Upload folder using huggingface_hub

mistral7b_dpo_sft_ftx0.01/README.md

@@ -0,0 +1,64 @@
+---
+library_name: peft
+license: other
+base_model: mistralai/Mistral-7B-v0.1
+tags:
+- base_model:adapter:mistralai/Mistral-7B-v0.1
+- llama-factory
+- lora
+- transformers
+pipeline_tag: text-generation
+model-index:
+- name: mistral7b_dpo_sft_ftx0.01
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# mistral7b_dpo_sft_ftx0.01
+
+This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the shp_dpo_train dataset.
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 1e-06
+- train_batch_size: 8
+- eval_batch_size: 8
+- seed: 42
+- distributed_type: multi-GPU
+- num_devices: 8
+- gradient_accumulation_steps: 8
+- total_train_batch_size: 512
+- total_eval_batch_size: 64
+- optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_steps: 100
+- num_epochs: 1.0
+
+### Training results
+
+
+
+### Framework versions
+
+- PEFT 0.17.1
+- Transformers 4.57.1
+- Pytorch 2.10.0+cu128
+- Datasets 4.0.0
+- Tokenizers 0.22.2

mistral7b_dpo_sft_ftx0.01/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": [
+    "gate_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj",
+    "q_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

mistral7b_dpo_sft_ftx0.01/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10a84c2822f397ff4d454a9121475c17342620ee1e58d0e4b0de6bcee3a79d00
+size 42002584

mistral7b_dpo_sft_ftx0.01/all_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "total_flos": 7.072674331683193e+18,
+    "train_loss": 0.6978096469324462,
+    "train_runtime": 4339.2504,
+    "train_samples_per_second": 23.045,
+    "train_steps_per_second": 0.045
+}

mistral7b_dpo_sft_ftx0.01/chat_template.jinja

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/README.md

@@ -0,0 +1,208 @@
+---
+base_model: mistralai/Mistral-7B-v0.1
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:mistralai/Mistral-7B-v0.1
+- llama-factory
+- lora
+- transformers
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.17.1

mistral7b_dpo_sft_ftx0.01/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": [
+    "gate_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "v_proj",
+    "q_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

mistral7b_dpo_sft_ftx0.01/checkpoint-196/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10a84c2822f397ff4d454a9121475c17342620ee1e58d0e4b0de6bcee3a79d00
+size 42002584

mistral7b_dpo_sft_ftx0.01/checkpoint-196/chat_template.jinja

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

mistral7b_dpo_sft_ftx0.01/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:cf402ce0c4cd4e9fba898df4a70732bf58681cf0ccb27b931a7dc754801f30a5
+size 31467333

mistral7b_dpo_sft_ftx0.01/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 sha256:8cd3f96733670ef0f799017b64cfc4696c40ba05cf34ac3ef0d58a3bbf9e7f12
+size 31467333

mistral7b_dpo_sft_ftx0.01/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 sha256:558e2e01b57bdd8381b0cc0dc913bf75811d5f549059229147520d7e0c4aeb42
+size 31467333

mistral7b_dpo_sft_ftx0.01/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 sha256:0a79b3debe715365e3b40d4858e7af01ba0f9345f75866a7dc63ad4dd5bf7989
+size 31467397

mistral7b_dpo_sft_ftx0.01/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:b88d5a39b1c559b3becc742cea837c1d20482332b10251d6f019771bfd9bad54
+size 31467397

mistral7b_dpo_sft_ftx0.01/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 sha256:43d1cb778d86e04672821c554dddd443981f5cec4998222c5570edfabd223a90
+size 31467397

mistral7b_dpo_sft_ftx0.01/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 sha256:2a13b65069b4bedd9903c303f98afb602125e31d05eeb9fdd236058a1268efe9
+size 31467397

mistral7b_dpo_sft_ftx0.01/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:e782817f8586ea3b91789ae94256575fe160e2cb376d88300c016011b00030a2
+size 31467397

mistral7b_dpo_sft_ftx0.01/checkpoint-196/global_step196/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5020e3d9922dba59b4b2aa12898a35e771b9ea6d76e1f22239ccf9198a36e65c
+size 42288409

