Upload folder using huggingface_hub

mistral7b_dpo_sft_ftx0.1/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_ftx0.1
+  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_ftx0.1
+
+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_ftx0.1/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": [
+    "k_proj",
+    "o_proj",
+    "v_proj",
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

mistral7b_dpo_sft_ftx0.1/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92b05d7a04b0800599ade807d33ea04002858abdba3d22f8cc03134290b7412c
+size 42002584

mistral7b_dpo_sft_ftx0.1/all_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "total_flos": 7.072674331683193e+18,
+    "train_loss": 0.9197707787460211,
+    "train_runtime": 4344.4262,
+    "train_samples_per_second": 23.018,
+    "train_steps_per_second": 0.045
+}

mistral7b_dpo_sft_ftx0.1/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_ftx0.1/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_ftx0.1/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": [
+    "k_proj",
+    "o_proj",
+    "v_proj",
+    "q_proj",
+    "gate_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

mistral7b_dpo_sft_ftx0.1/checkpoint-196/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92b05d7a04b0800599ade807d33ea04002858abdba3d22f8cc03134290b7412c
+size 42002584

mistral7b_dpo_sft_ftx0.1/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_ftx0.1/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 sha256:1682fcc5c15e22901ac3e2e074ce411f84df4ee11052aa9456465c851fbc3307
+size 31467333

mistral7b_dpo_sft_ftx0.1/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 sha256:386d453529991a18c5f752a355feba53b0906db848c1090b739ff411cc49dce0
+size 31467333

mistral7b_dpo_sft_ftx0.1/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:c86b42e79cf1a013217a517f36692dae058c78875cdbb4af8d274f0b0a167694
+size 31467333

mistral7b_dpo_sft_ftx0.1/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 sha256:6ba690104af617cabeb3fe31945c280d3c66c56114999cc1356282fc877c11b4
+size 31467397

mistral7b_dpo_sft_ftx0.1/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 sha256:2633d6a389f7abd8ec86013055a43de487090f23b07e24653f186657fce60884
+size 31467397

mistral7b_dpo_sft_ftx0.1/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:ef455a251c9be7376d954c448288dd78e95c772bac3fc45ae9a0a679e8d955b9
+size 31467397

mistral7b_dpo_sft_ftx0.1/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:a24aa2d857ed3d3c7e468805571d85bc5ac5c6960acd48c67ffb86153502a3e1
+size 31467397

mistral7b_dpo_sft_ftx0.1/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 sha256:75585606d97f4b5e31c2fdb495481c7016c47ba285a93c6c481c2a65caec4ce8
+size 31467397

mistral7b_dpo_sft_ftx0.1/checkpoint-196/global_step196/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a36afe2f62288b3a72806067e97ae03392c8237360a452670f9344958f4325e6
+size 42288409

mistral7b_dpo_sft_ftx0.1/checkpoint-196/latest

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

mistral7b_dpo_sft_ftx0.1/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_ftx0.1/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_ftx0.1/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_ftx0.1/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_ftx0.1/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_ftx0.1/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_ftx0.1/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_ftx0.1/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_ftx0.1/checkpoint-196/scheduler.pt

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

mistral7b_dpo_sft_ftx0.1/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_ftx0.1/checkpoint-196/tokenizer.model

