      "rewards/accuracies": 0.659375011920929,
+      "rewards/chosen": 0.6304783225059509,
+      "rewards/margins": 0.14480477571487427,
+      "rewards/rejected": 0.48567354679107666,
+      "stability/repetition_rate_mean": 0.758017361164093,
+      "stability/response_length_mean": 356.3999938964844,
+      "stability/response_length_std": 207.94882202148438,
+      "stability/response_length_var": 43242.7109375,
+      "stability/token_entropy_mean": 5.864062309265137,
+      "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/checkpoint-196/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fd7da70af52dd0c8e6abdd746742ad21462766c9009a0d2861e53318335e8da
+size 8017

mistral7b_dpo_sft/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/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/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": 6.987500190734863, "stability/repetition_rate_mean": 0.918988049030304, "step": 10}
+{"stability/response_length_mean": 345.70001220703125, "stability/response_length_std": 212.67974853515625, "stability/response_length_var": 45232.6796875, "stability/token_entropy_mean": 6.317187309265137, "stability/repetition_rate_mean": 0.852084755897522, "step": 15}
+{"stability/response_length_mean": 739.4000244140625, "stability/response_length_std": 281.282470703125, "stability/response_length_var": 79119.8203125, "stability/token_entropy_mean": 6.940625190734863, "stability/repetition_rate_mean": 0.934249997138977, "step": 20}
+{"stability/response_length_mean": 658.5999755859375, "stability/response_length_std": 288.0155029296875, "stability/response_length_var": 82952.9296875, "stability/token_entropy_mean": 7.03125, "stability/repetition_rate_mean": 0.9178038835525513, "step": 25}
+{"stability/response_length_mean": 356.0, "stability/response_length_std": 249.6766815185547, "stability/response_length_var": 62338.4453125, "stability/token_entropy_mean": 6.321875095367432, "stability/repetition_rate_mean": 0.8267838358879089, "step": 30}
+{"stability/response_length_mean": 606.2000122070312, "stability/response_length_std": 394.1366882324219, "stability/response_length_var": 155343.734375, "stability/token_entropy_mean": 7.153124809265137, "stability/repetition_rate_mean": 0.9363252520561218, "step": 35}
+{"stability/response_length_mean": 547.2000122070312, "stability/response_length_std": 290.8897399902344, "stability/response_length_var": 84616.84375, "stability/token_entropy_mean": 6.974999904632568, "stability/repetition_rate_mean": 0.9261367917060852, "step": 40}
+{"stability/response_length_mean": 366.5, "stability/response_length_std": 118.86243438720703, "stability/response_length_var": 14128.27734375, "stability/token_entropy_mean": 6.696875095367432, "stability/repetition_rate_mean": 0.8869589567184448, "step": 45}
+{"stability/response_length_mean": 384.1000061035156, "stability/response_length_std": 178.4809112548828, "stability/response_length_var": 31855.43359375, "stability/token_entropy_mean": 6.493750095367432, "stability/repetition_rate_mean": 0.8679786920547485, "step": 50}
+{"stability/response_length_mean": 332.8999938964844, "stability/response_length_std": 137.39193725585938, "stability/response_length_var": 18876.544921875, "stability/token_entropy_mean": 6.981249809265137, "stability/repetition_rate_mean": 0.9476622343063354, "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": 0.8755375146865845, "step": 60}
+{"stability/response_length_mean": 306.3999938964844, "stability/response_length_std": 141.7236785888672, "stability/response_length_var": 20085.599609375, "stability/token_entropy_mean": 6.787499904632568, "stability/repetition_rate_mean": 0.9074932336807251, "step": 65}
+{"stability/response_length_mean": 472.70001220703125, "stability/response_length_std": 159.89791870117188, "stability/response_length_var": 25567.34375, "stability/token_entropy_mean": 7.224999904632568, "stability/repetition_rate_mean": 0.970089316368103, "step": 70}
+{"stability/response_length_mean": 527.0, "stability/response_length_std": 293.7500915527344, "stability/response_length_var": 86289.109375, "stability/token_entropy_mean": 6.067187309265137, "stability/repetition_rate_mean": 0.9063584208488464, "step": 75}
+{"stability/response_length_mean": 635.5, "stability/response_length_std": 262.53900146484375, "stability/response_length_var": 68926.71875, "stability/token_entropy_mean": 7.193749904632568, "stability/repetition_rate_mean": 0.945895791053772, "step": 80}
+{"stability/response_length_mean": 600.0999755859375, "stability/response_length_std": 349.2030334472656, "stability/response_length_var": 121942.765625, "stability/token_entropy_mean": 6.728906154632568, "stability/repetition_rate_mean": 0.9032164812088013, "step": 85}
+{"stability/response_length_mean": 692.9000244140625, "stability/response_length_std": 127.31452941894531, "stability/response_length_var": 16208.9892578125, "stability/token_entropy_mean": 7.096875190734863, "stability/repetition_rate_mean": 0.9750186204910278, "step": 90}
+{"stability/response_length_mean": 727.0, "stability/response_length_std": 335.0764465332031, "stability/response_length_var": 112276.21875, "stability/token_entropy_mean": 7.3125, "stability/repetition_rate_mean": 0.9472225904464722, "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": 0.9518136978149414, "step": 100}
+{"stability/response_length_mean": 669.5, "stability/response_length_std": 271.60888671875, "stability/response_length_var": 73771.390625, "stability/token_entropy_mean": 6.586718559265137, "stability/repetition_rate_mean": 0.8962067365646362, "step": 105}
+{"stability/response_length_mean": 255.60000610351562, "stability/response_length_std": 114.30874633789062, "stability/response_length_var": 13066.4892578125, "stability/token_entropy_mean": 6.981249809265137, "stability/repetition_rate_mean": 0.9286434054374695, "step": 110}
+{"stability/response_length_mean": 499.29998779296875, "stability/response_length_std": 282.31072998046875, "stability/response_length_var": 79699.34375, "stability/token_entropy_mean": 6.537499904632568, "stability/repetition_rate_mean": 0.8308541178703308, "step": 115}
+{"stability/response_length_mean": 363.6000061035156, "stability/response_length_std": 146.0701446533203, "stability/response_length_var": 21336.48828125, "stability/token_entropy_mean": 6.934374809265137, "stability/repetition_rate_mean": 0.9338710904121399, "step": 120}
+{"stability/response_length_mean": 433.79998779296875, "stability/response_length_std": 126.92412567138672, "stability/response_length_var": 16109.7333984375, "stability/token_entropy_mean": 6.699999809265137, "stability/repetition_rate_mean": 0.8942571878433228, "step": 125}
+{"stability/response_length_mean": 296.1000061035156, "stability/response_length_std": 167.77593994140625, "stability/response_length_var": 28148.767578125, "stability/token_entropy_mean": 6.131249904632568, "stability/repetition_rate_mean": 0.8264604806900024, "step": 130}
+{"stability/response_length_mean": 596.4000244140625, "stability/response_length_std": 291.9129333496094, "stability/response_length_var": 85213.15625, "stability/token_entropy_mean": 6.550000190734863, "stability/repetition_rate_mean": 0.8641007542610168, "step": 135}
+{"stability/response_length_mean": 173.89999389648438, "stability/response_length_std": 70.74908447265625, "stability/response_length_var": 5005.43310546875, "stability/token_entropy_mean": 6.659375190734863, "stability/repetition_rate_mean": 0.8578893542289734, "step": 140}
+{"stability/response_length_mean": 720.2000122070312, "stability/response_length_std": 284.2842102050781, "stability/response_length_var": 80817.5078125, "stability/token_entropy_mean": 7.115624904632568, "stability/repetition_rate_mean": 0.9656392931938171, "step": 145}
+{"stability/response_length_mean": 654.2000122070312, "stability/response_length_std": 372.5482177734375, "stability/response_length_var": 138792.171875, "stability/token_entropy_mean": 6.520312309265137, "stability/repetition_rate_mean": 0.8474888801574707, "step": 150}
+{"stability/response_length_mean": 502.3999938964844, "stability/response_length_std": 200.0795440673828, "stability/response_length_var": 40031.8203125, "stability/token_entropy_mean": 6.5390625, "stability/repetition_rate_mean": 0.8770626187324524, "step": 155}
+{"stability/response_length_mean": 870.4000244140625, "stability/response_length_std": 147.6581573486328, "stability/response_length_var": 21802.93359375, "stability/token_entropy_mean": 7.256249904632568, "stability/repetition_rate_mean": 0.9664923548698425, "step": 160}
+{"stability/response_length_mean": 547.7000122070312, "stability/response_length_std": 249.60325622558594, "stability/response_length_var": 62301.7890625, "stability/token_entropy_mean": 6.637499809265137, "stability/repetition_rate_mean": 0.8816134333610535, "step": 165}
+{"stability/response_length_mean": 385.70001220703125, "stability/response_length_std": 143.96685791015625, "stability/response_length_var": 20726.455078125, "stability/token_entropy_mean": 6.84375, "stability/repetition_rate_mean": 0.9127447009086609, "step": 170}
+{"stability/response_length_mean": 460.6000061035156, "stability/response_length_std": 165.9894256591797, "stability/response_length_var": 27552.48828125, "stability/token_entropy_mean": 6.890625, "stability/repetition_rate_mean": 0.9286333322525024, "step": 175}
