Model save

README.md

@@ -1,90 +1,85 @@
 ---
 license: mit
-library_name: "trl"
+library_name: peft
 tags:
-- KTO
-- WeniGPT
+- trl
+- kto
+- generated_from_trainer
 base_model: HuggingFaceH4/zephyr-7b-beta
 model-index:
-- name: Weni/WeniGPT-QA-Zephyr-7B-4.0.2-KTO
+- name: WeniGPT-QA-Zephyr-7B-4.0.2-KTO
   results: []
-language: ['pt']
 ---
 
-# Weni/WeniGPT-QA-Zephyr-7B-4.0.2-KTO
+<!-- 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. -->
 
-This model is a fine-tuned version of [HuggingFaceH4/zephyr-7b-beta] on the dataset Weni/WeniGPT-QA-Binarized-1.2.0 with the KTO trainer. It is part of the WeniGPT project for [Weni](https://weni.ai/).
+# WeniGPT-QA-Zephyr-7B-4.0.2-KTO
 
+This model is a fine-tuned version of [HuggingFaceH4/zephyr-7b-beta](https://huggingface.co/HuggingFaceH4/zephyr-7b-beta) on the None dataset.
 It achieves the following results on the evaluation set:
-{'eval_loss': 0.01734423078596592, 'eval_runtime': 33.4092, 'eval_samples_per_second': 14.008, 'eval_steps_per_second': 0.898, 'epoch': 2.91}
+- Loss: 0.0052
+- Rewards/chosen: 6.8557
+- Rewards/rejected: -40.2341
+- Rewards/margins: 47.0897
+- Kl: 0.0
+- Logps/chosen: -138.8257
+- Logps/rejected: -603.3300
 
-## Intended uses & limitations
-
-This model has not been trained to avoid specific intructions. 
-
-## Training procedure
-
-Finetuning was done on the model HuggingFaceH4/zephyr-7b-beta with the following prompt:
-
-```
----------------------
-Question:
-<|system|>
-Você é um médico tratando um paciente com amnésia. Para responder as perguntas do paciente, você irá ler um texto anteriormente para se contextualizar. Se você trouxer informações desconhecidas, fora do texto lido, poderá deixar o paciente confuso. Se o paciente fizer uma questão sobre informações não presentes no texto, você precisa responder de forma educada que você não tem informação suficiente para responder, pois se tentar responder, pode trazer informações que não ajudarão o paciente recuperar sua memória. Lembre, se não estiver no texto, você precisa responder de forma educada que você não tem informação suficiente para responder. Precisamos ajudar o paciente.
-<|user|>
-Contexto: {context}
-
-Questão: {question}</s>
-<|assistant|>
+## Model description
 
+More information needed
 
+## Intended uses & limitations
 
----------------------
-Response:
-{response}</s>
+More information needed
 
+## Training and evaluation data
 
----------------------
+More information needed
 
-```
+## Training procedure
 
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
 - learning_rate: 0.0002
-- per_device_train_batch_size: 4
-- per_device_eval_batch_size: 2
+- train_batch_size: 4
+- eval_batch_size: 4
+- seed: 42
 - gradient_accumulation_steps: 8
-- num_gpus: 8
-- total_train_batch_size: 256
-- optimizer: AdamW
-- lr_scheduler_type: cosine
-- num_steps: 48
-- quantization_type: bitsandbytes
-- LoRA: ("\n  - bits: 4\n  - use_exllama: True\n  - device_map: auto\n  - use_cache: False\n  - lora_r: 16\n  - lora_alpha: 32\n  - lora_dropout: 0.05\n  - bias: none\n  - target_modules: ['q_proj', 'k_proj', 'v_proj', 'o_proj']\n  - task_type: CAUSAL_LM",)
+- total_train_batch_size: 32
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: linear
+- lr_scheduler_warmup_ratio: 0.03
+- training_steps: 786
+- mixed_precision_training: Native AMP
 
 ### Training results
 
+| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/margins | Kl  | Logps/chosen | Logps/rejected |
+|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:---------------:|:---:|:------------:|:--------------:|
+| 0.1911        | 0.38  | 50   | 0.0198          | 5.0250         | -23.0117         | 28.0367         | 0.0 | -157.1321    | -431.1063      |
+| 0.0525        | 0.76  | 100  | 0.0155          | 5.6300         | -30.1143         | 35.7443         | 0.0 | -151.0826    | -502.1325      |
+| 0.0684        | 1.14  | 150  | 0.0104          | 6.1959         | -32.9347         | 39.1306         | 0.0 | -145.4235    | -530.3358      |
+| 0.0253        | 1.52  | 200  | 0.0126          | 5.9530         | -28.8204         | 34.7734         | 0.0 | -147.8525    | -489.1933      |
+| 0.0177        | 1.9   | 250  | 0.0163          | 6.4453         | -53.9510         | 60.3963         | 0.0 | -142.9294    | -740.4993      |
+| 0.0126        | 2.28  | 300  | 0.0206          | 6.2002         | -65.0484         | 71.2486         | 0.0 | -145.3807    | -851.4736      |
+| 0.007         | 2.66  | 350  | 0.0068          | 6.6872         | -41.5169         | 48.2041         | 0.0 | -140.5102    | -616.1578      |
+| 0.0085        | 3.04  | 400  | 0.0076          | 6.7677         | -52.6223         | 59.3899         | 0.0 | -139.7057    | -727.2117      |
+| 0.0057        | 3.43  | 450  | 0.0060          | 6.5419         | -38.1888         | 44.7308         | 0.0 | -141.9630    | -582.8775      |
+| 0.0029        | 3.81  | 500  | 0.0061          | 6.7190         | -37.0418         | 43.7608         | 0.0 | -140.1928    | -571.4076      |
+| 0.003         | 4.19  | 550  | 0.0062          | 6.8462         | -36.6584         | 43.5045         | 0.0 | -138.9207    | -567.5727      |
+| 0.0023        | 4.57  | 600  | 0.0058          | 6.8631         | -38.2151         | 45.0782         | 0.0 | -138.7514    | -583.1401      |
+| 0.0028        | 4.95  | 650  | 0.0050          | 6.8261         | -40.4010         | 47.2271         | 0.0 | -139.1210    | -604.9990      |
+| 0.0029        | 5.33  | 700  | 0.0052          | 6.8557         | -40.2341         | 47.0897         | 0.0 | -138.8257    | -603.3300      |
+| 0.0022        | 5.71  | 750  | 0.0052          | 6.8628         | -40.4610         | 47.3239         | 0.0 | -138.7541    | -605.5995      |
+
+
 ### Framework versions
 
-- transformers==4.39.1
-- datasets==2.18.0
-- peft==0.10.0
-- safetensors==0.4.2
-- evaluate==0.4.1
-- bitsandbytes==0.43
-- huggingface_hub==0.20.3
-- seqeval==1.2.2
-- optimum==1.17.1
-- auto-gptq==0.7.1
-- gpustat==1.1.1
-- deepspeed==0.14.0
-- wandb==0.16.3
-- trl==0.8.1
-- accelerate==0.28.0
-- coloredlogs==15.0.1
-- traitlets==5.14.1
-- autoawq@https://github.com/casper-hansen/AutoAWQ/releases/download/v0.2.0/autoawq-0.2.0+cu118-cp310-cp310-linux_x86_64.whl
-
-### Hardware
-- Cloud provided: runpod.io
+- PEFT 0.10.0
+- Transformers 4.39.1
+- Pytorch 2.1.0+cu118
+- Datasets 2.18.0
+- Tokenizers 0.15.2