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

mistral7b_dpo_sft_ftx0.1/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_ftx0.1/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": 1.7173521518707275,
+      "learning_rate": 4e-08,
+      "logits/chosen": -2.2728710174560547,
+      "logits/rejected": -2.123058557510376,
+      "logps/chosen": -375.069580078125,
+      "logps/rejected": -210.8708038330078,
+      "loss": 0.9396,
+      "metrics/advantage_var": 0.040927786380052567,
+      "rewards/accuracies": 0.29960939288139343,
+      "rewards/chosen": 0.0005876531358808279,
+      "rewards/margins": 0.00023021931701805443,
+      "rewards/rejected": 0.00035743374610319734,
+      "step": 5
+    },
+    {
+      "epoch": 0.05118362124120281,
+      "grad_norm": 1.8015280961990356,
+      "learning_rate": 9e-08,
+      "logits/chosen": -2.2679998874664307,
+      "logits/rejected": -2.1166672706604004,
+      "logps/chosen": -382.03143310546875,
+      "logps/rejected": -202.44522094726562,
+      "loss": 0.9379,
+      "metrics/advantage_var": 0.07131563127040863,
+      "rewards/accuracies": 0.4902343451976776,
+      "rewards/chosen": 0.0011766888201236725,
+      "rewards/margins": -4.512531450018287e-05,
+      "rewards/rejected": 0.0012218140764161944,
+      "stability/repetition_rate_mean": 0.9183257222175598,
+      "stability/response_length_mean": 602.7999877929688,
+      "stability/response_length_std": 295.1842346191406,
+      "stability/response_length_var": 87133.734375,
+      "stability/token_entropy_mean": 6.990624904632568,
+      "step": 10
+    },
+    {
+      "epoch": 0.07677543186180422,
+      "grad_norm": 1.8550324440002441,
+      "learning_rate": 1.4e-07,
+      "logits/chosen": -2.260781764984131,
+      "logits/rejected": -2.099287509918213,
+      "logps/chosen": -389.2155456542969,
+      "logps/rejected": -207.05043029785156,
+      "loss": 0.9389,
+      "metrics/advantage_var": 0.07388237863779068,
+      "rewards/accuracies": 0.48750001192092896,
+      "rewards/chosen": 0.0010101307416334748,
+      "rewards/margins": -0.00031050166580826044,
+      "rewards/rejected": 0.0013206324074417353,
+      "stability/repetition_rate_mean": 0.8462336659431458,
+      "stability/response_length_mean": 345.70001220703125,
+      "stability/response_length_std": 212.67974853515625,
+      "stability/response_length_var": 45232.6796875,
+      "stability/token_entropy_mean": 6.323437690734863,
+      "step": 15
+    },
+    {
+      "epoch": 0.10236724248240563,
+      "grad_norm": 1.7850894927978516,
+      "learning_rate": 1.8999999999999998e-07,
+      "logits/chosen": -2.2612733840942383,
+      "logits/rejected": -2.1143851280212402,
+      "logps/chosen": -368.721435546875,
+      "logps/rejected": -198.57066345214844,
+      "loss": 0.9424,
+      "metrics/advantage_var": 0.0697961151599884,
+      "rewards/accuracies": 0.4957031011581421,
+      "rewards/chosen": 0.0002682031481526792,
+      "rewards/margins": -0.00014453462790697813,
+      "rewards/rejected": 0.0004127377178519964,
+      "stability/repetition_rate_mean": 0.9284495115280151,
+      "stability/response_length_mean": 739.4000244140625,
+      "stability/response_length_std": 281.282470703125,
+      "stability/response_length_var": 79119.8203125,
+      "stability/token_entropy_mean": 6.9375,
+      "step": 20
+    },
+    {
+      "epoch": 0.12795905310300704,
+      "grad_norm": 1.7784866094589233,
+      "learning_rate": 2.4e-07,
+      "logits/chosen": -2.277261257171631,
+      "logits/rejected": -2.1251189708709717,
+      "logps/chosen": -378.1658020019531,
+      "logps/rejected": -205.85794067382812,
+      "loss": 0.9376,
+      "metrics/advantage_var": 0.07424516975879669,
+      "rewards/accuracies": 0.4957031309604645,
+      "rewards/chosen": 0.0019111786969006062,
+      "rewards/margins": -7.651938358321786e-05,
+      "rewards/rejected": 0.0019876982551068068,
+      "stability/repetition_rate_mean": 0.9178038835525513,
+      "stability/response_length_mean": 658.5999755859375,
+      "stability/response_length_std": 288.0155029296875,
+      "stability/response_length_var": 82952.9296875,
+      "stability/token_entropy_mean": 7.028124809265137,
+      "step": 25
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 1.8653020858764648,
+      "learning_rate": 2.9e-07,
+      "logits/chosen": -2.2706894874572754,
+      "logits/rejected": -2.1234047412872314,
+      "logps/chosen": -373.90582275390625,
+      "logps/rejected": -208.9296417236328,
+      "loss": 0.941,
+      "metrics/advantage_var": 0.07175552845001221,
+      "rewards/accuracies": 0.512890636920929,
+      "rewards/chosen": 0.0024493553210049868,
+      "rewards/margins": 0.000800004811026156,
+      "rewards/rejected": 0.0016493506263941526,
+      "stability/repetition_rate_mean": 0.812473475933075,
+      "stability/response_length_mean": 356.0,
+      "stability/response_length_std": 249.6766815185547,
+      "stability/response_length_var": 62338.4453125,
+      "stability/token_entropy_mean": 6.231249809265137,
+      "step": 30
+    },
+    {
+      "epoch": 0.17914267434420986,
+      "grad_norm": 1.7568234205245972,
+      "learning_rate": 3.4000000000000003e-07,
+      "logits/chosen": -2.2611916065216064,
+      "logits/rejected": -2.119868516921997,
+      "logps/chosen": -366.31524658203125,
+      "logps/rejected": -203.98394775390625,
+      "loss": 0.9412,
+      "metrics/advantage_var": 0.06888910382986069,
+      "rewards/accuracies": 0.509765625,
+      "rewards/chosen": 0.0030620130710303783,
+      "rewards/margins": 0.0006677248165942729,
+      "rewards/rejected": 0.0023942883126437664,
+      "stability/repetition_rate_mean": 0.9296051859855652,
+      "stability/response_length_mean": 606.2000122070312,
+      "stability/response_length_std": 394.1366882324219,
+      "stability/response_length_var": 155343.734375,
+      "stability/token_entropy_mean": 7.099999904632568,
+      "step": 35
+    },
+    {
+      "epoch": 0.20473448496481125,
+      "grad_norm": 1.845935344696045,
+      "learning_rate": 3.8999999999999997e-07,
+      "logits/chosen": -2.2764148712158203,
+      "logits/rejected": -2.1227190494537354,
+      "logps/chosen": -369.178955078125,
+      "logps/rejected": -197.36029052734375,
+      "loss": 0.9407,
+      "metrics/advantage_var": 0.06902168691158295,
+      "rewards/accuracies": 0.49335938692092896,
+      "rewards/chosen": 0.0040647732093930244,
+      "rewards/margins": -0.0005382935632951558,
+      "rewards/rejected": 0.004603066947311163,
+      "stability/repetition_rate_mean": 0.8679192662239075,
+      "stability/response_length_mean": 547.2000122070312,
+      "stability/response_length_std": 290.8897399902344,
+      "stability/response_length_var": 84616.84375,
+      "stability/token_entropy_mean": 6.5078125,
+      "step": 40
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 1.855995774269104,
+      "learning_rate": 4.3999999999999997e-07,
+      "logits/chosen": -2.2577130794525146,
+      "logits/rejected": -2.108164072036743,
+      "logps/chosen": -385.1685485839844,
+      "logps/rejected": -212.5677490234375,
+      "loss": 0.9411,
+      "metrics/advantage_var": 0.07766187936067581,
+      "rewards/accuracies": 0.5074218511581421,
+      "rewards/chosen": 0.00504249706864357,
+      "rewards/margins": 0.0010317217092961073,
+      "rewards/rejected": 0.004010775592178106,
+      "stability/repetition_rate_mean": 0.8887706995010376,
+      "stability/response_length_mean": 366.5,
+      "stability/response_length_std": 118.86243438720703,
+      "stability/response_length_var": 14128.27734375,
+      "stability/token_entropy_mean": 6.690625190734863,
+      "step": 45
+    },
+    {
+      "epoch": 0.2559181062060141,
+      "grad_norm": 1.8825205564498901,
+      "learning_rate": 4.9e-07,
+      "logits/chosen": -2.2646327018737793,
+      "logits/rejected": -2.1097517013549805,
+      "logps/chosen": -378.27996826171875,
+      "logps/rejected": -207.4878387451172,
+      "loss": 0.9416,
+      "metrics/advantage_var": 0.07472482323646545,
+      "rewards/accuracies": 0.538281261920929,
+      "rewards/chosen": 0.007781261578202248,
+      "rewards/margins": 0.0034829857759177685,
+      "rewards/rejected": 0.004298275802284479,
+      "stability/repetition_rate_mean": 0.8785834312438965,
+      "stability/response_length_mean": 384.1000061035156,
+      "stability/response_length_std": 178.4809112548828,
+      "stability/response_length_var": 31855.43359375,
+      "stability/token_entropy_mean": 6.540625095367432,
+      "step": 50
+    },
+    {
+      "epoch": 0.28150991682661547,
+      "grad_norm": 1.9857012033462524,
+      "learning_rate": 5.4e-07,
+      "logits/chosen": -2.2532308101654053,
+      "logits/rejected": -2.094627618789673,
+      "logps/chosen": -393.7790222167969,
+      "logps/rejected": -210.345947265625,
+      "loss": 0.9363,
+      "metrics/advantage_var": 0.07802015542984009,
+      "rewards/accuracies": 0.536328136920929,
+      "rewards/chosen": 0.010103564709424973,
+      "rewards/margins": 0.0037730818148702383,
+      "rewards/rejected": 0.006330482661724091,
+      "stability/repetition_rate_mean": 0.9481571912765503,
+      "stability/response_length_mean": 332.8999938964844,
+      "stability/response_length_std": 137.39193725585938,
+      "stability/response_length_var": 18876.544921875,
+      "stability/token_entropy_mean": 7.012499809265137,
+      "step": 55
+    },
+    {
+      "epoch": 0.30710172744721687,
+      "grad_norm": 1.8655455112457275,
+      "learning_rate": 5.9e-07,
+      "logits/chosen": -2.2731332778930664,
+      "logits/rejected": -2.111682891845703,
+      "logps/chosen": -387.6856994628906,
+      "logps/rejected": -206.78878784179688,
+      "loss": 0.9388,
+      "metrics/advantage_var": 0.07866128534078598,
+      "rewards/accuracies": 0.5300781726837158,
+      "rewards/chosen": 0.012091020122170448,
+      "rewards/margins": 0.003263443009927869,
+      "rewards/rejected": 0.008827576413750648,
+      "stability/repetition_rate_mean": 0.8726905584335327,
+      "stability/response_length_mean": 655.2999877929688,
+      "stability/response_length_std": 282.005126953125,
+      "stability/response_length_var": 79526.8984375,
+      "stability/token_entropy_mean": 6.596875190734863,
+      "step": 60
+    },
+    {
+      "epoch": 0.3326935380678183,
+      "grad_norm": 1.729257583618164,
+      "learning_rate": 6.4e-07,
+      "logits/chosen": -2.251981496810913,
+      "logits/rejected": -2.0987205505371094,
+      "logps/chosen": -377.4482116699219,
+      "logps/rejected": -204.36764526367188,
+      "loss": 0.9428,
+      "metrics/advantage_var": 0.08284494280815125,
+      "rewards/accuracies": 0.5609374642372131,
+      "rewards/chosen": 0.015137546695768833,
+      "rewards/margins": 0.0056471494026482105,
+      "rewards/rejected": 0.009490396827459335,
+      "stability/repetition_rate_mean": 0.9237826466560364,
+      "stability/response_length_mean": 306.3999938964844,
+      "stability/response_length_std": 141.7236785888672,
+      "stability/response_length_var": 20085.599609375,
+      "stability/token_entropy_mean": 6.884375095367432,
+      "step": 65
+    },
+    {
+      "epoch": 0.3582853486884197,
+      "grad_norm": 2.0435707569122314,
+      "learning_rate": 6.9e-07,
+      "logits/chosen": -2.2713375091552734,
+      "logits/rejected": -2.1065561771392822,
+      "logps/chosen": -393.7109375,
+      "logps/rejected": -206.14852905273438,
+      "loss": 0.9333,
+      "metrics/advantage_var": 0.09623891115188599,
+      "rewards/accuracies": 0.5589843988418579,
+      "rewards/chosen": 0.01978578045964241,
+      "rewards/margins": 0.007668572943657637,
+      "rewards/rejected": 0.012117205187678337,
+      "stability/repetition_rate_mean": 0.9703803062438965,
+      "stability/response_length_mean": 472.70001220703125,
+      "stability/response_length_std": 159.89791870117188,
+      "stability/response_length_var": 25567.34375,
+      "stability/token_entropy_mean": 7.21875,
+      "step": 70
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 1.9249153137207031,
+      "learning_rate": 7.4e-07,
+      "logits/chosen": -2.275832414627075,
+      "logits/rejected": -2.1167995929718018,
+      "logps/chosen": -381.0315856933594,
+      "logps/rejected": -207.89634704589844,
+      "loss": 0.9371,
+      "metrics/advantage_var": 0.10018797218799591,
+      "rewards/accuracies": 0.564453125,
+      "rewards/chosen": 0.021552283316850662,
+      "rewards/margins": 0.007498453836888075,
+      "rewards/rejected": 0.014053828082978725,
+      "stability/repetition_rate_mean": 0.9106122851371765,
+      "stability/response_length_mean": 527.0,
+      "stability/response_length_std": 293.7500915527344,
+      "stability/response_length_var": 86289.109375,
+      "stability/token_entropy_mean": 6.084765434265137,
+      "step": 75
+    },
+    {
+      "epoch": 0.4094689699296225,
+      "grad_norm": 1.8595386743545532,
+      "learning_rate": 7.9e-07,
+      "logits/chosen": -2.26131272315979,
+      "logits/rejected": -2.103710174560547,
+      "logps/chosen": -382.3757629394531,
+      "logps/rejected": -208.9224853515625,
+      "loss": 0.9341,
+      "metrics/advantage_var": 0.11523017287254333,
+      "rewards/accuracies": 0.568359375,
+      "rewards/chosen": 0.02573232911527157,
+      "rewards/margins": 0.008440744131803513,
+      "rewards/rejected": 0.017291583120822906,
+      "stability/repetition_rate_mean": 0.9519937634468079,
+      "stability/response_length_mean": 635.5,
+      "stability/response_length_std": 262.53900146484375,
+      "stability/response_length_var": 68926.71875,
+      "stability/token_entropy_mean": 7.215624809265137,
+      "step": 80
+    },
+    {
+      "epoch": 0.4350607805502239,
+      "grad_norm": 1.8945162296295166,
+      "learning_rate": 8.399999999999999e-07,
+      "logits/chosen": -2.258488178253174,
+      "logits/rejected": -2.10145902633667,
+      "logps/chosen": -387.2498779296875,
+      "logps/rejected": -208.5669708251953,
+      "loss": 0.9353,
+      "metrics/advantage_var": 0.155354842543602,
+      "rewards/accuracies": 0.590624988079071,
+      "rewards/chosen": 0.034161973744630814,
+      "rewards/margins": 0.012971531599760056,
+      "rewards/rejected": 0.021190444007515907,
+      "stability/repetition_rate_mean": 0.9050097465515137,
+      "stability/response_length_mean": 600.0999755859375,
+      "stability/response_length_std": 349.2030334472656,
+      "stability/response_length_var": 121942.765625,
+      "stability/token_entropy_mean": 6.73828125,
+      "step": 85
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 1.9289894104003906,
+      "learning_rate": 8.9e-07,
+      "logits/chosen": -2.261462688446045,
+      "logits/rejected": -2.1133339405059814,
+      "logps/chosen": -375.13140869140625,
+      "logps/rejected": -206.13841247558594,
+      "loss": 0.933,
+      "metrics/advantage_var": 0.17288456857204437,
+      "rewards/accuracies": 0.6050781607627869,
+      "rewards/chosen": 0.039468325674533844,
+      "rewards/margins": 0.014231967739760876,
+      "rewards/rejected": 0.025236357003450394,
+      "stability/repetition_rate_mean": 0.9758660197257996,
+      "stability/response_length_mean": 692.9000244140625,
+      "stability/response_length_std": 127.31452941894531,
+      "stability/response_length_var": 16208.9892578125,
+      "stability/token_entropy_mean": 7.115624904632568,
+      "step": 90
+    },
+    {
+      "epoch": 0.48624440179142675,
+      "grad_norm": 1.8516160249710083,
+      "learning_rate": 9.399999999999999e-07,
+      "logits/chosen": -2.2591757774353027,
+      "logits/rejected": -2.1126770973205566,
+      "logps/chosen": -366.7373046875,
+      "logps/rejected": -203.33721923828125,
+      "loss": 0.9297,
+      "metrics/advantage_var": 0.21673114597797394,
+      "rewards/accuracies": 0.62890625,
+      "rewards/chosen": 0.047232918441295624,
+      "rewards/margins": 0.01979774609208107,
+      "rewards/rejected": 0.027435174211859703,
+      "stability/repetition_rate_mean": 0.9478110074996948,
+      "stability/response_length_mean": 727.0,
+      "stability/response_length_std": 335.0764465332031,
+      "stability/response_length_var": 112276.21875,
+      "stability/token_entropy_mean": 7.303124904632568,
+      "step": 95
+    },
+    {
+      "epoch": 0.5118362124120281,
+      "grad_norm": 1.857904076576233,