+{"stability/response_length_mean": 716.2000122070312, "stability/response_length_std": 329.3815002441406, "stability/response_length_var": 108492.1796875, "stability/token_entropy_mean": 6.537499904632568, "stability/repetition_rate_mean": 0.8930152654647827, "step": 180}
+{"stability/response_length_mean": 486.3999938964844, "stability/response_length_std": 213.8624725341797, "stability/response_length_var": 45737.15625, "stability/token_entropy_mean": 6.557812690734863, "stability/repetition_rate_mean": 0.8797545433044434, "step": 185}
+{"stability/response_length_mean": 736.7999877929688, "stability/response_length_std": 246.57106018066406, "stability/response_length_var": 60797.2890625, "stability/token_entropy_mean": 6.739062309265137, "stability/repetition_rate_mean": 0.8902595639228821, "step": 190}
+{"stability/response_length_mean": 356.3999938964844, "stability/response_length_std": 207.94882202148438, "stability/response_length_var": 43242.7109375, "stability/token_entropy_mean": 5.864062309265137, "stability/repetition_rate_mean": 0.758017361164093, "step": 195}

mistral7b_dpo_sft/tokenizer.model

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

mistral7b_dpo_sft/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/train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "total_flos": 7.072674331683193e+18,
+    "train_loss": 1.8980152759016777,
+    "train_runtime": 4342.782,
+    "train_samples_per_second": 23.027,
+    "train_steps_per_second": 0.045
+}

mistral7b_dpo_sft/trainer_log.jsonl

@@ -0,0 +1,40 @@
+{"current_steps": 5, "total_steps": 196, "loss": 1.9254, "accuracy": 0.29804688692092896, "lr": 4e-08, "epoch": 0.025591810620601407, "percentage": 2.55, "elapsed_time": "0:01:49", "remaining_time": "1:09:41", "rewards/chosen": 0.0010315654799342155, "rewards/rejected": 0.0005278826574794948, "rewards/accuracies": 0.29804688692092896, "rewards/margins": 0.0005036827642470598, "logps/chosen": -375.065185546875, "logps/rejected": -210.8690948486328, "logits/chosen": -2.272939682006836, "logits/rejected": -2.1233606338500977}
+{"current_steps": 10, "total_steps": 196, "loss": 1.9158, "accuracy": 0.5109375715255737, "lr": 9e-08, "epoch": 0.05118362124120281, "percentage": 5.1, "elapsed_time": "0:03:40", "remaining_time": "1:08:27", "rewards/chosen": 0.0012701328378170729, "rewards/rejected": 0.0006854852545075119, "rewards/accuracies": 0.5109375715255737, "rewards/margins": 0.0005846476415172219, "logps/chosen": -382.0304870605469, "logps/rejected": -202.4506072998047, "logits/chosen": -2.2681591510772705, "logits/rejected": -2.1167702674865723, "stability/response_length_mean": 602.7999877929688, "stability/response_length_std": 295.1842346191406, "stability/response_length_var": 87133.734375, "stability/token_entropy_mean": 6.987500190734863, "stability/repetition_rate_mean": 0.918988049030304}
+{"current_steps": 15, "total_steps": 196, "loss": 1.9201, "accuracy": 0.501171886920929, "lr": 1.4e-07, "epoch": 0.07677543186180422, "percentage": 7.65, "elapsed_time": "0:05:31", "remaining_time": "1:06:38", "rewards/chosen": 0.0017871337477117777, "rewards/rejected": 0.0012534803245216608, "rewards/accuracies": 0.501171886920929, "rewards/margins": 0.0005336535396054387, "logps/chosen": -389.207763671875, "logps/rejected": -207.05108642578125, "logits/chosen": -2.2606420516967773, "logits/rejected": -2.0991973876953125, "stability/response_length_mean": 345.70001220703125, "stability/response_length_std": 212.67974853515625, "stability/response_length_var": 45232.6796875, "stability/token_entropy_mean": 6.317187309265137, "stability/repetition_rate_mean": 0.852084755897522}
+{"current_steps": 20, "total_steps": 196, "loss": 1.9381, "accuracy": 0.49726563692092896, "lr": 1.8999999999999998e-07, "epoch": 0.10236724248240563, "percentage": 10.2, "elapsed_time": "0:07:22", "remaining_time": "1:04:52", "rewards/chosen": 0.001441984553821385, "rewards/rejected": 0.0014423017855733633, "rewards/accuracies": 0.49726563692092896, "rewards/margins": -3.1705712899565697e-07, "logps/chosen": -368.709716796875, "logps/rejected": -198.56033325195312, "logits/chosen": -2.2611918449401855, "logits/rejected": -2.114309549331665, "stability/response_length_mean": 739.4000244140625, "stability/response_length_std": 281.282470703125, "stability/response_length_var": 79119.8203125, "stability/token_entropy_mean": 6.940625190734863, "stability/repetition_rate_mean": 0.934249997138977}
+{"current_steps": 25, "total_steps": 196, "loss": 1.9146, "accuracy": 0.49296876788139343, "lr": 2.4e-07, "epoch": 0.12795905310300704, "percentage": 12.76, "elapsed_time": "0:09:13", "remaining_time": "1:03:07", "rewards/chosen": 0.0017121940618380904, "rewards/rejected": 0.00247232336550951, "rewards/accuracies": 0.49296876788139343, "rewards/margins": -0.0007601291872560978, "logps/chosen": -378.1678161621094, "logps/rejected": -205.85311889648438, "logits/chosen": -2.2772860527038574, "logits/rejected": -2.125190019607544, "stability/response_length_mean": 658.5999755859375, "stability/response_length_std": 288.0155029296875, "stability/response_length_var": 82952.9296875, "stability/token_entropy_mean": 7.03125, "stability/repetition_rate_mean": 0.9178038835525513}
+{"current_steps": 30, "total_steps": 196, "loss": 1.9331, "accuracy": 0.516406238079071, "lr": 2.9e-07, "epoch": 0.15355086372360843, "percentage": 15.31, "elapsed_time": "0:11:04", "remaining_time": "1:01:17", "rewards/chosen": 0.0037445530761033297, "rewards/rejected": 0.002405015053227544, "rewards/accuracies": 0.516406238079071, "rewards/margins": 0.0013395380228757858, "logps/chosen": -373.8929443359375, "logps/rejected": -208.92208862304688, "logits/chosen": -2.270566940307617, "logits/rejected": -2.123317241668701, "stability/response_length_mean": 356.0, "stability/response_length_std": 249.6766815185547, "stability/response_length_var": 62338.4453125, "stability/token_entropy_mean": 6.321875095367432, "stability/repetition_rate_mean": 0.8267838358879089}
+{"current_steps": 35, "total_steps": 196, "loss": 1.9336, "accuracy": 0.510937511920929, "lr": 3.4000000000000003e-07, "epoch": 0.17914267434420986, "percentage": 17.86, "elapsed_time": "0:12:56", "remaining_time": "0:59:30", "rewards/chosen": 0.004964467138051987, "rewards/rejected": 0.0038215997628867626, "rewards/accuracies": 0.510937511920929, "rewards/margins": 0.0011428671423345804, "logps/chosen": -366.2962341308594, "logps/rejected": -203.96969604492188, "logits/chosen": -2.261219024658203, "logits/rejected": -2.1199522018432617, "stability/response_length_mean": 606.2000122070312, "stability/response_length_std": 394.1366882324219, "stability/response_length_var": 155343.734375, "stability/token_entropy_mean": 7.153124809265137, "stability/repetition_rate_mean": 0.9363252520561218}
+{"current_steps": 40, "total_steps": 196, "loss": 1.9288, "accuracy": 0.4937499761581421, "lr": 3.8999999999999997e-07, "epoch": 0.20473448496481125, "percentage": 20.41, "elapsed_time": "0:14:47", "remaining_time": "0:57:43", "rewards/chosen": 0.005836769472807646, "rewards/rejected": 0.006121065001934767, "rewards/accuracies": 0.4937499761581421, "rewards/margins": -0.0002842957037501037, "logps/chosen": -369.1612243652344, "logps/rejected": -197.34510803222656, "logits/chosen": -2.276848554611206, "logits/rejected": -2.123274087905884, "stability/response_length_mean": 547.2000122070312, "stability/response_length_std": 290.8897399902344, "stability/response_length_var": 84616.84375, "stability/token_entropy_mean": 6.974999904632568, "stability/repetition_rate_mean": 0.9261367917060852}
+{"current_steps": 45, "total_steps": 196, "loss": 1.9343, "accuracy": 0.49335938692092896, "lr": 4.3999999999999997e-07, "epoch": 0.23032629558541268, "percentage": 22.96, "elapsed_time": "0:16:38", "remaining_time": "0:55:49", "rewards/chosen": 0.008032082580029964, "rewards/rejected": 0.007876048795878887, "rewards/accuracies": 0.49335938692092896, "rewards/margins": 0.00015603320207446814, "logps/chosen": -385.138671875, "logps/rejected": -212.5291290283203, "logits/chosen": -2.2577290534973145, "logits/rejected": -2.1079652309417725, "stability/response_length_mean": 366.5, "stability/response_length_std": 118.86243438720703, "stability/response_length_var": 14128.27734375, "stability/token_entropy_mean": 6.696875095367432, "stability/repetition_rate_mean": 0.8869589567184448}
+{"current_steps": 50, "total_steps": 196, "loss": 1.9417, "accuracy": 0.524609386920929, "lr": 4.9e-07, "epoch": 0.2559181062060141, "percentage": 25.51, "elapsed_time": "0:18:28", "remaining_time": "0:53:57", "rewards/chosen": 0.011302120983600616, "rewards/rejected": 0.00818617083132267, "rewards/accuracies": 0.524609386920929, "rewards/margins": 0.00311595038510859, "logps/chosen": -378.2447509765625, "logps/rejected": -207.44895935058594, "logits/chosen": -2.265531539916992, "logits/rejected": -2.1107473373413086, "stability/response_length_mean": 384.1000061035156, "stability/response_length_std": 178.4809112548828, "stability/response_length_var": 31855.43359375, "stability/token_entropy_mean": 6.493750095367432, "stability/repetition_rate_mean": 0.8679786920547485}