mistral7b_dpo_sft_ftx0.01/checkpoint-196/latest

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/rng_state_0.pth

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/rng_state_1.pth

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/rng_state_2.pth

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/rng_state_3.pth

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/rng_state_4.pth

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/rng_state_5.pth

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/rng_state_6.pth

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/rng_state_7.pth

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/scheduler.pt

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/special_tokens_map.json

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/tokenizer.model

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

mistral7b_dpo_sft_ftx0.01/checkpoint-196/tokenizer_config.json

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

mistral7b_dpo_sft_ftx0.01/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": 1.7010759115219116,
+      "learning_rate": 4e-08,
+      "logits/chosen": -2.272961139678955,
+      "logits/rejected": -2.1232829093933105,
+      "logps/chosen": -375.06817626953125,
+      "logps/rejected": -210.86959838867188,
+      "loss": 0.7178,
+      "metrics/advantage_var": 0.04110218212008476,
+      "rewards/accuracies": 0.30351561307907104,
+      "rewards/chosen": 0.0007280350546352565,
+      "rewards/margins": 0.0002506318560335785,
+      "rewards/rejected": 0.00047740322770550847,
+      "step": 5
+    },
+    {
+      "epoch": 0.05118362124120281,
+      "grad_norm": 1.7759603261947632,
+      "learning_rate": 9e-08,
+      "logits/chosen": -2.2679052352905273,
+      "logits/rejected": -2.1166000366210938,
+      "logps/chosen": -382.0252380371094,
+      "logps/rejected": -202.4450225830078,
+      "loss": 0.7175,
+      "metrics/advantage_var": 0.07497945427894592,
+      "rewards/accuracies": 0.4945312440395355,
+      "rewards/chosen": 0.0017963653663173318,
+      "rewards/margins": 0.0005524610751308501,
+      "rewards/rejected": 0.0012439042329788208,
+      "stability/repetition_rate_mean": 0.9197490811347961,
+      "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": 1.8210687637329102,
+      "learning_rate": 1.4e-07,
+      "logits/chosen": -2.2607593536376953,
+      "logits/rejected": -2.0992488861083984,
+      "logps/chosen": -389.2146911621094,
+      "logps/rejected": -207.05316162109375,
+      "loss": 0.7179,
+      "metrics/advantage_var": 0.07240394502878189,
+      "rewards/accuracies": 0.49882811307907104,
+      "rewards/chosen": 0.0010917808394879103,
+      "rewards/margins": 4.640953557100147e-05,
+      "rewards/rejected": 0.0010453711729496717,
+      "stability/repetition_rate_mean": 0.8429937362670898,
+      "stability/response_length_mean": 345.70001220703125,
+      "stability/response_length_std": 212.67974853515625,
+      "stability/response_length_var": 45232.6796875,
+      "stability/token_entropy_mean": 6.348437309265137,
+      "step": 15
+    },
+    {
+      "epoch": 0.10236724248240563,
+      "grad_norm": 1.7627458572387695,
+      "learning_rate": 1.8999999999999998e-07,
+      "logits/chosen": -2.2612171173095703,
+      "logits/rejected": -2.114251136779785,
+      "logps/chosen": -368.7218933105469,
+      "logps/rejected": -198.56837463378906,
+      "loss": 0.7184,
+      "metrics/advantage_var": 0.06786168366670609,
+      "rewards/accuracies": 0.4976562559604645,
+      "rewards/chosen": 0.00022399029694497585,
+      "rewards/margins": -0.0004166103317402303,
+      "rewards/rejected": 0.0006406005704775453,
+      "stability/repetition_rate_mean": 0.9278882741928101,
+      "stability/response_length_mean": 739.4000244140625,
+      "stability/response_length_std": 281.282470703125,
+      "stability/response_length_var": 79119.8203125,
+      "stability/token_entropy_mean": 6.946875095367432,
+      "step": 20
+    },
+    {
+      "epoch": 0.12795905310300704,
+      "grad_norm": 1.7385414838790894,
+      "learning_rate": 2.4e-07,
+      "logits/chosen": -2.2770488262176514,
+      "logits/rejected": -2.124872922897339,
+      "logps/chosen": -378.17303466796875,
+      "logps/rejected": -205.866943359375,
+      "loss": 0.7177,
+      "metrics/advantage_var": 0.06800527125597,
+      "rewards/accuracies": 0.490234375,
+      "rewards/chosen": 0.0011880360543727875,
+      "rewards/margins": 9.972632688004524e-05,
+      "rewards/rejected": 0.0010883098002523184,
+      "stability/repetition_rate_mean": 0.8768361806869507,
+      "stability/response_length_mean": 658.5999755859375,
+      "stability/response_length_std": 288.0155029296875,
+      "stability/response_length_var": 82952.9296875,
+      "stability/token_entropy_mean": 6.731249809265137,
+      "step": 25
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 1.8636959791183472,
+      "learning_rate": 2.9e-07,
+      "logits/chosen": -2.270461082458496,
+      "logits/rejected": -2.123213768005371,
+      "logps/chosen": -373.9102478027344,
+      "logps/rejected": -208.93922424316406,
+      "loss": 0.7175,
+      "metrics/advantage_var": 0.07215224206447601,
+      "rewards/accuracies": 0.5085937976837158,
+      "rewards/chosen": 0.0020091882906854153,
+      "rewards/margins": 0.0013196640647947788,
+      "rewards/rejected": 0.0006895243423059583,
+      "stability/repetition_rate_mean": 0.8138343095779419,
+      "stability/response_length_mean": 356.0,
+      "stability/response_length_std": 249.6766815185547,
+      "stability/response_length_var": 62338.4453125,
+      "stability/token_entropy_mean": 6.196875095367432,
+      "step": 30
+    },
+    {
+      "epoch": 0.17914267434420986,
+      "grad_norm": 1.7182040214538574,
+      "learning_rate": 3.4000000000000003e-07,
+      "logits/chosen": -2.261341094970703,
+      "logits/rejected": -2.120121955871582,
+      "logps/chosen": -366.32391357421875,
+      "logps/rejected": -203.99069213867188,
+      "loss": 0.7179,
+      "metrics/advantage_var": 0.07069019973278046,
+      "rewards/accuracies": 0.518359363079071,
+      "rewards/chosen": 0.0021961305756121874,
+      "rewards/margins": 0.0004749999789055437,
+      "rewards/rejected": 0.0017211305676028132,
+      "stability/repetition_rate_mean": 0.9360000491142273,
+      "stability/response_length_mean": 606.2000122070312,
+      "stability/response_length_std": 394.1366882324219,
+      "stability/response_length_var": 155343.734375,
+      "stability/token_entropy_mean": 7.150000095367432,
+      "step": 35
+    },
+    {
+      "epoch": 0.20473448496481125,
+      "grad_norm": 1.8097240924835205,
+      "learning_rate": 3.8999999999999997e-07,
+      "logits/chosen": -2.276620864868164,
+      "logits/rejected": -2.1228880882263184,
+      "logps/chosen": -369.1826477050781,
+      "logps/rejected": -197.37864685058594,
+      "loss": 0.7176,
+      "metrics/advantage_var": 0.07206886261701584,
+      "rewards/accuracies": 0.5093749761581421,
+      "rewards/chosen": 0.0036926856264472008,
+      "rewards/margins": 0.0009252781164832413,
+      "rewards/rejected": 0.0027674080338329077,
+      "stability/repetition_rate_mean": 0.8702197074890137,
+      "stability/response_length_mean": 547.2000122070312,
+      "stability/response_length_std": 290.8897399902344,
+      "stability/response_length_var": 84616.84375,
+      "stability/token_entropy_mean": 6.526562690734863,
+      "step": 40
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 1.8226284980773926,
+      "learning_rate": 4.3999999999999997e-07,
+      "logits/chosen": -2.2571909427642822,
+      "logits/rejected": -2.1073906421661377,
+      "logps/chosen": -385.17181396484375,
+      "logps/rejected": -212.56410217285156,
+      "loss": 0.718,
+      "metrics/advantage_var": 0.07308375090360641,
+      "rewards/accuracies": 0.513671875,
+      "rewards/chosen": 0.0047181700356304646,
+      "rewards/margins": 0.00034062692429870367,
+      "rewards/rejected": 0.004377542994916439,
+      "stability/repetition_rate_mean": 0.8764154314994812,
+      "stability/response_length_mean": 366.5,
+      "stability/response_length_std": 118.86243438720703,
+      "stability/response_length_var": 14128.27734375,
+      "stability/token_entropy_mean": 6.640625,
+      "step": 45
+    },
+    {
+      "epoch": 0.2559181062060141,
+      "grad_norm": 1.8556437492370605,
+      "learning_rate": 4.9e-07,
+      "logits/chosen": -2.2644577026367188,
+      "logits/rejected": -2.109606981277466,
+      "logps/chosen": -378.29864501953125,
+      "logps/rejected": -207.4927978515625,
+      "loss": 0.7173,
+      "metrics/advantage_var": 0.07552963495254517,
+      "rewards/accuracies": 0.520703136920929,
+      "rewards/chosen": 0.005913693457841873,
+      "rewards/margins": 0.0021115851122885942,
+      "rewards/rejected": 0.0038021078798919916,
+      "stability/repetition_rate_mean": 0.8770252466201782,
+      "stability/response_length_mean": 384.1000061035156,
+      "stability/response_length_std": 178.4809112548828,
+      "stability/response_length_var": 31855.43359375,
+      "stability/token_entropy_mean": 6.543749809265137,
+      "step": 50
+    },
+    {
+      "epoch": 0.28150991682661547,
+      "grad_norm": 1.9488743543624878,
+      "learning_rate": 5.4e-07,
+      "logits/chosen": -2.25301456451416,
+      "logits/rejected": -2.094449758529663,
+      "logps/chosen": -393.7789306640625,
+      "logps/rejected": -210.35589599609375,
+      "loss": 0.7154,
+      "metrics/advantage_var": 0.07901743054389954,
+      "rewards/accuracies": 0.5429688096046448,
+      "rewards/chosen": 0.010107376612722874,
+      "rewards/margins": 0.004774035885930061,
+      "rewards/rejected": 0.005333342123776674,
+      "stability/repetition_rate_mean": 0.9362810850143433,
+      "stability/response_length_mean": 332.8999938964844,
+      "stability/response_length_std": 137.39193725585938,
+      "stability/response_length_var": 18876.544921875,
+      "stability/token_entropy_mean": 6.903124809265137,
+      "step": 55
+    },
+    {
+      "epoch": 0.30710172744721687,
+      "grad_norm": 1.8305964469909668,
+      "learning_rate": 5.9e-07,
+      "logits/chosen": -2.2730324268341064,
+      "logits/rejected": -2.111363410949707,
+      "logps/chosen": -387.6959228515625,
+      "logps/rejected": -206.80484008789062,
+      "loss": 0.7161,
+      "metrics/advantage_var": 0.08022293448448181,
+      "rewards/accuracies": 0.5257812738418579,
+      "rewards/chosen": 0.01106810662895441,
+      "rewards/margins": 0.003843404119834304,
+      "rewards/rejected": 0.007224703207612038,
+      "stability/repetition_rate_mean": 0.8751958608627319,
+      "stability/response_length_mean": 655.2999877929688,
+      "stability/response_length_std": 282.005126953125,
+      "stability/response_length_var": 79526.8984375,
+      "stability/token_entropy_mean": 6.609375,
+      "step": 60
+    },
+    {
+      "epoch": 0.3326935380678183,
+      "grad_norm": 1.6957237720489502,
+      "learning_rate": 6.4e-07,
+      "logits/chosen": -2.251922369003296,
+      "logits/rejected": -2.0987448692321777,
+      "logps/chosen": -377.4503173828125,
+      "logps/rejected": -204.38616943359375,
+      "loss": 0.715,
+      "metrics/advantage_var": 0.08649040758609772,
+      "rewards/accuracies": 0.5859375,
+      "rewards/chosen": 0.014928197488188744,
+      "rewards/margins": 0.007287245709449053,
+      "rewards/rejected": 0.007640953175723553,
+      "stability/repetition_rate_mean": 0.9006223678588867,
+      "stability/response_length_mean": 306.3999938964844,
+      "stability/response_length_std": 141.7236785888672,
+      "stability/response_length_var": 20085.599609375,
+      "stability/token_entropy_mean": 6.818749904632568,
+      "step": 65
+    },
+    {
+      "epoch": 0.3582853486884197,
+      "grad_norm": 1.9992988109588623,
+      "learning_rate": 6.9e-07,
+      "logits/chosen": -2.2710888385772705,
+      "logits/rejected": -2.106233835220337,
+      "logps/chosen": -393.7522277832031,
+      "logps/rejected": -206.18975830078125,
+      "loss": 0.7139,
+      "metrics/advantage_var": 0.09332478046417236,
+      "rewards/accuracies": 0.5785155892372131,
+      "rewards/chosen": 0.015660293400287628,
+      "rewards/margins": 0.007665838580578566,
+      "rewards/rejected": 0.007994452491402626,
+      "stability/repetition_rate_mean": 0.9697188138961792,
+      "stability/response_length_mean": 472.70001220703125,
+      "stability/response_length_std": 159.89791870117188,
+      "stability/response_length_var": 25567.34375,
+      "stability/token_entropy_mean": 7.199999809265137,
+      "step": 70
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 1.8864988088607788,
+      "learning_rate": 7.4e-07,
+      "logits/chosen": -2.2754428386688232,
+      "logits/rejected": -2.1162445545196533,
+      "logps/chosen": -381.0527648925781,
+      "logps/rejected": -207.933837890625,
+      "loss": 0.7136,
+      "metrics/advantage_var": 0.10101056843996048,
+      "rewards/accuracies": 0.5796874761581421,
+      "rewards/chosen": 0.019434239715337753,
+      "rewards/margins": 0.009132225066423416,
+      "rewards/rejected": 0.010302015580236912,
+      "stability/repetition_rate_mean": 0.9078971147537231,
+      "stability/response_length_mean": 527.0,
+      "stability/response_length_std": 293.7500915527344,
+      "stability/response_length_var": 86289.109375,
+      "stability/token_entropy_mean": 6.140625,
+      "step": 75
+    },
+    {
+      "epoch": 0.4094689699296225,
+      "grad_norm": 1.8260862827301025,
+      "learning_rate": 7.9e-07,
+      "logits/chosen": -2.260484218597412,
+      "logits/rejected": -2.10276198387146,
+      "logps/chosen": -382.4151611328125,
+      "logps/rejected": -208.9755096435547,
+      "loss": 0.713,
+      "metrics/advantage_var": 0.1154346615076065,
+      "rewards/accuracies": 0.583984375,
+      "rewards/chosen": 0.02179381437599659,
+      "rewards/margins": 0.009805195033550262,
+      "rewards/rejected": 0.011988619342446327,
+      "stability/repetition_rate_mean": 0.9465335011482239,
+      "stability/response_length_mean": 635.5,
+      "stability/response_length_std": 262.53900146484375,
+      "stability/response_length_var": 68926.71875,
+      "stability/token_entropy_mean": 7.199999809265137,
+      "step": 80
+    },
+    {
+      "epoch": 0.4350607805502239,
+      "grad_norm": 1.871209740638733,
+      "learning_rate": 8.399999999999999e-07,
+      "logits/chosen": -2.257699966430664,
+      "logits/rejected": -2.1006553173065186,
+      "logps/chosen": -387.3171081542969,
+      "logps/rejected": -208.62698364257812,
+      "loss": 0.7122,
+      "metrics/advantage_var": 0.14266015589237213,
+      "rewards/accuracies": 0.5902344584465027,
+      "rewards/chosen": 0.02743753232061863,
+      "rewards/margins": 0.012249825522303581,
+      "rewards/rejected": 0.015187706798315048,
+      "stability/repetition_rate_mean": 0.9052130579948425,
+      "stability/response_length_mean": 600.0999755859375,
+      "stability/response_length_std": 349.2030334472656,
+      "stability/response_length_var": 121942.765625,
+      "stability/token_entropy_mean": 6.737500190734863,
+      "step": 85
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 1.8958905935287476,
+      "learning_rate": 8.9e-07,
+      "logits/chosen": -2.261382818222046,
+      "logits/rejected": -2.113365888595581,
+      "logps/chosen": -375.1956481933594,
+      "logps/rejected": -206.21043395996094,
+      "loss": 0.7107,
+      "metrics/advantage_var": 0.16143371164798737,
+      "rewards/accuracies": 0.6230468153953552,
+      "rewards/chosen": 0.033041536808013916,
+      "rewards/margins": 0.015009062364697456,
+      "rewards/rejected": 0.01803247444331646,
+      "stability/repetition_rate_mean": 0.9765706062316895,
+      "stability/response_length_mean": 692.9000244140625,
+      "stability/response_length_std": 127.31452941894531,
+      "stability/response_length_var": 16208.9892578125,
+      "stability/token_entropy_mean": 7.106249809265137,
+      "step": 90
+    },
+    {
+      "epoch": 0.48624440179142675,
+      "grad_norm": 1.8168452978134155,
+      "learning_rate": 9.399999999999999e-07,
+      "logits/chosen": -2.258915901184082,
+      "logits/rejected": -2.112346887588501,
+      "logps/chosen": -366.824462890625,
+      "logps/rejected": -203.4202423095703,
+      "loss": 0.7085,
+      "metrics/advantage_var": 0.20107638835906982,
+      "rewards/accuracies": 0.6175781488418579,
+      "rewards/chosen": 0.038514357060194016,
+      "rewards/margins": 0.019381558522582054,
+      "rewards/rejected": 0.01913280040025711,
+      "stability/repetition_rate_mean": 0.9561265707015991,
+      "stability/response_length_mean": 727.0,
+      "stability/response_length_std": 335.0764465332031,
+      "stability/response_length_var": 112276.21875,
+      "stability/token_entropy_mean": 7.340624809265137,
+      "step": 95
+    },
+    {
+      "epoch": 0.5118362124120281,
+      "grad_norm": 1.8184173107147217,
