Model save

README.md

@@ -0,0 +1,75 @@
+---
+license: apache-2.0
+library_name: peft
+tags:
+- trl
+- dpo
+- generated_from_trainer
+base_model: alignment-handbook/zephyr-7b-sft-full
+model-index:
+- name: nash_dpo_merge_iter_5
+  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. -->
+
+# nash_dpo_merge_iter_5
+
+This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.5499
+- Rewards/chosen: -0.7963
+- Rewards/rejected: -1.3580
+- Rewards/accuracies: 0.7260
+- Rewards/margins: 0.5617
+- Logps/rejected: -406.1575
+- Logps/chosen: -371.4769
+- Logits/rejected: -0.1370
+- Logits/chosen: -0.7194
+
+## 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: 5e-06
+- train_batch_size: 2
+- eval_batch_size: 2
+- seed: 42
+- distributed_type: multi-GPU
+- num_devices: 4
+- gradient_accumulation_steps: 16
+- total_train_batch_size: 128
+- total_eval_batch_size: 8
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_ratio: 0.1
+- num_epochs: 1
+
+### Training results
+
+| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
+|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
+| 0.5681        | 0.51  | 100  | 0.5499          | -0.7963        | -1.3580          | 0.7260             | 0.5617          | -406.1575      | -371.4769    | -0.1370         | -0.7194       |
+
+
+### Framework versions
+
+- PEFT 0.7.1
+- Transformers 4.36.2
+- Pytorch 2.1.2+cu121
+- Datasets 2.14.6
+- Tokenizers 0.15.2

adapter_config.json

@@ -0,0 +1,31 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "alignment-handbook/zephyr-7b-sft-full",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "up_proj",
+    "gate_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM"
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2669fb4137094bb8cfb22b84ad4337963df2fc7d3b784d0f3c66d4866b2a0c0c
+size 335605144

all_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "train_loss": 0.5790139436721802,
+    "train_runtime": 8829.5489,
+    "train_samples": 25000,
+    "train_samples_per_second": 2.831,
+    "train_steps_per_second": 0.022
+}

special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "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": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -0,0 +1,43 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "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>",
+  "chat_template": "{% for message in messages %}\n{% if message['role'] == 'user' %}\n{{ '<|user|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'system' %}\n{{ '<|system|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'assistant' %}\n{{ '<|assistant|>\n'  + message['content'] + eos_token }}\n{% endif %}\n{% if loop.last and add_generation_prompt %}\n{{ '<|assistant|>' }}\n{% endif %}\n{% endfor %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": true,
+  "model_max_length": 2048,
+  "pad_token": "</s>",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "train_loss": 0.5790139436721802,
+    "train_runtime": 8829.5489,
+    "train_samples": 25000,
+    "train_samples_per_second": 2.831,
+    "train_steps_per_second": 0.022
+}