+      "learning_rate": 9.9e-07,
+      "logits/chosen": -2.275299072265625,
+      "logits/rejected": -2.119931936264038,
+      "logps/chosen": -384.3828430175781,
+      "logps/rejected": -210.50686645507812,
+      "loss": 0.9271,
+      "metrics/advantage_var": 0.3037579357624054,
+      "rewards/accuracies": 0.6441406607627869,
+      "rewards/chosen": 0.05749749764800072,
+      "rewards/margins": 0.02403966337442398,
+      "rewards/rejected": 0.033457837998867035,
+      "stability/repetition_rate_mean": 0.9512196779251099,
+      "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": 1.7671213150024414,
+      "learning_rate": 9.999815554351224e-07,
+      "logits/chosen": -2.2592649459838867,
+      "logits/rejected": -2.108891248703003,
+      "logps/chosen": -377.8589172363281,
+      "logps/rejected": -209.54722595214844,
+      "loss": 0.9294,
+      "metrics/advantage_var": 0.38598760962486267,
+      "rewards/accuracies": 0.6253905892372131,
+      "rewards/chosen": 0.06676852703094482,
+      "rewards/margins": 0.026257099583745003,
+      "rewards/rejected": 0.04051142930984497,
+      "stability/repetition_rate_mean": 0.8934451937675476,
+      "stability/response_length_mean": 669.5,
+      "stability/response_length_std": 271.60888671875,
+      "stability/response_length_var": 73771.390625,
+      "stability/token_entropy_mean": 6.578125,
+      "step": 105
+    },
+    {
+      "epoch": 0.5630198336532309,
+      "grad_norm": 1.689988374710083,
+      "learning_rate": 9.999066267225444e-07,
+      "logits/chosen": -2.275810956954956,
+      "logits/rejected": -2.118212938308716,
+      "logps/chosen": -385.0805358886719,
+      "logps/rejected": -206.55056762695312,
+      "loss": 0.9258,
+      "metrics/advantage_var": 0.4739305377006531,
+      "rewards/accuracies": 0.6488281488418579,
+      "rewards/chosen": 0.07808978110551834,
+      "rewards/margins": 0.030354248359799385,
+      "rewards/rejected": 0.04773552715778351,
+      "stability/repetition_rate_mean": 0.9327242970466614,
+      "stability/response_length_mean": 255.60000610351562,
+      "stability/response_length_std": 114.30874633789062,
+      "stability/response_length_var": 13066.4892578125,
+      "stability/token_entropy_mean": 6.978125095367432,
+      "step": 110
+    },
+    {
+      "epoch": 0.5886116442738324,
+      "grad_norm": 1.8266487121582031,
+      "learning_rate": 9.997740697079592e-07,
+      "logits/chosen": -2.2522411346435547,
+      "logits/rejected": -2.1116788387298584,
+      "logps/chosen": -372.5317687988281,
+      "logps/rejected": -213.750732421875,
+      "loss": 0.9241,
+      "metrics/advantage_var": 0.6695685386657715,
+      "rewards/accuracies": 0.6449218392372131,
+      "rewards/chosen": 0.09085610508918762,
+      "rewards/margins": 0.036989450454711914,
+      "rewards/rejected": 0.05386665090918541,
+      "stability/repetition_rate_mean": 0.8290537595748901,
+      "stability/response_length_mean": 499.29998779296875,
+      "stability/response_length_std": 282.31072998046875,
+      "stability/response_length_var": 79699.34375,
+      "stability/token_entropy_mean": 6.53125,
+      "step": 115
+    },
+    {
+      "epoch": 0.6142034548944337,
+      "grad_norm": 1.8164492845535278,
+      "learning_rate": 9.995838996722914e-07,
+      "logits/chosen": -2.249505043029785,
+      "logits/rejected": -2.104017734527588,
+      "logps/chosen": -367.5647888183594,
+      "logps/rejected": -207.3728790283203,
+      "loss": 0.92,
+      "metrics/advantage_var": 0.8198556303977966,
+      "rewards/accuracies": 0.6578125357627869,
+      "rewards/chosen": 0.1038859412074089,
+      "rewards/margins": 0.04292682185769081,
+      "rewards/rejected": 0.06095912307500839,
+      "stability/repetition_rate_mean": 0.9244946241378784,
+      "stability/response_length_mean": 363.6000061035156,
+      "stability/response_length_std": 146.0701446533203,
+      "stability/response_length_var": 21336.48828125,
+      "stability/token_entropy_mean": 6.881249904632568,
+      "step": 120
+    },
+    {
+      "epoch": 0.6397952655150352,
+      "grad_norm": 1.8048653602600098,
+      "learning_rate": 9.993361385379874e-07,
+      "logits/chosen": -2.253950595855713,
+      "logits/rejected": -2.0959606170654297,
+      "logps/chosen": -385.4555358886719,
+      "logps/rejected": -204.85488891601562,
+      "loss": 0.9179,
+      "metrics/advantage_var": 1.1633703708648682,
+      "rewards/accuracies": 0.666015625,
+      "rewards/chosen": 0.11797237396240234,
+      "rewards/margins": 0.04964681342244148,
+      "rewards/rejected": 0.06832555681467056,
+      "stability/repetition_rate_mean": 0.8939984440803528,
+      "stability/response_length_mean": 433.79998779296875,
+      "stability/response_length_std": 126.92412567138672,
+      "stability/response_length_var": 16109.7333984375,
+      "stability/token_entropy_mean": 6.703125,
+      "step": 125
+    },
+    {
+      "epoch": 0.6653870761356366,
+      "grad_norm": 1.89629328250885,
+      "learning_rate": 9.99030814866488e-07,
+      "logits/chosen": -2.246368169784546,
+      "logits/rejected": -2.097729206085205,
+      "logps/chosen": -375.3144836425781,
+      "logps/rejected": -203.45748901367188,
+      "loss": 0.9137,
+      "metrics/advantage_var": 1.2405811548233032,
+      "rewards/accuracies": 0.663281261920929,
+      "rewards/chosen": 0.12805037200450897,
+      "rewards/margins": 0.05569981038570404,
+      "rewards/rejected": 0.07235056161880493,
+      "stability/repetition_rate_mean": 0.8216736912727356,
+      "stability/response_length_mean": 296.1000061035156,
+      "stability/response_length_std": 167.77593994140625,
+      "stability/response_length_var": 28148.767578125,
+      "stability/token_entropy_mean": 6.053124904632568,
+      "step": 130
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 1.762459397315979,
+      "learning_rate": 9.986679638549358e-07,
+      "logits/chosen": -2.268277645111084,
+      "logits/rejected": -2.108889102935791,
+      "logps/chosen": -387.0592041015625,
+      "logps/rejected": -213.4181365966797,
+      "loss": 0.9111,
+      "metrics/advantage_var": 1.49413001537323,
+      "rewards/accuracies": 0.671093761920929,
+      "rewards/chosen": 0.14280582964420319,
+      "rewards/margins": 0.06248578429222107,
+      "rewards/rejected": 0.08032004535198212,
+      "stability/repetition_rate_mean": 0.8815739750862122,
+      "stability/response_length_mean": 596.4000244140625,
+      "stability/response_length_std": 291.9129333496094,
+      "stability/response_length_var": 85213.15625,
+      "stability/token_entropy_mean": 6.671875,
+      "step": 135
+    },
+    {
+      "epoch": 0.7165706973768394,
+      "grad_norm": 1.8487615585327148,
+      "learning_rate": 9.982476273321175e-07,
+      "logits/chosen": -2.2565722465515137,
+      "logits/rejected": -2.1014785766601562,
+      "logps/chosen": -388.67364501953125,
+      "logps/rejected": -210.8116912841797,
+      "loss": 0.907,
+      "metrics/advantage_var": 1.983569622039795,
+      "rewards/accuracies": 0.6847655773162842,
+      "rewards/chosen": 0.1552000790834427,
+      "rewards/margins": 0.07189314067363739,
+      "rewards/rejected": 0.0833069309592247,
+      "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": 1.7706032991409302,
+      "learning_rate": 9.977698537536417e-07,
+      "logits/chosen": -2.268200635910034,
+      "logits/rejected": -2.1011595726013184,
+      "logps/chosen": -387.3850402832031,
+      "logps/rejected": -203.22686767578125,
+      "loss": 0.9045,
+      "metrics/advantage_var": 2.342012882232666,
+      "rewards/accuracies": 0.6761718392372131,
+      "rewards/chosen": 0.17279884219169617,
+      "rewards/margins": 0.07742971926927567,
+      "rewards/rejected": 0.09536914527416229,
+      "stability/repetition_rate_mean": 0.9656392931938171,
+      "stability/response_length_mean": 720.2000122070312,
+      "stability/response_length_std": 284.2842102050781,
+      "stability/response_length_var": 80817.5078125,
+      "stability/token_entropy_mean": 7.112500190734863,
+      "step": 145
+    },
+    {
+      "epoch": 0.7677543186180422,
+      "grad_norm": 1.8285086154937744,
+      "learning_rate": 9.972346981963546e-07,
+      "logits/chosen": -2.2355778217315674,
+      "logits/rejected": -2.0934510231018066,
+      "logps/chosen": -373.40240478515625,
+      "logps/rejected": -210.79360961914062,
+      "loss": 0.9046,
+      "metrics/advantage_var": 3.491899013519287,
+      "rewards/accuracies": 0.65234375,
+      "rewards/chosen": 0.18629421293735504,
+      "rewards/margins": 0.08447620272636414,
+      "rewards/rejected": 0.1018180176615715,
+      "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": 1.866870641708374,
+      "learning_rate": 9.966422223519885e-07,
+      "logits/chosen": -2.261730670928955,
+      "logits/rejected": -2.1110422611236572,
+      "logps/chosen": -376.8971862792969,
+      "logps/rejected": -205.9225311279297,
+      "loss": 0.8992,
+      "metrics/advantage_var": 3.0832598209381104,
+      "rewards/accuracies": 0.6664062142372131,
+      "rewards/chosen": 0.20672671496868134,
+      "rewards/margins": 0.09230801463127136,
+      "rewards/rejected": 0.11441870778799057,
+      "stability/repetition_rate_mean": 0.8748709559440613,
+      "stability/response_length_mean": 502.3999938964844,
+      "stability/response_length_std": 200.0795440673828,
+      "stability/response_length_var": 40031.8203125,
+      "stability/token_entropy_mean": 6.542187690734863,
+      "step": 155
+    },
+    {
+      "epoch": 0.818937939859245,
+      "grad_norm": 1.7998372316360474,
+      "learning_rate": 9.959924945200523e-07,
+      "logits/chosen": -2.255324602127075,
+      "logits/rejected": -2.0892727375030518,
+      "logps/chosen": -382.9584655761719,
+      "logps/rejected": -206.74269104003906,
+      "loss": 0.8939,
+      "metrics/advantage_var": 4.283844470977783,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.22003380954265594,
+      "rewards/margins": 0.10556791722774506,
+      "rewards/rejected": 0.11446589976549149,
+      "stability/repetition_rate_mean": 0.9658440351486206,
+      "stability/response_length_mean": 870.4000244140625,
+      "stability/response_length_std": 147.6581573486328,
+      "stability/response_length_var": 21802.93359375,
+      "stability/token_entropy_mean": 7.243750095367432,
+      "step": 160
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.6668412685394287,
+      "learning_rate": 9.952855895999568e-07,
+      "logits/chosen": -2.240656614303589,
+      "logits/rejected": -2.093395948410034,
+      "logps/chosen": -361.86102294921875,
+      "logps/rejected": -203.08079528808594,
+      "loss": 0.8928,
+      "metrics/advantage_var": 4.067698001861572,
+      "rewards/accuracies": 0.6761718988418579,
+      "rewards/chosen": 0.22143030166625977,
+      "rewards/margins": 0.10688450187444687,
+      "rewards/rejected": 0.11454581469297409,
+      "stability/repetition_rate_mean": 0.8792880177497864,
+      "stability/response_length_mean": 547.7000122070312,
+      "stability/response_length_std": 249.60325622558594,
+      "stability/response_length_var": 62301.7890625,
+      "stability/token_entropy_mean": 6.625,
+      "step": 165
+    },
+    {
+      "epoch": 0.8701215611004478,
+      "grad_norm": 1.7253059148788452,
+      "learning_rate": 9.9452158908238e-07,
+      "logits/chosen": -2.2658770084381104,
+      "logits/rejected": -2.100456714630127,
+      "logps/chosen": -376.3424377441406,
+      "logps/rejected": -203.1697235107422,
+      "loss": 0.8864,
+      "metrics/advantage_var": 5.073078632354736,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": 0.24319852888584137,
+      "rewards/margins": 0.12121755629777908,
+      "rewards/rejected": 0.12198095768690109,
+      "stability/repetition_rate_mean": 0.9187561869621277,
+      "stability/response_length_mean": 385.70001220703125,
+      "stability/response_length_std": 143.96685791015625,
+      "stability/response_length_var": 20726.455078125,
+      "stability/token_entropy_mean": 6.868750095367432,
+      "step": 170
+    },
+    {
+      "epoch": 0.8957133717210493,
+      "grad_norm": 1.8084509372711182,
+      "learning_rate": 9.937005810398745e-07,
+      "logits/chosen": -2.2514612674713135,
+      "logits/rejected": -2.096510410308838,
+      "logps/chosen": -379.8197021484375,
+      "logps/rejected": -211.5930633544922,
+      "loss": 0.8856,
+      "metrics/advantage_var": 6.549341678619385,
+      "rewards/accuracies": 0.6929687261581421,
+      "rewards/chosen": 0.2569131851196289,
+      "rewards/margins": 0.12991970777511597,
+      "rewards/rejected": 0.12699344754219055,
+      "stability/repetition_rate_mean": 0.9395610690116882,
+      "stability/response_length_mean": 460.6000061035156,
+      "stability/response_length_std": 165.9894256591797,
+      "stability/response_length_var": 27552.48828125,
+      "stability/token_entropy_mean": 6.9375,
+      "step": 175
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 1.7756394147872925,
+      "learning_rate": 9.928226601167138e-07,
+      "logits/chosen": -2.250619411468506,
+      "logits/rejected": -2.0897774696350098,
+      "logps/chosen": -386.3166809082031,
+      "logps/rejected": -212.235107421875,
+      "loss": 0.8842,
+      "metrics/advantage_var": 5.9247918128967285,
+      "rewards/accuracies": 0.6890625357627869,
+      "rewards/chosen": 0.2764429748058319,
+      "rewards/margins": 0.13361024856567383,
+      "rewards/rejected": 0.1428327113389969,
+      "stability/repetition_rate_mean": 0.8915449976921082,
+      "stability/response_length_mean": 716.2000122070312,
+      "stability/response_length_std": 329.3815002441406,
+      "stability/response_length_var": 108492.1796875,
+      "stability/token_entropy_mean": 6.525000095367432,
+      "step": 180
+    },
+    {
+      "epoch": 0.946896992962252,
+      "grad_norm": 1.6765127182006836,
+      "learning_rate": 9.918879275179817e-07,
+      "logits/chosen": -2.252854824066162,
+      "logits/rejected": -2.1023361682891846,
+      "logps/chosen": -373.72027587890625,
+      "logps/rejected": -209.63833618164062,
+      "loss": 0.8819,
+      "metrics/advantage_var": 7.567996978759766,
+      "rewards/accuracies": 0.6832031011581421,
+      "rewards/chosen": 0.2906308174133301,
+      "rewards/margins": 0.1405182033777237,
+      "rewards/rejected": 0.1501125991344452,
+      "stability/repetition_rate_mean": 0.8790209889411926,
+      "stability/response_length_mean": 486.3999938964844,
+      "stability/response_length_std": 213.8624725341797,
+      "stability/response_length_var": 45737.15625,
+      "stability/token_entropy_mean": 6.574999809265137,
+      "step": 185
+    },
+    {
+      "epoch": 0.9724888035828535,
+      "grad_norm": 1.5650509595870972,
+      "learning_rate": 9.90896490997906e-07,
+      "logits/chosen": -2.2370643615722656,
+      "logits/rejected": -2.0819854736328125,
+      "logps/chosen": -376.81488037109375,
+      "logps/rejected": -205.4518280029297,
+      "loss": 0.8751,
+      "metrics/advantage_var": 8.249796867370605,
+      "rewards/accuracies": 0.6980469226837158,
+      "rewards/chosen": 0.30180665850639343,
+      "rewards/margins": 0.1598372757434845,
+      "rewards/rejected": 0.14196938276290894,
+      "stability/repetition_rate_mean": 0.8905978202819824,
+      "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": 1.658704400062561,
+      "learning_rate": 9.898484648474362e-07,
+      "logits/chosen": -2.2535288333892822,
+      "logits/rejected": -2.100834369659424,
+      "logps/chosen": -360.88079833984375,
+      "logps/rejected": -207.50393676757812,
+      "loss": 0.8762,
+      "metrics/advantage_var": 8.585060119628906,
+      "rewards/accuracies": 0.6851562857627869,
+      "rewards/chosen": 0.3115439713001251,
+      "rewards/margins": 0.16069400310516357,
+      "rewards/rejected": 0.15084993839263916,
+      "stability/repetition_rate_mean": 0.7758646011352539,
+      "stability/response_length_mean": 356.3999938964844,
+      "stability/response_length_std": 207.94882202148438,
+      "stability/response_length_var": 43242.7109375,
+      "stability/token_entropy_mean": 5.946875095367432,
+      "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_ftx0.1/checkpoint-196/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47caba11889928f4a482733e328ad85abe61ff25b18bfec1963dc94f55545974
+size 8017