+      "learning_rate": 9.9e-07,
+      "logits/chosen": -2.2744827270507812,
+      "logits/rejected": -2.119138479232788,
+      "logps/chosen": -384.4900817871094,
+      "logps/rejected": -210.60585021972656,
+      "loss": 0.7067,
+      "metrics/advantage_var": 0.2732086777687073,
+      "rewards/accuracies": 0.6441406607627869,
+      "rewards/chosen": 0.046771928668022156,
+      "rewards/margins": 0.023212522268295288,
+      "rewards/rejected": 0.023559408262372017,
+      "stability/repetition_rate_mean": 0.9514176249504089,
+      "stability/response_length_mean": 419.0,
+      "stability/response_length_std": 232.0225067138672,
+      "stability/response_length_var": 53834.4453125,
+      "stability/token_entropy_mean": 7.068749904632568,
+      "step": 100
+    },
+    {
+      "epoch": 0.5374280230326296,
+      "grad_norm": 1.741348385810852,
+      "learning_rate": 9.999815554351224e-07,
+      "logits/chosen": -2.2587363719940186,
+      "logits/rejected": -2.1084041595458984,
+      "logps/chosen": -377.98284912109375,
+      "logps/rejected": -209.66653442382812,
+      "loss": 0.7059,
+      "metrics/advantage_var": 0.34950339794158936,
+      "rewards/accuracies": 0.6378905773162842,
+      "rewards/chosen": 0.05437100678682327,
+      "rewards/margins": 0.025791026651859283,
+      "rewards/rejected": 0.02857998013496399,
+      "stability/repetition_rate_mean": 0.9007232785224915,
+      "stability/response_length_mean": 669.5,
+      "stability/response_length_std": 271.60888671875,
+      "stability/response_length_var": 73771.390625,
+      "stability/token_entropy_mean": 6.55859375,
+      "step": 105
+    },
+    {
+      "epoch": 0.5630198336532309,
+      "grad_norm": 1.6754577159881592,
+      "learning_rate": 9.999066267225444e-07,
+      "logits/chosen": -2.275355339050293,
+      "logits/rejected": -2.117722988128662,
+      "logps/chosen": -385.252197265625,
+      "logps/rejected": -206.70867919921875,
+      "loss": 0.7043,
+      "metrics/advantage_var": 0.4181045889854431,
+      "rewards/accuracies": 0.6539062261581421,
+      "rewards/chosen": 0.060924846678972244,
+      "rewards/margins": 0.028999093919992447,
+      "rewards/rejected": 0.0319257527589798,
+      "stability/repetition_rate_mean": 0.9306775331497192,
+      "stability/response_length_mean": 255.60000610351562,
+      "stability/response_length_std": 114.30874633789062,
+      "stability/response_length_var": 13066.4892578125,
+      "stability/token_entropy_mean": 6.971875190734863,
+      "step": 110
+    },
+    {
+      "epoch": 0.5886116442738324,
+      "grad_norm": 1.795372724533081,
+      "learning_rate": 9.997740697079592e-07,
+      "logits/chosen": -2.251579523086548,
+      "logits/rejected": -2.1110000610351562,
+      "logps/chosen": -372.7075500488281,
+      "logps/rejected": -213.91139221191406,
+      "loss": 0.7016,
+      "metrics/advantage_var": 0.6075757145881653,
+      "rewards/accuracies": 0.651171863079071,
+      "rewards/chosen": 0.07327413558959961,
+      "rewards/margins": 0.03547437861561775,
+      "rewards/rejected": 0.037799760699272156,
+      "stability/repetition_rate_mean": 0.8305716514587402,
+      "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": 1.7747018337249756,
+      "learning_rate": 9.995838996722914e-07,
+      "logits/chosen": -2.248891592025757,
+      "logits/rejected": -2.1032941341400146,
+      "logps/chosen": -367.7718200683594,
+      "logps/rejected": -207.56582641601562,
+      "loss": 0.6987,
+      "metrics/advantage_var": 0.7378088235855103,
+      "rewards/accuracies": 0.662109375,
+      "rewards/chosen": 0.08318294584751129,
+      "rewards/margins": 0.04151991754770279,
+      "rewards/rejected": 0.0416630320250988,
+      "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": 1.7948890924453735,
+      "learning_rate": 9.993361385379874e-07,
+      "logits/chosen": -2.2534642219543457,
+      "logits/rejected": -2.0952932834625244,
+      "logps/chosen": -385.6851806640625,
+      "logps/rejected": -205.0888671875,
+      "loss": 0.6951,
+      "metrics/advantage_var": 1.040814995765686,
+      "rewards/accuracies": 0.672656238079071,
+      "rewards/chosen": 0.09500810503959656,
+      "rewards/margins": 0.05008320510387421,
+      "rewards/rejected": 0.04492490366101265,
+      "stability/repetition_rate_mean": 0.9095968008041382,
+      "stability/response_length_mean": 433.79998779296875,
+      "stability/response_length_std": 126.92412567138672,
+      "stability/response_length_var": 16109.7333984375,
+      "stability/token_entropy_mean": 6.753125190734863,
+      "step": 125
+    },
+    {
+      "epoch": 0.6653870761356366,
+      "grad_norm": 1.8776071071624756,
+      "learning_rate": 9.99030814866488e-07,
+      "logits/chosen": -2.2454466819763184,
+      "logits/rejected": -2.0966694355010986,
+      "logps/chosen": -375.5842590332031,
+      "logps/rejected": -203.7355499267578,
+      "loss": 0.692,
+      "metrics/advantage_var": 1.1227996349334717,
+      "rewards/accuracies": 0.673828125,
+      "rewards/chosen": 0.10106992721557617,
+      "rewards/margins": 0.056524716317653656,
+      "rewards/rejected": 0.044545214623212814,
+      "stability/repetition_rate_mean": 0.8208531141281128,
+      "stability/response_length_mean": 296.1000061035156,
+      "stability/response_length_std": 167.77593994140625,
+      "stability/response_length_var": 28148.767578125,
+      "stability/token_entropy_mean": 6.074999809265137,
+      "step": 130
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 1.7475597858428955,
+      "learning_rate": 9.986679638549358e-07,
+      "logits/chosen": -2.2671430110931396,
+      "logits/rejected": -2.107732057571411,
+      "logps/chosen": -387.3439636230469,
+      "logps/rejected": -213.7014923095703,
+      "loss": 0.6897,
+      "metrics/advantage_var": 1.4129046201705933,
+      "rewards/accuracies": 0.6683593988418579,
+      "rewards/chosen": 0.11433131992816925,
+      "rewards/margins": 0.062347739934921265,
+      "rewards/rejected": 0.05198357626795769,
+      "stability/repetition_rate_mean": 0.8598278760910034,
+      "stability/response_length_mean": 596.4000244140625,
+      "stability/response_length_std": 291.9129333496094,
+      "stability/response_length_var": 85213.15625,
+      "stability/token_entropy_mean": 6.5234375,
+      "step": 135
+    },
+    {
+      "epoch": 0.7165706973768394,
+      "grad_norm": 1.8308006525039673,
+      "learning_rate": 9.982476273321175e-07,
+      "logits/chosen": -2.255606174468994,
+      "logits/rejected": -2.100269079208374,
+      "logps/chosen": -389.0058898925781,
+      "logps/rejected": -211.13906860351562,
+      "loss": 0.686,
+      "metrics/advantage_var": 1.8579171895980835,
+      "rewards/accuracies": 0.689453125,
+      "rewards/chosen": 0.1219753548502922,
+      "rewards/margins": 0.07140406221151352,
+      "rewards/rejected": 0.05057128518819809,
+      "stability/repetition_rate_mean": 0.8619750142097473,
+      "stability/response_length_mean": 173.89999389648438,
+      "stability/response_length_std": 70.74908447265625,
+      "stability/response_length_var": 5005.43310546875,
+      "stability/token_entropy_mean": 6.634375095367432,
+      "step": 140
+    },
+    {
+      "epoch": 0.7421625079974408,
+      "grad_norm": 1.7569092512130737,
+      "learning_rate": 9.977698537536417e-07,
+      "logits/chosen": -2.267023801803589,
+      "logits/rejected": -2.0997886657714844,
+      "logps/chosen": -387.7537841796875,
+      "logps/rejected": -203.59815979003906,
+      "loss": 0.6834,
+      "metrics/advantage_var": 2.1575894355773926,
+      "rewards/accuracies": 0.68359375,
+      "rewards/chosen": 0.13592004776000977,
+      "rewards/margins": 0.07768016308546066,
+      "rewards/rejected": 0.0582398846745491,
+      "stability/repetition_rate_mean": 0.960322380065918,
+      "stability/response_length_mean": 720.2000122070312,
+      "stability/response_length_std": 284.2842102050781,
+      "stability/response_length_var": 80817.5078125,
+      "stability/token_entropy_mean": 7.084374904632568,
+      "step": 145
+    },
+    {
+      "epoch": 0.7677543186180422,
+      "grad_norm": 1.8092952966690063,
+      "learning_rate": 9.972346981963546e-07,
+      "logits/chosen": -2.234330415725708,
+      "logits/rejected": -2.092057704925537,
+      "logps/chosen": -373.80157470703125,
+      "logps/rejected": -211.20767211914062,
+      "loss": 0.6816,
+      "metrics/advantage_var": 3.304790735244751,
+      "rewards/accuracies": 0.6605468392372131,
+      "rewards/chosen": 0.14637623727321625,
+      "rewards/margins": 0.08596397191286087,
+      "rewards/rejected": 0.06041226163506508,
+      "stability/repetition_rate_mean": 0.8549944162368774,
+      "stability/response_length_mean": 654.2000122070312,
+      "stability/response_length_std": 372.5482177734375,
+      "stability/response_length_var": 138792.171875,
+      "stability/token_entropy_mean": 6.556250095367432,
+      "step": 150
+    },
+    {
+      "epoch": 0.7933461292386437,
+      "grad_norm": 1.8317517042160034,
+      "learning_rate": 9.966422223519885e-07,
+      "logits/chosen": -2.260176658630371,
+      "logits/rejected": -2.1092514991760254,
+      "logps/chosen": -377.3446350097656,
+      "logps/rejected": -206.38644409179688,
+      "loss": 0.677,
+      "metrics/advantage_var": 2.97253680229187,
+      "rewards/accuracies": 0.680468738079071,
+      "rewards/chosen": 0.16198468208312988,
+      "rewards/margins": 0.09395644068717957,
+      "rewards/rejected": 0.06802822649478912,
+      "stability/repetition_rate_mean": 0.8795069456100464,
+      "stability/response_length_mean": 502.3999938964844,
+      "stability/response_length_std": 200.0795440673828,
+      "stability/response_length_var": 40031.8203125,
+      "stability/token_entropy_mean": 6.557812690734863,
+      "step": 155
+    },
+    {
+      "epoch": 0.818937939859245,
+      "grad_norm": 1.7747522592544556,
+      "learning_rate": 9.959924945200523e-07,
+      "logits/chosen": -2.253509759902954,
+      "logits/rejected": -2.086948871612549,
+      "logps/chosen": -383.45159912109375,
+      "logps/rejected": -207.2310333251953,
+      "loss": 0.6736,
+      "metrics/advantage_var": 4.025539398193359,
+      "rewards/accuracies": 0.6894532442092896,
+      "rewards/chosen": 0.1707187443971634,
+      "rewards/margins": 0.10508601367473602,
+      "rewards/rejected": 0.06563273072242737,
+      "stability/repetition_rate_mean": 0.969073474407196,
+      "stability/response_length_mean": 870.4000244140625,
+      "stability/response_length_std": 147.6581573486328,
+      "stability/response_length_var": 21802.93359375,
+      "stability/token_entropy_mean": 7.228125095367432,
+      "step": 160
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.641774296760559,
+      "learning_rate": 9.952855895999568e-07,
+      "logits/chosen": -2.238173484802246,
+      "logits/rejected": -2.0906169414520264,
+      "logps/chosen": -362.37127685546875,
+      "logps/rejected": -203.5992431640625,
+      "loss": 0.6721,
+      "metrics/advantage_var": 3.921074867248535,
+      "rewards/accuracies": 0.6953125,
+      "rewards/chosen": 0.17040367424488068,
+      "rewards/margins": 0.10770220309495926,
+      "rewards/rejected": 0.06270146369934082,
+      "stability/repetition_rate_mean": 0.8834701776504517,
+      "stability/response_length_mean": 547.7000122070312,
+      "stability/response_length_std": 249.60325622558594,
+      "stability/response_length_var": 62301.7890625,
+      "stability/token_entropy_mean": 6.631249904632568,
+      "step": 165
+    },
+    {
+      "epoch": 0.8701215611004478,
+      "grad_norm": 1.7060259580612183,
+      "learning_rate": 9.9452158908238e-07,
+      "logits/chosen": -2.2633023262023926,
+      "logits/rejected": -2.0972681045532227,
+      "logps/chosen": -376.9095153808594,
+      "logps/rejected": -203.75167846679688,
+      "loss": 0.6663,
+      "metrics/advantage_var": 4.847568988800049,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": 0.18648947775363922,
+      "rewards/margins": 0.12270568311214447,
+      "rewards/rejected": 0.06378380209207535,
+      "stability/repetition_rate_mean": 0.9116168022155762,
+      "stability/response_length_mean": 385.70001220703125,
+      "stability/response_length_std": 143.96685791015625,
+      "stability/response_length_var": 20726.455078125,
+      "stability/token_entropy_mean": 6.831250190734863,
+      "step": 170
+    },
+    {
+      "epoch": 0.8957133717210493,
+      "grad_norm": 1.7979154586791992,
+      "learning_rate": 9.937005810398745e-07,
+      "logits/chosen": -2.249255895614624,
+      "logits/rejected": -2.093829870223999,
+      "logps/chosen": -380.42779541015625,
+      "logps/rejected": -212.22317504882812,
+      "loss": 0.6642,
+      "metrics/advantage_var": 6.434874534606934,
+      "rewards/accuracies": 0.6957030892372131,
+      "rewards/chosen": 0.19610357284545898,
+      "rewards/margins": 0.13212153315544128,
+      "rewards/rejected": 0.0639820322394371,
+      "stability/repetition_rate_mean": 0.9416834712028503,
+      "stability/response_length_mean": 460.6000061035156,
+      "stability/response_length_std": 165.9894256591797,
+      "stability/response_length_var": 27552.48828125,
+      "stability/token_entropy_mean": 6.956250190734863,
+      "step": 175
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 2.1531741619110107,
+      "learning_rate": 9.928226601167138e-07,
+      "logits/chosen": -2.248629093170166,
+      "logits/rejected": -2.0873544216156006,
+      "logps/chosen": -386.9737854003906,
+      "logps/rejected": -212.9161376953125,
+      "loss": 0.6621,
+      "metrics/advantage_var": 5.855362415313721,
+      "rewards/accuracies": 0.6867187023162842,
+      "rewards/chosen": 0.21073000133037567,
+      "rewards/margins": 0.13600066304206848,
+      "rewards/rejected": 0.07472933828830719,
+      "stability/repetition_rate_mean": 0.8946305513381958,
+      "stability/response_length_mean": 716.2000122070312,
+      "stability/response_length_std": 329.3815002441406,
+      "stability/response_length_var": 108492.1796875,
+      "stability/token_entropy_mean": 6.5234375,
+      "step": 180
+    },
+    {
+      "epoch": 0.946896992962252,
+      "grad_norm": 1.6736973524093628,
+      "learning_rate": 9.918879275179817e-07,
+      "logits/chosen": -2.250131130218506,
+      "logits/rejected": -2.099020481109619,
+      "logps/chosen": -374.459716796875,
+      "logps/rejected": -210.3748016357422,
+      "loss": 0.6621,
+      "metrics/advantage_var": 7.431878566741943,
+      "rewards/accuracies": 0.68359375,
+      "rewards/chosen": 0.216688334941864,
+      "rewards/margins": 0.14022144675254822,
+      "rewards/rejected": 0.07646691799163818,
+      "stability/repetition_rate_mean": 0.8750947713851929,
+      "stability/response_length_mean": 486.3999938964844,
+      "stability/response_length_std": 213.8624725341797,
+      "stability/response_length_var": 45737.15625,
+      "stability/token_entropy_mean": 6.5625,
+      "step": 185
+    },
+    {
+      "epoch": 0.9724888035828535,
+      "grad_norm": 1.5515198707580566,
+      "learning_rate": 9.90896490997906e-07,
+      "logits/chosen": -2.234440326690674,
+      "logits/rejected": -2.0786972045898438,
+      "logps/chosen": -377.5950622558594,
+      "logps/rejected": -206.24668884277344,
+      "loss": 0.6539,
+      "metrics/advantage_var": 8.211545944213867,
+      "rewards/accuracies": 0.7039062976837158,
+      "rewards/chosen": 0.22378642857074738,
+      "rewards/margins": 0.1613050103187561,
+      "rewards/rejected": 0.062481436878442764,
+      "stability/repetition_rate_mean": 0.8891245722770691,
+      "stability/response_length_mean": 736.7999877929688,
+      "stability/response_length_std": 246.57106018066406,
+      "stability/response_length_var": 60797.2890625,
+      "stability/token_entropy_mean": 6.714062690734863,
+      "step": 190
+    },
+    {
+      "epoch": 0.9980806142034548,
+      "grad_norm": 1.6709520816802979,
+      "learning_rate": 9.898484648474362e-07,
+      "logits/chosen": -2.2508366107940674,
+      "logits/rejected": -2.0976295471191406,
+      "logps/chosen": -361.7326965332031,
+      "logps/rejected": -208.36793518066406,
+      "loss": 0.6545,
+      "metrics/advantage_var": 8.507060050964355,
+      "rewards/accuracies": 0.6992188096046448,
+      "rewards/chosen": 0.22635439038276672,
+      "rewards/margins": 0.16190506517887115,
+      "rewards/rejected": 0.06444933265447617,
+      "stability/repetition_rate_mean": 0.7645308375358582,
+      "stability/response_length_mean": 356.3999938964844,
+      "stability/response_length_std": 207.94882202148438,
+      "stability/response_length_var": 43242.7109375,
+      "stability/token_entropy_mean": 5.896874904632568,
+      "step": 195
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 196,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 200,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.072674331683193e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