+{"current_steps": 55, "total_steps": 196, "loss": 1.9151, "accuracy": 0.5433593988418579, "lr": 5.4e-07, "epoch": 0.28150991682661547, "percentage": 28.06, "elapsed_time": "0:20:19", "remaining_time": "0:52:05", "rewards/chosen": 0.015990909188985825, "rewards/rejected": 0.011889317072927952, "rewards/accuracies": 0.5433593988418579, "rewards/margins": 0.00410159258171916, "logps/chosen": -393.7201232910156, "logps/rejected": -210.2903289794922, "logits/chosen": -2.2538602352142334, "logits/rejected": -2.095405101776123, "stability/response_length_mean": 332.8999938964844, "stability/response_length_std": 137.39193725585938, "stability/response_length_var": 18876.544921875, "stability/token_entropy_mean": 6.981249809265137, "stability/repetition_rate_mean": 0.9476622343063354}
+{"current_steps": 60, "total_steps": 196, "loss": 1.9258, "accuracy": 0.5367187857627869, "lr": 5.9e-07, "epoch": 0.30710172744721687, "percentage": 30.61, "elapsed_time": "0:22:11", "remaining_time": "0:50:17", "rewards/chosen": 0.01970873773097992, "rewards/rejected": 0.014687979593873024, "rewards/accuracies": 0.5367187857627869, "rewards/margins": 0.005020757671445608, "logps/chosen": -387.6095275878906, "logps/rejected": -206.73020935058594, "logits/chosen": -2.27402400970459, "logits/rejected": -2.1127142906188965, "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": 0.8755375146865845}
+{"current_steps": 65, "total_steps": 196, "loss": 1.9516, "accuracy": 0.5406250357627869, "lr": 6.4e-07, "epoch": 0.3326935380678183, "percentage": 33.16, "elapsed_time": "0:24:01", "remaining_time": "0:48:24", "rewards/chosen": 0.023161547258496284, "rewards/rejected": 0.01849152334034443, "rewards/accuracies": 0.5406250357627869, "rewards/margins": 0.004670025315135717, "logps/chosen": -377.36798095703125, "logps/rejected": -204.2776336669922, "logits/chosen": -2.2534067630767822, "logits/rejected": -2.1004819869995117, "stability/response_length_mean": 306.3999938964844, "stability/response_length_std": 141.7236785888672, "stability/response_length_var": 20085.599609375, "stability/token_entropy_mean": 6.787499904632568, "stability/repetition_rate_mean": 0.9074932336807251}
+{"current_steps": 70, "total_steps": 196, "loss": 1.907, "accuracy": 0.571093738079071, "lr": 6.9e-07, "epoch": 0.3582853486884197, "percentage": 35.71, "elapsed_time": "0:25:52", "remaining_time": "0:46:33", "rewards/chosen": 0.030948856845498085, "rewards/rejected": 0.022783391177654266, "rewards/accuracies": 0.571093738079071, "rewards/margins": 0.008165466599166393, "logps/chosen": -393.5992736816406, "logps/rejected": -206.0418701171875, "logits/chosen": -2.2728006839752197, "logits/rejected": -2.1082444190979004, "stability/response_length_mean": 472.70001220703125, "stability/response_length_std": 159.89791870117188, "stability/response_length_var": 25567.34375, "stability/token_entropy_mean": 7.224999904632568, "stability/repetition_rate_mean": 0.970089316368103}
+{"current_steps": 75, "total_steps": 196, "loss": 1.9254, "accuracy": 0.5691406726837158, "lr": 7.4e-07, "epoch": 0.3838771593090211, "percentage": 38.27, "elapsed_time": "0:27:42", "remaining_time": "0:44:42", "rewards/chosen": 0.0375763475894928, "rewards/rejected": 0.0298237893730402, "rewards/accuracies": 0.5691406726837158, "rewards/margins": 0.007752560544759035, "logps/chosen": -380.871337890625, "logps/rejected": -207.7386474609375, "logits/chosen": -2.27734637260437, "logits/rejected": -2.1185312271118164, "stability/response_length_mean": 527.0, "stability/response_length_std": 293.7500915527344, "stability/response_length_var": 86289.109375, "stability/token_entropy_mean": 6.067187309265137, "stability/repetition_rate_mean": 0.9063584208488464}
+{"current_steps": 80, "total_steps": 196, "loss": 1.9115, "accuracy": 0.5769531726837158, "lr": 7.9e-07, "epoch": 0.4094689699296225, "percentage": 40.82, "elapsed_time": "0:29:34", "remaining_time": "0:42:52", "rewards/chosen": 0.04558176547288895, "rewards/rejected": 0.036044586449861526, "rewards/accuracies": 0.5769531726837158, "rewards/margins": 0.009537179954349995, "logps/chosen": -382.1772766113281, "logps/rejected": -208.73495483398438, "logits/chosen": -2.262965440750122, "logits/rejected": -2.1055397987365723, "stability/response_length_mean": 635.5, "stability/response_length_std": 262.53900146484375, "stability/response_length_var": 68926.71875, "stability/token_entropy_mean": 7.193749904632568, "stability/repetition_rate_mean": 0.945895791053772}
+{"current_steps": 85, "total_steps": 196, "loss": 1.9267, "accuracy": 0.5796875357627869, "lr": 8.399999999999999e-07, "epoch": 0.4350607805502239, "percentage": 43.37, "elapsed_time": "0:31:26", "remaining_time": "0:41:03", "rewards/chosen": 0.056618086993694305, "rewards/rejected": 0.045003630220890045, "rewards/accuracies": 0.5796875357627869, "rewards/margins": 0.011614463292062283, "logps/chosen": -387.02532958984375, "logps/rejected": -208.32882690429688, "logits/chosen": -2.2605109214782715, "logits/rejected": -2.1037368774414062, "stability/response_length_mean": 600.0999755859375, "stability/response_length_std": 349.2030334472656, "stability/response_length_var": 121942.765625, "stability/token_entropy_mean": 6.728906154632568, "stability/repetition_rate_mean": 0.9032164812088013}
+{"current_steps": 90, "total_steps": 196, "loss": 1.9156, "accuracy": 0.608593761920929, "lr": 8.9e-07, "epoch": 0.46065259117082535, "percentage": 45.92, "elapsed_time": "0:33:17", "remaining_time": "0:39:13", "rewards/chosen": 0.06897309422492981, "rewards/rejected": 0.05248956382274628, "rewards/accuracies": 0.608593761920929, "rewards/margins": 0.016483532264828682, "logps/chosen": -374.8363342285156, "logps/rejected": -205.86585998535156, "logits/chosen": -2.263827323913574, "logits/rejected": -2.1160788536071777, "stability/response_length_mean": 692.9000244140625, "stability/response_length_std": 127.31452941894531, "stability/response_length_var": 16208.9892578125, "stability/token_entropy_mean": 7.096875190734863, "stability/repetition_rate_mean": 0.9750186204910278}
+{"current_steps": 95, "total_steps": 196, "loss": 1.91, "accuracy": 0.6187500357627869, "lr": 9.399999999999999e-07, "epoch": 0.48624440179142675, "percentage": 48.47, "elapsed_time": "0:35:09", "remaining_time": "0:37:22", "rewards/chosen": 0.08120815455913544, "rewards/rejected": 0.060329943895339966, "rewards/accuracies": 0.6187500357627869, "rewards/margins": 0.020878203213214874, "logps/chosen": -366.39752197265625, "logps/rejected": -203.00827026367188, "logits/chosen": -2.2615463733673096, "logits/rejected": -2.115648031234741, "stability/response_length_mean": 727.0, "stability/response_length_std": 335.0764465332031, "stability/response_length_var": 112276.21875, "stability/token_entropy_mean": 7.3125, "stability/repetition_rate_mean": 0.9472225904464722}
+{"current_steps": 100, "total_steps": 196, "loss": 1.9049, "accuracy": 0.6289063096046448, "lr": 9.9e-07, "epoch": 0.5118362124120281, "percentage": 51.02, "elapsed_time": "0:37:00", "remaining_time": "0:35:31", "rewards/chosen": 0.09820398688316345, "rewards/rejected": 0.07367560267448425, "rewards/accuracies": 0.6289063096046448, "rewards/margins": 0.024528376758098602, "logps/chosen": -383.9757995605469, "logps/rejected": -210.1046905517578, "logits/chosen": -2.2777926921844482, "logits/rejected": -2.123100757598877, "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": 0.9518136978149414}
+{"current_steps": 105, "total_steps": 196, "loss": 1.9191, "accuracy": 0.6265625357627869, "lr": 9.999815554351224e-07, "epoch": 0.5374280230326296, "percentage": 53.57, "elapsed_time": "0:38:52", "remaining_time": "0:33:41", "rewards/chosen": 0.116055428981781, "rewards/rejected": 0.08972139656543732, "rewards/accuracies": 0.6265625357627869, "rewards/margins": 0.02633403241634369, "logps/chosen": -377.36602783203125, "logps/rejected": -209.0551300048828, "logits/chosen": -2.2625222206115723, "logits/rejected": -2.112849473953247, "stability/response_length_mean": 669.5, "stability/response_length_std": 271.60888671875, "stability/response_length_var": 73771.390625, "stability/token_entropy_mean": 6.586718559265137, "stability/repetition_rate_mean": 0.8962067365646362}
+{"current_steps": 110, "total_steps": 196, "loss": 1.9087, "accuracy": 0.6167968511581421, "lr": 9.999066267225444e-07, "epoch": 0.5630198336532309, "percentage": 56.12, "elapsed_time": "0:40:41", "remaining_time": "0:31:49", "rewards/chosen": 0.13622210919857025, "rewards/rejected": 0.10789017379283905, "rewards/accuracies": 0.6167968511581421, "rewards/margins": 0.0283319354057312, "logps/chosen": -384.4992370605469, "logps/rejected": -205.9490203857422, "logits/chosen": -2.279107093811035, "logits/rejected": -2.122042655944824, "stability/response_length_mean": 255.60000610351562, "stability/response_length_std": 114.30874633789062, "stability/response_length_var": 13066.4892578125, "stability/token_entropy_mean": 6.981249809265137, "stability/repetition_rate_mean": 0.9286434054374695}
+{"current_steps": 115, "total_steps": 196, "loss": 1.9089, "accuracy": 0.6238281726837158, "lr": 9.997740697079592e-07, "epoch": 0.5886116442738324, "percentage": 58.67, "elapsed_time": "0:42:33", "remaining_time": "0:29:58", "rewards/chosen": 0.16183851659297943, "rewards/rejected": 0.12316557765007019, "rewards/accuracies": 0.6238281726837158, "rewards/margins": 0.03867293521761894, "logps/chosen": -371.8219299316406, "logps/rejected": -213.05772399902344, "logits/chosen": -2.2564876079559326, "logits/rejected": -2.1166574954986572, "stability/response_length_mean": 499.29998779296875, "stability/response_length_std": 282.31072998046875, "stability/response_length_var": 79699.34375, "stability/token_entropy_mean": 6.537499904632568, "stability/repetition_rate_mean": 0.8308541178703308}