mistral7b_dpo_sft_ftx0.1/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_ftx0.1/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_ftx0.1/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": 6.990624904632568, "stability/repetition_rate_mean": 0.9183257222175598, "step": 10}
+{"stability/response_length_mean": 345.70001220703125, "stability/response_length_std": 212.67974853515625, "stability/response_length_var": 45232.6796875, "stability/token_entropy_mean": 6.323437690734863, "stability/repetition_rate_mean": 0.8462336659431458, "step": 15}
+{"stability/response_length_mean": 739.4000244140625, "stability/response_length_std": 281.282470703125, "stability/response_length_var": 79119.8203125, "stability/token_entropy_mean": 6.9375, "stability/repetition_rate_mean": 0.9284495115280151, "step": 20}
+{"stability/response_length_mean": 658.5999755859375, "stability/response_length_std": 288.0155029296875, "stability/response_length_var": 82952.9296875, "stability/token_entropy_mean": 7.028124809265137, "stability/repetition_rate_mean": 0.9178038835525513, "step": 25}
+{"stability/response_length_mean": 356.0, "stability/response_length_std": 249.6766815185547, "stability/response_length_var": 62338.4453125, "stability/token_entropy_mean": 6.231249809265137, "stability/repetition_rate_mean": 0.812473475933075, "step": 30}
+{"stability/response_length_mean": 606.2000122070312, "stability/response_length_std": 394.1366882324219, "stability/response_length_var": 155343.734375, "stability/token_entropy_mean": 7.099999904632568, "stability/repetition_rate_mean": 0.9296051859855652, "step": 35}
+{"stability/response_length_mean": 547.2000122070312, "stability/response_length_std": 290.8897399902344, "stability/response_length_var": 84616.84375, "stability/token_entropy_mean": 6.5078125, "stability/repetition_rate_mean": 0.8679192662239075, "step": 40}
+{"stability/response_length_mean": 366.5, "stability/response_length_std": 118.86243438720703, "stability/response_length_var": 14128.27734375, "stability/token_entropy_mean": 6.690625190734863, "stability/repetition_rate_mean": 0.8887706995010376, "step": 45}
+{"stability/response_length_mean": 384.1000061035156, "stability/response_length_std": 178.4809112548828, "stability/response_length_var": 31855.43359375, "stability/token_entropy_mean": 6.540625095367432, "stability/repetition_rate_mean": 0.8785834312438965, "step": 50}
+{"stability/response_length_mean": 332.8999938964844, "stability/response_length_std": 137.39193725585938, "stability/response_length_var": 18876.544921875, "stability/token_entropy_mean": 7.012499809265137, "stability/repetition_rate_mean": 0.9481571912765503, "step": 55}
+{"stability/response_length_mean": 655.2999877929688, "stability/response_length_std": 282.005126953125, "stability/response_length_var": 79526.8984375, "stability/token_entropy_mean": 6.596875190734863, "stability/repetition_rate_mean": 0.8726905584335327, "step": 60}
+{"stability/response_length_mean": 306.3999938964844, "stability/response_length_std": 141.7236785888672, "stability/response_length_var": 20085.599609375, "stability/token_entropy_mean": 6.884375095367432, "stability/repetition_rate_mean": 0.9237826466560364, "step": 65}
+{"stability/response_length_mean": 472.70001220703125, "stability/response_length_std": 159.89791870117188, "stability/response_length_var": 25567.34375, "stability/token_entropy_mean": 7.21875, "stability/repetition_rate_mean": 0.9703803062438965, "step": 70}
+{"stability/response_length_mean": 527.0, "stability/response_length_std": 293.7500915527344, "stability/response_length_var": 86289.109375, "stability/token_entropy_mean": 6.084765434265137, "stability/repetition_rate_mean": 0.9106122851371765, "step": 75}
+{"stability/response_length_mean": 635.5, "stability/response_length_std": 262.53900146484375, "stability/response_length_var": 68926.71875, "stability/token_entropy_mean": 7.215624809265137, "stability/repetition_rate_mean": 0.9519937634468079, "step": 80}
+{"stability/response_length_mean": 600.0999755859375, "stability/response_length_std": 349.2030334472656, "stability/response_length_var": 121942.765625, "stability/token_entropy_mean": 6.73828125, "stability/repetition_rate_mean": 0.9050097465515137, "step": 85}
+{"stability/response_length_mean": 692.9000244140625, "stability/response_length_std": 127.31452941894531, "stability/response_length_var": 16208.9892578125, "stability/token_entropy_mean": 7.115624904632568, "stability/repetition_rate_mean": 0.9758660197257996, "step": 90}
+{"stability/response_length_mean": 727.0, "stability/response_length_std": 335.0764465332031, "stability/response_length_var": 112276.21875, "stability/token_entropy_mean": 7.303124904632568, "stability/repetition_rate_mean": 0.9478110074996948, "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": 0.9512196779251099, "step": 100}
+{"stability/response_length_mean": 669.5, "stability/response_length_std": 271.60888671875, "stability/response_length_var": 73771.390625, "stability/token_entropy_mean": 6.578125, "stability/repetition_rate_mean": 0.8934451937675476, "step": 105}
+{"stability/response_length_mean": 255.60000610351562, "stability/response_length_std": 114.30874633789062, "stability/response_length_var": 13066.4892578125, "stability/token_entropy_mean": 6.978125095367432, "stability/repetition_rate_mean": 0.9327242970466614, "step": 110}
+{"stability/response_length_mean": 499.29998779296875, "stability/response_length_std": 282.31072998046875, "stability/response_length_var": 79699.34375, "stability/token_entropy_mean": 6.53125, "stability/repetition_rate_mean": 0.8290537595748901, "step": 115}
+{"stability/response_length_mean": 363.6000061035156, "stability/response_length_std": 146.0701446533203, "stability/response_length_var": 21336.48828125, "stability/token_entropy_mean": 6.881249904632568, "stability/repetition_rate_mean": 0.9244946241378784, "step": 120}
+{"stability/response_length_mean": 433.79998779296875, "stability/response_length_std": 126.92412567138672, "stability/response_length_var": 16109.7333984375, "stability/token_entropy_mean": 6.703125, "stability/repetition_rate_mean": 0.8939984440803528, "step": 125}
+{"stability/response_length_mean": 296.1000061035156, "stability/response_length_std": 167.77593994140625, "stability/response_length_var": 28148.767578125, "stability/token_entropy_mean": 6.053124904632568, "stability/repetition_rate_mean": 0.8216736912727356, "step": 130}
+{"stability/response_length_mean": 596.4000244140625, "stability/response_length_std": 291.9129333496094, "stability/response_length_var": 85213.15625, "stability/token_entropy_mean": 6.671875, "stability/repetition_rate_mean": 0.8815739750862122, "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": 7.112500190734863, "stability/repetition_rate_mean": 0.9656392931938171, "step": 145}
+{"stability/response_length_mean": 654.2000122070312, "stability/response_length_std": 372.5482177734375, "stability/response_length_var": 138792.171875, "stability/token_entropy_mean": 6.556250095367432, "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": 6.542187690734863, "stability/repetition_rate_mean": 0.8748709559440613, "step": 155}
+{"stability/response_length_mean": 870.4000244140625, "stability/response_length_std": 147.6581573486328, "stability/response_length_var": 21802.93359375, "stability/token_entropy_mean": 7.243750095367432, "stability/repetition_rate_mean": 0.9658440351486206, "step": 160}
+{"stability/response_length_mean": 547.7000122070312, "stability/response_length_std": 249.60325622558594, "stability/response_length_var": 62301.7890625, "stability/token_entropy_mean": 6.625, "stability/repetition_rate_mean": 0.8792880177497864, "step": 165}
+{"stability/response_length_mean": 385.70001220703125, "stability/response_length_std": 143.96685791015625, "stability/response_length_var": 20726.455078125, "stability/token_entropy_mean": 6.868750095367432, "stability/repetition_rate_mean": 0.9187561869621277, "step": 170}
+{"stability/response_length_mean": 460.6000061035156, "stability/response_length_std": 165.9894256591797, "stability/response_length_var": 27552.48828125, "stability/token_entropy_mean": 6.9375, "stability/repetition_rate_mean": 0.9395610690116882, "step": 175}
+{"stability/response_length_mean": 716.2000122070312, "stability/response_length_std": 329.3815002441406, "stability/response_length_var": 108492.1796875, "stability/token_entropy_mean": 6.525000095367432, "stability/repetition_rate_mean": 0.8915449976921082, "step": 180}
+{"stability/response_length_mean": 486.3999938964844, "stability/response_length_std": 213.8624725341797, "stability/response_length_var": 45737.15625, "stability/token_entropy_mean": 6.574999809265137, "stability/repetition_rate_mean": 0.8790209889411926, "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": 0.8905978202819824, "step": 190}
+{"stability/response_length_mean": 356.3999938964844, "stability/response_length_std": 207.94882202148438, "stability/response_length_var": 43242.7109375, "stability/token_entropy_mean": 5.946875095367432, "stability/repetition_rate_mean": 0.7758646011352539, "step": 195}