mistral7b_dpo_sft_ftx0.01/checkpoint-196/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1a19b35383c5d0923ed5719f6321bcbbdad041809e90dbc9f6d7f7a5fde3b88
+size 8017

mistral7b_dpo_sft_ftx0.01/checkpoint-196/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

mistral7b_dpo_sft_ftx0.01/special_tokens_map.json

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

mistral7b_dpo_sft_ftx0.01/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": 0.9197490811347961, "step": 10}
+{"stability/response_length_mean": 345.70001220703125, "stability/response_length_std": 212.67974853515625, "stability/response_length_var": 45232.6796875, "stability/token_entropy_mean": 6.348437309265137, "stability/repetition_rate_mean": 0.8429937362670898, "step": 15}
+{"stability/response_length_mean": 739.4000244140625, "stability/response_length_std": 281.282470703125, "stability/response_length_var": 79119.8203125, "stability/token_entropy_mean": 6.946875095367432, "stability/repetition_rate_mean": 0.9278882741928101, "step": 20}
+{"stability/response_length_mean": 658.5999755859375, "stability/response_length_std": 288.0155029296875, "stability/response_length_var": 82952.9296875, "stability/token_entropy_mean": 6.731249809265137, "stability/repetition_rate_mean": 0.8768361806869507, "step": 25}
+{"stability/response_length_mean": 356.0, "stability/response_length_std": 249.6766815185547, "stability/response_length_var": 62338.4453125, "stability/token_entropy_mean": 6.196875095367432, "stability/repetition_rate_mean": 0.8138343095779419, "step": 30}
+{"stability/response_length_mean": 606.2000122070312, "stability/response_length_std": 394.1366882324219, "stability/response_length_var": 155343.734375, "stability/token_entropy_mean": 7.150000095367432, "stability/repetition_rate_mean": 0.9360000491142273, "step": 35}
+{"stability/response_length_mean": 547.2000122070312, "stability/response_length_std": 290.8897399902344, "stability/response_length_var": 84616.84375, "stability/token_entropy_mean": 6.526562690734863, "stability/repetition_rate_mean": 0.8702197074890137, "step": 40}
+{"stability/response_length_mean": 366.5, "stability/response_length_std": 118.86243438720703, "stability/response_length_var": 14128.27734375, "stability/token_entropy_mean": 6.640625, "stability/repetition_rate_mean": 0.8764154314994812, "step": 45}
+{"stability/response_length_mean": 384.1000061035156, "stability/response_length_std": 178.4809112548828, "stability/response_length_var": 31855.43359375, "stability/token_entropy_mean": 6.543749809265137, "stability/repetition_rate_mean": 0.8770252466201782, "step": 50}
+{"stability/response_length_mean": 332.8999938964844, "stability/response_length_std": 137.39193725585938, "stability/response_length_var": 18876.544921875, "stability/token_entropy_mean": 6.903124809265137, "stability/repetition_rate_mean": 0.9362810850143433, "step": 55}
+{"stability/response_length_mean": 655.2999877929688, "stability/response_length_std": 282.005126953125, "stability/response_length_var": 79526.8984375, "stability/token_entropy_mean": 6.609375, "stability/repetition_rate_mean": 0.8751958608627319, "step": 60}
+{"stability/response_length_mean": 306.3999938964844, "stability/response_length_std": 141.7236785888672, "stability/response_length_var": 20085.599609375, "stability/token_entropy_mean": 6.818749904632568, "stability/repetition_rate_mean": 0.9006223678588867, "step": 65}
+{"stability/response_length_mean": 472.70001220703125, "stability/response_length_std": 159.89791870117188, "stability/response_length_var": 25567.34375, "stability/token_entropy_mean": 7.199999809265137, "stability/repetition_rate_mean": 0.9697188138961792, "step": 70}
+{"stability/response_length_mean": 527.0, "stability/response_length_std": 293.7500915527344, "stability/response_length_var": 86289.109375, "stability/token_entropy_mean": 6.140625, "stability/repetition_rate_mean": 0.9078971147537231, "step": 75}
+{"stability/response_length_mean": 635.5, "stability/response_length_std": 262.53900146484375, "stability/response_length_var": 68926.71875, "stability/token_entropy_mean": 7.199999809265137, "stability/repetition_rate_mean": 0.9465335011482239, "step": 80}
+{"stability/response_length_mean": 600.0999755859375, "stability/response_length_std": 349.2030334472656, "stability/response_length_var": 121942.765625, "stability/token_entropy_mean": 6.737500190734863, "stability/repetition_rate_mean": 0.9052130579948425, "step": 85}
+{"stability/response_length_mean": 692.9000244140625, "stability/response_length_std": 127.31452941894531, "stability/response_length_var": 16208.9892578125, "stability/token_entropy_mean": 7.106249809265137, "stability/repetition_rate_mean": 0.9765706062316895, "step": 90}
+{"stability/response_length_mean": 727.0, "stability/response_length_std": 335.0764465332031, "stability/response_length_var": 112276.21875, "stability/token_entropy_mean": 7.340624809265137, "stability/repetition_rate_mean": 0.9561265707015991, "step": 95}
+{"stability/response_length_mean": 419.0, "stability/response_length_std": 232.0225067138672, "stability/response_length_var": 53834.4453125, "stability/token_entropy_mean": 7.068749904632568, "stability/repetition_rate_mean": 0.9514176249504089, "step": 100}
+{"stability/response_length_mean": 669.5, "stability/response_length_std": 271.60888671875, "stability/response_length_var": 73771.390625, "stability/token_entropy_mean": 6.55859375, "stability/repetition_rate_mean": 0.9007232785224915, "step": 105}
+{"stability/response_length_mean": 255.60000610351562, "stability/response_length_std": 114.30874633789062, "stability/response_length_var": 13066.4892578125, "stability/token_entropy_mean": 6.971875190734863, "stability/repetition_rate_mean": 0.9306775331497192, "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": 0.8305716514587402, "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": 6.753125190734863, "stability/repetition_rate_mean": 0.9095968008041382, "step": 125}
+{"stability/response_length_mean": 296.1000061035156, "stability/response_length_std": 167.77593994140625, "stability/response_length_var": 28148.767578125, "stability/token_entropy_mean": 6.074999809265137, "stability/repetition_rate_mean": 0.8208531141281128, "step": 130}
+{"stability/response_length_mean": 596.4000244140625, "stability/response_length_std": 291.9129333496094, "stability/response_length_var": 85213.15625, "stability/token_entropy_mean": 6.5234375, "stability/repetition_rate_mean": 0.8598278760910034, "step": 135}
+{"stability/response_length_mean": 173.89999389648438, "stability/response_length_std": 70.74908447265625, "stability/response_length_var": 5005.43310546875, "stability/token_entropy_mean": 6.634375095367432, "stability/repetition_rate_mean": 0.8619750142097473, "step": 140}
+{"stability/response_length_mean": 720.2000122070312, "stability/response_length_std": 284.2842102050781, "stability/response_length_var": 80817.5078125, "stability/token_entropy_mean": 7.084374904632568, "stability/repetition_rate_mean": 0.960322380065918, "step": 145}
+{"stability/response_length_mean": 654.2000122070312, "stability/response_length_std": 372.5482177734375, "stability/response_length_var": 138792.171875, "stability/token_entropy_mean": 6.556250095367432, "stability/repetition_rate_mean": 0.8549944162368774, "step": 150}
+{"stability/response_length_mean": 502.3999938964844, "stability/response_length_std": 200.0795440673828, "stability/response_length_var": 40031.8203125, "stability/token_entropy_mean": 6.557812690734863, "stability/repetition_rate_mean": 0.8795069456100464, "step": 155}
+{"stability/response_length_mean": 870.4000244140625, "stability/response_length_std": 147.6581573486328, "stability/response_length_var": 21802.93359375, "stability/token_entropy_mean": 7.228125095367432, "stability/repetition_rate_mean": 0.969073474407196, "step": 160}
+{"stability/response_length_mean": 547.7000122070312, "stability/response_length_std": 249.60325622558594, "stability/response_length_var": 62301.7890625, "stability/token_entropy_mean": 6.631249904632568, "stability/repetition_rate_mean": 0.8834701776504517, "step": 165}
+{"stability/response_length_mean": 385.70001220703125, "stability/response_length_std": 143.96685791015625, "stability/response_length_var": 20726.455078125, "stability/token_entropy_mean": 6.831250190734863, "stability/repetition_rate_mean": 0.9116168022155762, "step": 170}
+{"stability/response_length_mean": 460.6000061035156, "stability/response_length_std": 165.9894256591797, "stability/response_length_var": 27552.48828125, "stability/token_entropy_mean": 6.956250190734863, "stability/repetition_rate_mean": 0.9416834712028503, "step": 175}
+{"stability/response_length_mean": 716.2000122070312, "stability/response_length_std": 329.3815002441406, "stability/response_length_var": 108492.1796875, "stability/token_entropy_mean": 6.5234375, "stability/repetition_rate_mean": 0.8946305513381958, "step": 180}
+{"stability/response_length_mean": 486.3999938964844, "stability/response_length_std": 213.8624725341797, "stability/response_length_var": 45737.15625, "stability/token_entropy_mean": 6.5625, "stability/repetition_rate_mean": 0.8750947713851929, "step": 185}
+{"stability/response_length_mean": 736.7999877929688, "stability/response_length_std": 246.57106018066406, "stability/response_length_var": 60797.2890625, "stability/token_entropy_mean": 6.714062690734863, "stability/repetition_rate_mean": 0.8891245722770691, "step": 190}
+{"stability/response_length_mean": 356.3999938964844, "stability/response_length_std": 207.94882202148438, "stability/response_length_var": 43242.7109375, "stability/token_entropy_mean": 5.896874904632568, "stability/repetition_rate_mean": 0.7645308375358582, "step": 195}

mistral7b_dpo_sft_ftx0.01/tokenizer.model

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

mistral7b_dpo_sft_ftx0.01/tokenizer_config.json

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

mistral7b_dpo_sft_ftx0.01/train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "total_flos": 7.072674331683193e+18,
+    "train_loss": 0.6978096469324462,
+    "train_runtime": 4339.2504,
+    "train_samples_per_second": 23.045,
+    "train_steps_per_second": 0.045
+}