+{"current_steps": 120, "total_steps": 196, "loss": 1.8995, "accuracy": 0.6390625238418579, "lr": 9.995838996722914e-07, "epoch": 0.6142034548944337, "percentage": 61.22, "elapsed_time": "0:44:23", "remaining_time": "0:28:06", "rewards/chosen": 0.18527083098888397, "rewards/rejected": 0.1433136910200119, "rewards/accuracies": 0.6390625238418579, "rewards/margins": 0.04195715859532356, "logps/chosen": -366.7509460449219, "logps/rejected": -206.54931640625, "logits/chosen": -2.253622055053711, "logits/rejected": -2.108788251876831, "stability/response_length_mean": 363.6000061035156, "stability/response_length_std": 146.0701446533203, "stability/response_length_var": 21336.48828125, "stability/token_entropy_mean": 6.934374809265137, "stability/repetition_rate_mean": 0.9338710904121399}
+{"current_steps": 125, "total_steps": 196, "loss": 1.8987, "accuracy": 0.6304687857627869, "lr": 9.993361385379874e-07, "epoch": 0.6397952655150352, "percentage": 63.78, "elapsed_time": "0:46:13", "remaining_time": "0:26:15", "rewards/chosen": 0.21179263293743134, "rewards/rejected": 0.16302187740802765, "rewards/accuracies": 0.6304687857627869, "rewards/margins": 0.04877074807882309, "logps/chosen": -384.51739501953125, "logps/rejected": -203.9079132080078, "logits/chosen": -2.2593374252319336, "logits/rejected": -2.1024961471557617, "stability/response_length_mean": 433.79998779296875, "stability/response_length_std": 126.92412567138672, "stability/response_length_var": 16109.7333984375, "stability/token_entropy_mean": 6.699999809265137, "stability/repetition_rate_mean": 0.8942571878433228}
+{"current_steps": 130, "total_steps": 196, "loss": 1.8884, "accuracy": 0.639843761920929, "lr": 9.99030814866488e-07, "epoch": 0.6653870761356366, "percentage": 66.33, "elapsed_time": "0:48:03", "remaining_time": "0:24:24", "rewards/chosen": 0.23221388459205627, "rewards/rejected": 0.17882166802883148, "rewards/accuracies": 0.639843761920929, "rewards/margins": 0.0533922053873539, "logps/chosen": -374.2728271484375, "logps/rejected": -202.3927764892578, "logits/chosen": -2.251539468765259, "logits/rejected": -2.103938579559326, "stability/response_length_mean": 296.1000061035156, "stability/response_length_std": 167.77593994140625, "stability/response_length_var": 28148.767578125, "stability/token_entropy_mean": 6.131249904632568, "stability/repetition_rate_mean": 0.8264604806900024}
+{"current_steps": 135, "total_steps": 196, "loss": 1.8861, "accuracy": 0.647265613079071, "lr": 9.986679638549358e-07, "epoch": 0.690978886756238, "percentage": 68.88, "elapsed_time": "0:49:55", "remaining_time": "0:22:33", "rewards/chosen": 0.25935378670692444, "rewards/rejected": 0.19871726632118225, "rewards/accuracies": 0.647265613079071, "rewards/margins": 0.06063650920987129, "logps/chosen": -385.8937072753906, "logps/rejected": -212.23419189453125, "logits/chosen": -2.273317813873291, "logits/rejected": -2.1149117946624756, "stability/response_length_mean": 596.4000244140625, "stability/response_length_std": 291.9129333496094, "stability/response_length_var": 85213.15625, "stability/token_entropy_mean": 6.550000190734863, "stability/repetition_rate_mean": 0.8641007542610168}
+{"current_steps": 140, "total_steps": 196, "loss": 1.8781, "accuracy": 0.6578124761581421, "lr": 9.982476273321175e-07, "epoch": 0.7165706973768394, "percentage": 71.43, "elapsed_time": "0:51:45", "remaining_time": "0:20:42", "rewards/chosen": 0.28835996985435486, "rewards/rejected": 0.2175232321023941, "rewards/accuracies": 0.6578124761581421, "rewards/margins": 0.07083673030138016, "logps/chosen": -387.342041015625, "logps/rejected": -209.46954345703125, "logits/chosen": -2.262234926223755, "logits/rejected": -2.1079959869384766, "stability/response_length_mean": 173.89999389648438, "stability/response_length_std": 70.74908447265625, "stability/response_length_var": 5005.43310546875, "stability/token_entropy_mean": 6.659375190734863, "stability/repetition_rate_mean": 0.8578893542289734}
+{"current_steps": 145, "total_steps": 196, "loss": 1.8752, "accuracy": 0.6351562738418579, "lr": 9.977698537536417e-07, "epoch": 0.7421625079974408, "percentage": 73.98, "elapsed_time": "0:53:37", "remaining_time": "0:18:51", "rewards/chosen": 0.32083240151405334, "rewards/rejected": 0.24812880158424377, "rewards/accuracies": 0.6351562738418579, "rewards/margins": 0.07270360738039017, "logps/chosen": -385.9046630859375, "logps/rejected": -201.69927978515625, "logits/chosen": -2.2734627723693848, "logits/rejected": -2.1078710556030273, "stability/response_length_mean": 720.2000122070312, "stability/response_length_std": 284.2842102050781, "stability/response_length_var": 80817.5078125, "stability/token_entropy_mean": 7.115624904632568, "stability/repetition_rate_mean": 0.9656392931938171}
+{"current_steps": 150, "total_steps": 196, "loss": 1.8773, "accuracy": 0.6312500238418579, "lr": 9.972346981963546e-07, "epoch": 0.7677543186180422, "percentage": 76.53, "elapsed_time": "0:55:29", "remaining_time": "0:17:00", "rewards/chosen": 0.3538295328617096, "rewards/rejected": 0.27351611852645874, "rewards/accuracies": 0.6312500238418579, "rewards/margins": 0.08031345158815384, "logps/chosen": -371.7270812988281, "logps/rejected": -209.07662963867188, "logits/chosen": -2.2425670623779297, "logits/rejected": -2.1018757820129395, "stability/response_length_mean": 654.2000122070312, "stability/response_length_std": 372.5482177734375, "stability/response_length_var": 138792.171875, "stability/token_entropy_mean": 6.520312309265137, "stability/repetition_rate_mean": 0.8474888801574707}
+{"current_steps": 155, "total_steps": 196, "loss": 1.867, "accuracy": 0.641406238079071, "lr": 9.966422223519885e-07, "epoch": 0.7933461292386437, "percentage": 79.08, "elapsed_time": "0:57:20", "remaining_time": "0:15:10", "rewards/chosen": 0.39468881487846375, "rewards/rejected": 0.30538496375083923, "rewards/accuracies": 0.641406238079071, "rewards/margins": 0.08930384367704391, "logps/chosen": -375.017578125, "logps/rejected": -204.01287841796875, "logits/chosen": -2.2688398361206055, "logits/rejected": -2.1201179027557373, "stability/response_length_mean": 502.3999938964844, "stability/response_length_std": 200.0795440673828, "stability/response_length_var": 40031.8203125, "stability/token_entropy_mean": 6.5390625, "stability/repetition_rate_mean": 0.8770626187324524}
+{"current_steps": 160, "total_steps": 196, "loss": 1.8557, "accuracy": 0.65625, "lr": 9.959924945200523e-07, "epoch": 0.818937939859245, "percentage": 81.63, "elapsed_time": "0:59:11", "remaining_time": "0:13:19", "rewards/chosen": 0.42312800884246826, "rewards/rejected": 0.3233042359352112, "rewards/accuracies": 0.65625, "rewards/margins": 0.09982379525899887, "logps/chosen": -380.92755126953125, "logps/rejected": -204.65432739257812, "logits/chosen": -2.2627499103546143, "logits/rejected": -2.098644971847534, "stability/response_length_mean": 870.4000244140625, "stability/response_length_std": 147.6581573486328, "stability/response_length_var": 21802.93359375, "stability/token_entropy_mean": 7.256249904632568, "stability/repetition_rate_mean": 0.9664923548698425}
+{"current_steps": 165, "total_steps": 196, "loss": 1.8541, "accuracy": 0.6617187857627869, "lr": 9.952855895999568e-07, "epoch": 0.8445297504798465, "percentage": 84.18, "elapsed_time": "1:01:02", "remaining_time": "0:11:28", "rewards/chosen": 0.43628647923469543, "rewards/rejected": 0.3360297679901123, "rewards/accuracies": 0.6617187857627869, "rewards/margins": 0.10025672614574432, "logps/chosen": -359.7124938964844, "logps/rejected": -200.86595153808594, "logits/chosen": -2.249159812927246, "logits/rejected": -2.1040854454040527, "stability/response_length_mean": 547.7000122070312, "stability/response_length_std": 249.60325622558594, "stability/response_length_var": 62301.7890625, "stability/token_entropy_mean": 6.637499809265137, "stability/repetition_rate_mean": 0.8816134333610535}
+{"current_steps": 170, "total_steps": 196, "loss": 1.8434, "accuracy": 0.657421886920929, "lr": 9.9452158908238e-07, "epoch": 0.8701215611004478, "percentage": 86.73, "elapsed_time": "1:02:53", "remaining_time": "0:09:37", "rewards/chosen": 0.4749736785888672, "rewards/rejected": 0.36358642578125, "rewards/accuracies": 0.657421886920929, "rewards/margins": 0.11138726025819778, "logps/chosen": -374.024658203125, "logps/rejected": -200.75364685058594, "logits/chosen": -2.2745604515075684, "logits/rejected": -2.1118056774139404, "stability/response_length_mean": 385.70001220703125, "stability/response_length_std": 143.96685791015625, "stability/response_length_var": 20726.455078125, "stability/token_entropy_mean": 6.84375, "stability/repetition_rate_mean": 0.9127447009086609}
+{"current_steps": 175, "total_steps": 196, "loss": 1.8446, "accuracy": 0.6503906846046448, "lr": 9.937005810398745e-07, "epoch": 0.8957133717210493, "percentage": 89.29, "elapsed_time": "1:04:44", "remaining_time": "0:07:46", "rewards/chosen": 0.5074092149734497, "rewards/rejected": 0.3859958052635193, "rewards/accuracies": 0.6503906846046448, "rewards/margins": 0.1214134693145752, "logps/chosen": -377.3147277832031, "logps/rejected": -209.00302124023438, "logits/chosen": -2.2613742351531982, "logits/rejected": -2.1090588569641113, "stability/response_length_mean": 460.6000061035156, "stability/response_length_std": 165.9894256591797, "stability/response_length_var": 27552.48828125, "stability/token_entropy_mean": 6.890625, "stability/repetition_rate_mean": 0.9286333322525024}