mistral7b_dpo_sft_ftx0.1/tokenizer.model

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

mistral7b_dpo_sft_ftx0.1/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_ftx0.1/train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "total_flos": 7.072674331683193e+18,
+    "train_loss": 0.9197707787460211,
+    "train_runtime": 4344.4262,
+    "train_samples_per_second": 23.018,
+    "train_steps_per_second": 0.045
+}

mistral7b_dpo_sft_ftx0.1/trainer_log.jsonl

@@ -0,0 +1,40 @@
+{"current_steps": 5, "total_steps": 196, "loss": 0.9396, "accuracy": 0.29960939288139343, "lr": 4e-08, "epoch": 0.025591810620601407, "percentage": 2.55, "elapsed_time": "0:01:49", "remaining_time": "1:09:33", "rewards/chosen": 0.0005876531358808279, "rewards/rejected": 0.00035743374610319734, "rewards/accuracies": 0.29960939288139343, "rewards/margins": 0.00023021931701805443, "logps/chosen": -375.069580078125, "logps/rejected": -210.8708038330078, "logits/chosen": -2.2728710174560547, "logits/rejected": -2.123058557510376}
+{"current_steps": 10, "total_steps": 196, "loss": 0.9379, "accuracy": 0.4902343451976776, "lr": 9e-08, "epoch": 0.05118362124120281, "percentage": 5.1, "elapsed_time": "0:03:40", "remaining_time": "1:08:23", "rewards/chosen": 0.0011766888201236725, "rewards/rejected": 0.0012218140764161944, "rewards/accuracies": 0.4902343451976776, "rewards/margins": -4.512531450018287e-05, "logps/chosen": -382.03143310546875, "logps/rejected": -202.44522094726562, "logits/chosen": -2.2679998874664307, "logits/rejected": -2.1166672706604004, "stability/response_length_mean": 602.7999877929688, "stability/response_length_std": 295.1842346191406, "stability/response_length_var": 87133.734375, "stability/token_entropy_mean": 6.990624904632568, "stability/repetition_rate_mean": 0.9183257222175598}
+{"current_steps": 15, "total_steps": 196, "loss": 0.9389, "accuracy": 0.48750001192092896, "lr": 1.4e-07, "epoch": 0.07677543186180422, "percentage": 7.65, "elapsed_time": "0:05:31", "remaining_time": "1:06:36", "rewards/chosen": 0.0010101307416334748, "rewards/rejected": 0.0013206324074417353, "rewards/accuracies": 0.48750001192092896, "rewards/margins": -0.00031050166580826044, "logps/chosen": -389.2155456542969, "logps/rejected": -207.05043029785156, "logits/chosen": -2.260781764984131, "logits/rejected": -2.099287509918213, "stability/response_length_mean": 345.70001220703125, "stability/response_length_std": 212.67974853515625, "stability/response_length_var": 45232.6796875, "stability/token_entropy_mean": 6.323437690734863, "stability/repetition_rate_mean": 0.8462336659431458}
+{"current_steps": 20, "total_steps": 196, "loss": 0.9424, "accuracy": 0.4957031011581421, "lr": 1.8999999999999998e-07, "epoch": 0.10236724248240563, "percentage": 10.2, "elapsed_time": "0:07:22", "remaining_time": "1:04:50", "rewards/chosen": 0.0002682031481526792, "rewards/rejected": 0.0004127377178519964, "rewards/accuracies": 0.4957031011581421, "rewards/margins": -0.00014453462790697813, "logps/chosen": -368.721435546875, "logps/rejected": -198.57066345214844, "logits/chosen": -2.2612733840942383, "logits/rejected": -2.1143851280212402, "stability/response_length_mean": 739.4000244140625, "stability/response_length_std": 281.282470703125, "stability/response_length_var": 79119.8203125, "stability/token_entropy_mean": 6.9375, "stability/repetition_rate_mean": 0.9284495115280151}
+{"current_steps": 25, "total_steps": 196, "loss": 0.9376, "accuracy": 0.4957031309604645, "lr": 2.4e-07, "epoch": 0.12795905310300704, "percentage": 12.76, "elapsed_time": "0:09:13", "remaining_time": "1:03:06", "rewards/chosen": 0.0019111786969006062, "rewards/rejected": 0.0019876982551068068, "rewards/accuracies": 0.4957031309604645, "rewards/margins": -7.651938358321786e-05, "logps/chosen": -378.1658020019531, "logps/rejected": -205.85794067382812, "logits/chosen": -2.277261257171631, "logits/rejected": -2.1251189708709717, "stability/response_length_mean": 658.5999755859375, "stability/response_length_std": 288.0155029296875, "stability/response_length_var": 82952.9296875, "stability/token_entropy_mean": 7.028124809265137, "stability/repetition_rate_mean": 0.9178038835525513}
+{"current_steps": 30, "total_steps": 196, "loss": 0.941, "accuracy": 0.512890636920929, "lr": 2.9e-07, "epoch": 0.15355086372360843, "percentage": 15.31, "elapsed_time": "0:11:04", "remaining_time": "1:01:16", "rewards/chosen": 0.0024493553210049868, "rewards/rejected": 0.0016493506263941526, "rewards/accuracies": 0.512890636920929, "rewards/margins": 0.000800004811026156, "logps/chosen": -373.90582275390625, "logps/rejected": -208.9296417236328, "logits/chosen": -2.2706894874572754, "logits/rejected": -2.1234047412872314, "stability/response_length_mean": 356.0, "stability/response_length_std": 249.6766815185547, "stability/response_length_var": 62338.4453125, "stability/token_entropy_mean": 6.231249809265137, "stability/repetition_rate_mean": 0.812473475933075}
+{"current_steps": 35, "total_steps": 196, "loss": 0.9412, "accuracy": 0.509765625, "lr": 3.4000000000000003e-07, "epoch": 0.17914267434420986, "percentage": 17.86, "elapsed_time": "0:12:55", "remaining_time": "0:59:29", "rewards/chosen": 0.0030620130710303783, "rewards/rejected": 0.0023942883126437664, "rewards/accuracies": 0.509765625, "rewards/margins": 0.0006677248165942729, "logps/chosen": -366.31524658203125, "logps/rejected": -203.98394775390625, "logits/chosen": -2.2611916065216064, "logits/rejected": -2.119868516921997, "stability/response_length_mean": 606.2000122070312, "stability/response_length_std": 394.1366882324219, "stability/response_length_var": 155343.734375, "stability/token_entropy_mean": 7.099999904632568, "stability/repetition_rate_mean": 0.9296051859855652}
+{"current_steps": 40, "total_steps": 196, "loss": 0.9407, "accuracy": 0.49335938692092896, "lr": 3.8999999999999997e-07, "epoch": 0.20473448496481125, "percentage": 20.41, "elapsed_time": "0:14:47", "remaining_time": "0:57:42", "rewards/chosen": 0.0040647732093930244, "rewards/rejected": 0.004603066947311163, "rewards/accuracies": 0.49335938692092896, "rewards/margins": -0.0005382935632951558, "logps/chosen": -369.178955078125, "logps/rejected": -197.36029052734375, "logits/chosen": -2.2764148712158203, "logits/rejected": -2.1227190494537354, "stability/response_length_mean": 547.2000122070312, "stability/response_length_std": 290.8897399902344, "stability/response_length_var": 84616.84375, "stability/token_entropy_mean": 6.5078125, "stability/repetition_rate_mean": 0.8679192662239075}
+{"current_steps": 45, "total_steps": 196, "loss": 0.9411, "accuracy": 0.5074218511581421, "lr": 4.3999999999999997e-07, "epoch": 0.23032629558541268, "percentage": 22.96, "elapsed_time": "0:16:38", "remaining_time": "0:55:49", "rewards/chosen": 0.00504249706864357, "rewards/rejected": 0.004010775592178106, "rewards/accuracies": 0.5074218511581421, "rewards/margins": 0.0010317217092961073, "logps/chosen": -385.1685485839844, "logps/rejected": -212.5677490234375, "logits/chosen": -2.2577130794525146, "logits/rejected": -2.108164072036743, "stability/response_length_mean": 366.5, "stability/response_length_std": 118.86243438720703, "stability/response_length_var": 14128.27734375, "stability/token_entropy_mean": 6.690625190734863, "stability/repetition_rate_mean": 0.8887706995010376}
+{"current_steps": 50, "total_steps": 196, "loss": 0.9416, "accuracy": 0.538281261920929, "lr": 4.9e-07, "epoch": 0.2559181062060141, "percentage": 25.51, "elapsed_time": "0:18:28", "remaining_time": "0:53:57", "rewards/chosen": 0.007781261578202248, "rewards/rejected": 0.004298275802284479, "rewards/accuracies": 0.538281261920929, "rewards/margins": 0.0034829857759177685, "logps/chosen": -378.27996826171875, "logps/rejected": -207.4878387451172, "logits/chosen": -2.2646327018737793, "logits/rejected": -2.1097517013549805, "stability/response_length_mean": 384.1000061035156, "stability/response_length_std": 178.4809112548828, "stability/response_length_var": 31855.43359375, "stability/token_entropy_mean": 6.540625095367432, "stability/repetition_rate_mean": 0.8785834312438965}
+{"current_steps": 55, "total_steps": 196, "loss": 0.9363, "accuracy": 0.536328136920929, "lr": 5.4e-07, "epoch": 0.28150991682661547, "percentage": 28.06, "elapsed_time": "0:20:19", "remaining_time": "0:52:05", "rewards/chosen": 0.010103564709424973, "rewards/rejected": 0.006330482661724091, "rewards/accuracies": 0.536328136920929, "rewards/margins": 0.0037730818148702383, "logps/chosen": -393.7790222167969, "logps/rejected": -210.345947265625, "logits/chosen": -2.2532308101654053, "logits/rejected": -2.094627618789673, "stability/response_length_mean": 332.8999938964844, "stability/response_length_std": 137.39193725585938, "stability/response_length_var": 18876.544921875, "stability/token_entropy_mean": 7.012499809265137, "stability/repetition_rate_mean": 0.9481571912765503}
+{"current_steps": 60, "total_steps": 196, "loss": 0.9388, "accuracy": 0.5300781726837158, "lr": 5.9e-07, "epoch": 0.30710172744721687, "percentage": 30.61, "elapsed_time": "0:22:10", "remaining_time": "0:50:16", "rewards/chosen": 0.012091020122170448, "rewards/rejected": 0.008827576413750648, "rewards/accuracies": 0.5300781726837158, "rewards/margins": 0.003263443009927869, "logps/chosen": -387.6856994628906, "logps/rejected": -206.78878784179688, "logits/chosen": -2.2731332778930664, "logits/rejected": -2.111682891845703, "stability/response_length_mean": 655.2999877929688, "stability/response_length_std": 282.005126953125, "stability/response_length_var": 79526.8984375, "stability/token_entropy_mean": 6.596875190734863, "stability/repetition_rate_mean": 0.8726905584335327}
+{"current_steps": 65, "total_steps": 196, "loss": 0.9428, "accuracy": 0.5609374642372131, "lr": 6.4e-07, "epoch": 0.3326935380678183, "percentage": 33.16, "elapsed_time": "0:24:01", "remaining_time": "0:48:24", "rewards/chosen": 0.015137546695768833, "rewards/rejected": 0.009490396827459335, "rewards/accuracies": 0.5609374642372131, "rewards/margins": 0.0056471494026482105, "logps/chosen": -377.4482116699219, "logps/rejected": -204.36764526367188, "logits/chosen": -2.251981496810913, "logits/rejected": -2.0987205505371094, "stability/response_length_mean": 306.3999938964844, "stability/response_length_std": 141.7236785888672, "stability/response_length_var": 20085.599609375, "stability/token_entropy_mean": 6.884375095367432, "stability/repetition_rate_mean": 0.9237826466560364}
+{"current_steps": 70, "total_steps": 196, "loss": 0.9333, "accuracy": 0.5589843988418579, "lr": 6.9e-07, "epoch": 0.3582853486884197, "percentage": 35.71, "elapsed_time": "0:25:51", "remaining_time": "0:46:33", "rewards/chosen": 0.01978578045964241, "rewards/rejected": 0.012117205187678337, "rewards/accuracies": 0.5589843988418579, "rewards/margins": 0.007668572943657637, "logps/chosen": -393.7109375, "logps/rejected": -206.14852905273438, "logits/chosen": -2.2713375091552734, "logits/rejected": -2.1065561771392822, "stability/response_length_mean": 472.70001220703125, "stability/response_length_std": 159.89791870117188, "stability/response_length_var": 25567.34375, "stability/token_entropy_mean": 7.21875, "stability/repetition_rate_mean": 0.9703803062438965}
+{"current_steps": 75, "total_steps": 196, "loss": 0.9371, "accuracy": 0.564453125, "lr": 7.4e-07, "epoch": 0.3838771593090211, "percentage": 38.27, "elapsed_time": "0:27:42", "remaining_time": "0:44:42", "rewards/chosen": 0.021552283316850662, "rewards/rejected": 0.014053828082978725, "rewards/accuracies": 0.564453125, "rewards/margins": 0.007498453836888075, "logps/chosen": -381.0315856933594, "logps/rejected": -207.89634704589844, "logits/chosen": -2.275832414627075, "logits/rejected": -2.1167995929718018, "stability/response_length_mean": 527.0, "stability/response_length_std": 293.7500915527344, "stability/response_length_var": 86289.109375, "stability/token_entropy_mean": 6.084765434265137, "stability/repetition_rate_mean": 0.9106122851371765}
+{"current_steps": 80, "total_steps": 196, "loss": 0.9341, "accuracy": 0.568359375, "lr": 7.9e-07, "epoch": 0.4094689699296225, "percentage": 40.82, "elapsed_time": "0:29:34", "remaining_time": "0:42:52", "rewards/chosen": 0.02573232911527157, "rewards/rejected": 0.017291583120822906, "rewards/accuracies": 0.568359375, "rewards/margins": 0.008440744131803513, "logps/chosen": -382.3757629394531, "logps/rejected": -208.9224853515625, "logits/chosen": -2.26131272315979, "logits/rejected": -2.103710174560547, "stability/response_length_mean": 635.5, "stability/response_length_std": 262.53900146484375, "stability/response_length_var": 68926.71875, "stability/token_entropy_mean": 7.215624809265137, "stability/repetition_rate_mean": 0.9519937634468079}
+{"current_steps": 85, "total_steps": 196, "loss": 0.9353, "accuracy": 0.590624988079071, "lr": 8.399999999999999e-07, "epoch": 0.4350607805502239, "percentage": 43.37, "elapsed_time": "0:31:26", "remaining_time": "0:41:03", "rewards/chosen": 0.034161973744630814, "rewards/rejected": 0.021190444007515907, "rewards/accuracies": 0.590624988079071, "rewards/margins": 0.012971531599760056, "logps/chosen": -387.2498779296875, "logps/rejected": -208.5669708251953, "logits/chosen": -2.258488178253174, "logits/rejected": -2.10145902633667, "stability/response_length_mean": 600.0999755859375, "stability/response_length_std": 349.2030334472656, "stability/response_length_var": 121942.765625, "stability/token_entropy_mean": 6.73828125, "stability/repetition_rate_mean": 0.9050097465515137}
+{"current_steps": 90, "total_steps": 196, "loss": 0.933, "accuracy": 0.6050781607627869, "lr": 8.9e-07, "epoch": 0.46065259117082535, "percentage": 45.92, "elapsed_time": "0:33:17", "remaining_time": "0:39:13", "rewards/chosen": 0.039468325674533844, "rewards/rejected": 0.025236357003450394, "rewards/accuracies": 0.6050781607627869, "rewards/margins": 0.014231967739760876, "logps/chosen": -375.13140869140625, "logps/rejected": -206.13841247558594, "logits/chosen": -2.261462688446045, "logits/rejected": -2.1133339405059814, "stability/response_length_mean": 692.9000244140625, "stability/response_length_std": 127.31452941894531, "stability/response_length_var": 16208.9892578125, "stability/token_entropy_mean": 7.115624904632568, "stability/repetition_rate_mean": 0.9758660197257996}
+{"current_steps": 95, "total_steps": 196, "loss": 0.9297, "accuracy": 0.62890625, "lr": 9.399999999999999e-07, "epoch": 0.48624440179142675, "percentage": 48.47, "elapsed_time": "0:35:09", "remaining_time": "0:37:22", "rewards/chosen": 0.047232918441295624, "rewards/rejected": 0.027435174211859703, "rewards/accuracies": 0.62890625, "rewards/margins": 0.01979774609208107, "logps/chosen": -366.7373046875, "logps/rejected": -203.33721923828125, "logits/chosen": -2.2591757774353027, "logits/rejected": -2.1126770973205566, "stability/response_length_mean": 727.0, "stability/response_length_std": 335.0764465332031, "stability/response_length_var": 112276.21875, "stability/token_entropy_mean": 7.303124904632568, "stability/repetition_rate_mean": 0.9478110074996948}