mistral7b_dpo_sft_ftx0.01/trainer_log.jsonl

@@ -0,0 +1,40 @@
+{"current_steps": 5, "total_steps": 196, "loss": 0.7178, "accuracy": 0.30351561307907104, "lr": 4e-08, "epoch": 0.025591810620601407, "percentage": 2.55, "elapsed_time": "0:01:47", "remaining_time": "1:08:42", "rewards/chosen": 0.0007280350546352565, "rewards/rejected": 0.00047740322770550847, "rewards/accuracies": 0.30351561307907104, "rewards/margins": 0.0002506318560335785, "logps/chosen": -375.06817626953125, "logps/rejected": -210.86959838867188, "logits/chosen": -2.272961139678955, "logits/rejected": -2.1232829093933105}
+{"current_steps": 10, "total_steps": 196, "loss": 0.7175, "accuracy": 0.4945312440395355, "lr": 9e-08, "epoch": 0.05118362124120281, "percentage": 5.1, "elapsed_time": "0:03:39", "remaining_time": "1:07:59", "rewards/chosen": 0.0017963653663173318, "rewards/rejected": 0.0012439042329788208, "rewards/accuracies": 0.4945312440395355, "rewards/margins": 0.0005524610751308501, "logps/chosen": -382.0252380371094, "logps/rejected": -202.4450225830078, "logits/chosen": -2.2679052352905273, "logits/rejected": -2.1166000366210938, "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": 0.9197490811347961}
+{"current_steps": 15, "total_steps": 196, "loss": 0.7179, "accuracy": 0.49882811307907104, "lr": 1.4e-07, "epoch": 0.07677543186180422, "percentage": 7.65, "elapsed_time": "0:05:29", "remaining_time": "1:06:20", "rewards/chosen": 0.0010917808394879103, "rewards/rejected": 0.0010453711729496717, "rewards/accuracies": 0.49882811307907104, "rewards/margins": 4.640953557100147e-05, "logps/chosen": -389.2146911621094, "logps/rejected": -207.05316162109375, "logits/chosen": -2.2607593536376953, "logits/rejected": -2.0992488861083984, "stability/response_length_mean": 345.70001220703125, "stability/response_length_std": 212.67974853515625, "stability/response_length_var": 45232.6796875, "stability/token_entropy_mean": 6.348437309265137, "stability/repetition_rate_mean": 0.8429937362670898}
+{"current_steps": 20, "total_steps": 196, "loss": 0.7184, "accuracy": 0.4976562559604645, "lr": 1.8999999999999998e-07, "epoch": 0.10236724248240563, "percentage": 10.2, "elapsed_time": "0:07:20", "remaining_time": "1:04:38", "rewards/chosen": 0.00022399029694497585, "rewards/rejected": 0.0006406005704775453, "rewards/accuracies": 0.4976562559604645, "rewards/margins": -0.0004166103317402303, "logps/chosen": -368.7218933105469, "logps/rejected": -198.56837463378906, "logits/chosen": -2.2612171173095703, "logits/rejected": -2.114251136779785, "stability/response_length_mean": 739.4000244140625, "stability/response_length_std": 281.282470703125, "stability/response_length_var": 79119.8203125, "stability/token_entropy_mean": 6.946875095367432, "stability/repetition_rate_mean": 0.9278882741928101}
+{"current_steps": 25, "total_steps": 196, "loss": 0.7177, "accuracy": 0.490234375, "lr": 2.4e-07, "epoch": 0.12795905310300704, "percentage": 12.76, "elapsed_time": "0:09:12", "remaining_time": "1:02:56", "rewards/chosen": 0.0011880360543727875, "rewards/rejected": 0.0010883098002523184, "rewards/accuracies": 0.490234375, "rewards/margins": 9.972632688004524e-05, "logps/chosen": -378.17303466796875, "logps/rejected": -205.866943359375, "logits/chosen": -2.2770488262176514, "logits/rejected": -2.124872922897339, "stability/response_length_mean": 658.5999755859375, "stability/response_length_std": 288.0155029296875, "stability/response_length_var": 82952.9296875, "stability/token_entropy_mean": 6.731249809265137, "stability/repetition_rate_mean": 0.8768361806869507}
+{"current_steps": 30, "total_steps": 196, "loss": 0.7175, "accuracy": 0.5085937976837158, "lr": 2.9e-07, "epoch": 0.15355086372360843, "percentage": 15.31, "elapsed_time": "0:11:03", "remaining_time": "1:01:09", "rewards/chosen": 0.0020091882906854153, "rewards/rejected": 0.0006895243423059583, "rewards/accuracies": 0.5085937976837158, "rewards/margins": 0.0013196640647947788, "logps/chosen": -373.9102478027344, "logps/rejected": -208.93922424316406, "logits/chosen": -2.270461082458496, "logits/rejected": -2.123213768005371, "stability/response_length_mean": 356.0, "stability/response_length_std": 249.6766815185547, "stability/response_length_var": 62338.4453125, "stability/token_entropy_mean": 6.196875095367432, "stability/repetition_rate_mean": 0.8138343095779419}
+{"current_steps": 35, "total_steps": 196, "loss": 0.7179, "accuracy": 0.518359363079071, "lr": 3.4000000000000003e-07, "epoch": 0.17914267434420986, "percentage": 17.86, "elapsed_time": "0:12:54", "remaining_time": "0:59:23", "rewards/chosen": 0.0021961305756121874, "rewards/rejected": 0.0017211305676028132, "rewards/accuracies": 0.518359363079071, "rewards/margins": 0.0004749999789055437, "logps/chosen": -366.32391357421875, "logps/rejected": -203.99069213867188, "logits/chosen": -2.261341094970703, "logits/rejected": -2.120121955871582, "stability/response_length_mean": 606.2000122070312, "stability/response_length_std": 394.1366882324219, "stability/response_length_var": 155343.734375, "stability/token_entropy_mean": 7.150000095367432, "stability/repetition_rate_mean": 0.9360000491142273}
+{"current_steps": 40, "total_steps": 196, "loss": 0.7176, "accuracy": 0.5093749761581421, "lr": 3.8999999999999997e-07, "epoch": 0.20473448496481125, "percentage": 20.41, "elapsed_time": "0:14:46", "remaining_time": "0:57:37", "rewards/chosen": 0.0036926856264472008, "rewards/rejected": 0.0027674080338329077, "rewards/accuracies": 0.5093749761581421, "rewards/margins": 0.0009252781164832413, "logps/chosen": -369.1826477050781, "logps/rejected": -197.37864685058594, "logits/chosen": -2.276620864868164, "logits/rejected": -2.1228880882263184, "stability/response_length_mean": 547.2000122070312, "stability/response_length_std": 290.8897399902344, "stability/response_length_var": 84616.84375, "stability/token_entropy_mean": 6.526562690734863, "stability/repetition_rate_mean": 0.8702197074890137}
+{"current_steps": 45, "total_steps": 196, "loss": 0.718, "accuracy": 0.513671875, "lr": 4.3999999999999997e-07, "epoch": 0.23032629558541268, "percentage": 22.96, "elapsed_time": "0:16:36", "remaining_time": "0:55:45", "rewards/chosen": 0.0047181700356304646, "rewards/rejected": 0.004377542994916439, "rewards/accuracies": 0.513671875, "rewards/margins": 0.00034062692429870367, "logps/chosen": -385.17181396484375, "logps/rejected": -212.56410217285156, "logits/chosen": -2.2571909427642822, "logits/rejected": -2.1073906421661377, "stability/response_length_mean": 366.5, "stability/response_length_std": 118.86243438720703, "stability/response_length_var": 14128.27734375, "stability/token_entropy_mean": 6.640625, "stability/repetition_rate_mean": 0.8764154314994812}
+{"current_steps": 50, "total_steps": 196, "loss": 0.7173, "accuracy": 0.520703136920929, "lr": 4.9e-07, "epoch": 0.2559181062060141, "percentage": 25.51, "elapsed_time": "0:18:27", "remaining_time": "0:53:53", "rewards/chosen": 0.005913693457841873, "rewards/rejected": 0.0038021078798919916, "rewards/accuracies": 0.520703136920929, "rewards/margins": 0.0021115851122885942, "logps/chosen": -378.29864501953125, "logps/rejected": -207.4927978515625, "logits/chosen": -2.2644577026367188, "logits/rejected": -2.109606981277466, "stability/response_length_mean": 384.1000061035156, "stability/response_length_std": 178.4809112548828, "stability/response_length_var": 31855.43359375, "stability/token_entropy_mean": 6.543749809265137, "stability/repetition_rate_mean": 0.8770252466201782}
+{"current_steps": 55, "total_steps": 196, "loss": 0.7154, "accuracy": 0.5429688096046448, "lr": 5.4e-07, "epoch": 0.28150991682661547, "percentage": 28.06, "elapsed_time": "0:20:18", "remaining_time": "0:52:02", "rewards/chosen": 0.010107376612722874, "rewards/rejected": 0.005333342123776674, "rewards/accuracies": 0.5429688096046448, "rewards/margins": 0.004774035885930061, "logps/chosen": -393.7789306640625, "logps/rejected": -210.35589599609375, "logits/chosen": -2.25301456451416, "logits/rejected": -2.094449758529663, "stability/response_length_mean": 332.8999938964844, "stability/response_length_std": 137.39193725585938, "stability/response_length_var": 18876.544921875, "stability/token_entropy_mean": 6.903124809265137, "stability/repetition_rate_mean": 0.9362810850143433}
+{"current_steps": 60, "total_steps": 196, "loss": 0.7161, "accuracy": 0.5257812738418579, "lr": 5.9e-07, "epoch": 0.30710172744721687, "percentage": 30.61, "elapsed_time": "0:22:09", "remaining_time": "0:50:14", "rewards/chosen": 0.01106810662895441, "rewards/rejected": 0.007224703207612038, "rewards/accuracies": 0.5257812738418579, "rewards/margins": 0.003843404119834304, "logps/chosen": -387.6959228515625, "logps/rejected": -206.80484008789062, "logits/chosen": -2.2730324268341064, "logits/rejected": -2.111363410949707, "stability/response_length_mean": 655.2999877929688, "stability/response_length_std": 282.005126953125, "stability/response_length_var": 79526.8984375, "stability/token_entropy_mean": 6.609375, "stability/repetition_rate_mean": 0.8751958608627319}
+{"current_steps": 65, "total_steps": 196, "loss": 0.715, "accuracy": 0.5859375, "lr": 6.4e-07, "epoch": 0.3326935380678183, "percentage": 33.16, "elapsed_time": "0:24:00", "remaining_time": "0:48:22", "rewards/chosen": 0.014928197488188744, "rewards/rejected": 0.007640953175723553, "rewards/accuracies": 0.5859375, "rewards/margins": 0.007287245709449053, "logps/chosen": -377.4503173828125, "logps/rejected": -204.38616943359375, "logits/chosen": -2.251922369003296, "logits/rejected": -2.0987448692321777, "stability/response_length_mean": 306.3999938964844, "stability/response_length_std": 141.7236785888672, "stability/response_length_var": 20085.599609375, "stability/token_entropy_mean": 6.818749904632568, "stability/repetition_rate_mean": 0.9006223678588867}
+{"current_steps": 70, "total_steps": 196, "loss": 0.7139, "accuracy": 0.5785155892372131, "lr": 6.9e-07, "epoch": 0.3582853486884197, "percentage": 35.71, "elapsed_time": "0:25:50", "remaining_time": "0:46:31", "rewards/chosen": 0.015660293400287628, "rewards/rejected": 0.007994452491402626, "rewards/accuracies": 0.5785155892372131, "rewards/margins": 0.007665838580578566, "logps/chosen": -393.7522277832031, "logps/rejected": -206.18975830078125, "logits/chosen": -2.2710888385772705, "logits/rejected": -2.106233835220337, "stability/response_length_mean": 472.70001220703125, "stability/response_length_std": 159.89791870117188, "stability/response_length_var": 25567.34375, "stability/token_entropy_mean": 7.199999809265137, "stability/repetition_rate_mean": 0.9697188138961792}
+{"current_steps": 75, "total_steps": 196, "loss": 0.7136, "accuracy": 0.5796874761581421, "lr": 7.4e-07, "epoch": 0.3838771593090211, "percentage": 38.27, "elapsed_time": "0:27:41", "remaining_time": "0:44:40", "rewards/chosen": 0.019434239715337753, "rewards/rejected": 0.010302015580236912, "rewards/accuracies": 0.5796874761581421, "rewards/margins": 0.009132225066423416, "logps/chosen": -381.0527648925781, "logps/rejected": -207.933837890625, "logits/chosen": -2.2754428386688232, "logits/rejected": -2.1162445545196533, "stability/response_length_mean": 527.0, "stability/response_length_std": 293.7500915527344, "stability/response_length_var": 86289.109375, "stability/token_entropy_mean": 6.140625, "stability/repetition_rate_mean": 0.9078971147537231}
+{"current_steps": 80, "total_steps": 196, "loss": 0.713, "accuracy": 0.583984375, "lr": 7.9e-07, "epoch": 0.4094689699296225, "percentage": 40.82, "elapsed_time": "0:29:33", "remaining_time": "0:42:50", "rewards/chosen": 0.02179381437599659, "rewards/rejected": 0.011988619342446327, "rewards/accuracies": 0.583984375, "rewards/margins": 0.009805195033550262, "logps/chosen": -382.4151611328125, "logps/rejected": -208.9755096435547, "logits/chosen": -2.260484218597412, "logits/rejected": -2.10276198387146, "stability/response_length_mean": 635.5, "stability/response_length_std": 262.53900146484375, "stability/response_length_var": 68926.71875, "stability/token_entropy_mean": 7.199999809265137, "stability/repetition_rate_mean": 0.9465335011482239}
+{"current_steps": 85, "total_steps": 196, "loss": 0.7122, "accuracy": 0.5902344584465027, "lr": 8.399999999999999e-07, "epoch": 0.4350607805502239, "percentage": 43.37, "elapsed_time": "0:31:25", "remaining_time": "0:41:01", "rewards/chosen": 0.02743753232061863, "rewards/rejected": 0.015187706798315048, "rewards/accuracies": 0.5902344584465027, "rewards/margins": 0.012249825522303581, "logps/chosen": -387.3171081542969, "logps/rejected": -208.62698364257812, "logits/chosen": -2.257699966430664, "logits/rejected": -2.1006553173065186, "stability/response_length_mean": 600.0999755859375, "stability/response_length_std": 349.2030334472656, "stability/response_length_var": 121942.765625, "stability/token_entropy_mean": 6.737500190734863, "stability/repetition_rate_mean": 0.9052130579948425}
+{"current_steps": 90, "total_steps": 196, "loss": 0.7107, "accuracy": 0.6230468153953552, "lr": 8.9e-07, "epoch": 0.46065259117082535, "percentage": 45.92, "elapsed_time": "0:33:16", "remaining_time": "0:39:11", "rewards/chosen": 0.033041536808013916, "rewards/rejected": 0.01803247444331646, "rewards/accuracies": 0.6230468153953552, "rewards/margins": 0.015009062364697456, "logps/chosen": -375.1956481933594, "logps/rejected": -206.21043395996094, "logits/chosen": -2.261382818222046, "logits/rejected": -2.113365888595581, "stability/response_length_mean": 692.9000244140625, "stability/response_length_std": 127.31452941894531, "stability/response_length_var": 16208.9892578125, "stability/token_entropy_mean": 7.106249809265137, "stability/repetition_rate_mean": 0.9765706062316895}
+{"current_steps": 95, "total_steps": 196, "loss": 0.7085, "accuracy": 0.6175781488418579, "lr": 9.399999999999999e-07, "epoch": 0.48624440179142675, "percentage": 48.47, "elapsed_time": "0:35:08", "remaining_time": "0:37:21", "rewards/chosen": 0.038514357060194016, "rewards/rejected": 0.01913280040025711, "rewards/accuracies": 0.6175781488418579, "rewards/margins": 0.019381558522582054, "logps/chosen": -366.824462890625, "logps/rejected": -203.4202423095703, "logits/chosen": -2.258915901184082, "logits/rejected": -2.112346887588501, "stability/response_length_mean": 727.0, "stability/response_length_std": 335.0764465332031, "stability/response_length_var": 112276.21875, "stability/token_entropy_mean": 7.340624809265137, "stability/repetition_rate_mean": 0.9561265707015991}
+{"current_steps": 100, "total_steps": 196, "loss": 0.7067, "accuracy": 0.6441406607627869, "lr": 9.9e-07, "epoch": 0.5118362124120281, "percentage": 51.02, "elapsed_time": "0:36:59", "remaining_time": "0:35:30", "rewards/chosen": 0.046771928668022156, "rewards/rejected": 0.023559408262372017, "rewards/accuracies": 0.6441406607627869, "rewards/margins": 0.023212522268295288, "logps/chosen": -384.4900817871094, "logps/rejected": -210.60585021972656, "logits/chosen": -2.2744827270507812, "logits/rejected": -2.119138479232788, "stability/response_length_mean": 419.0, "stability/response_length_std": 232.0225067138672, "stability/response_length_var": 53834.4453125, "stability/token_entropy_mean": 7.068749904632568, "stability/repetition_rate_mean": 0.9514176249504089}