+{"current_steps": 180, "total_steps": 196, "loss": 1.8445, "accuracy": 0.6527343988418579, "lr": 9.928226601167138e-07, "epoch": 0.9213051823416507, "percentage": 91.84, "elapsed_time": "1:06:36", "remaining_time": "0:05:55", "rewards/chosen": 0.5428277254104614, "rewards/rejected": 0.4194636940956116, "rewards/accuracies": 0.6527343988418579, "rewards/margins": 0.12336398661136627, "logps/chosen": -383.6528625488281, "logps/rejected": -209.46878051757812, "logits/chosen": -2.261464834213257, "logits/rejected": -2.1034388542175293, "stability/response_length_mean": 716.2000122070312, "stability/response_length_std": 329.3815002441406, "stability/response_length_var": 108492.1796875, "stability/token_entropy_mean": 6.537499904632568, "stability/repetition_rate_mean": 0.8930152654647827}
+{"current_steps": 185, "total_steps": 196, "loss": 1.8358, "accuracy": 0.6503906846046448, "lr": 9.918879275179817e-07, "epoch": 0.946896992962252, "percentage": 94.39, "elapsed_time": "1:08:27", "remaining_time": "0:04:04", "rewards/chosen": 0.5830380916595459, "rewards/rejected": 0.45366230607032776, "rewards/accuracies": 0.6503906846046448, "rewards/margins": 0.12937577068805695, "logps/chosen": -370.79620361328125, "logps/rejected": -206.6028289794922, "logits/chosen": -2.264568328857422, "logits/rejected": -2.1175734996795654, "stability/response_length_mean": 486.3999938964844, "stability/response_length_std": 213.8624725341797, "stability/response_length_var": 45737.15625, "stability/token_entropy_mean": 6.557812690734863, "stability/repetition_rate_mean": 0.8797545433044434}
+{"current_steps": 190, "total_steps": 196, "loss": 1.8332, "accuracy": 0.6656249761581421, "lr": 9.90896490997906e-07, "epoch": 0.9724888035828535, "percentage": 96.94, "elapsed_time": "1:10:18", "remaining_time": "0:02:13", "rewards/chosen": 0.6052470803260803, "rewards/rejected": 0.4624924063682556, "rewards/accuracies": 0.6656249761581421, "rewards/margins": 0.1427546888589859, "logps/chosen": -373.78045654296875, "logps/rejected": -202.24656677246094, "logits/chosen": -2.2499892711639404, "logits/rejected": -2.098752737045288, "stability/response_length_mean": 736.7999877929688, "stability/response_length_std": 246.57106018066406, "stability/response_length_var": 60797.2890625, "stability/token_entropy_mean": 6.739062309265137, "stability/repetition_rate_mean": 0.8902595639228821}
+{"current_steps": 195, "total_steps": 196, "loss": 1.8325, "accuracy": 0.659375011920929, "lr": 9.898484648474362e-07, "epoch": 0.9980806142034548, "percentage": 99.49, "elapsed_time": "1:12:08", "remaining_time": "0:00:22", "rewards/chosen": 0.6304783225059509, "rewards/rejected": 0.48567354679107666, "rewards/accuracies": 0.659375011920929, "rewards/margins": 0.14480477571487427, "logps/chosen": -357.6914367675781, "logps/rejected": -204.1557159423828, "logits/chosen": -2.267423391342163, "logits/rejected": -2.1186468601226807, "stability/response_length_mean": 356.3999938964844, "stability/response_length_std": 207.94882202148438, "stability/response_length_var": 43242.7109375, "stability/token_entropy_mean": 5.864062309265137, "stability/repetition_rate_mean": 0.758017361164093}
+{"current_steps": 196, "total_steps": 196, "epoch": 1.0, "percentage": 100.0, "elapsed_time": "1:12:22", "remaining_time": "0:00:00"}

mistral7b_dpo_sft/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": 1.914654016494751,
+      "learning_rate": 4e-08,
+      "logits/chosen": -2.272939682006836,
+      "logits/rejected": -2.1233606338500977,
+      "logps/chosen": -375.065185546875,
+      "logps/rejected": -210.8690948486328,
+      "loss": 1.9254,
+      "metrics/advantage_var": 0.04025906324386597,
+      "rewards/accuracies": 0.29804688692092896,
+      "rewards/chosen": 0.0010315654799342155,
+      "rewards/margins": 0.0005036827642470598,
+      "rewards/rejected": 0.0005278826574794948,
+      "step": 5
+    },
+    {
+      "epoch": 0.05118362124120281,
+      "grad_norm": 1.9914286136627197,
+      "learning_rate": 9e-08,
+      "logits/chosen": -2.2681591510772705,
+      "logits/rejected": -2.1167702674865723,
+      "logps/chosen": -382.0304870605469,
+      "logps/rejected": -202.4506072998047,
+      "loss": 1.9158,
+      "metrics/advantage_var": 0.07169340550899506,
+      "rewards/accuracies": 0.5109375715255737,
+      "rewards/chosen": 0.0012701328378170729,
+      "rewards/margins": 0.0005846476415172219,
+      "rewards/rejected": 0.0006854852545075119,
+      "stability/repetition_rate_mean": 0.918988049030304,
+      "stability/response_length_mean": 602.7999877929688,
+      "stability/response_length_std": 295.1842346191406,
+      "stability/response_length_var": 87133.734375,
+      "stability/token_entropy_mean": 6.987500190734863,
+      "step": 10
+    },
+    {
+      "epoch": 0.07677543186180422,
+      "grad_norm": 2.024305582046509,
+      "learning_rate": 1.4e-07,
+      "logits/chosen": -2.2606420516967773,
+      "logits/rejected": -2.0991973876953125,
+      "logps/chosen": -389.207763671875,
+      "logps/rejected": -207.05108642578125,
+      "loss": 1.9201,
+      "metrics/advantage_var": 0.07511603087186813,
+      "rewards/accuracies": 0.501171886920929,
+      "rewards/chosen": 0.0017871337477117777,
+      "rewards/margins": 0.0005336535396054387,
+      "rewards/rejected": 0.0012534803245216608,
+      "stability/repetition_rate_mean": 0.852084755897522,
+      "stability/response_length_mean": 345.70001220703125,
+      "stability/response_length_std": 212.67974853515625,
+      "stability/response_length_var": 45232.6796875,
+      "stability/token_entropy_mean": 6.317187309265137,
+      "step": 15
+    },
+    {
+      "epoch": 0.10236724248240563,
+      "grad_norm": 1.9839270114898682,
+      "learning_rate": 1.8999999999999998e-07,
+      "logits/chosen": -2.2611918449401855,
+      "logits/rejected": -2.114309549331665,
+      "logps/chosen": -368.709716796875,
+      "logps/rejected": -198.56033325195312,
+      "loss": 1.9381,
+      "metrics/advantage_var": 0.06882387399673462,
+      "rewards/accuracies": 0.49726563692092896,
+      "rewards/chosen": 0.001441984553821385,
+      "rewards/margins": -3.1705712899565697e-07,
+      "rewards/rejected": 0.0014423017855733633,
+      "stability/repetition_rate_mean": 0.934249997138977,
+      "stability/response_length_mean": 739.4000244140625,
+      "stability/response_length_std": 281.282470703125,
+      "stability/response_length_var": 79119.8203125,
+      "stability/token_entropy_mean": 6.940625190734863,
+      "step": 20
+    },
+    {
+      "epoch": 0.12795905310300704,
+      "grad_norm": 2.0085482597351074,
+      "learning_rate": 2.4e-07,
+      "logits/chosen": -2.2772860527038574,
+      "logits/rejected": -2.125190019607544,
+      "logps/chosen": -378.1678161621094,
+      "logps/rejected": -205.85311889648438,
+      "loss": 1.9146,
+      "metrics/advantage_var": 0.06963051110506058,
+      "rewards/accuracies": 0.49296876788139343,
+      "rewards/chosen": 0.0017121940618380904,
+      "rewards/margins": -0.0007601291872560978,
+      "rewards/rejected": 0.00247232336550951,
+      "stability/repetition_rate_mean": 0.9178038835525513,
+      "stability/response_length_mean": 658.5999755859375,
+      "stability/response_length_std": 288.0155029296875,
+      "stability/response_length_var": 82952.9296875,
+      "stability/token_entropy_mean": 7.03125,
+      "step": 25
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 2.009305477142334,
+      "learning_rate": 2.9e-07,
+      "logits/chosen": -2.270566940307617,
+      "logits/rejected": -2.123317241668701,
+      "logps/chosen": -373.8929443359375,
+      "logps/rejected": -208.92208862304688,
+      "loss": 1.9331,
+      "metrics/advantage_var": 0.07247471809387207,
+      "rewards/accuracies": 0.516406238079071,
+      "rewards/chosen": 0.0037445530761033297,
+      "rewards/margins": 0.0013395380228757858,
+      "rewards/rejected": 0.002405015053227544,
+      "stability/repetition_rate_mean": 0.8267838358879089,
+      "stability/response_length_mean": 356.0,
+      "stability/response_length_std": 249.6766815185547,
+      "stability/response_length_var": 62338.4453125,
+      "stability/token_entropy_mean": 6.321875095367432,
+      "step": 30
+    },
+    {
+      "epoch": 0.17914267434420986,
+      "grad_norm": 1.9789156913757324,
+      "learning_rate": 3.4000000000000003e-07,
+      "logits/chosen": -2.261219024658203,
+      "logits/rejected": -2.1199522018432617,
+      "logps/chosen": -366.2962341308594,
+      "logps/rejected": -203.96969604492188,
+      "loss": 1.9336,
+      "metrics/advantage_var": 0.07087455689907074,
+      "rewards/accuracies": 0.510937511920929,
+      "rewards/chosen": 0.004964467138051987,
+      "rewards/margins": 0.0011428671423345804,
+      "rewards/rejected": 0.0038215997628867626,
+      "stability/repetition_rate_mean": 0.9363252520561218,
+      "stability/response_length_mean": 606.2000122070312,
+      "stability/response_length_std": 394.1366882324219,