+{"current_steps": 100, "total_steps": 196, "loss": 0.9271, "accuracy": 0.6441406607627869, "lr": 9.9e-07, "epoch": 0.5118362124120281, "percentage": 51.02, "elapsed_time": "0:37:00", "remaining_time": "0:35:31", "rewards/chosen": 0.05749749764800072, "rewards/rejected": 0.033457837998867035, "rewards/accuracies": 0.6441406607627869, "rewards/margins": 0.02403966337442398, "logps/chosen": -384.3828430175781, "logps/rejected": -210.50686645507812, "logits/chosen": -2.275299072265625, "logits/rejected": -2.119931936264038, "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": 0.9512196779251099}
+{"current_steps": 105, "total_steps": 196, "loss": 0.9294, "accuracy": 0.6253905892372131, "lr": 9.999815554351224e-07, "epoch": 0.5374280230326296, "percentage": 53.57, "elapsed_time": "0:38:52", "remaining_time": "0:33:41", "rewards/chosen": 0.06676852703094482, "rewards/rejected": 0.04051142930984497, "rewards/accuracies": 0.6253905892372131, "rewards/margins": 0.026257099583745003, "logps/chosen": -377.8589172363281, "logps/rejected": -209.54722595214844, "logits/chosen": -2.2592649459838867, "logits/rejected": -2.108891248703003, "stability/response_length_mean": 669.5, "stability/response_length_std": 271.60888671875, "stability/response_length_var": 73771.390625, "stability/token_entropy_mean": 6.578125, "stability/repetition_rate_mean": 0.8934451937675476}
+{"current_steps": 110, "total_steps": 196, "loss": 0.9258, "accuracy": 0.6488281488418579, "lr": 9.999066267225444e-07, "epoch": 0.5630198336532309, "percentage": 56.12, "elapsed_time": "0:40:42", "remaining_time": "0:31:49", "rewards/chosen": 0.07808978110551834, "rewards/rejected": 0.04773552715778351, "rewards/accuracies": 0.6488281488418579, "rewards/margins": 0.030354248359799385, "logps/chosen": -385.0805358886719, "logps/rejected": -206.55056762695312, "logits/chosen": -2.275810956954956, "logits/rejected": -2.118212938308716, "stability/response_length_mean": 255.60000610351562, "stability/response_length_std": 114.30874633789062, "stability/response_length_var": 13066.4892578125, "stability/token_entropy_mean": 6.978125095367432, "stability/repetition_rate_mean": 0.9327242970466614}
+{"current_steps": 115, "total_steps": 196, "loss": 0.9241, "accuracy": 0.6449218392372131, "lr": 9.997740697079592e-07, "epoch": 0.5886116442738324, "percentage": 58.67, "elapsed_time": "0:42:33", "remaining_time": "0:29:58", "rewards/chosen": 0.09085610508918762, "rewards/rejected": 0.05386665090918541, "rewards/accuracies": 0.6449218392372131, "rewards/margins": 0.036989450454711914, "logps/chosen": -372.5317687988281, "logps/rejected": -213.750732421875, "logits/chosen": -2.2522411346435547, "logits/rejected": -2.1116788387298584, "stability/response_length_mean": 499.29998779296875, "stability/response_length_std": 282.31072998046875, "stability/response_length_var": 79699.34375, "stability/token_entropy_mean": 6.53125, "stability/repetition_rate_mean": 0.8290537595748901}
+{"current_steps": 120, "total_steps": 196, "loss": 0.92, "accuracy": 0.6578125357627869, "lr": 9.995838996722914e-07, "epoch": 0.6142034548944337, "percentage": 61.22, "elapsed_time": "0:44:23", "remaining_time": "0:28:07", "rewards/chosen": 0.1038859412074089, "rewards/rejected": 0.06095912307500839, "rewards/accuracies": 0.6578125357627869, "rewards/margins": 0.04292682185769081, "logps/chosen": -367.5647888183594, "logps/rejected": -207.3728790283203, "logits/chosen": -2.249505043029785, "logits/rejected": -2.104017734527588, "stability/response_length_mean": 363.6000061035156, "stability/response_length_std": 146.0701446533203, "stability/response_length_var": 21336.48828125, "stability/token_entropy_mean": 6.881249904632568, "stability/repetition_rate_mean": 0.9244946241378784}
+{"current_steps": 125, "total_steps": 196, "loss": 0.9179, "accuracy": 0.666015625, "lr": 9.993361385379874e-07, "epoch": 0.6397952655150352, "percentage": 63.78, "elapsed_time": "0:46:13", "remaining_time": "0:26:15", "rewards/chosen": 0.11797237396240234, "rewards/rejected": 0.06832555681467056, "rewards/accuracies": 0.666015625, "rewards/margins": 0.04964681342244148, "logps/chosen": -385.4555358886719, "logps/rejected": -204.85488891601562, "logits/chosen": -2.253950595855713, "logits/rejected": -2.0959606170654297, "stability/response_length_mean": 433.79998779296875, "stability/response_length_std": 126.92412567138672, "stability/response_length_var": 16109.7333984375, "stability/token_entropy_mean": 6.703125, "stability/repetition_rate_mean": 0.8939984440803528}
+{"current_steps": 130, "total_steps": 196, "loss": 0.9137, "accuracy": 0.663281261920929, "lr": 9.99030814866488e-07, "epoch": 0.6653870761356366, "percentage": 66.33, "elapsed_time": "0:48:04", "remaining_time": "0:24:24", "rewards/chosen": 0.12805037200450897, "rewards/rejected": 0.07235056161880493, "rewards/accuracies": 0.663281261920929, "rewards/margins": 0.05569981038570404, "logps/chosen": -375.3144836425781, "logps/rejected": -203.45748901367188, "logits/chosen": -2.246368169784546, "logits/rejected": -2.097729206085205, "stability/response_length_mean": 296.1000061035156, "stability/response_length_std": 167.77593994140625, "stability/response_length_var": 28148.767578125, "stability/token_entropy_mean": 6.053124904632568, "stability/repetition_rate_mean": 0.8216736912727356}
+{"current_steps": 135, "total_steps": 196, "loss": 0.9111, "accuracy": 0.671093761920929, "lr": 9.986679638549358e-07, "epoch": 0.690978886756238, "percentage": 68.88, "elapsed_time": "0:49:56", "remaining_time": "0:22:33", "rewards/chosen": 0.14280582964420319, "rewards/rejected": 0.08032004535198212, "rewards/accuracies": 0.671093761920929, "rewards/margins": 0.06248578429222107, "logps/chosen": -387.0592041015625, "logps/rejected": -213.4181365966797, "logits/chosen": -2.268277645111084, "logits/rejected": -2.108889102935791, "stability/response_length_mean": 596.4000244140625, "stability/response_length_std": 291.9129333496094, "stability/response_length_var": 85213.15625, "stability/token_entropy_mean": 6.671875, "stability/repetition_rate_mean": 0.8815739750862122}
+{"current_steps": 140, "total_steps": 196, "loss": 0.907, "accuracy": 0.6847655773162842, "lr": 9.982476273321175e-07, "epoch": 0.7165706973768394, "percentage": 71.43, "elapsed_time": "0:51:46", "remaining_time": "0:20:42", "rewards/chosen": 0.1552000790834427, "rewards/rejected": 0.0833069309592247, "rewards/accuracies": 0.6847655773162842, "rewards/margins": 0.07189314067363739, "logps/chosen": -388.67364501953125, "logps/rejected": -210.8116912841797, "logits/chosen": -2.2565722465515137, "logits/rejected": -2.1014785766601562, "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": 0.9045, "accuracy": 0.6761718392372131, "lr": 9.977698537536417e-07, "epoch": 0.7421625079974408, "percentage": 73.98, "elapsed_time": "0:53:38", "remaining_time": "0:18:51", "rewards/chosen": 0.17279884219169617, "rewards/rejected": 0.09536914527416229, "rewards/accuracies": 0.6761718392372131, "rewards/margins": 0.07742971926927567, "logps/chosen": -387.3850402832031, "logps/rejected": -203.22686767578125, "logits/chosen": -2.268200635910034, "logits/rejected": -2.1011595726013184, "stability/response_length_mean": 720.2000122070312, "stability/response_length_std": 284.2842102050781, "stability/response_length_var": 80817.5078125, "stability/token_entropy_mean": 7.112500190734863, "stability/repetition_rate_mean": 0.9656392931938171}
+{"current_steps": 150, "total_steps": 196, "loss": 0.9046, "accuracy": 0.65234375, "lr": 9.972346981963546e-07, "epoch": 0.7677543186180422, "percentage": 76.53, "elapsed_time": "0:55:29", "remaining_time": "0:17:01", "rewards/chosen": 0.18629421293735504, "rewards/rejected": 0.1018180176615715, "rewards/accuracies": 0.65234375, "rewards/margins": 0.08447620272636414, "logps/chosen": -373.40240478515625, "logps/rejected": -210.79360961914062, "logits/chosen": -2.2355778217315674, "logits/rejected": -2.0934510231018066, "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": 0.8992, "accuracy": 0.6664062142372131, "lr": 9.966422223519885e-07, "epoch": 0.7933461292386437, "percentage": 79.08, "elapsed_time": "0:57:21", "remaining_time": "0:15:10", "rewards/chosen": 0.20672671496868134, "rewards/rejected": 0.11441870778799057, "rewards/accuracies": 0.6664062142372131, "rewards/margins": 0.09230801463127136, "logps/chosen": -376.8971862792969, "logps/rejected": -205.9225311279297, "logits/chosen": -2.261730670928955, "logits/rejected": -2.1110422611236572, "stability/response_length_mean": 502.3999938964844, "stability/response_length_std": 200.0795440673828, "stability/response_length_var": 40031.8203125, "stability/token_entropy_mean": 6.542187690734863, "stability/repetition_rate_mean": 0.8748709559440613}
+{"current_steps": 160, "total_steps": 196, "loss": 0.8939, "accuracy": 0.6875, "lr": 9.959924945200523e-07, "epoch": 0.818937939859245, "percentage": 81.63, "elapsed_time": "0:59:12", "remaining_time": "0:13:19", "rewards/chosen": 0.22003380954265594, "rewards/rejected": 0.11446589976549149, "rewards/accuracies": 0.6875, "rewards/margins": 0.10556791722774506, "logps/chosen": -382.9584655761719, "logps/rejected": -206.74269104003906, "logits/chosen": -2.255324602127075, "logits/rejected": -2.0892727375030518, "stability/response_length_mean": 870.4000244140625, "stability/response_length_std": 147.6581573486328, "stability/response_length_var": 21802.93359375, "stability/token_entropy_mean": 7.243750095367432, "stability/repetition_rate_mean": 0.9658440351486206}
+{"current_steps": 165, "total_steps": 196, "loss": 0.8928, "accuracy": 0.6761718988418579, "lr": 9.952855895999568e-07, "epoch": 0.8445297504798465, "percentage": 84.18, "elapsed_time": "1:01:03", "remaining_time": "0:11:28", "rewards/chosen": 0.22143030166625977, "rewards/rejected": 0.11454581469297409, "rewards/accuracies": 0.6761718988418579, "rewards/margins": 0.10688450187444687, "logps/chosen": -361.86102294921875, "logps/rejected": -203.08079528808594, "logits/chosen": -2.240656614303589, "logits/rejected": -2.093395948410034, "stability/response_length_mean": 547.7000122070312, "stability/response_length_std": 249.60325622558594, "stability/response_length_var": 62301.7890625, "stability/token_entropy_mean": 6.625, "stability/repetition_rate_mean": 0.8792880177497864}
+{"current_steps": 170, "total_steps": 196, "loss": 0.8864, "accuracy": 0.6937500238418579, "lr": 9.9452158908238e-07, "epoch": 0.8701215611004478, "percentage": 86.73, "elapsed_time": "1:02:54", "remaining_time": "0:09:37", "rewards/chosen": 0.24319852888584137, "rewards/rejected": 0.12198095768690109, "rewards/accuracies": 0.6937500238418579, "rewards/margins": 0.12121755629777908, "logps/chosen": -376.3424377441406, "logps/rejected": -203.1697235107422, "logits/chosen": -2.2658770084381104, "logits/rejected": -2.100456714630127, "stability/response_length_mean": 385.70001220703125, "stability/response_length_std": 143.96685791015625, "stability/response_length_var": 20726.455078125, "stability/token_entropy_mean": 6.868750095367432, "stability/repetition_rate_mean": 0.9187561869621277}
+{"current_steps": 175, "total_steps": 196, "loss": 0.8856, "accuracy": 0.6929687261581421, "lr": 9.937005810398745e-07, "epoch": 0.8957133717210493, "percentage": 89.29, "elapsed_time": "1:04:45", "remaining_time": "0:07:46", "rewards/chosen": 0.2569131851196289, "rewards/rejected": 0.12699344754219055, "rewards/accuracies": 0.6929687261581421, "rewards/margins": 0.12991970777511597, "logps/chosen": -379.8197021484375, "logps/rejected": -211.5930633544922, "logits/chosen": -2.2514612674713135, "logits/rejected": -2.096510410308838, "stability/response_length_mean": 460.6000061035156, "stability/response_length_std": 165.9894256591797, "stability/response_length_var": 27552.48828125, "stability/token_entropy_mean": 6.9375, "stability/repetition_rate_mean": 0.9395610690116882}
+{"current_steps": 180, "total_steps": 196, "loss": 0.8842, "accuracy": 0.6890625357627869, "lr": 9.928226601167138e-07, "epoch": 0.9213051823416507, "percentage": 91.84, "elapsed_time": "1:06:37", "remaining_time": "0:05:55", "rewards/chosen": 0.2764429748058319, "rewards/rejected": 0.1428327113389969, "rewards/accuracies": 0.6890625357627869, "rewards/margins": 0.13361024856567383, "logps/chosen": -386.3166809082031, "logps/rejected": -212.235107421875, "logits/chosen": -2.250619411468506, "logits/rejected": -2.0897774696350098, "stability/response_length_mean": 716.2000122070312, "stability/response_length_std": 329.3815002441406, "stability/response_length_var": 108492.1796875, "stability/token_entropy_mean": 6.525000095367432, "stability/repetition_rate_mean": 0.8915449976921082}
+{"current_steps": 185, "total_steps": 196, "loss": 0.8819, "accuracy": 0.6832031011581421, "lr": 9.918879275179817e-07, "epoch": 0.946896992962252, "percentage": 94.39, "elapsed_time": "1:08:28", "remaining_time": "0:04:04", "rewards/chosen": 0.2906308174133301, "rewards/rejected": 0.1501125991344452, "rewards/accuracies": 0.6832031011581421, "rewards/margins": 0.1405182033777237, "logps/chosen": -373.72027587890625, "logps/rejected": -209.63833618164062, "logits/chosen": -2.252854824066162, "logits/rejected": -2.1023361682891846, "stability/response_length_mean": 486.3999938964844, "stability/response_length_std": 213.8624725341797, "stability/response_length_var": 45737.15625, "stability/token_entropy_mean": 6.574999809265137, "stability/repetition_rate_mean": 0.8790209889411926}
+{"current_steps": 190, "total_steps": 196, "loss": 0.8751, "accuracy": 0.6980469226837158, "lr": 9.90896490997906e-07, "epoch": 0.9724888035828535, "percentage": 96.94, "elapsed_time": "1:10:20", "remaining_time": "0:02:13", "rewards/chosen": 0.30180665850639343, "rewards/rejected": 0.14196938276290894, "rewards/accuracies": 0.6980469226837158, "rewards/margins": 0.1598372757434845, "logps/chosen": -376.81488037109375, "logps/rejected": -205.4518280029297, "logits/chosen": -2.2370643615722656, "logits/rejected": -2.0819854736328125, "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": 0.8905978202819824}
+{"current_steps": 195, "total_steps": 196, "loss": 0.8762, "accuracy": 0.6851562857627869, "lr": 9.898484648474362e-07, "epoch": 0.9980806142034548, "percentage": 99.49, "elapsed_time": "1:12:09", "remaining_time": "0:00:22", "rewards/chosen": 0.3115439713001251, "rewards/rejected": 0.15084993839263916, "rewards/accuracies": 0.6851562857627869, "rewards/margins": 0.16069400310516357, "logps/chosen": -360.88079833984375, "logps/rejected": -207.50393676757812, "logits/chosen": -2.2535288333892822, "logits/rejected": -2.100834369659424, "stability/response_length_mean": 356.3999938964844, "stability/response_length_std": 207.94882202148438, "stability/response_length_var": 43242.7109375, "stability/token_entropy_mean": 5.946875095367432, "stability/repetition_rate_mean": 0.7758646011352539}
+{"current_steps": 196, "total_steps": 196, "epoch": 1.0, "percentage": 100.0, "elapsed_time": "1:12:24", "remaining_time": "0:00:00"}