+{"current_steps": 105, "total_steps": 196, "loss": 0.7059, "accuracy": 0.6378905773162842, "lr": 9.999815554351224e-07, "epoch": 0.5374280230326296, "percentage": 53.57, "elapsed_time": "0:38:50", "remaining_time": "0:33:40", "rewards/chosen": 0.05437100678682327, "rewards/rejected": 0.02857998013496399, "rewards/accuracies": 0.6378905773162842, "rewards/margins": 0.025791026651859283, "logps/chosen": -377.98284912109375, "logps/rejected": -209.66653442382812, "logits/chosen": -2.2587363719940186, "logits/rejected": -2.1084041595458984, "stability/response_length_mean": 669.5, "stability/response_length_std": 271.60888671875, "stability/response_length_var": 73771.390625, "stability/token_entropy_mean": 6.55859375, "stability/repetition_rate_mean": 0.9007232785224915}
+{"current_steps": 110, "total_steps": 196, "loss": 0.7043, "accuracy": 0.6539062261581421, "lr": 9.999066267225444e-07, "epoch": 0.5630198336532309, "percentage": 56.12, "elapsed_time": "0:40:40", "remaining_time": "0:31:48", "rewards/chosen": 0.060924846678972244, "rewards/rejected": 0.0319257527589798, "rewards/accuracies": 0.6539062261581421, "rewards/margins": 0.028999093919992447, "logps/chosen": -385.252197265625, "logps/rejected": -206.70867919921875, "logits/chosen": -2.275355339050293, "logits/rejected": -2.117722988128662, "stability/response_length_mean": 255.60000610351562, "stability/response_length_std": 114.30874633789062, "stability/response_length_var": 13066.4892578125, "stability/token_entropy_mean": 6.971875190734863, "stability/repetition_rate_mean": 0.9306775331497192}
+{"current_steps": 115, "total_steps": 196, "loss": 0.7016, "accuracy": 0.651171863079071, "lr": 9.997740697079592e-07, "epoch": 0.5886116442738324, "percentage": 58.67, "elapsed_time": "0:42:32", "remaining_time": "0:29:57", "rewards/chosen": 0.07327413558959961, "rewards/rejected": 0.037799760699272156, "rewards/accuracies": 0.651171863079071, "rewards/margins": 0.03547437861561775, "logps/chosen": -372.7075500488281, "logps/rejected": -213.91139221191406, "logits/chosen": -2.251579523086548, "logits/rejected": -2.1110000610351562, "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": 0.8305716514587402}
+{"current_steps": 120, "total_steps": 196, "loss": 0.6987, "accuracy": 0.662109375, "lr": 9.995838996722914e-07, "epoch": 0.6142034548944337, "percentage": 61.22, "elapsed_time": "0:44:22", "remaining_time": "0:28:06", "rewards/chosen": 0.08318294584751129, "rewards/rejected": 0.0416630320250988, "rewards/accuracies": 0.662109375, "rewards/margins": 0.04151991754770279, "logps/chosen": -367.7718200683594, "logps/rejected": -207.56582641601562, "logits/chosen": -2.248891592025757, "logits/rejected": -2.1032941341400146, "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": 0.6951, "accuracy": 0.672656238079071, "lr": 9.993361385379874e-07, "epoch": 0.6397952655150352, "percentage": 63.78, "elapsed_time": "0:46:12", "remaining_time": "0:26:14", "rewards/chosen": 0.09500810503959656, "rewards/rejected": 0.04492490366101265, "rewards/accuracies": 0.672656238079071, "rewards/margins": 0.05008320510387421, "logps/chosen": -385.6851806640625, "logps/rejected": -205.0888671875, "logits/chosen": -2.2534642219543457, "logits/rejected": -2.0952932834625244, "stability/response_length_mean": 433.79998779296875, "stability/response_length_std": 126.92412567138672, "stability/response_length_var": 16109.7333984375, "stability/token_entropy_mean": 6.753125190734863, "stability/repetition_rate_mean": 0.9095968008041382}
+{"current_steps": 130, "total_steps": 196, "loss": 0.692, "accuracy": 0.673828125, "lr": 9.99030814866488e-07, "epoch": 0.6653870761356366, "percentage": 66.33, "elapsed_time": "0:48:02", "remaining_time": "0:24:23", "rewards/chosen": 0.10106992721557617, "rewards/rejected": 0.044545214623212814, "rewards/accuracies": 0.673828125, "rewards/margins": 0.056524716317653656, "logps/chosen": -375.5842590332031, "logps/rejected": -203.7355499267578, "logits/chosen": -2.2454466819763184, "logits/rejected": -2.0966694355010986, "stability/response_length_mean": 296.1000061035156, "stability/response_length_std": 167.77593994140625, "stability/response_length_var": 28148.767578125, "stability/token_entropy_mean": 6.074999809265137, "stability/repetition_rate_mean": 0.8208531141281128}
+{"current_steps": 135, "total_steps": 196, "loss": 0.6897, "accuracy": 0.6683593988418579, "lr": 9.986679638549358e-07, "epoch": 0.690978886756238, "percentage": 68.88, "elapsed_time": "0:49:54", "remaining_time": "0:22:32", "rewards/chosen": 0.11433131992816925, "rewards/rejected": 0.05198357626795769, "rewards/accuracies": 0.6683593988418579, "rewards/margins": 0.062347739934921265, "logps/chosen": -387.3439636230469, "logps/rejected": -213.7014923095703, "logits/chosen": -2.2671430110931396, "logits/rejected": -2.107732057571411, "stability/response_length_mean": 596.4000244140625, "stability/response_length_std": 291.9129333496094, "stability/response_length_var": 85213.15625, "stability/token_entropy_mean": 6.5234375, "stability/repetition_rate_mean": 0.8598278760910034}
+{"current_steps": 140, "total_steps": 196, "loss": 0.686, "accuracy": 0.689453125, "lr": 9.982476273321175e-07, "epoch": 0.7165706973768394, "percentage": 71.43, "elapsed_time": "0:51:43", "remaining_time": "0:20:41", "rewards/chosen": 0.1219753548502922, "rewards/rejected": 0.05057128518819809, "rewards/accuracies": 0.689453125, "rewards/margins": 0.07140406221151352, "logps/chosen": -389.0058898925781, "logps/rejected": -211.13906860351562, "logits/chosen": -2.255606174468994, "logits/rejected": -2.100269079208374, "stability/response_length_mean": 173.89999389648438, "stability/response_length_std": 70.74908447265625, "stability/response_length_var": 5005.43310546875, "stability/token_entropy_mean": 6.634375095367432, "stability/repetition_rate_mean": 0.8619750142097473}
+{"current_steps": 145, "total_steps": 196, "loss": 0.6834, "accuracy": 0.68359375, "lr": 9.977698537536417e-07, "epoch": 0.7421625079974408, "percentage": 73.98, "elapsed_time": "0:53:35", "remaining_time": "0:18:50", "rewards/chosen": 0.13592004776000977, "rewards/rejected": 0.0582398846745491, "rewards/accuracies": 0.68359375, "rewards/margins": 0.07768016308546066, "logps/chosen": -387.7537841796875, "logps/rejected": -203.59815979003906, "logits/chosen": -2.267023801803589, "logits/rejected": -2.0997886657714844, "stability/response_length_mean": 720.2000122070312, "stability/response_length_std": 284.2842102050781, "stability/response_length_var": 80817.5078125, "stability/token_entropy_mean": 7.084374904632568, "stability/repetition_rate_mean": 0.960322380065918}
+{"current_steps": 150, "total_steps": 196, "loss": 0.6816, "accuracy": 0.6605468392372131, "lr": 9.972346981963546e-07, "epoch": 0.7677543186180422, "percentage": 76.53, "elapsed_time": "0:55:27", "remaining_time": "0:17:00", "rewards/chosen": 0.14637623727321625, "rewards/rejected": 0.06041226163506508, "rewards/accuracies": 0.6605468392372131, "rewards/margins": 0.08596397191286087, "logps/chosen": -373.80157470703125, "logps/rejected": -211.20767211914062, "logits/chosen": -2.234330415725708, "logits/rejected": -2.092057704925537, "stability/response_length_mean": 654.2000122070312, "stability/response_length_std": 372.5482177734375, "stability/response_length_var": 138792.171875, "stability/token_entropy_mean": 6.556250095367432, "stability/repetition_rate_mean": 0.8549944162368774}
+{"current_steps": 155, "total_steps": 196, "loss": 0.677, "accuracy": 0.680468738079071, "lr": 9.966422223519885e-07, "epoch": 0.7933461292386437, "percentage": 79.08, "elapsed_time": "0:57:18", "remaining_time": "0:15:09", "rewards/chosen": 0.16198468208312988, "rewards/rejected": 0.06802822649478912, "rewards/accuracies": 0.680468738079071, "rewards/margins": 0.09395644068717957, "logps/chosen": -377.3446350097656, "logps/rejected": -206.38644409179688, "logits/chosen": -2.260176658630371, "logits/rejected": -2.1092514991760254, "stability/response_length_mean": 502.3999938964844, "stability/response_length_std": 200.0795440673828, "stability/response_length_var": 40031.8203125, "stability/token_entropy_mean": 6.557812690734863, "stability/repetition_rate_mean": 0.8795069456100464}
+{"current_steps": 160, "total_steps": 196, "loss": 0.6736, "accuracy": 0.6894532442092896, "lr": 9.959924945200523e-07, "epoch": 0.818937939859245, "percentage": 81.63, "elapsed_time": "0:59:09", "remaining_time": "0:13:18", "rewards/chosen": 0.1707187443971634, "rewards/rejected": 0.06563273072242737, "rewards/accuracies": 0.6894532442092896, "rewards/margins": 0.10508601367473602, "logps/chosen": -383.45159912109375, "logps/rejected": -207.2310333251953, "logits/chosen": -2.253509759902954, "logits/rejected": -2.086948871612549, "stability/response_length_mean": 870.4000244140625, "stability/response_length_std": 147.6581573486328, "stability/response_length_var": 21802.93359375, "stability/token_entropy_mean": 7.228125095367432, "stability/repetition_rate_mean": 0.969073474407196}
+{"current_steps": 165, "total_steps": 196, "loss": 0.6721, "accuracy": 0.6953125, "lr": 9.952855895999568e-07, "epoch": 0.8445297504798465, "percentage": 84.18, "elapsed_time": "1:01:00", "remaining_time": "0:11:27", "rewards/chosen": 0.17040367424488068, "rewards/rejected": 0.06270146369934082, "rewards/accuracies": 0.6953125, "rewards/margins": 0.10770220309495926, "logps/chosen": -362.37127685546875, "logps/rejected": -203.5992431640625, "logits/chosen": -2.238173484802246, "logits/rejected": -2.0906169414520264, "stability/response_length_mean": 547.7000122070312, "stability/response_length_std": 249.60325622558594, "stability/response_length_var": 62301.7890625, "stability/token_entropy_mean": 6.631249904632568, "stability/repetition_rate_mean": 0.8834701776504517}
+{"current_steps": 170, "total_steps": 196, "loss": 0.6663, "accuracy": 0.706250011920929, "lr": 9.9452158908238e-07, "epoch": 0.8701215611004478, "percentage": 86.73, "elapsed_time": "1:02:51", "remaining_time": "0:09:36", "rewards/chosen": 0.18648947775363922, "rewards/rejected": 0.06378380209207535, "rewards/accuracies": 0.706250011920929, "rewards/margins": 0.12270568311214447, "logps/chosen": -376.9095153808594, "logps/rejected": -203.75167846679688, "logits/chosen": -2.2633023262023926, "logits/rejected": -2.0972681045532227, "stability/response_length_mean": 385.70001220703125, "stability/response_length_std": 143.96685791015625, "stability/response_length_var": 20726.455078125, "stability/token_entropy_mean": 6.831250190734863, "stability/repetition_rate_mean": 0.9116168022155762}
+{"current_steps": 175, "total_steps": 196, "loss": 0.6642, "accuracy": 0.6957030892372131, "lr": 9.937005810398745e-07, "epoch": 0.8957133717210493, "percentage": 89.29, "elapsed_time": "1:04:41", "remaining_time": "0:07:45", "rewards/chosen": 0.19610357284545898, "rewards/rejected": 0.0639820322394371, "rewards/accuracies": 0.6957030892372131, "rewards/margins": 0.13212153315544128, "logps/chosen": -380.42779541015625, "logps/rejected": -212.22317504882812, "logits/chosen": -2.249255895614624, "logits/rejected": -2.093829870223999, "stability/response_length_mean": 460.6000061035156, "stability/response_length_std": 165.9894256591797, "stability/response_length_var": 27552.48828125, "stability/token_entropy_mean": 6.956250190734863, "stability/repetition_rate_mean": 0.9416834712028503}
+{"current_steps": 180, "total_steps": 196, "loss": 0.6621, "accuracy": 0.6867187023162842, "lr": 9.928226601167138e-07, "epoch": 0.9213051823416507, "percentage": 91.84, "elapsed_time": "1:06:33", "remaining_time": "0:05:54", "rewards/chosen": 0.21073000133037567, "rewards/rejected": 0.07472933828830719, "rewards/accuracies": 0.6867187023162842, "rewards/margins": 0.13600066304206848, "logps/chosen": -386.9737854003906, "logps/rejected": -212.9161376953125, "logits/chosen": -2.248629093170166, "logits/rejected": -2.0873544216156006, "stability/response_length_mean": 716.2000122070312, "stability/response_length_std": 329.3815002441406, "stability/response_length_var": 108492.1796875, "stability/token_entropy_mean": 6.5234375, "stability/repetition_rate_mean": 0.8946305513381958}
+{"current_steps": 185, "total_steps": 196, "loss": 0.6621, "accuracy": 0.68359375, "lr": 9.918879275179817e-07, "epoch": 0.946896992962252, "percentage": 94.39, "elapsed_time": "1:08:24", "remaining_time": "0:04:04", "rewards/chosen": 0.216688334941864, "rewards/rejected": 0.07646691799163818, "rewards/accuracies": 0.68359375, "rewards/margins": 0.14022144675254822, "logps/chosen": -374.459716796875, "logps/rejected": -210.3748016357422, "logits/chosen": -2.250131130218506, "logits/rejected": -2.099020481109619, "stability/response_length_mean": 486.3999938964844, "stability/response_length_std": 213.8624725341797, "stability/response_length_var": 45737.15625, "stability/token_entropy_mean": 6.5625, "stability/repetition_rate_mean": 0.8750947713851929}
+{"current_steps": 190, "total_steps": 196, "loss": 0.6539, "accuracy": 0.7039062976837158, "lr": 9.90896490997906e-07, "epoch": 0.9724888035828535, "percentage": 96.94, "elapsed_time": "1:10:15", "remaining_time": "0:02:13", "rewards/chosen": 0.22378642857074738, "rewards/rejected": 0.062481436878442764, "rewards/accuracies": 0.7039062976837158, "rewards/margins": 0.1613050103187561, "logps/chosen": -377.5950622558594, "logps/rejected": -206.24668884277344, "logits/chosen": -2.234440326690674, "logits/rejected": -2.0786972045898438, "stability/response_length_mean": 736.7999877929688, "stability/response_length_std": 246.57106018066406, "stability/response_length_var": 60797.2890625, "stability/token_entropy_mean": 6.714062690734863, "stability/repetition_rate_mean": 0.8891245722770691}
+{"current_steps": 195, "total_steps": 196, "loss": 0.6545, "accuracy": 0.6992188096046448, "lr": 9.898484648474362e-07, "epoch": 0.9980806142034548, "percentage": 99.49, "elapsed_time": "1:12:04", "remaining_time": "0:00:22", "rewards/chosen": 0.22635439038276672, "rewards/rejected": 0.06444933265447617, "rewards/accuracies": 0.6992188096046448, "rewards/margins": 0.16190506517887115, "logps/chosen": -361.7326965332031, "logps/rejected": -208.36793518066406, "logits/chosen": -2.2508366107940674, "logits/rejected": -2.0976295471191406, "stability/response_length_mean": 356.3999938964844, "stability/response_length_std": 207.94882202148438, "stability/response_length_var": 43242.7109375, "stability/token_entropy_mean": 5.896874904632568, "stability/repetition_rate_mean": 0.7645308375358582}
+{"current_steps": 196, "total_steps": 196, "epoch": 1.0, "percentage": 100.0, "elapsed_time": "1:12:19", "remaining_time": "0:00:00"}