+      "stability/response_length_var": 155343.734375,
+      "stability/token_entropy_mean": 7.153124809265137,
+      "step": 35
+    },
+    {
+      "epoch": 0.20473448496481125,
+      "grad_norm": 2.0971174240112305,
+      "learning_rate": 3.8999999999999997e-07,
+      "logits/chosen": -2.276848554611206,
+      "logits/rejected": -2.123274087905884,
+      "logps/chosen": -369.1612243652344,
+      "logps/rejected": -197.34510803222656,
+      "loss": 1.9288,
+      "metrics/advantage_var": 0.07022758573293686,
+      "rewards/accuracies": 0.4937499761581421,
+      "rewards/chosen": 0.005836769472807646,
+      "rewards/margins": -0.0002842957037501037,
+      "rewards/rejected": 0.006121065001934767,
+      "stability/repetition_rate_mean": 0.9261367917060852,
+      "stability/response_length_mean": 547.2000122070312,
+      "stability/response_length_std": 290.8897399902344,
+      "stability/response_length_var": 84616.84375,
+      "stability/token_entropy_mean": 6.974999904632568,
+      "step": 40
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 2.0916647911071777,
+      "learning_rate": 4.3999999999999997e-07,
+      "logits/chosen": -2.2577290534973145,
+      "logits/rejected": -2.1079652309417725,
+      "logps/chosen": -385.138671875,
+      "logps/rejected": -212.5291290283203,
+      "loss": 1.9343,
+      "metrics/advantage_var": 0.07318010926246643,
+      "rewards/accuracies": 0.49335938692092896,
+      "rewards/chosen": 0.008032082580029964,
+      "rewards/margins": 0.00015603320207446814,
+      "rewards/rejected": 0.007876048795878887,
+      "stability/repetition_rate_mean": 0.8869589567184448,
+      "stability/response_length_mean": 366.5,
+      "stability/response_length_std": 118.86243438720703,
+      "stability/response_length_var": 14128.27734375,
+      "stability/token_entropy_mean": 6.696875095367432,
+      "step": 45
+    },
+    {
+      "epoch": 0.2559181062060141,
+      "grad_norm": 2.067622423171997,
+      "learning_rate": 4.9e-07,
+      "logits/chosen": -2.265531539916992,
+      "logits/rejected": -2.1107473373413086,
+      "logps/chosen": -378.2447509765625,
+      "logps/rejected": -207.44895935058594,
+      "loss": 1.9417,
+      "metrics/advantage_var": 0.07468117028474808,
+      "rewards/accuracies": 0.524609386920929,
+      "rewards/chosen": 0.011302120983600616,
+      "rewards/margins": 0.00311595038510859,
+      "rewards/rejected": 0.00818617083132267,
+      "stability/repetition_rate_mean": 0.8679786920547485,
+      "stability/response_length_mean": 384.1000061035156,
+      "stability/response_length_std": 178.4809112548828,
+      "stability/response_length_var": 31855.43359375,
+      "stability/token_entropy_mean": 6.493750095367432,
+      "step": 50
+    },
+    {
+      "epoch": 0.28150991682661547,
+      "grad_norm": 2.2453255653381348,
+      "learning_rate": 5.4e-07,
+      "logits/chosen": -2.2538602352142334,
+      "logits/rejected": -2.095405101776123,
+      "logps/chosen": -393.7201232910156,
+      "logps/rejected": -210.2903289794922,
+      "loss": 1.9151,
+      "metrics/advantage_var": 0.08342276513576508,
+      "rewards/accuracies": 0.5433593988418579,
+      "rewards/chosen": 0.015990909188985825,
+      "rewards/margins": 0.00410159258171916,
+      "rewards/rejected": 0.011889317072927952,
+      "stability/repetition_rate_mean": 0.9476622343063354,
+      "stability/response_length_mean": 332.8999938964844,
+      "stability/response_length_std": 137.39193725585938,
+      "stability/response_length_var": 18876.544921875,
+      "stability/token_entropy_mean": 6.981249809265137,
+      "step": 55
+    },
+    {
+      "epoch": 0.30710172744721687,
+      "grad_norm": 2.1323249340057373,
+      "learning_rate": 5.9e-07,
+      "logits/chosen": -2.27402400970459,
+      "logits/rejected": -2.1127142906188965,
+      "logps/chosen": -387.6095275878906,
+      "logps/rejected": -206.73020935058594,
+      "loss": 1.9258,
+      "metrics/advantage_var": 0.08840229362249374,
+      "rewards/accuracies": 0.5367187857627869,
+      "rewards/chosen": 0.01970873773097992,
+      "rewards/margins": 0.005020757671445608,
+      "rewards/rejected": 0.014687979593873024,
+      "stability/repetition_rate_mean": 0.8755375146865845,
+      "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": 1.914339542388916,
+      "learning_rate": 6.4e-07,
+      "logits/chosen": -2.2534067630767822,
+      "logits/rejected": -2.1004819869995117,
+      "logps/chosen": -377.36798095703125,
+      "logps/rejected": -204.2776336669922,
+      "loss": 1.9516,
+      "metrics/advantage_var": 0.0927530974149704,
+      "rewards/accuracies": 0.5406250357627869,
+      "rewards/chosen": 0.023161547258496284,
+      "rewards/margins": 0.004670025315135717,
+      "rewards/rejected": 0.01849152334034443,
+      "stability/repetition_rate_mean": 0.9074932336807251,
+      "stability/response_length_mean": 306.3999938964844,
+      "stability/response_length_std": 141.7236785888672,
+      "stability/response_length_var": 20085.599609375,
+      "stability/token_entropy_mean": 6.787499904632568,
+      "step": 65
+    },
+    {
+      "epoch": 0.3582853486884197,
+      "grad_norm": 2.256079912185669,
+      "learning_rate": 6.9e-07,
+      "logits/chosen": -2.2728006839752197,
+      "logits/rejected": -2.1082444190979004,
+      "logps/chosen": -393.5992736816406,
+      "logps/rejected": -206.0418701171875,
+      "loss": 1.907,
+      "metrics/advantage_var": 0.10806500911712646,
+      "rewards/accuracies": 0.571093738079071,
+      "rewards/chosen": 0.030948856845498085,
+      "rewards/margins": 0.008165466599166393,
+      "rewards/rejected": 0.022783391177654266,
+      "stability/repetition_rate_mean": 0.970089316368103,
+      "stability/response_length_mean": 472.70001220703125,
+      "stability/response_length_std": 159.89791870117188,
+      "stability/response_length_var": 25567.34375,
+      "stability/token_entropy_mean": 7.224999904632568,
+      "step": 70
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 2.153575897216797,
+      "learning_rate": 7.4e-07,
+      "logits/chosen": -2.27734637260437,
+      "logits/rejected": -2.1185312271118164,
+      "logps/chosen": -380.871337890625,
+      "logps/rejected": -207.7386474609375,
+      "loss": 1.9254,
+      "metrics/advantage_var": 0.12323446571826935,
+      "rewards/accuracies": 0.5691406726837158,
+      "rewards/chosen": 0.0375763475894928,
+      "rewards/margins": 0.007752560544759035,
+      "rewards/rejected": 0.0298237893730402,
+      "stability/repetition_rate_mean": 0.9063584208488464,
+      "stability/response_length_mean": 527.0,
+      "stability/response_length_std": 293.7500915527344,
+      "stability/response_length_var": 86289.109375,
+      "stability/token_entropy_mean": 6.067187309265137,
+      "step": 75
+    },
+    {
+      "epoch": 0.4094689699296225,
+      "grad_norm": 2.0587069988250732,
+      "learning_rate": 7.9e-07,
+      "logits/chosen": -2.262965440750122,
+      "logits/rejected": -2.1055397987365723,
+      "logps/chosen": -382.1772766113281,
+      "logps/rejected": -208.73495483398438,
+      "loss": 1.9115,
+      "metrics/advantage_var": 0.14562615752220154,
+      "rewards/accuracies": 0.5769531726837158,
+      "rewards/chosen": 0.04558176547288895,
+      "rewards/margins": 0.009537179954349995,
+      "rewards/rejected": 0.036044586449861526,
+      "stability/repetition_rate_mean": 0.945895791053772,
+      "stability/response_length_mean": 635.5,
+      "stability/response_length_std": 262.53900146484375,
+      "stability/response_length_var": 68926.71875,
+      "stability/token_entropy_mean": 7.193749904632568,
+      "step": 80
+    },
+    {
+      "epoch": 0.4350607805502239,
+      "grad_norm": 2.1112332344055176,
+      "learning_rate": 8.399999999999999e-07,
+      "logits/chosen": -2.2605109214782715,
+      "logits/rejected": -2.1037368774414062,
+      "logps/chosen": -387.02532958984375,
+      "logps/rejected": -208.32882690429688,
+      "loss": 1.9267,
+      "metrics/advantage_var": 0.19675803184509277,
+      "rewards/accuracies": 0.5796875357627869,
+      "rewards/chosen": 0.056618086993694305,
+      "rewards/margins": 0.011614463292062283,
+      "rewards/rejected": 0.045003630220890045,
+      "stability/repetition_rate_mean": 0.9032164812088013,
+      "stability/response_length_mean": 600.0999755859375,
+      "stability/response_length_std": 349.2030334472656,
+      "stability/response_length_var": 121942.765625,
+      "stability/token_entropy_mean": 6.728906154632568,
+      "step": 85
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 2.1287057399749756,
+      "learning_rate": 8.9e-07,
+      "logits/chosen": -2.263827323913574,
+      "logits/rejected": -2.1160788536071777,
+      "logps/chosen": -374.8363342285156,
+      "logps/rejected": -205.86585998535156,
+      "loss": 1.9156,
+      "metrics/advantage_var": 0.23069456219673157,
+      "rewards/accuracies": 0.608593761920929,
+      "rewards/chosen": 0.06897309422492981,
+      "rewards/margins": 0.016483532264828682,
+      "rewards/rejected": 0.05248956382274628,
+      "stability/repetition_rate_mean": 0.9750186204910278,
+      "stability/response_length_mean": 692.9000244140625,
+      "stability/response_length_std": 127.31452941894531,
+      "stability/response_length_var": 16208.9892578125,