trainer_state.json

@@ -0,0 +1,326 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9984,
+  "eval_steps": 100,
+  "global_step": 195,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5000000000000004e-07,
+      "logits/chosen": -2.380741834640503,
+      "logits/rejected": -2.4268531799316406,
+      "logps/chosen": -248.77545166015625,
+      "logps/rejected": -290.06207275390625,
+      "loss": 0.6931,
+      "rewards/accuracies": 0.0,
+      "rewards/chosen": 0.0,
+      "rewards/margins": 0.0,
+      "rewards/rejected": 0.0,
+      "step": 1
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.5e-06,
+      "logits/chosen": -2.4806137084960938,
+      "logits/rejected": -2.3910207748413086,
+      "logps/chosen": -312.885986328125,
+      "logps/rejected": -292.342041015625,
+      "loss": 0.6903,
+      "rewards/accuracies": 0.5138888955116272,
+      "rewards/chosen": -0.009276159107685089,
+      "rewards/margins": 0.005926927551627159,
+      "rewards/rejected": -0.015203084796667099,
+      "step": 10
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 5e-06,
+      "logits/chosen": -2.246102809906006,
+      "logits/rejected": -2.1860694885253906,
+      "logps/chosen": -315.72320556640625,
+      "logps/rejected": -313.44061279296875,
+      "loss": 0.6611,
+      "rewards/accuracies": 0.6499999761581421,
+      "rewards/chosen": -0.13601551949977875,
+      "rewards/margins": 0.06950711458921432,
+      "rewards/rejected": -0.20552262663841248,
+      "step": 20
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.959823971496575e-06,
+      "logits/chosen": -1.6343454122543335,
+      "logits/rejected": -1.3590177297592163,
+      "logps/chosen": -325.590087890625,
+      "logps/rejected": -323.31927490234375,
+      "loss": 0.6178,
+      "rewards/accuracies": 0.65625,
+      "rewards/chosen": -0.2257387638092041,
+      "rewards/margins": 0.23529870808124542,
+      "rewards/rejected": -0.46103745698928833,
+      "step": 30
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.8405871765993435e-06,
+      "logits/chosen": -1.110471487045288,
+      "logits/rejected": -0.7200212478637695,
+      "logps/chosen": -354.31494140625,
+      "logps/rejected": -372.0474853515625,
+      "loss": 0.5912,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.489678293466568,
+      "rewards/margins": 0.3759506046772003,
+      "rewards/rejected": -0.8656288385391235,
+      "step": 40
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.646121984004666e-06,
+      "logits/chosen": -0.8880270719528198,
+      "logits/rejected": -0.7102452516555786,
+      "logps/chosen": -333.6441345214844,
+      "logps/rejected": -384.5443115234375,
+      "loss": 0.5748,
+      "rewards/accuracies": 0.690625011920929,
+      "rewards/chosen": -0.38810840249061584,
+      "rewards/margins": 0.47071677446365356,
+      "rewards/rejected": -0.8588252067565918,
+      "step": 50
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.382678665009028e-06,
+      "logits/chosen": -0.7485916614532471,
+      "logits/rejected": -0.3218249976634979,
+      "logps/chosen": -340.6280822753906,
+      "logps/rejected": -402.73248291015625,
+      "loss": 0.575,
+      "rewards/accuracies": 0.6812499761581421,
+      "rewards/chosen": -0.5559459924697876,
+      "rewards/margins": 0.50255286693573,
+      "rewards/rejected": -1.058498740196228,
+      "step": 60
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.058724504646834e-06,
+      "logits/chosen": -0.5164636969566345,
+      "logits/rejected": -0.4610891342163086,
+      "logps/chosen": -332.63995361328125,
+      "logps/rejected": -399.91632080078125,
+      "loss": 0.5625,
+      "rewards/accuracies": 0.6937500238418579,
+      "rewards/chosen": -0.5889195203781128,
+      "rewards/margins": 0.5268474221229553,
+      "rewards/rejected": -1.1157668828964233,
+      "step": 70
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.684671656182497e-06,
+      "logits/chosen": -0.8454337120056152,
+      "logits/rejected": -0.3965529501438141,
+      "logps/chosen": -324.0966796875,
+      "logps/rejected": -363.40777587890625,
+      "loss": 0.5678,
+      "rewards/accuracies": 0.703125,
+      "rewards/chosen": -0.5458763837814331,
+      "rewards/margins": 0.5120958685874939,
+      "rewards/rejected": -1.0579721927642822,
+      "step": 80
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.272542485937369e-06,
+      "logits/chosen": -1.001415491104126,
+      "logits/rejected": -0.700812041759491,
+      "logps/chosen": -294.38006591796875,
+      "logps/rejected": -334.1767883300781,
+      "loss": 0.5911,
+      "rewards/accuracies": 0.6968749761581421,
+      "rewards/chosen": -0.2868848145008087,
+      "rewards/margins": 0.4894043803215027,
+      "rewards/rejected": -0.7762892246246338,
+      "step": 90
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.835583164544139e-06,
+      "logits/chosen": -0.6290949583053589,
+      "logits/rejected": -0.2771238684654236,
+      "logps/chosen": -365.6756896972656,
+      "logps/rejected": -423.69659423828125,
+      "loss": 0.5681,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": -0.7756337523460388,
+      "rewards/margins": 0.49269890785217285,
+      "rewards/rejected": -1.268332600593567,
+      "step": 100
+    },
+    {
+      "epoch": 0.51,