mistral7b_dpo_sft_ftx0.1/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": 1.7173521518707275,
+      "learning_rate": 4e-08,
+      "logits/chosen": -2.2728710174560547,
+      "logits/rejected": -2.123058557510376,
+      "logps/chosen": -375.069580078125,
+      "logps/rejected": -210.8708038330078,
+      "loss": 0.9396,
+      "metrics/advantage_var": 0.040927786380052567,
+      "rewards/accuracies": 0.29960939288139343,
+      "rewards/chosen": 0.0005876531358808279,
+      "rewards/margins": 0.00023021931701805443,
+      "rewards/rejected": 0.00035743374610319734,
+      "step": 5
+    },
+    {
+      "epoch": 0.05118362124120281,
+      "grad_norm": 1.8015280961990356,
+      "learning_rate": 9e-08,
+      "logits/chosen": -2.2679998874664307,
+      "logits/rejected": -2.1166672706604004,
+      "logps/chosen": -382.03143310546875,
+      "logps/rejected": -202.44522094726562,
+      "loss": 0.9379,
+      "metrics/advantage_var": 0.07131563127040863,
+      "rewards/accuracies": 0.4902343451976776,
+      "rewards/chosen": 0.0011766888201236725,
+      "rewards/margins": -4.512531450018287e-05,
+      "rewards/rejected": 0.0012218140764161944,
+      "stability/repetition_rate_mean": 0.9183257222175598,
+      "stability/response_length_mean": 602.7999877929688,
+      "stability/response_length_std": 295.1842346191406,
+      "stability/response_length_var": 87133.734375,
+      "stability/token_entropy_mean": 6.990624904632568,
+      "step": 10
+    },
+    {
+      "epoch": 0.07677543186180422,
+      "grad_norm": 1.8550324440002441,
+      "learning_rate": 1.4e-07,
+      "logits/chosen": -2.260781764984131,
+      "logits/rejected": -2.099287509918213,
+      "logps/chosen": -389.2155456542969,
+      "logps/rejected": -207.05043029785156,
+      "loss": 0.9389,
+      "metrics/advantage_var": 0.07388237863779068,
+      "rewards/accuracies": 0.48750001192092896,
+      "rewards/chosen": 0.0010101307416334748,
+      "rewards/margins": -0.00031050166580826044,
+      "rewards/rejected": 0.0013206324074417353,
+      "stability/repetition_rate_mean": 0.8462336659431458,
+      "stability/response_length_mean": 345.70001220703125,
+      "stability/response_length_std": 212.67974853515625,
+      "stability/response_length_var": 45232.6796875,
+      "stability/token_entropy_mean": 6.323437690734863,
+      "step": 15
+    },
+    {
+      "epoch": 0.10236724248240563,
+      "grad_norm": 1.7850894927978516,
+      "learning_rate": 1.8999999999999998e-07,
+      "logits/chosen": -2.2612733840942383,
+      "logits/rejected": -2.1143851280212402,
+      "logps/chosen": -368.721435546875,
+      "logps/rejected": -198.57066345214844,
+      "loss": 0.9424,
+      "metrics/advantage_var": 0.0697961151599884,
+      "rewards/accuracies": 0.4957031011581421,
+      "rewards/chosen": 0.0002682031481526792,
+      "rewards/margins": -0.00014453462790697813,
+      "rewards/rejected": 0.0004127377178519964,
+      "stability/repetition_rate_mean": 0.9284495115280151,
+      "stability/response_length_mean": 739.4000244140625,
+      "stability/response_length_std": 281.282470703125,
+      "stability/response_length_var": 79119.8203125,
+      "stability/token_entropy_mean": 6.9375,
+      "step": 20
+    },
+    {
+      "epoch": 0.12795905310300704,
+      "grad_norm": 1.7784866094589233,
+      "learning_rate": 2.4e-07,
+      "logits/chosen": -2.277261257171631,
+      "logits/rejected": -2.1251189708709717,
+      "logps/chosen": -378.1658020019531,
+      "logps/rejected": -205.85794067382812,
+      "loss": 0.9376,
+      "metrics/advantage_var": 0.07424516975879669,
+      "rewards/accuracies": 0.4957031309604645,
+      "rewards/chosen": 0.0019111786969006062,
+      "rewards/margins": -7.651938358321786e-05,
+      "rewards/rejected": 0.0019876982551068068,
+      "stability/repetition_rate_mean": 0.9178038835525513,
+      "stability/response_length_mean": 658.5999755859375,
+      "stability/response_length_std": 288.0155029296875,
+      "stability/response_length_var": 82952.9296875,
+      "stability/token_entropy_mean": 7.028124809265137,
+      "step": 25
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 1.8653020858764648,
+      "learning_rate": 2.9e-07,
+      "logits/chosen": -2.2706894874572754,
+      "logits/rejected": -2.1234047412872314,
+      "logps/chosen": -373.90582275390625,
+      "logps/rejected": -208.9296417236328,
+      "loss": 0.941,
+      "metrics/advantage_var": 0.07175552845001221,
+      "rewards/accuracies": 0.512890636920929,
+      "rewards/chosen": 0.0024493553210049868,
+      "rewards/margins": 0.000800004811026156,
+      "rewards/rejected": 0.0016493506263941526,
+      "stability/repetition_rate_mean": 0.812473475933075,
+      "stability/response_length_mean": 356.0,
+      "stability/response_length_std": 249.6766815185547,
+      "stability/response_length_var": 62338.4453125,
+      "stability/token_entropy_mean": 6.231249809265137,
+      "step": 30
+    },
+    {
+      "epoch": 0.17914267434420986,
+      "grad_norm": 1.7568234205245972,
+      "learning_rate": 3.4000000000000003e-07,
+      "logits/chosen": -2.2611916065216064,
+      "logits/rejected": -2.119868516921997,
+      "logps/chosen": -366.31524658203125,
+      "logps/rejected": -203.98394775390625,
+      "loss": 0.9412,
+      "metrics/advantage_var": 0.06888910382986069,
+      "rewards/accuracies": 0.509765625,
+      "rewards/chosen": 0.0030620130710303783,
+      "rewards/margins": 0.0006677248165942729,
+      "rewards/rejected": 0.0023942883126437664,
+      "stability/repetition_rate_mean": 0.9296051859855652,
+      "stability/response_length_mean": 606.2000122070312,
+      "stability/response_length_std": 394.1366882324219,
+      "stability/response_length_var": 155343.734375,
+      "stability/token_entropy_mean": 7.099999904632568,
+      "step": 35
+    },
+    {
+      "epoch": 0.20473448496481125,
+      "grad_norm": 1.845935344696045,
+      "learning_rate": 3.8999999999999997e-07,
+      "logits/chosen": -2.2764148712158203,
+      "logits/rejected": -2.1227190494537354,
+      "logps/chosen": -369.178955078125,
+      "logps/rejected": -197.36029052734375,
+      "loss": 0.9407,
+      "metrics/advantage_var": 0.06902168691158295,
+      "rewards/accuracies": 0.49335938692092896,
+      "rewards/chosen": 0.0040647732093930244,
+      "rewards/margins": -0.0005382935632951558,
+      "rewards/rejected": 0.004603066947311163,
+      "stability/repetition_rate_mean": 0.8679192662239075,
+      "stability/response_length_mean": 547.2000122070312,
+      "stability/response_length_std": 290.8897399902344,
+      "stability/response_length_var": 84616.84375,
+      "stability/token_entropy_mean": 6.5078125,
+      "step": 40
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 1.855995774269104,
+      "learning_rate": 4.3999999999999997e-07,
+      "logits/chosen": -2.2577130794525146,
+      "logits/rejected": -2.108164072036743,
+      "logps/chosen": -385.1685485839844,
+      "logps/rejected": -212.5677490234375,
+      "loss": 0.9411,
+      "metrics/advantage_var": 0.07766187936067581,
+      "rewards/accuracies": 0.5074218511581421,
+      "rewards/chosen": 0.00504249706864357,
+      "rewards/margins": 0.0010317217092961073,
+      "rewards/rejected": 0.004010775592178106,
+      "stability/repetition_rate_mean": 0.8887706995010376,
+      "stability/response_length_mean": 366.5,
+      "stability/response_length_std": 118.86243438720703,
+      "stability/response_length_var": 14128.27734375,
+      "stability/token_entropy_mean": 6.690625190734863,
+      "step": 45
+    },
+    {
+      "epoch": 0.2559181062060141,
+      "grad_norm": 1.8825205564498901,
+      "learning_rate": 4.9e-07,
+      "logits/chosen": -2.2646327018737793,
+      "logits/rejected": -2.1097517013549805,
+      "logps/chosen": -378.27996826171875,
+      "logps/rejected": -207.4878387451172,
+      "loss": 0.9416,
+      "metrics/advantage_var": 0.07472482323646545,
+      "rewards/accuracies": 0.538281261920929,
+      "rewards/chosen": 0.007781261578202248,
+      "rewards/margins": 0.0034829857759177685,
+      "rewards/rejected": 0.004298275802284479,
+      "stability/repetition_rate_mean": 0.8785834312438965,
+      "stability/response_length_mean": 384.1000061035156,
+      "stability/response_length_std": 178.4809112548828,
+      "stability/response_length_var": 31855.43359375,
+      "stability/token_entropy_mean": 6.540625095367432,
+      "step": 50
+    },
+    {
+      "epoch": 0.28150991682661547,
+      "grad_norm": 1.9857012033462524,
+      "learning_rate": 5.4e-07,
+      "logits/chosen": -2.2532308101654053,
+      "logits/rejected": -2.094627618789673,
+      "logps/chosen": -393.7790222167969,
+      "logps/rejected": -210.345947265625,
+      "loss": 0.9363,
+      "metrics/advantage_var": 0.07802015542984009,
+      "rewards/accuracies": 0.536328136920929,
+      "rewards/chosen": 0.010103564709424973,
+      "rewards/margins": 0.0037730818148702383,
+      "rewards/rejected": 0.006330482661724091,
+      "stability/repetition_rate_mean": 0.9481571912765503,
+      "stability/response_length_mean": 332.8999938964844,
+      "stability/response_length_std": 137.39193725585938,
+      "stability/response_length_var": 18876.544921875,
+      "stability/token_entropy_mean": 7.012499809265137,
+      "step": 55
+    },
+    {
+      "epoch": 0.30710172744721687,
+      "grad_norm": 1.8655455112457275,
+      "learning_rate": 5.9e-07,
+      "logits/chosen": -2.2731332778930664,
+      "logits/rejected": -2.111682891845703,
+      "logps/chosen": -387.6856994628906,
+      "logps/rejected": -206.78878784179688,
+      "loss": 0.9388,
+      "metrics/advantage_var": 0.07866128534078598,
+      "rewards/accuracies": 0.5300781726837158,
+      "rewards/chosen": 0.012091020122170448,
+      "rewards/margins": 0.003263443009927869,
+      "rewards/rejected": 0.008827576413750648,
+      "stability/repetition_rate_mean": 0.8726905584335327,
+      "stability/response_length_mean": 655.2999877929688,
+      "stability/response_length_std": 282.005126953125,
+      "stability/response_length_var": 79526.8984375,
+      "stability/token_entropy_mean": 6.596875190734863,
+      "step": 60
+    },
+    {
+      "epoch": 0.3326935380678183,
+      "grad_norm": 1.729257583618164,
+      "learning_rate": 6.4e-07,
+      "logits/chosen": -2.251981496810913,
+      "logits/rejected": -2.0987205505371094,
+      "logps/chosen": -377.4482116699219,
+      "logps/rejected": -204.36764526367188,
+      "loss": 0.9428,
+      "metrics/advantage_var": 0.08284494280815125,
+      "rewards/accuracies": 0.5609374642372131,
+      "rewards/chosen": 0.015137546695768833,
+      "rewards/margins": 0.0056471494026482105,
+      "rewards/rejected": 0.009490396827459335,
+      "stability/repetition_rate_mean": 0.9237826466560364,
+      "stability/response_length_mean": 306.3999938964844,
+      "stability/response_length_std": 141.7236785888672,
+      "stability/response_length_var": 20085.599609375,
+      "stability/token_entropy_mean": 6.884375095367432,
+      "step": 65
+    },
+    {
+      "epoch": 0.3582853486884197,
+      "grad_norm": 2.0435707569122314,
+      "learning_rate": 6.9e-07,
+      "logits/chosen": -2.2713375091552734,
+      "logits/rejected": -2.1065561771392822,
+      "logps/chosen": -393.7109375,
+      "logps/rejected": -206.14852905273438,
+      "loss": 0.9333,
+      "metrics/advantage_var": 0.09623891115188599,
+      "rewards/accuracies": 0.5589843988418579,
+      "rewards/chosen": 0.01978578045964241,
+      "rewards/margins": 0.007668572943657637,
+      "rewards/rejected": 0.012117205187678337,
+      "stability/repetition_rate_mean": 0.9703803062438965,
+      "stability/response_length_mean": 472.70001220703125,
+      "stability/response_length_std": 159.89791870117188,
+      "stability/response_length_var": 25567.34375,
+      "stability/token_entropy_mean": 7.21875,
+      "step": 70
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 1.9249153137207031,
+      "learning_rate": 7.4e-07,
+      "logits/chosen": -2.275832414627075,
+      "logits/rejected": -2.1167995929718018,
+      "logps/chosen": -381.0315856933594,
+      "logps/rejected": -207.89634704589844,
+      "loss": 0.9371,
+      "metrics/advantage_var": 0.10018797218799591,
+      "rewards/accuracies": 0.564453125,
+      "rewards/chosen": 0.021552283316850662,
+      "rewards/margins": 0.007498453836888075,
+      "rewards/rejected": 0.014053828082978725,
+      "stability/repetition_rate_mean": 0.9106122851371765,
+      "stability/response_length_mean": 527.0,
+      "stability/response_length_std": 293.7500915527344,
+      "stability/response_length_var": 86289.109375,
+      "stability/token_entropy_mean": 6.084765434265137,
+      "step": 75
+    },
+    {
+      "epoch": 0.4094689699296225,
+      "grad_norm": 1.8595386743545532,
+      "learning_rate": 7.9e-07,
+      "logits/chosen": -2.26131272315979,
+      "logits/rejected": -2.103710174560547,
+      "logps/chosen": -382.3757629394531,
+      "logps/rejected": -208.9224853515625,
+      "loss": 0.9341,
+      "metrics/advantage_var": 0.11523017287254333,
+      "rewards/accuracies": 0.568359375,
+      "rewards/chosen": 0.02573232911527157,
+      "rewards/margins": 0.008440744131803513,
+      "rewards/rejected": 0.017291583120822906,
+      "stability/repetition_rate_mean": 0.9519937634468079,
+      "stability/response_length_mean": 635.5,
+      "stability/response_length_std": 262.53900146484375,
+      "stability/response_length_var": 68926.71875,
+      "stability/token_entropy_mean": 7.215624809265137,
+      "step": 80
+    },
+    {
+      "epoch": 0.4350607805502239,
+      "grad_norm": 1.8945162296295166,
+      "learning_rate": 8.399999999999999e-07,
+      "logits/chosen": -2.258488178253174,
+      "logits/rejected": -2.10145902633667,
+      "logps/chosen": -387.2498779296875,
+      "logps/rejected": -208.5669708251953,
+      "loss": 0.9353,
+      "metrics/advantage_var": 0.155354842543602,
+      "rewards/accuracies": 0.590624988079071,
+      "rewards/chosen": 0.034161973744630814,
+      "rewards/margins": 0.012971531599760056,
+      "rewards/rejected": 0.021190444007515907,
+      "stability/repetition_rate_mean": 0.9050097465515137,
+      "stability/response_length_mean": 600.0999755859375,
+      "stability/response_length_std": 349.2030334472656,
+      "stability/response_length_var": 121942.765625,
+      "stability/token_entropy_mean": 6.73828125,
+      "step": 85
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 1.9289894104003906,
+      "learning_rate": 8.9e-07,
+      "logits/chosen": -2.261462688446045,
+      "logits/rejected": -2.1133339405059814,
+      "logps/chosen": -375.13140869140625,
+      "logps/rejected": -206.13841247558594,
+      "loss": 0.933,
+      "metrics/advantage_var": 0.17288456857204437,
+      "rewards/accuracies": 0.6050781607627869,
+      "rewards/chosen": 0.039468325674533844,
+      "rewards/margins": 0.014231967739760876,
+      "rewards/rejected": 0.025236357003450394,
+      "stability/repetition_rate_mean": 0.9758660197257996,
+      "stability/response_length_mean": 692.9000244140625,
+      "stability/response_length_std": 127.31452941894531,
+      "stability/response_length_var": 16208.9892578125,
+      "stability/token_entropy_mean": 7.115624904632568,
+      "step": 90
+    },
+    {
+      "epoch": 0.48624440179142675,
+      "grad_norm": 1.8516160249710083,
+      "learning_rate": 9.399999999999999e-07,
+      "logits/chosen": -2.2591757774353027,
+      "logits/rejected": -2.1126770973205566,
+      "logps/chosen": -366.7373046875,
+      "logps/rejected": -203.33721923828125,
+      "loss": 0.9297,
+      "metrics/advantage_var": 0.21673114597797394,
+      "rewards/accuracies": 0.62890625,
+      "rewards/chosen": 0.047232918441295624,
+      "rewards/margins": 0.01979774609208107,
+      "rewards/rejected": 0.027435174211859703,
+      "stability/repetition_rate_mean": 0.9478110074996948,
+      "stability/response_length_mean": 727.0,
+      "stability/response_length_std": 335.0764465332031,
+      "stability/response_length_var": 112276.21875,
+      "stability/token_entropy_mean": 7.303124904632568,
+      "step": 95
+    },
+    {
+      "epoch": 0.5118362124120281,
+      "grad_norm": 1.857904076576233,
+      "learning_rate": 9.9e-07,
+      "logits/chosen": -2.275299072265625,
+      "logits/rejected": -2.119931936264038,
+      "logps/chosen": -384.3828430175781,