+      "stability/token_entropy_mean": 7.096875190734863,
+      "step": 90
+    },
+    {
+      "epoch": 0.48624440179142675,
+      "grad_norm": 2.0764377117156982,
+      "learning_rate": 9.399999999999999e-07,
+      "logits/chosen": -2.2615463733673096,
+      "logits/rejected": -2.115648031234741,
+      "logps/chosen": -366.39752197265625,
+      "logps/rejected": -203.00827026367188,
+      "loss": 1.91,
+      "metrics/advantage_var": 0.3129298686981201,
+      "rewards/accuracies": 0.6187500357627869,
+      "rewards/chosen": 0.08120815455913544,
+      "rewards/margins": 0.020878203213214874,
+      "rewards/rejected": 0.060329943895339966,
+      "stability/repetition_rate_mean": 0.9472225904464722,
+      "stability/response_length_mean": 727.0,
+      "stability/response_length_std": 335.0764465332031,
+      "stability/response_length_var": 112276.21875,
+      "stability/token_entropy_mean": 7.3125,
+      "step": 95
+    },
+    {
+      "epoch": 0.5118362124120281,
+      "grad_norm": 2.0339155197143555,
+      "learning_rate": 9.9e-07,
+      "logits/chosen": -2.2777926921844482,
+      "logits/rejected": -2.123100757598877,
+      "logps/chosen": -383.9757995605469,
+      "logps/rejected": -210.1046905517578,
+      "loss": 1.9049,
+      "metrics/advantage_var": 0.4217294156551361,
+      "rewards/accuracies": 0.6289063096046448,
+      "rewards/chosen": 0.09820398688316345,
+      "rewards/margins": 0.024528376758098602,
+      "rewards/rejected": 0.07367560267448425,
+      "stability/repetition_rate_mean": 0.9518136978149414,
+      "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": 1.9553194046020508,
+      "learning_rate": 9.999815554351224e-07,
+      "logits/chosen": -2.2625222206115723,
+      "logits/rejected": -2.112849473953247,
+      "logps/chosen": -377.36602783203125,
+      "logps/rejected": -209.0551300048828,
+      "loss": 1.9191,
+      "metrics/advantage_var": 0.5498611330986023,
+      "rewards/accuracies": 0.6265625357627869,
+      "rewards/chosen": 0.116055428981781,
+      "rewards/margins": 0.02633403241634369,
+      "rewards/rejected": 0.08972139656543732,
+      "stability/repetition_rate_mean": 0.8962067365646362,
+      "stability/response_length_mean": 669.5,
+      "stability/response_length_std": 271.60888671875,
+      "stability/response_length_var": 73771.390625,
+      "stability/token_entropy_mean": 6.586718559265137,
+      "step": 105
+    },
+    {
+      "epoch": 0.5630198336532309,
+      "grad_norm": 1.9055349826812744,
+      "learning_rate": 9.999066267225444e-07,
+      "logits/chosen": -2.279107093811035,
+      "logits/rejected": -2.122042655944824,
+      "logps/chosen": -384.4992370605469,
+      "logps/rejected": -205.9490203857422,
+      "loss": 1.9087,
+      "metrics/advantage_var": 0.6977776885032654,
+      "rewards/accuracies": 0.6167968511581421,
+      "rewards/chosen": 0.13622210919857025,
+      "rewards/margins": 0.0283319354057312,
+      "rewards/rejected": 0.10789017379283905,
+      "stability/repetition_rate_mean": 0.9286434054374695,
+      "stability/response_length_mean": 255.60000610351562,
+      "stability/response_length_std": 114.30874633789062,
+      "stability/response_length_var": 13066.4892578125,
+      "stability/token_entropy_mean": 6.981249809265137,
+      "step": 110
+    },
+    {
+      "epoch": 0.5886116442738324,
+      "grad_norm": 1.9901360273361206,
+      "learning_rate": 9.997740697079592e-07,
+      "logits/chosen": -2.2564876079559326,
+      "logits/rejected": -2.1166574954986572,
+      "logps/chosen": -371.8219299316406,
+      "logps/rejected": -213.05772399902344,
+      "loss": 1.9089,
+      "metrics/advantage_var": 0.9774861931800842,
+      "rewards/accuracies": 0.6238281726837158,
+      "rewards/chosen": 0.16183851659297943,
+      "rewards/margins": 0.03867293521761894,
+      "rewards/rejected": 0.12316557765007019,
+      "stability/repetition_rate_mean": 0.8308541178703308,
+      "stability/response_length_mean": 499.29998779296875,
+      "stability/response_length_std": 282.31072998046875,
+      "stability/response_length_var": 79699.34375,
+      "stability/token_entropy_mean": 6.537499904632568,
+      "step": 115
+    },
+    {
+      "epoch": 0.6142034548944337,
+      "grad_norm": 1.9900635480880737,
+      "learning_rate": 9.995838996722914e-07,
+      "logits/chosen": -2.253622055053711,
+      "logits/rejected": -2.108788251876831,
+      "logps/chosen": -366.7509460449219,
+      "logps/rejected": -206.54931640625,
+      "loss": 1.8995,
+      "metrics/advantage_var": 1.1828315258026123,
+      "rewards/accuracies": 0.6390625238418579,
+      "rewards/chosen": 0.18527083098888397,
+      "rewards/margins": 0.04195715859532356,
+      "rewards/rejected": 0.1433136910200119,
+      "stability/repetition_rate_mean": 0.9338710904121399,
+      "stability/response_length_mean": 363.6000061035156,
+      "stability/response_length_std": 146.0701446533203,
+      "stability/response_length_var": 21336.48828125,
+      "stability/token_entropy_mean": 6.934374809265137,
+      "step": 120
+    },
+    {
+      "epoch": 0.6397952655150352,
+      "grad_norm": 1.9649677276611328,
+      "learning_rate": 9.993361385379874e-07,
+      "logits/chosen": -2.2593374252319336,
+      "logits/rejected": -2.1024961471557617,
+      "logps/chosen": -384.51739501953125,
+      "logps/rejected": -203.9079132080078,
+      "loss": 1.8987,
+      "metrics/advantage_var": 1.6206504106521606,
+      "rewards/accuracies": 0.6304687857627869,
+      "rewards/chosen": 0.21179263293743134,
+      "rewards/margins": 0.04877074807882309,
+      "rewards/rejected": 0.16302187740802765,
+      "stability/repetition_rate_mean": 0.8942571878433228,
+      "stability/response_length_mean": 433.79998779296875,
+      "stability/response_length_std": 126.92412567138672,
+      "stability/response_length_var": 16109.7333984375,
+      "stability/token_entropy_mean": 6.699999809265137,
+      "step": 125
+    },
+    {
+      "epoch": 0.6653870761356366,
+      "grad_norm": 2.029193162918091,
+      "learning_rate": 9.99030814866488e-07,
+      "logits/chosen": -2.251539468765259,
+      "logits/rejected": -2.103938579559326,
+      "logps/chosen": -374.2728271484375,
+      "logps/rejected": -202.3927764892578,
+      "loss": 1.8884,
+      "metrics/advantage_var": 1.7878669500350952,
+      "rewards/accuracies": 0.639843761920929,
+      "rewards/chosen": 0.23221388459205627,
+      "rewards/margins": 0.0533922053873539,
+      "rewards/rejected": 0.17882166802883148,
+      "stability/repetition_rate_mean": 0.8264604806900024,
+      "stability/response_length_mean": 296.1000061035156,
+      "stability/response_length_std": 167.77593994140625,
+      "stability/response_length_var": 28148.767578125,
+      "stability/token_entropy_mean": 6.131249904632568,
+      "step": 130
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 1.9165802001953125,
+      "learning_rate": 9.986679638549358e-07,
+      "logits/chosen": -2.273317813873291,
+      "logits/rejected": -2.1149117946624756,
+      "logps/chosen": -385.8937072753906,
+      "logps/rejected": -212.23419189453125,
+      "loss": 1.8861,
+      "metrics/advantage_var": 2.076320171356201,
+      "rewards/accuracies": 0.647265613079071,
+      "rewards/chosen": 0.25935378670692444,
+      "rewards/margins": 0.06063650920987129,
+      "rewards/rejected": 0.19871726632118225,
+      "stability/repetition_rate_mean": 0.8641007542610168,
+      "stability/response_length_mean": 596.4000244140625,
+      "stability/response_length_std": 291.9129333496094,
+      "stability/response_length_var": 85213.15625,
+      "stability/token_entropy_mean": 6.550000190734863,
+      "step": 135
+    },
+    {
+      "epoch": 0.7165706973768394,
+      "grad_norm": 2.0632386207580566,
+      "learning_rate": 9.982476273321175e-07,
+      "logits/chosen": -2.262234926223755,
+      "logits/rejected": -2.1079959869384766,
+      "logps/chosen": -387.342041015625,
+      "logps/rejected": -209.46954345703125,
+      "loss": 1.8781,
+      "metrics/advantage_var": 2.7123324871063232,
+      "rewards/accuracies": 0.6578124761581421,
+      "rewards/chosen": 0.28835996985435486,
+      "rewards/margins": 0.07083673030138016,
+      "rewards/rejected": 0.2175232321023941,
+      "stability/repetition_rate_mean": 0.8578893542289734,
+      "stability/response_length_mean": 173.89999389648438,
+      "stability/response_length_std": 70.74908447265625,
+      "stability/response_length_var": 5005.43310546875,
+      "stability/token_entropy_mean": 6.659375190734863,
+      "step": 140
+    },
+    {
+      "epoch": 0.7421625079974408,
+      "grad_norm": 1.8859580755233765,
+      "learning_rate": 9.977698537536417e-07,
+      "logits/chosen": -2.2734627723693848,
+      "logits/rejected": -2.1078710556030273,
+      "logps/chosen": -385.9046630859375,
+      "logps/rejected": -201.69927978515625,
+      "loss": 1.8752,
+      "metrics/advantage_var": 3.1621975898742676,
+      "rewards/accuracies": 0.6351562738418579,
+      "rewards/chosen": 0.32083240151405334,
+      "rewards/margins": 0.07270360738039017,
+      "rewards/rejected": 0.24812880158424377,
+      "stability/repetition_rate_mean": 0.9656392931938171,
+      "stability/response_length_mean": 720.2000122070312,