mistral7b_dpo_sft_ftx0.01/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": 1.7010759115219116,
+      "learning_rate": 4e-08,
+      "logits/chosen": -2.272961139678955,
+      "logits/rejected": -2.1232829093933105,
+      "logps/chosen": -375.06817626953125,
+      "logps/rejected": -210.86959838867188,
+      "loss": 0.7178,
+      "metrics/advantage_var": 0.04110218212008476,
+      "rewards/accuracies": 0.30351561307907104,
+      "rewards/chosen": 0.0007280350546352565,
+      "rewards/margins": 0.0002506318560335785,
+      "rewards/rejected": 0.00047740322770550847,
+      "step": 5
+    },
+    {
+      "epoch": 0.05118362124120281,
+      "grad_norm": 1.7759603261947632,
+      "learning_rate": 9e-08,
+      "logits/chosen": -2.2679052352905273,
+      "logits/rejected": -2.1166000366210938,
+      "logps/chosen": -382.0252380371094,
+      "logps/rejected": -202.4450225830078,
+      "loss": 0.7175,
+      "metrics/advantage_var": 0.07497945427894592,
+      "rewards/accuracies": 0.4945312440395355,
+      "rewards/chosen": 0.0017963653663173318,
+      "rewards/margins": 0.0005524610751308501,
+      "rewards/rejected": 0.0012439042329788208,
+      "stability/repetition_rate_mean": 0.9197490811347961,
+      "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": 1.8210687637329102,
+      "learning_rate": 1.4e-07,
+      "logits/chosen": -2.2607593536376953,
+      "logits/rejected": -2.0992488861083984,
+      "logps/chosen": -389.2146911621094,
+      "logps/rejected": -207.05316162109375,
+      "loss": 0.7179,
+      "metrics/advantage_var": 0.07240394502878189,
+      "rewards/accuracies": 0.49882811307907104,
+      "rewards/chosen": 0.0010917808394879103,
+      "rewards/margins": 4.640953557100147e-05,
+      "rewards/rejected": 0.0010453711729496717,
+      "stability/repetition_rate_mean": 0.8429937362670898,
+      "stability/response_length_mean": 345.70001220703125,
+      "stability/response_length_std": 212.67974853515625,
+      "stability/response_length_var": 45232.6796875,
+      "stability/token_entropy_mean": 6.348437309265137,
+      "step": 15
+    },
+    {
+      "epoch": 0.10236724248240563,
+      "grad_norm": 1.7627458572387695,
+      "learning_rate": 1.8999999999999998e-07,
+      "logits/chosen": -2.2612171173095703,
+      "logits/rejected": -2.114251136779785,
+      "logps/chosen": -368.7218933105469,
+      "logps/rejected": -198.56837463378906,
+      "loss": 0.7184,
+      "metrics/advantage_var": 0.06786168366670609,
+      "rewards/accuracies": 0.4976562559604645,
+      "rewards/chosen": 0.00022399029694497585,
+      "rewards/margins": -0.0004166103317402303,
+      "rewards/rejected": 0.0006406005704775453,
+      "stability/repetition_rate_mean": 0.9278882741928101,
+      "stability/response_length_mean": 739.4000244140625,
+      "stability/response_length_std": 281.282470703125,
+      "stability/response_length_var": 79119.8203125,
+      "stability/token_entropy_mean": 6.946875095367432,
+      "step": 20
+    },
+    {
+      "epoch": 0.12795905310300704,
+      "grad_norm": 1.7385414838790894,
+      "learning_rate": 2.4e-07,
+      "logits/chosen": -2.2770488262176514,
+      "logits/rejected": -2.124872922897339,
+      "logps/chosen": -378.17303466796875,
+      "logps/rejected": -205.866943359375,
+      "loss": 0.7177,
+      "metrics/advantage_var": 0.06800527125597,
+      "rewards/accuracies": 0.490234375,
+      "rewards/chosen": 0.0011880360543727875,
+      "rewards/margins": 9.972632688004524e-05,
+      "rewards/rejected": 0.0010883098002523184,
+      "stability/repetition_rate_mean": 0.8768361806869507,
+      "stability/response_length_mean": 658.5999755859375,
+      "stability/response_length_std": 288.0155029296875,
+      "stability/response_length_var": 82952.9296875,
+      "stability/token_entropy_mean": 6.731249809265137,
+      "step": 25
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 1.8636959791183472,
+      "learning_rate": 2.9e-07,
+      "logits/chosen": -2.270461082458496,
+      "logits/rejected": -2.123213768005371,
+      "logps/chosen": -373.9102478027344,
+      "logps/rejected": -208.93922424316406,
+      "loss": 0.7175,
+      "metrics/advantage_var": 0.07215224206447601,
+      "rewards/accuracies": 0.5085937976837158,
+      "rewards/chosen": 0.0020091882906854153,
+      "rewards/margins": 0.0013196640647947788,
+      "rewards/rejected": 0.0006895243423059583,
+      "stability/repetition_rate_mean": 0.8138343095779419,
+      "stability/response_length_mean": 356.0,
+      "stability/response_length_std": 249.6766815185547,
+      "stability/response_length_var": 62338.4453125,
+      "stability/token_entropy_mean": 6.196875095367432,
+      "step": 30
+    },
+    {
+      "epoch": 0.17914267434420986,
+      "grad_norm": 1.7182040214538574,
+      "learning_rate": 3.4000000000000003e-07,
+      "logits/chosen": -2.261341094970703,
+      "logits/rejected": -2.120121955871582,
+      "logps/chosen": -366.32391357421875,
+      "logps/rejected": -203.99069213867188,
+      "loss": 0.7179,
+      "metrics/advantage_var": 0.07069019973278046,
+      "rewards/accuracies": 0.518359363079071,
+      "rewards/chosen": 0.0021961305756121874,
+      "rewards/margins": 0.0004749999789055437,
+      "rewards/rejected": 0.0017211305676028132,
+      "stability/repetition_rate_mean": 0.9360000491142273,
+      "stability/response_length_mean": 606.2000122070312,
+      "stability/response_length_std": 394.1366882324219,
+      "stability/response_length_var": 155343.734375,
+      "stability/token_entropy_mean": 7.150000095367432,
+      "step": 35
+    },
+    {
+      "epoch": 0.20473448496481125,
+      "grad_norm": 1.8097240924835205,
+      "learning_rate": 3.8999999999999997e-07,
+      "logits/chosen": -2.276620864868164,
+      "logits/rejected": -2.1228880882263184,
+      "logps/chosen": -369.1826477050781,
+      "logps/rejected": -197.37864685058594,
+      "loss": 0.7176,
+      "metrics/advantage_var": 0.07206886261701584,
+      "rewards/accuracies": 0.5093749761581421,
+      "rewards/chosen": 0.0036926856264472008,
+      "rewards/margins": 0.0009252781164832413,
+      "rewards/rejected": 0.0027674080338329077,
+      "stability/repetition_rate_mean": 0.8702197074890137,
+      "stability/response_length_mean": 547.2000122070312,
+      "stability/response_length_std": 290.8897399902344,
+      "stability/response_length_var": 84616.84375,
+      "stability/token_entropy_mean": 6.526562690734863,
+      "step": 40
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 1.8226284980773926,
+      "learning_rate": 4.3999999999999997e-07,
+      "logits/chosen": -2.2571909427642822,
+      "logits/rejected": -2.1073906421661377,
+      "logps/chosen": -385.17181396484375,
+      "logps/rejected": -212.56410217285156,
+      "loss": 0.718,
+      "metrics/advantage_var": 0.07308375090360641,
+      "rewards/accuracies": 0.513671875,
+      "rewards/chosen": 0.0047181700356304646,
+      "rewards/margins": 0.00034062692429870367,
+      "rewards/rejected": 0.004377542994916439,
+      "stability/repetition_rate_mean": 0.8764154314994812,
+      "stability/response_length_mean": 366.5,
+      "stability/response_length_std": 118.86243438720703,
+      "stability/response_length_var": 14128.27734375,
+      "stability/token_entropy_mean": 6.640625,
+      "step": 45
+    },
+    {
+      "epoch": 0.2559181062060141,
+      "grad_norm": 1.8556437492370605,
+      "learning_rate": 4.9e-07,
+      "logits/chosen": -2.2644577026367188,
+      "logits/rejected": -2.109606981277466,
+      "logps/chosen": -378.29864501953125,
+      "logps/rejected": -207.4927978515625,
+      "loss": 0.7173,
+      "metrics/advantage_var": 0.07552963495254517,
+      "rewards/accuracies": 0.520703136920929,
+      "rewards/chosen": 0.005913693457841873,
+      "rewards/margins": 0.0021115851122885942,
+      "rewards/rejected": 0.0038021078798919916,
+      "stability/repetition_rate_mean": 0.8770252466201782,
+      "stability/response_length_mean": 384.1000061035156,
+      "stability/response_length_std": 178.4809112548828,
+      "stability/response_length_var": 31855.43359375,
+      "stability/token_entropy_mean": 6.543749809265137,
+      "step": 50
+    },
+    {
+      "epoch": 0.28150991682661547,
+      "grad_norm": 1.9488743543624878,
+      "learning_rate": 5.4e-07,
+      "logits/chosen": -2.25301456451416,
+      "logits/rejected": -2.094449758529663,
+      "logps/chosen": -393.7789306640625,
+      "logps/rejected": -210.35589599609375,
+      "loss": 0.7154,
+      "metrics/advantage_var": 0.07901743054389954,
+      "rewards/accuracies": 0.5429688096046448,
+      "rewards/chosen": 0.010107376612722874,
+      "rewards/margins": 0.004774035885930061,
+      "rewards/rejected": 0.005333342123776674,
+      "stability/repetition_rate_mean": 0.9362810850143433,
+      "stability/response_length_mean": 332.8999938964844,
+      "stability/response_length_std": 137.39193725585938,
+      "stability/response_length_var": 18876.544921875,
+      "stability/token_entropy_mean": 6.903124809265137,
+      "step": 55
+    },
+    {
+      "epoch": 0.30710172744721687,
+      "grad_norm": 1.8305964469909668,
+      "learning_rate": 5.9e-07,
+      "logits/chosen": -2.2730324268341064,
+      "logits/rejected": -2.111363410949707,
+      "logps/chosen": -387.6959228515625,
+      "logps/rejected": -206.80484008789062,
+      "loss": 0.7161,
+      "metrics/advantage_var": 0.08022293448448181,
+      "rewards/accuracies": 0.5257812738418579,
+      "rewards/chosen": 0.01106810662895441,
+      "rewards/margins": 0.003843404119834304,
+      "rewards/rejected": 0.007224703207612038,
+      "stability/repetition_rate_mean": 0.8751958608627319,
+      "stability/response_length_mean": 655.2999877929688,
+      "stability/response_length_std": 282.005126953125,
+      "stability/response_length_var": 79526.8984375,
+      "stability/token_entropy_mean": 6.609375,
+      "step": 60
+    },
+    {
+      "epoch": 0.3326935380678183,
+      "grad_norm": 1.6957237720489502,
+      "learning_rate": 6.4e-07,
+      "logits/chosen": -2.251922369003296,
+      "logits/rejected": -2.0987448692321777,
+      "logps/chosen": -377.4503173828125,
+      "logps/rejected": -204.38616943359375,
+      "loss": 0.715,
+      "metrics/advantage_var": 0.08649040758609772,
+      "rewards/accuracies": 0.5859375,
+      "rewards/chosen": 0.014928197488188744,
+      "rewards/margins": 0.007287245709449053,
+      "rewards/rejected": 0.007640953175723553,
+      "stability/repetition_rate_mean": 0.9006223678588867,
+      "stability/response_length_mean": 306.3999938964844,
+      "stability/response_length_std": 141.7236785888672,
+      "stability/response_length_var": 20085.599609375,
+      "stability/token_entropy_mean": 6.818749904632568,
+      "step": 65
+    },
+    {
+      "epoch": 0.3582853486884197,
+      "grad_norm": 1.9992988109588623,
+      "learning_rate": 6.9e-07,
+      "logits/chosen": -2.2710888385772705,
+      "logits/rejected": -2.106233835220337,
+      "logps/chosen": -393.7522277832031,
+      "logps/rejected": -206.18975830078125,
+      "loss": 0.7139,
+      "metrics/advantage_var": 0.09332478046417236,
+      "rewards/accuracies": 0.5785155892372131,
+      "rewards/chosen": 0.015660293400287628,
+      "rewards/margins": 0.007665838580578566,
+      "rewards/rejected": 0.007994452491402626,
+      "stability/repetition_rate_mean": 0.9697188138961792,
+      "stability/response_length_mean": 472.70001220703125,
+      "stability/response_length_std": 159.89791870117188,
+      "stability/response_length_var": 25567.34375,
+      "stability/token_entropy_mean": 7.199999809265137,
+      "step": 70
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 1.8864988088607788,
+      "learning_rate": 7.4e-07,
+      "logits/chosen": -2.2754428386688232,
+      "logits/rejected": -2.1162445545196533,
+      "logps/chosen": -381.0527648925781,
+      "logps/rejected": -207.933837890625,
+      "loss": 0.7136,
+      "metrics/advantage_var": 0.10101056843996048,
+      "rewards/accuracies": 0.5796874761581421,
+      "rewards/chosen": 0.019434239715337753,
+      "rewards/margins": 0.009132225066423416,
+      "rewards/rejected": 0.010302015580236912,
+      "stability/repetition_rate_mean": 0.9078971147537231,
+      "stability/response_length_mean": 527.0,
+      "stability/response_length_std": 293.7500915527344,
+      "stability/response_length_var": 86289.109375,
+      "stability/token_entropy_mean": 6.140625,
+      "step": 75
+    },
+    {
+      "epoch": 0.4094689699296225,
+      "grad_norm": 1.8260862827301025,
+      "learning_rate": 7.9e-07,
+      "logits/chosen": -2.260484218597412,
+      "logits/rejected": -2.10276198387146,
+      "logps/chosen": -382.4151611328125,
+      "logps/rejected": -208.9755096435547,
+      "loss": 0.713,
+      "metrics/advantage_var": 0.1154346615076065,
+      "rewards/accuracies": 0.583984375,
+      "rewards/chosen": 0.02179381437599659,
+      "rewards/margins": 0.009805195033550262,
+      "rewards/rejected": 0.011988619342446327,
+      "stability/repetition_rate_mean": 0.9465335011482239,
+      "stability/response_length_mean": 635.5,
+      "stability/response_length_std": 262.53900146484375,
+      "stability/response_length_var": 68926.71875,
+      "stability/token_entropy_mean": 7.199999809265137,
+      "step": 80
+    },
+    {
+      "epoch": 0.4350607805502239,
+      "grad_norm": 1.871209740638733,
+      "learning_rate": 8.399999999999999e-07,
+      "logits/chosen": -2.257699966430664,
+      "logits/rejected": -2.1006553173065186,
+      "logps/chosen": -387.3171081542969,
+      "logps/rejected": -208.62698364257812,
+      "loss": 0.7122,
+      "metrics/advantage_var": 0.14266015589237213,
+      "rewards/accuracies": 0.5902344584465027,
+      "rewards/chosen": 0.02743753232061863,
+      "rewards/margins": 0.012249825522303581,
+      "rewards/rejected": 0.015187706798315048,
+      "stability/repetition_rate_mean": 0.9052130579948425,
+      "stability/response_length_mean": 600.0999755859375,
+      "stability/response_length_std": 349.2030334472656,
+      "stability/response_length_var": 121942.765625,
+      "stability/token_entropy_mean": 6.737500190734863,
+      "step": 85
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 1.8958905935287476,
+      "learning_rate": 8.9e-07,
+      "logits/chosen": -2.261382818222046,
+      "logits/rejected": -2.113365888595581,
+      "logps/chosen": -375.1956481933594,
+      "logps/rejected": -206.21043395996094,
+      "loss": 0.7107,
+      "metrics/advantage_var": 0.16143371164798737,
+      "rewards/accuracies": 0.6230468153953552,
+      "rewards/chosen": 0.033041536808013916,
+      "rewards/margins": 0.015009062364697456,
+      "rewards/rejected": 0.01803247444331646,
+      "stability/repetition_rate_mean": 0.9765706062316895,
+      "stability/response_length_mean": 692.9000244140625,
+      "stability/response_length_std": 127.31452941894531,
+      "stability/response_length_var": 16208.9892578125,
+      "stability/token_entropy_mean": 7.106249809265137,
+      "step": 90
+    },
+    {
+      "epoch": 0.48624440179142675,
+      "grad_norm": 1.8168452978134155,
+      "learning_rate": 9.399999999999999e-07,
+      "logits/chosen": -2.258915901184082,
+      "logits/rejected": -2.112346887588501,
+      "logps/chosen": -366.824462890625,
+      "logps/rejected": -203.4202423095703,
+      "loss": 0.7085,
+      "metrics/advantage_var": 0.20107638835906982,
+      "rewards/accuracies": 0.6175781488418579,
+      "rewards/chosen": 0.038514357060194016,
+      "rewards/margins": 0.019381558522582054,
+      "rewards/rejected": 0.01913280040025711,
+      "stability/repetition_rate_mean": 0.9561265707015991,
+      "stability/response_length_mean": 727.0,
+      "stability/response_length_std": 335.0764465332031,
+      "stability/response_length_var": 112276.21875,
+      "stability/token_entropy_mean": 7.340624809265137,
+      "step": 95
+    },
+    {
+      "epoch": 0.5118362124120281,
+      "grad_norm": 1.8184173107147217,
+      "learning_rate": 9.9e-07,
+      "logits/chosen": -2.2744827270507812,
+      "logits/rejected": -2.119138479232788,
+      "logps/chosen": -384.4900817871094,
+      "logps/rejected": -210.60585021972656,
+      "loss": 0.7067,
+      "metrics/advantage_var": 0.2732086777687073,
+      "rewards/accuracies": 0.6441406607627869,
+      "rewards/chosen": 0.046771928668022156,
+      "rewards/margins": 0.023212522268295288,
+      "rewards/rejected": 0.023559408262372017,
+      "stability/repetition_rate_mean": 0.9514176249504089,
+      "stability/response_length_mean": 419.0,
+      "stability/response_length_std": 232.0225067138672,
+      "stability/response_length_var": 53834.4453125,