+      "logps/rejected": -210.50686645507812,
+      "loss": 0.9271,
+      "metrics/advantage_var": 0.3037579357624054,
+      "rewards/accuracies": 0.6441406607627869,
+      "rewards/chosen": 0.05749749764800072,
+      "rewards/margins": 0.02403966337442398,
+      "rewards/rejected": 0.033457837998867035,
+      "stability/repetition_rate_mean": 0.9512196779251099,
+      "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": 1.7671213150024414,
+      "learning_rate": 9.999815554351224e-07,
+      "logits/chosen": -2.2592649459838867,
+      "logits/rejected": -2.108891248703003,
+      "logps/chosen": -377.8589172363281,
+      "logps/rejected": -209.54722595214844,
+      "loss": 0.9294,
+      "metrics/advantage_var": 0.38598760962486267,
+      "rewards/accuracies": 0.6253905892372131,
+      "rewards/chosen": 0.06676852703094482,
+      "rewards/margins": 0.026257099583745003,
+      "rewards/rejected": 0.04051142930984497,
+      "stability/repetition_rate_mean": 0.8934451937675476,
+      "stability/response_length_mean": 669.5,
+      "stability/response_length_std": 271.60888671875,
+      "stability/response_length_var": 73771.390625,
+      "stability/token_entropy_mean": 6.578125,
+      "step": 105
+    },
+    {
+      "epoch": 0.5630198336532309,
+      "grad_norm": 1.689988374710083,
+      "learning_rate": 9.999066267225444e-07,
+      "logits/chosen": -2.275810956954956,
+      "logits/rejected": -2.118212938308716,
+      "logps/chosen": -385.0805358886719,
+      "logps/rejected": -206.55056762695312,
+      "loss": 0.9258,
+      "metrics/advantage_var": 0.4739305377006531,
+      "rewards/accuracies": 0.6488281488418579,
+      "rewards/chosen": 0.07808978110551834,
+      "rewards/margins": 0.030354248359799385,
+      "rewards/rejected": 0.04773552715778351,
+      "stability/repetition_rate_mean": 0.9327242970466614,
+      "stability/response_length_mean": 255.60000610351562,
+      "stability/response_length_std": 114.30874633789062,
+      "stability/response_length_var": 13066.4892578125,
+      "stability/token_entropy_mean": 6.978125095367432,
+      "step": 110
+    },
+    {
+      "epoch": 0.5886116442738324,
+      "grad_norm": 1.8266487121582031,
+      "learning_rate": 9.997740697079592e-07,
+      "logits/chosen": -2.2522411346435547,
+      "logits/rejected": -2.1116788387298584,
+      "logps/chosen": -372.5317687988281,
+      "logps/rejected": -213.750732421875,
+      "loss": 0.9241,
+      "metrics/advantage_var": 0.6695685386657715,
+      "rewards/accuracies": 0.6449218392372131,
+      "rewards/chosen": 0.09085610508918762,
+      "rewards/margins": 0.036989450454711914,
+      "rewards/rejected": 0.05386665090918541,
+      "stability/repetition_rate_mean": 0.8290537595748901,
+      "stability/response_length_mean": 499.29998779296875,
+      "stability/response_length_std": 282.31072998046875,
+      "stability/response_length_var": 79699.34375,
+      "stability/token_entropy_mean": 6.53125,
+      "step": 115
+    },
+    {
+      "epoch": 0.6142034548944337,
+      "grad_norm": 1.8164492845535278,
+      "learning_rate": 9.995838996722914e-07,
+      "logits/chosen": -2.249505043029785,
+      "logits/rejected": -2.104017734527588,
+      "logps/chosen": -367.5647888183594,
+      "logps/rejected": -207.3728790283203,
+      "loss": 0.92,
+      "metrics/advantage_var": 0.8198556303977966,
+      "rewards/accuracies": 0.6578125357627869,
+      "rewards/chosen": 0.1038859412074089,
+      "rewards/margins": 0.04292682185769081,
+      "rewards/rejected": 0.06095912307500839,
+      "stability/repetition_rate_mean": 0.9244946241378784,
+      "stability/response_length_mean": 363.6000061035156,
+      "stability/response_length_std": 146.0701446533203,
+      "stability/response_length_var": 21336.48828125,
+      "stability/token_entropy_mean": 6.881249904632568,
+      "step": 120
+    },
+    {
+      "epoch": 0.6397952655150352,
+      "grad_norm": 1.8048653602600098,
+      "learning_rate": 9.993361385379874e-07,
+      "logits/chosen": -2.253950595855713,
+      "logits/rejected": -2.0959606170654297,
+      "logps/chosen": -385.4555358886719,
+      "logps/rejected": -204.85488891601562,
+      "loss": 0.9179,
+      "metrics/advantage_var": 1.1633703708648682,
+      "rewards/accuracies": 0.666015625,
+      "rewards/chosen": 0.11797237396240234,
+      "rewards/margins": 0.04964681342244148,
+      "rewards/rejected": 0.06832555681467056,
+      "stability/repetition_rate_mean": 0.8939984440803528,
+      "stability/response_length_mean": 433.79998779296875,
+      "stability/response_length_std": 126.92412567138672,
+      "stability/response_length_var": 16109.7333984375,
+      "stability/token_entropy_mean": 6.703125,
+      "step": 125
+    },
+    {
+      "epoch": 0.6653870761356366,
+      "grad_norm": 1.89629328250885,
+      "learning_rate": 9.99030814866488e-07,
+      "logits/chosen": -2.246368169784546,
+      "logits/rejected": -2.097729206085205,
+      "logps/chosen": -375.3144836425781,
+      "logps/rejected": -203.45748901367188,
+      "loss": 0.9137,
+      "metrics/advantage_var": 1.2405811548233032,
+      "rewards/accuracies": 0.663281261920929,
+      "rewards/chosen": 0.12805037200450897,
+      "rewards/margins": 0.05569981038570404,
+      "rewards/rejected": 0.07235056161880493,
+      "stability/repetition_rate_mean": 0.8216736912727356,
+      "stability/response_length_mean": 296.1000061035156,
+      "stability/response_length_std": 167.77593994140625,
+      "stability/response_length_var": 28148.767578125,
+      "stability/token_entropy_mean": 6.053124904632568,
+      "step": 130
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 1.762459397315979,
+      "learning_rate": 9.986679638549358e-07,
+      "logits/chosen": -2.268277645111084,
+      "logits/rejected": -2.108889102935791,
+      "logps/chosen": -387.0592041015625,
+      "logps/rejected": -213.4181365966797,
+      "loss": 0.9111,
+      "metrics/advantage_var": 1.49413001537323,
+      "rewards/accuracies": 0.671093761920929,
+      "rewards/chosen": 0.14280582964420319,
+      "rewards/margins": 0.06248578429222107,
+      "rewards/rejected": 0.08032004535198212,
+      "stability/repetition_rate_mean": 0.8815739750862122,
+      "stability/response_length_mean": 596.4000244140625,
+      "stability/response_length_std": 291.9129333496094,
+      "stability/response_length_var": 85213.15625,
+      "stability/token_entropy_mean": 6.671875,
+      "step": 135
+    },
+    {
+      "epoch": 0.7165706973768394,
+      "grad_norm": 1.8487615585327148,
+      "learning_rate": 9.982476273321175e-07,
+      "logits/chosen": -2.2565722465515137,
+      "logits/rejected": -2.1014785766601562,
+      "logps/chosen": -388.67364501953125,
+      "logps/rejected": -210.8116912841797,
+      "loss": 0.907,
+      "metrics/advantage_var": 1.983569622039795,
+      "rewards/accuracies": 0.6847655773162842,
+      "rewards/chosen": 0.1552000790834427,
+      "rewards/margins": 0.07189314067363739,
+      "rewards/rejected": 0.0833069309592247,
+      "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": 1.7706032991409302,
+      "learning_rate": 9.977698537536417e-07,
+      "logits/chosen": -2.268200635910034,
+      "logits/rejected": -2.1011595726013184,
+      "logps/chosen": -387.3850402832031,
+      "logps/rejected": -203.22686767578125,
+      "loss": 0.9045,
+      "metrics/advantage_var": 2.342012882232666,
+      "rewards/accuracies": 0.6761718392372131,
+      "rewards/chosen": 0.17279884219169617,
+      "rewards/margins": 0.07742971926927567,
+      "rewards/rejected": 0.09536914527416229,
+      "stability/repetition_rate_mean": 0.9656392931938171,
+      "stability/response_length_mean": 720.2000122070312,
+      "stability/response_length_std": 284.2842102050781,
+      "stability/response_length_var": 80817.5078125,
+      "stability/token_entropy_mean": 7.112500190734863,
+      "step": 145
+    },
+    {
+      "epoch": 0.7677543186180422,
+      "grad_norm": 1.8285086154937744,
+      "learning_rate": 9.972346981963546e-07,
+      "logits/chosen": -2.2355778217315674,
+      "logits/rejected": -2.0934510231018066,
+      "logps/chosen": -373.40240478515625,
+      "logps/rejected": -210.79360961914062,
+      "loss": 0.9046,
+      "metrics/advantage_var": 3.491899013519287,
+      "rewards/accuracies": 0.65234375,
+      "rewards/chosen": 0.18629421293735504,
+      "rewards/margins": 0.08447620272636414,
+      "rewards/rejected": 0.1018180176615715,
+      "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": 1.866870641708374,
+      "learning_rate": 9.966422223519885e-07,
+      "logits/chosen": -2.261730670928955,
+      "logits/rejected": -2.1110422611236572,
+      "logps/chosen": -376.8971862792969,
+      "logps/rejected": -205.9225311279297,
+      "loss": 0.8992,
+      "metrics/advantage_var": 3.0832598209381104,
+      "rewards/accuracies": 0.6664062142372131,
+      "rewards/chosen": 0.20672671496868134,
+      "rewards/margins": 0.09230801463127136,
+      "rewards/rejected": 0.11441870778799057,
+      "stability/repetition_rate_mean": 0.8748709559440613,
+      "stability/response_length_mean": 502.3999938964844,
+      "stability/response_length_std": 200.0795440673828,
+      "stability/response_length_var": 40031.8203125,
+      "stability/token_entropy_mean": 6.542187690734863,
+      "step": 155
+    },
+    {
+      "epoch": 0.818937939859245,
+      "grad_norm": 1.7998372316360474,
+      "learning_rate": 9.959924945200523e-07,
+      "logits/chosen": -2.255324602127075,
+      "logits/rejected": -2.0892727375030518,
+      "logps/chosen": -382.9584655761719,
+      "logps/rejected": -206.74269104003906,
+      "loss": 0.8939,
+      "metrics/advantage_var": 4.283844470977783,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": 0.22003380954265594,
+      "rewards/margins": 0.10556791722774506,
+      "rewards/rejected": 0.11446589976549149,
+      "stability/repetition_rate_mean": 0.9658440351486206,
+      "stability/response_length_mean": 870.4000244140625,
+      "stability/response_length_std": 147.6581573486328,
+      "stability/response_length_var": 21802.93359375,
+      "stability/token_entropy_mean": 7.243750095367432,
+      "step": 160
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.6668412685394287,
+      "learning_rate": 9.952855895999568e-07,
+      "logits/chosen": -2.240656614303589,
+      "logits/rejected": -2.093395948410034,
+      "logps/chosen": -361.86102294921875,
+      "logps/rejected": -203.08079528808594,
+      "loss": 0.8928,
+      "metrics/advantage_var": 4.067698001861572,
+      "rewards/accuracies": 0.6761718988418579,
+      "rewards/chosen": 0.22143030166625977,
+      "rewards/margins": 0.10688450187444687,
+      "rewards/rejected": 0.11454581469297409,
+      "stability/repetition_rate_mean": 0.8792880177497864,
+      "stability/response_length_mean": 547.7000122070312,
+      "stability/response_length_std": 249.60325622558594,
+      "stability/response_length_var": 62301.7890625,
+      "stability/token_entropy_mean": 6.625,
+      "step": 165
+    },
+    {
+      "epoch": 0.8701215611004478,
+      "grad_norm": 1.7253059148788452,
+      "learning_rate": 9.9452158908238e-07,
+      "logits/chosen": -2.2658770084381104,
+      "logits/rejected": -2.100456714630127,
+      "logps/chosen": -376.3424377441406,
+      "logps/rejected": -203.1697235107422,
+      "loss": 0.8864,
+      "metrics/advantage_var": 5.073078632354736,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": 0.24319852888584137,
+      "rewards/margins": 0.12121755629777908,
+      "rewards/rejected": 0.12198095768690109,
+      "stability/repetition_rate_mean": 0.9187561869621277,
+      "stability/response_length_mean": 385.70001220703125,
+      "stability/response_length_std": 143.96685791015625,
+      "stability/response_length_var": 20726.455078125,
+      "stability/token_entropy_mean": 6.868750095367432,
+      "step": 170
+    },
+    {
+      "epoch": 0.8957133717210493,
+      "grad_norm": 1.8084509372711182,
+      "learning_rate": 9.937005810398745e-07,
+      "logits/chosen": -2.2514612674713135,
+      "logits/rejected": -2.096510410308838,
+      "logps/chosen": -379.8197021484375,
+      "logps/rejected": -211.5930633544922,
+      "loss": 0.8856,
+      "metrics/advantage_var": 6.549341678619385,
+      "rewards/accuracies": 0.6929687261581421,
+      "rewards/chosen": 0.2569131851196289,
+      "rewards/margins": 0.12991970777511597,
+      "rewards/rejected": 0.12699344754219055,
+      "stability/repetition_rate_mean": 0.9395610690116882,
+      "stability/response_length_mean": 460.6000061035156,
+      "stability/response_length_std": 165.9894256591797,
+      "stability/response_length_var": 27552.48828125,
+      "stability/token_entropy_mean": 6.9375,
+      "step": 175
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 1.7756394147872925,
+      "learning_rate": 9.928226601167138e-07,
+      "logits/chosen": -2.250619411468506,
+      "logits/rejected": -2.0897774696350098,
+      "logps/chosen": -386.3166809082031,
+      "logps/rejected": -212.235107421875,
+      "loss": 0.8842,
+      "metrics/advantage_var": 5.9247918128967285,
+      "rewards/accuracies": 0.6890625357627869,
+      "rewards/chosen": 0.2764429748058319,
+      "rewards/margins": 0.13361024856567383,
+      "rewards/rejected": 0.1428327113389969,
+      "stability/repetition_rate_mean": 0.8915449976921082,
+      "stability/response_length_mean": 716.2000122070312,
+      "stability/response_length_std": 329.3815002441406,
+      "stability/response_length_var": 108492.1796875,
+      "stability/token_entropy_mean": 6.525000095367432,
+      "step": 180
+    },
+    {
+      "epoch": 0.946896992962252,
+      "grad_norm": 1.6765127182006836,
+      "learning_rate": 9.918879275179817e-07,
+      "logits/chosen": -2.252854824066162,
+      "logits/rejected": -2.1023361682891846,
+      "logps/chosen": -373.72027587890625,
+      "logps/rejected": -209.63833618164062,
+      "loss": 0.8819,
+      "metrics/advantage_var": 7.567996978759766,
+      "rewards/accuracies": 0.6832031011581421,
+      "rewards/chosen": 0.2906308174133301,
+      "rewards/margins": 0.1405182033777237,
+      "rewards/rejected": 0.1501125991344452,
+      "stability/repetition_rate_mean": 0.8790209889411926,
+      "stability/response_length_mean": 486.3999938964844,
+      "stability/response_length_std": 213.8624725341797,
+      "stability/response_length_var": 45737.15625,
+      "stability/token_entropy_mean": 6.574999809265137,
+      "step": 185
+    },
+    {
+      "epoch": 0.9724888035828535,
+      "grad_norm": 1.5650509595870972,
+      "learning_rate": 9.90896490997906e-07,
+      "logits/chosen": -2.2370643615722656,
+      "logits/rejected": -2.0819854736328125,
+      "logps/chosen": -376.81488037109375,
+      "logps/rejected": -205.4518280029297,
+      "loss": 0.8751,
+      "metrics/advantage_var": 8.249796867370605,
+      "rewards/accuracies": 0.6980469226837158,
+      "rewards/chosen": 0.30180665850639343,
+      "rewards/margins": 0.1598372757434845,
+      "rewards/rejected": 0.14196938276290894,
+      "stability/repetition_rate_mean": 0.8905978202819824,
+      "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": 1.658704400062561,
+      "learning_rate": 9.898484648474362e-07,
+      "logits/chosen": -2.2535288333892822,
+      "logits/rejected": -2.100834369659424,
+      "logps/chosen": -360.88079833984375,
+      "logps/rejected": -207.50393676757812,
+      "loss": 0.8762,
+      "metrics/advantage_var": 8.585060119628906,
+      "rewards/accuracies": 0.6851562857627869,
+      "rewards/chosen": 0.3115439713001251,
+      "rewards/margins": 0.16069400310516357,
+      "rewards/rejected": 0.15084993839263916,
+      "stability/repetition_rate_mean": 0.7758646011352539,
+      "stability/response_length_mean": 356.3999938964844,
+      "stability/response_length_std": 207.94882202148438,
+      "stability/response_length_var": 43242.7109375,
+      "stability/token_entropy_mean": 5.946875095367432,
+      "step": 195
+    },
+    {
+      "epoch": 1.0,
+      "step": 196,
+      "total_flos": 7.072674331683193e+18,
+      "train_loss": 0.9197707787460211,
+      "train_runtime": 4344.4262,
+      "train_samples_per_second": 23.018,
+      "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_ftx0.1/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47caba11889928f4a482733e328ad85abe61ff25b18bfec1963dc94f55545974
+size 8017