+      "eval_logits/chosen": -0.7193729281425476,
+      "eval_logits/rejected": -0.1369708925485611,
+      "eval_logps/chosen": -371.4768981933594,
+      "eval_logps/rejected": -406.15753173828125,
+      "eval_loss": 0.5499268770217896,
+      "eval_rewards/accuracies": 0.7260000109672546,
+      "eval_rewards/chosen": -0.7962849736213684,
+      "eval_rewards/margins": 0.5616858005523682,
+      "eval_rewards/rejected": -1.3579708337783813,
+      "eval_runtime": 384.3714,
+      "eval_samples_per_second": 5.203,
+      "eval_steps_per_second": 0.65,
+      "step": 100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.3878379241237136e-06,
+      "logits/chosen": -0.5928317904472351,
+      "logits/rejected": -0.17278830707073212,
+      "logps/chosen": -362.83734130859375,
+      "logps/rejected": -395.8005065917969,
+      "loss": 0.5682,
+      "rewards/accuracies": 0.671875,
+      "rewards/chosen": -0.7755601406097412,
+      "rewards/margins": 0.4718254506587982,
+      "rewards/rejected": -1.2473855018615723,
+      "step": 110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9436976651092143e-06,
+      "logits/chosen": -0.7886568307876587,
+      "logits/rejected": -0.368556410074234,
+      "logps/chosen": -358.6850891113281,
+      "logps/rejected": -397.3220520019531,
+      "loss": 0.5481,
+      "rewards/accuracies": 0.7281249761581421,
+      "rewards/chosen": -0.6172876358032227,
+      "rewards/margins": 0.5609332919120789,
+      "rewards/rejected": -1.1782209873199463,
+      "step": 120
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5174374208651913e-06,
+      "logits/chosen": -0.7793182134628296,
+      "logits/rejected": -0.35960158705711365,
+      "logps/chosen": -349.0416564941406,
+      "logps/rejected": -410.9454040527344,
+      "loss": 0.5318,
+      "rewards/accuracies": 0.784375011920929,
+      "rewards/chosen": -0.5557482838630676,
+      "rewards/margins": 0.6911678314208984,
+      "rewards/rejected": -1.2469160556793213,
+      "step": 130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.122757546369744e-06,
+      "logits/chosen": -1.0120689868927002,
+      "logits/rejected": -0.4049127697944641,
+      "logps/chosen": -386.8887634277344,
+      "logps/rejected": -410.48797607421875,
+      "loss": 0.557,
+      "rewards/accuracies": 0.659375011920929,
+      "rewards/chosen": -0.5914907455444336,
+      "rewards/margins": 0.5310963988304138,
+      "rewards/rejected": -1.1225870847702026,
+      "step": 140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 7.723433775328385e-07,
+      "logits/chosen": -0.9131206274032593,
+      "logits/rejected": -0.45857247710227966,
+      "logps/chosen": -347.942626953125,
+      "logps/rejected": -417.690185546875,
+      "loss": 0.5635,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": -0.5880693197250366,
+      "rewards/margins": 0.5640393495559692,
+      "rewards/rejected": -1.1521086692810059,
+      "step": 150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.774575140626317e-07,
+      "logits/chosen": -0.7125102877616882,
+      "logits/rejected": -0.3716251850128174,
+      "logps/chosen": -357.5932312011719,
+      "logps/rejected": -408.4806823730469,
+      "loss": 0.5606,
+      "rewards/accuracies": 0.71875,
+      "rewards/chosen": -0.7405123710632324,
+      "rewards/margins": 0.5258091688156128,
+      "rewards/rejected": -1.2663214206695557,
+      "step": 160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.4757783024395244e-07,
+      "logits/chosen": -0.6120478510856628,
+      "logits/rejected": -0.32458120584487915,
+      "logps/chosen": -353.44525146484375,
+      "logps/rejected": -399.55194091796875,
+      "loss": 0.5724,
+      "rewards/accuracies": 0.6875,
+      "rewards/chosen": -0.6805830597877502,
+      "rewards/margins": 0.5673653483390808,
+      "rewards/rejected": -1.247948408126831,
+      "step": 170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.00928482603669e-08,
+      "logits/chosen": -0.9088006019592285,
+      "logits/rejected": -0.3657810091972351,
+      "logps/chosen": -385.3566589355469,
+      "logps/rejected": -408.7562561035156,
+      "loss": 0.545,
+      "rewards/accuracies": 0.734375,
+      "rewards/chosen": -0.6916705965995789,
+      "rewards/margins": 0.5680912137031555,
+      "rewards/rejected": -1.2597618103027344,
+      "step": 180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.006426501190233e-08,
+      "logits/chosen": -0.7876042723655701,
+      "logits/rejected": -0.3451666235923767,
+      "logps/chosen": -363.17401123046875,
+      "logps/rejected": -411.7339782714844,
+      "loss": 0.5615,
+      "rewards/accuracies": 0.715624988079071,
+      "rewards/chosen": -0.7461715936660767,
+      "rewards/margins": 0.5213161706924438,
+      "rewards/rejected": -1.26748788356781,
+      "step": 190
+    },
+    {
+      "epoch": 1.0,
+      "step": 195,
+      "total_flos": 0.0,
+      "train_loss": 0.5790139436721802,
+      "train_runtime": 8829.5489,
+      "train_samples_per_second": 2.831,
+      "train_steps_per_second": 0.022
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 195,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 100,
+  "total_flos": 0.0,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41f018254f48ec5e2352f1ac3be40af42f84c846a978d8aad7ff2e6fcc504552
+size 4920