+      "stability/token_entropy_mean": 7.068749904632568,
+      "step": 100
+    },
+    {
+      "epoch": 0.5374280230326296,
+      "grad_norm": 1.741348385810852,
+      "learning_rate": 9.999815554351224e-07,
+      "logits/chosen": -2.2587363719940186,
+      "logits/rejected": -2.1084041595458984,
+      "logps/chosen": -377.98284912109375,
+      "logps/rejected": -209.66653442382812,
+      "loss": 0.7059,
+      "metrics/advantage_var": 0.34950339794158936,
+      "rewards/accuracies": 0.6378905773162842,
+      "rewards/chosen": 0.05437100678682327,
+      "rewards/margins": 0.025791026651859283,
+      "rewards/rejected": 0.02857998013496399,
+      "stability/repetition_rate_mean": 0.9007232785224915,
+      "stability/response_length_mean": 669.5,
+      "stability/response_length_std": 271.60888671875,
+      "stability/response_length_var": 73771.390625,
+      "stability/token_entropy_mean": 6.55859375,
+      "step": 105
+    },
+    {
+      "epoch": 0.5630198336532309,
+      "grad_norm": 1.6754577159881592,
+      "learning_rate": 9.999066267225444e-07,
+      "logits/chosen": -2.275355339050293,
+      "logits/rejected": -2.117722988128662,
+      "logps/chosen": -385.252197265625,
+      "logps/rejected": -206.70867919921875,
+      "loss": 0.7043,
+      "metrics/advantage_var": 0.4181045889854431,
+      "rewards/accuracies": 0.6539062261581421,
+      "rewards/chosen": 0.060924846678972244,
+      "rewards/margins": 0.028999093919992447,
+      "rewards/rejected": 0.0319257527589798,
+      "stability/repetition_rate_mean": 0.9306775331497192,
+      "stability/response_length_mean": 255.60000610351562,
+      "stability/response_length_std": 114.30874633789062,
+      "stability/response_length_var": 13066.4892578125,
+      "stability/token_entropy_mean": 6.971875190734863,
+      "step": 110
+    },
+    {
+      "epoch": 0.5886116442738324,
+      "grad_norm": 1.795372724533081,
+      "learning_rate": 9.997740697079592e-07,
+      "logits/chosen": -2.251579523086548,
+      "logits/rejected": -2.1110000610351562,
+      "logps/chosen": -372.7075500488281,
+      "logps/rejected": -213.91139221191406,
+      "loss": 0.7016,
+      "metrics/advantage_var": 0.6075757145881653,
+      "rewards/accuracies": 0.651171863079071,
+      "rewards/chosen": 0.07327413558959961,
+      "rewards/margins": 0.03547437861561775,
+      "rewards/rejected": 0.037799760699272156,
+      "stability/repetition_rate_mean": 0.8305716514587402,
+      "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": 1.7747018337249756,
+      "learning_rate": 9.995838996722914e-07,
+      "logits/chosen": -2.248891592025757,
+      "logits/rejected": -2.1032941341400146,
+      "logps/chosen": -367.7718200683594,
+      "logps/rejected": -207.56582641601562,
+      "loss": 0.6987,
+      "metrics/advantage_var": 0.7378088235855103,
+      "rewards/accuracies": 0.662109375,
+      "rewards/chosen": 0.08318294584751129,
+      "rewards/margins": 0.04151991754770279,
+      "rewards/rejected": 0.0416630320250988,
+      "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": 1.7948890924453735,
+      "learning_rate": 9.993361385379874e-07,
+      "logits/chosen": -2.2534642219543457,
+      "logits/rejected": -2.0952932834625244,
+      "logps/chosen": -385.6851806640625,
+      "logps/rejected": -205.0888671875,
+      "loss": 0.6951,
+      "metrics/advantage_var": 1.040814995765686,
+      "rewards/accuracies": 0.672656238079071,
+      "rewards/chosen": 0.09500810503959656,
+      "rewards/margins": 0.05008320510387421,
+      "rewards/rejected": 0.04492490366101265,
+      "stability/repetition_rate_mean": 0.9095968008041382,
+      "stability/response_length_mean": 433.79998779296875,
+      "stability/response_length_std": 126.92412567138672,
+      "stability/response_length_var": 16109.7333984375,
+      "stability/token_entropy_mean": 6.753125190734863,
+      "step": 125
+    },
+    {
+      "epoch": 0.6653870761356366,
+      "grad_norm": 1.8776071071624756,
+      "learning_rate": 9.99030814866488e-07,
+      "logits/chosen": -2.2454466819763184,
+      "logits/rejected": -2.0966694355010986,
+      "logps/chosen": -375.5842590332031,
+      "logps/rejected": -203.7355499267578,
+      "loss": 0.692,
+      "metrics/advantage_var": 1.1227996349334717,
+      "rewards/accuracies": 0.673828125,
+      "rewards/chosen": 0.10106992721557617,
+      "rewards/margins": 0.056524716317653656,
+      "rewards/rejected": 0.044545214623212814,
+      "stability/repetition_rate_mean": 0.8208531141281128,
+      "stability/response_length_mean": 296.1000061035156,
+      "stability/response_length_std": 167.77593994140625,
+      "stability/response_length_var": 28148.767578125,
+      "stability/token_entropy_mean": 6.074999809265137,
+      "step": 130
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 1.7475597858428955,
+      "learning_rate": 9.986679638549358e-07,
+      "logits/chosen": -2.2671430110931396,
+      "logits/rejected": -2.107732057571411,
+      "logps/chosen": -387.3439636230469,
+      "logps/rejected": -213.7014923095703,
+      "loss": 0.6897,
+      "metrics/advantage_var": 1.4129046201705933,
+      "rewards/accuracies": 0.6683593988418579,
+      "rewards/chosen": 0.11433131992816925,
+      "rewards/margins": 0.062347739934921265,
+      "rewards/rejected": 0.05198357626795769,
+      "stability/repetition_rate_mean": 0.8598278760910034,
+      "stability/response_length_mean": 596.4000244140625,
+      "stability/response_length_std": 291.9129333496094,
+      "stability/response_length_var": 85213.15625,
+      "stability/token_entropy_mean": 6.5234375,
+      "step": 135
+    },
+    {
+      "epoch": 0.7165706973768394,
+      "grad_norm": 1.8308006525039673,
+      "learning_rate": 9.982476273321175e-07,
+      "logits/chosen": -2.255606174468994,
+      "logits/rejected": -2.100269079208374,
+      "logps/chosen": -389.0058898925781,
+      "logps/rejected": -211.13906860351562,
+      "loss": 0.686,
+      "metrics/advantage_var": 1.8579171895980835,
+      "rewards/accuracies": 0.689453125,
+      "rewards/chosen": 0.1219753548502922,
+      "rewards/margins": 0.07140406221151352,
+      "rewards/rejected": 0.05057128518819809,
+      "stability/repetition_rate_mean": 0.8619750142097473,
+      "stability/response_length_mean": 173.89999389648438,
+      "stability/response_length_std": 70.74908447265625,
+      "stability/response_length_var": 5005.43310546875,
+      "stability/token_entropy_mean": 6.634375095367432,
+      "step": 140
+    },
+    {
+      "epoch": 0.7421625079974408,
+      "grad_norm": 1.7569092512130737,
+      "learning_rate": 9.977698537536417e-07,
+      "logits/chosen": -2.267023801803589,
+      "logits/rejected": -2.0997886657714844,
+      "logps/chosen": -387.7537841796875,
+      "logps/rejected": -203.59815979003906,
+      "loss": 0.6834,
+      "metrics/advantage_var": 2.1575894355773926,
+      "rewards/accuracies": 0.68359375,
+      "rewards/chosen": 0.13592004776000977,
+      "rewards/margins": 0.07768016308546066,
+      "rewards/rejected": 0.0582398846745491,
+      "stability/repetition_rate_mean": 0.960322380065918,
+      "stability/response_length_mean": 720.2000122070312,
+      "stability/response_length_std": 284.2842102050781,
+      "stability/response_length_var": 80817.5078125,
+      "stability/token_entropy_mean": 7.084374904632568,
+      "step": 145
+    },
+    {
+      "epoch": 0.7677543186180422,
+      "grad_norm": 1.8092952966690063,
+      "learning_rate": 9.972346981963546e-07,
+      "logits/chosen": -2.234330415725708,
+      "logits/rejected": -2.092057704925537,
+      "logps/chosen": -373.80157470703125,
+      "logps/rejected": -211.20767211914062,
+      "loss": 0.6816,
+      "metrics/advantage_var": 3.304790735244751,
+      "rewards/accuracies": 0.6605468392372131,
+      "rewards/chosen": 0.14637623727321625,
+      "rewards/margins": 0.08596397191286087,
+      "rewards/rejected": 0.06041226163506508,
+      "stability/repetition_rate_mean": 0.8549944162368774,
+      "stability/response_length_mean": 654.2000122070312,
+      "stability/response_length_std": 372.5482177734375,
+      "stability/response_length_var": 138792.171875,
+      "stability/token_entropy_mean": 6.556250095367432,
+      "step": 150
+    },
+    {
+      "epoch": 0.7933461292386437,
+      "grad_norm": 1.8317517042160034,
+      "learning_rate": 9.966422223519885e-07,
+      "logits/chosen": -2.260176658630371,
+      "logits/rejected": -2.1092514991760254,
+      "logps/chosen": -377.3446350097656,
+      "logps/rejected": -206.38644409179688,
+      "loss": 0.677,
+      "metrics/advantage_var": 2.97253680229187,
+      "rewards/accuracies": 0.680468738079071,
+      "rewards/chosen": 0.16198468208312988,
+      "rewards/margins": 0.09395644068717957,
+      "rewards/rejected": 0.06802822649478912,
+      "stability/repetition_rate_mean": 0.8795069456100464,
+      "stability/response_length_mean": 502.3999938964844,
+      "stability/response_length_std": 200.0795440673828,
+      "stability/response_length_var": 40031.8203125,
+      "stability/token_entropy_mean": 6.557812690734863,
+      "step": 155
+    },
+    {
+      "epoch": 0.818937939859245,
+      "grad_norm": 1.7747522592544556,
+      "learning_rate": 9.959924945200523e-07,
+      "logits/chosen": -2.253509759902954,
+      "logits/rejected": -2.086948871612549,
+      "logps/chosen": -383.45159912109375,
+      "logps/rejected": -207.2310333251953,
+      "loss": 0.6736,
+      "metrics/advantage_var": 4.025539398193359,
+      "rewards/accuracies": 0.6894532442092896,
+      "rewards/chosen": 0.1707187443971634,
+      "rewards/margins": 0.10508601367473602,
+      "rewards/rejected": 0.06563273072242737,
+      "stability/repetition_rate_mean": 0.969073474407196,
+      "stability/response_length_mean": 870.4000244140625,
+      "stability/response_length_std": 147.6581573486328,
+      "stability/response_length_var": 21802.93359375,
+      "stability/token_entropy_mean": 7.228125095367432,
+      "step": 160
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.641774296760559,
+      "learning_rate": 9.952855895999568e-07,
+      "logits/chosen": -2.238173484802246,
+      "logits/rejected": -2.0906169414520264,
+      "logps/chosen": -362.37127685546875,
+      "logps/rejected": -203.5992431640625,
+      "loss": 0.6721,
+      "metrics/advantage_var": 3.921074867248535,
+      "rewards/accuracies": 0.6953125,
+      "rewards/chosen": 0.17040367424488068,
+      "rewards/margins": 0.10770220309495926,
+      "rewards/rejected": 0.06270146369934082,
+      "stability/repetition_rate_mean": 0.8834701776504517,
+      "stability/response_length_mean": 547.7000122070312,
+      "stability/response_length_std": 249.60325622558594,
+      "stability/response_length_var": 62301.7890625,
+      "stability/token_entropy_mean": 6.631249904632568,
+      "step": 165
+    },
+    {
+      "epoch": 0.8701215611004478,
+      "grad_norm": 1.7060259580612183,
+      "learning_rate": 9.9452158908238e-07,
+      "logits/chosen": -2.2633023262023926,
+      "logits/rejected": -2.0972681045532227,
+      "logps/chosen": -376.9095153808594,
+      "logps/rejected": -203.75167846679688,
+      "loss": 0.6663,
+      "metrics/advantage_var": 4.847568988800049,
+      "rewards/accuracies": 0.706250011920929,
+      "rewards/chosen": 0.18648947775363922,
+      "rewards/margins": 0.12270568311214447,
+      "rewards/rejected": 0.06378380209207535,
+      "stability/repetition_rate_mean": 0.9116168022155762,
+      "stability/response_length_mean": 385.70001220703125,
+      "stability/response_length_std": 143.96685791015625,
+      "stability/response_length_var": 20726.455078125,
+      "stability/token_entropy_mean": 6.831250190734863,
+      "step": 170
+    },
+    {
+      "epoch": 0.8957133717210493,
+      "grad_norm": 1.7979154586791992,
+      "learning_rate": 9.937005810398745e-07,
+      "logits/chosen": -2.249255895614624,
+      "logits/rejected": -2.093829870223999,
+      "logps/chosen": -380.42779541015625,
+      "logps/rejected": -212.22317504882812,
+      "loss": 0.6642,
+      "metrics/advantage_var": 6.434874534606934,
+      "rewards/accuracies": 0.6957030892372131,
+      "rewards/chosen": 0.19610357284545898,
+      "rewards/margins": 0.13212153315544128,
+      "rewards/rejected": 0.0639820322394371,
+      "stability/repetition_rate_mean": 0.9416834712028503,
+      "stability/response_length_mean": 460.6000061035156,
+      "stability/response_length_std": 165.9894256591797,
+      "stability/response_length_var": 27552.48828125,
+      "stability/token_entropy_mean": 6.956250190734863,
+      "step": 175
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 2.1531741619110107,
+      "learning_rate": 9.928226601167138e-07,
+      "logits/chosen": -2.248629093170166,
+      "logits/rejected": -2.0873544216156006,
+      "logps/chosen": -386.9737854003906,
+      "logps/rejected": -212.9161376953125,
+      "loss": 0.6621,
+      "metrics/advantage_var": 5.855362415313721,
+      "rewards/accuracies": 0.6867187023162842,
+      "rewards/chosen": 0.21073000133037567,
+      "rewards/margins": 0.13600066304206848,
+      "rewards/rejected": 0.07472933828830719,
+      "stability/repetition_rate_mean": 0.8946305513381958,
+      "stability/response_length_mean": 716.2000122070312,
+      "stability/response_length_std": 329.3815002441406,
+      "stability/response_length_var": 108492.1796875,
+      "stability/token_entropy_mean": 6.5234375,
+      "step": 180
+    },
+    {
+      "epoch": 0.946896992962252,
+      "grad_norm": 1.6736973524093628,
+      "learning_rate": 9.918879275179817e-07,
+      "logits/chosen": -2.250131130218506,
+      "logits/rejected": -2.099020481109619,
+      "logps/chosen": -374.459716796875,
+      "logps/rejected": -210.3748016357422,
+      "loss": 0.6621,
+      "metrics/advantage_var": 7.431878566741943,
+      "rewards/accuracies": 0.68359375,
+      "rewards/chosen": 0.216688334941864,
+      "rewards/margins": 0.14022144675254822,
+      "rewards/rejected": 0.07646691799163818,
+      "stability/repetition_rate_mean": 0.8750947713851929,
+      "stability/response_length_mean": 486.3999938964844,
+      "stability/response_length_std": 213.8624725341797,
+      "stability/response_length_var": 45737.15625,
+      "stability/token_entropy_mean": 6.5625,
+      "step": 185
+    },
+    {
+      "epoch": 0.9724888035828535,
+      "grad_norm": 1.5515198707580566,
+      "learning_rate": 9.90896490997906e-07,
+      "logits/chosen": -2.234440326690674,
+      "logits/rejected": -2.0786972045898438,
+      "logps/chosen": -377.5950622558594,
+      "logps/rejected": -206.24668884277344,
+      "loss": 0.6539,
+      "metrics/advantage_var": 8.211545944213867,
+      "rewards/accuracies": 0.7039062976837158,
+      "rewards/chosen": 0.22378642857074738,
+      "rewards/margins": 0.1613050103187561,
+      "rewards/rejected": 0.062481436878442764,
+      "stability/repetition_rate_mean": 0.8891245722770691,
+      "stability/response_length_mean": 736.7999877929688,
+      "stability/response_length_std": 246.57106018066406,
+      "stability/response_length_var": 60797.2890625,
+      "stability/token_entropy_mean": 6.714062690734863,
+      "step": 190
+    },
+    {
+      "epoch": 0.9980806142034548,
+      "grad_norm": 1.6709520816802979,
+      "learning_rate": 9.898484648474362e-07,
+      "logits/chosen": -2.2508366107940674,
+      "logits/rejected": -2.0976295471191406,
+      "logps/chosen": -361.7326965332031,
+      "logps/rejected": -208.36793518066406,
+      "loss": 0.6545,
+      "metrics/advantage_var": 8.507060050964355,
+      "rewards/accuracies": 0.6992188096046448,
+      "rewards/chosen": 0.22635439038276672,
+      "rewards/margins": 0.16190506517887115,
+      "rewards/rejected": 0.06444933265447617,
+      "stability/repetition_rate_mean": 0.7645308375358582,
+      "stability/response_length_mean": 356.3999938964844,
+      "stability/response_length_std": 207.94882202148438,
+      "stability/response_length_var": 43242.7109375,
+      "stability/token_entropy_mean": 5.896874904632568,
+      "step": 195
+    },
+    {
+      "epoch": 1.0,
+      "step": 196,
+      "total_flos": 7.072674331683193e+18,
+      "train_loss": 0.6978096469324462,
+      "train_runtime": 4339.2504,
+      "train_samples_per_second": 23.045,
+      "train_steps_per_second": 0.045
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 196,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 200,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.072674331683193e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

mistral7b_dpo_sft_ftx0.01/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1a19b35383c5d0923ed5719f6321bcbbdad041809e90dbc9f6d7f7a5fde3b88
+size 8017