+      "stability/response_length_std": 284.2842102050781,
+      "stability/response_length_var": 80817.5078125,
+      "stability/token_entropy_mean": 7.115624904632568,
+      "step": 145
+    },
+    {
+      "epoch": 0.7677543186180422,
+      "grad_norm": 1.9997695684432983,
+      "learning_rate": 9.972346981963546e-07,
+      "logits/chosen": -2.2425670623779297,
+      "logits/rejected": -2.1018757820129395,
+      "logps/chosen": -371.7270812988281,
+      "logps/rejected": -209.07662963867188,
+      "loss": 1.8773,
+      "metrics/advantage_var": 4.346500396728516,
+      "rewards/accuracies": 0.6312500238418579,
+      "rewards/chosen": 0.3538295328617096,
+      "rewards/margins": 0.08031345158815384,
+      "rewards/rejected": 0.27351611852645874,
+      "stability/repetition_rate_mean": 0.8474888801574707,
+      "stability/response_length_mean": 654.2000122070312,
+      "stability/response_length_std": 372.5482177734375,
+      "stability/response_length_var": 138792.171875,
+      "stability/token_entropy_mean": 6.520312309265137,
+      "step": 150
+    },
+    {
+      "epoch": 0.7933461292386437,
+      "grad_norm": 2.04394793510437,
+      "learning_rate": 9.966422223519885e-07,
+      "logits/chosen": -2.2688398361206055,
+      "logits/rejected": -2.1201179027557373,
+      "logps/chosen": -375.017578125,
+      "logps/rejected": -204.01287841796875,
+      "loss": 1.867,
+      "metrics/advantage_var": 4.019099235534668,
+      "rewards/accuracies": 0.641406238079071,
+      "rewards/chosen": 0.39468881487846375,
+      "rewards/margins": 0.08930384367704391,
+      "rewards/rejected": 0.30538496375083923,
+      "stability/repetition_rate_mean": 0.8770626187324524,
+      "stability/response_length_mean": 502.3999938964844,
+      "stability/response_length_std": 200.0795440673828,
+      "stability/response_length_var": 40031.8203125,
+      "stability/token_entropy_mean": 6.5390625,
+      "step": 155
+    },
+    {
+      "epoch": 0.818937939859245,
+      "grad_norm": 1.9942933320999146,
+      "learning_rate": 9.959924945200523e-07,
+      "logits/chosen": -2.2627499103546143,
+      "logits/rejected": -2.098644971847534,
+      "logps/chosen": -380.92755126953125,
+      "logps/rejected": -204.65432739257812,
+      "loss": 1.8557,
+      "metrics/advantage_var": 5.160045146942139,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": 0.42312800884246826,
+      "rewards/margins": 0.09982379525899887,
+      "rewards/rejected": 0.3233042359352112,
+      "stability/repetition_rate_mean": 0.9664923548698425,
+      "stability/response_length_mean": 870.4000244140625,
+      "stability/response_length_std": 147.6581573486328,
+      "stability/response_length_var": 21802.93359375,
+      "stability/token_entropy_mean": 7.256249904632568,
+      "step": 160
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.8497010469436646,
+      "learning_rate": 9.952855895999568e-07,
+      "logits/chosen": -2.249159812927246,
+      "logits/rejected": -2.1040854454040527,
+      "logps/chosen": -359.7124938964844,
+      "logps/rejected": -200.86595153808594,
+      "loss": 1.8541,
+      "metrics/advantage_var": 4.966158866882324,
+      "rewards/accuracies": 0.6617187857627869,
+      "rewards/chosen": 0.43628647923469543,
+      "rewards/margins": 0.10025672614574432,
+      "rewards/rejected": 0.3360297679901123,
+      "stability/repetition_rate_mean": 0.8816134333610535,
+      "stability/response_length_mean": 547.7000122070312,
+      "stability/response_length_std": 249.60325622558594,
+      "stability/response_length_var": 62301.7890625,
+      "stability/token_entropy_mean": 6.637499809265137,
+      "step": 165
+    },
+    {
+      "epoch": 0.8701215611004478,
+      "grad_norm": 1.8971326351165771,
+      "learning_rate": 9.9452158908238e-07,
+      "logits/chosen": -2.2745604515075684,
+      "logits/rejected": -2.1118056774139404,
+      "logps/chosen": -374.024658203125,
+      "logps/rejected": -200.75364685058594,
+      "loss": 1.8434,
+      "metrics/advantage_var": 6.0166192054748535,
+      "rewards/accuracies": 0.657421886920929,
+      "rewards/chosen": 0.4749736785888672,
+      "rewards/margins": 0.11138726025819778,
+      "rewards/rejected": 0.36358642578125,
+      "stability/repetition_rate_mean": 0.9127447009086609,
+      "stability/response_length_mean": 385.70001220703125,
+      "stability/response_length_std": 143.96685791015625,
+      "stability/response_length_var": 20726.455078125,
+      "stability/token_entropy_mean": 6.84375,
+      "step": 170
+    },
+    {
+      "epoch": 0.8957133717210493,
+      "grad_norm": 1.9797008037567139,
+      "learning_rate": 9.937005810398745e-07,
+      "logits/chosen": -2.2613742351531982,
+      "logits/rejected": -2.1090588569641113,
+      "logps/chosen": -377.3147277832031,
+      "logps/rejected": -209.00302124023438,
+      "loss": 1.8446,
+      "metrics/advantage_var": 7.684011936187744,
+      "rewards/accuracies": 0.6503906846046448,
+      "rewards/chosen": 0.5074092149734497,
+      "rewards/margins": 0.1214134693145752,
+      "rewards/rejected": 0.3859958052635193,
+      "stability/repetition_rate_mean": 0.9286333322525024,
+      "stability/response_length_mean": 460.6000061035156,
+      "stability/response_length_std": 165.9894256591797,
+      "stability/response_length_var": 27552.48828125,
+      "stability/token_entropy_mean": 6.890625,
+      "step": 175
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 1.9558175802230835,
+      "learning_rate": 9.928226601167138e-07,
+      "logits/chosen": -2.261464834213257,
+      "logits/rejected": -2.1034388542175293,
+      "logps/chosen": -383.6528625488281,
+      "logps/rejected": -209.46878051757812,
+      "loss": 1.8445,
+      "metrics/advantage_var": 6.964868068695068,
+      "rewards/accuracies": 0.6527343988418579,
+      "rewards/chosen": 0.5428277254104614,
+      "rewards/margins": 0.12336398661136627,
+      "rewards/rejected": 0.4194636940956116,
+      "stability/repetition_rate_mean": 0.8930152654647827,
+      "stability/response_length_mean": 716.2000122070312,
+      "stability/response_length_std": 329.3815002441406,
+      "stability/response_length_var": 108492.1796875,
+      "stability/token_entropy_mean": 6.537499904632568,
+      "step": 180
+    },
+    {
+      "epoch": 0.946896992962252,
+      "grad_norm": 1.8684709072113037,
+      "learning_rate": 9.918879275179817e-07,
+      "logits/chosen": -2.264568328857422,
+      "logits/rejected": -2.1175734996795654,
+      "logps/chosen": -370.79620361328125,
+      "logps/rejected": -206.6028289794922,
+      "loss": 1.8358,
+      "metrics/advantage_var": 8.66801643371582,
+      "rewards/accuracies": 0.6503906846046448,
+      "rewards/chosen": 0.5830380916595459,
+      "rewards/margins": 0.12937577068805695,
+      "rewards/rejected": 0.45366230607032776,
+      "stability/repetition_rate_mean": 0.8797545433044434,
+      "stability/response_length_mean": 486.3999938964844,
+      "stability/response_length_std": 213.8624725341797,
+      "stability/response_length_var": 45737.15625,
+      "stability/token_entropy_mean": 6.557812690734863,
+      "step": 185
+    },
+    {
+      "epoch": 0.9724888035828535,
+      "grad_norm": 1.716091513633728,
+      "learning_rate": 9.90896490997906e-07,
+      "logits/chosen": -2.2499892711639404,
+      "logits/rejected": -2.098752737045288,
+      "logps/chosen": -373.78045654296875,
+      "logps/rejected": -202.24656677246094,
+      "loss": 1.8332,
+      "metrics/advantage_var": 9.276333808898926,
+      "rewards/accuracies": 0.6656249761581421,
+      "rewards/chosen": 0.6052470803260803,
+      "rewards/margins": 0.1427546888589859,
+      "rewards/rejected": 0.4624924063682556,
+      "stability/repetition_rate_mean": 0.8902595639228821,
+      "stability/response_length_mean": 736.7999877929688,
+      "stability/response_length_std": 246.57106018066406,
+      "stability/response_length_var": 60797.2890625,
+      "stability/token_entropy_mean": 6.739062309265137,
+      "step": 190
+    },
+    {
+      "epoch": 0.9980806142034548,
+      "grad_norm": 1.9551644325256348,
+      "learning_rate": 9.898484648474362e-07,
+      "logits/chosen": -2.267423391342163,
+      "logits/rejected": -2.1186468601226807,
+      "logps/chosen": -357.6914367675781,
+      "logps/rejected": -204.1557159423828,
+      "loss": 1.8325,
+      "metrics/advantage_var": 9.100943565368652,
+      "rewards/accuracies": 0.659375011920929,
+      "rewards/chosen": 0.6304783225059509,
+      "rewards/margins": 0.14480477571487427,
+      "rewards/rejected": 0.48567354679107666,
+      "stability/repetition_rate_mean": 0.758017361164093,
+      "stability/response_length_mean": 356.3999938964844,
+      "stability/response_length_std": 207.94882202148438,
+      "stability/response_length_var": 43242.7109375,
+      "stability/token_entropy_mean": 5.864062309265137,
+      "step": 195
+    },
+    {
+      "epoch": 1.0,
+      "step": 196,
+      "total_flos": 7.072674331683193e+18,
+      "train_loss": 1.8980152759016777,
+      "train_runtime": 4342.782,
+      "train_samples_per_second": 23.027,
+      "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/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fd7da70af52dd0c8e6abdd746742ad21462766c9009a0d2861e53318335e8da
+size 8017