beamaia commited on
Commit
f2ae4b8
1 Parent(s): 0dd025c

Upload folder using huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +57 -41
README.md CHANGED
@@ -1,70 +1,86 @@
1
  ---
2
  license: mit
3
- library_name: peft
4
  tags:
5
- - trl
6
- - dpo
7
- - generated_from_trainer
8
- base_model: HuggingFaceH4/zephyr-7b-beta
9
  model-index:
10
- - name: WeniGPT-DPO-test
11
  results: []
 
12
  ---
13
 
14
- <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
- should probably proofread and complete it, then remove this comment. -->
16
 
17
- # WeniGPT-DPO-test
18
 
19
- This model is a fine-tuned version of [HuggingFaceH4/zephyr-7b-beta](https://huggingface.co/HuggingFaceH4/zephyr-7b-beta) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
- - Loss: 0.6931
22
- - Rewards/chosen: 0.0
23
- - Rewards/rejected: 0.0
24
- - Rewards/accuracies: 0.0
25
- - Rewards/margins: 0.0
26
- - Logps/rejected: -12.7530
27
- - Logps/chosen: -7.2418
28
- - Logits/rejected: -2.2130
29
- - Logits/chosen: -2.2118
30
 
31
- ## Model description
32
 
33
- More information needed
34
 
35
- ## Intended uses & limitations
36
 
37
- More information needed
38
 
39
- ## Training and evaluation data
 
 
 
 
 
 
40
 
41
- More information needed
 
 
 
 
 
 
42
 
43
- ## Training procedure
44
 
45
  ### Training hyperparameters
46
 
47
  The following hyperparameters were used during training:
48
  - learning_rate: 2e-05
49
- - train_batch_size: 2
50
- - eval_batch_size: 2
51
- - seed: 42
52
  - gradient_accumulation_steps: 2
 
53
  - total_train_batch_size: 4
54
- - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
55
- - lr_scheduler_type: linear
56
- - lr_scheduler_warmup_ratio: 0.1
57
- - training_steps: 1
58
- - mixed_precision_training: Native AMP
59
 
60
  ### Training results
61
 
62
-
63
-
64
  ### Framework versions
65
 
66
- - PEFT 0.8.2
67
- - Transformers 4.38.2
68
- - Pytorch 2.1.0+cu118
69
- - Datasets 2.17.1
70
- - Tokenizers 0.15.1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: mit
3
+ library_name: "trl"
4
  tags:
5
+ - DPO
6
+ - WeniGPT
7
+ base_model: Weni/WeniGPT-2.3.3-Zephyr-7B-zephyr-prompt-LLM_Base_2.0.3_SFT_reduction_variation
 
8
  model-index:
9
+ - name: Weni/WeniGPT-DPO-test
10
  results: []
11
+ language: ['pt']
12
  ---
13
 
14
+ # Weni/WeniGPT-DPO-test
 
15
 
16
+ This model is a fine-tuned version of [Weni/WeniGPT-2.3.3-Zephyr-7B-zephyr-prompt-LLM_Base_2.0.3_SFT_reduction_variation] on the dataset Weni/WeniGPT-QA-1.0.1_DPO with the DPO trainer. It is part of the WeniGPT project for [Weni](https://weni.ai/).
17
 
 
18
  It achieves the following results on the evaluation set:
19
+ {'eval_loss': 0.6931471228599548, 'eval_runtime': 214.1053, 'eval_samples_per_second': 1.485, 'eval_steps_per_second': 0.743, 'eval_rewards/chosen': 0.0, 'eval_rewards/rejected': 0.0, 'eval_rewards/accuracies': 0.0, 'eval_rewards/margins': 0.0, 'eval_logps/rejected': -12.752974510192871, 'eval_logps/chosen': -7.2417826652526855, 'eval_logits/rejected': -2.2130377292633057, 'eval_logits/chosen': -2.2118029594421387, 'epoch': 0.0}
 
 
 
 
 
 
 
 
20
 
21
+ ## Intended uses & limitations
22
 
23
+ This model has not been trained to avoid specific intructions.
24
 
25
+ ## Training procedure
26
 
27
+ Finetuning was done on the model Weni/WeniGPT-2.3.3-Zephyr-7B-zephyr-prompt-LLM_Base_2.0.3_SFT_reduction_variation with the following prompt:
28
 
29
+ ```
30
+ +---------------------
31
+ Pt:
32
+ ### Instruction:
33
+ 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.
34
+ </s>### Input:
35
+ TEXTO: {context}
36
 
37
+ PERGUNTA: {question}
38
+ </s>
39
+ ### Response:
40
+ RESPOSTA: {chosen_response}</s>
41
+ ### Response:
42
+ RESPOSTA: {rejected_response}</s>
43
+ ---------------------
44
 
45
+ ```
46
 
47
  ### Training hyperparameters
48
 
49
  The following hyperparameters were used during training:
50
  - learning_rate: 2e-05
51
+ - per_device_train_batch_size: 2
52
+ - per_device_eval_batch_size: 2
 
53
  - gradient_accumulation_steps: 2
54
+ - num_gpus: 1
55
  - total_train_batch_size: 4
56
+ - optimizer: AdamW
57
+ - lr_scheduler_type: cosine
58
+ - num_steps: 1
59
+ - quantization_type: bitsandbytes
60
+ - LoRA: ("\n - bits: 4\n - use_exllama: True\n - device_map: auto\n - use_cache: False\n - lora_r: 8\n - lora_alpha: 16\n - lora_dropout: 0.1\n - bias: none\n - target_modules: ['q_proj', 'k_proj', 'v_proj', 'o_proj']\n - task_type: CAUSAL_LM",)
61
 
62
  ### Training results
63
 
 
 
64
  ### Framework versions
65
 
66
+ - transformers==4.38.2
67
+ - datasets==2.17.1
68
+ - peft==0.8.2
69
+ - safetensors==0.4.2
70
+ - evaluate==0.4.1
71
+ - bitsandbytes==0.42
72
+ - huggingface_hub==0.20.3
73
+ - seqeval==1.2.2
74
+ - optimum==1.17.1
75
+ - auto-gptq==0.7.0
76
+ - gpustat==1.1.1
77
+ - deepspeed==0.13.2
78
+ - wandb==0.16.3
79
+ - trl==0.7.11
80
+ - accelerate==0.27.2
81
+ - coloredlogs==15.0.1
82
+ - traitlets==5.14.1
83
+ - autoawq@https://github.com/casper-hansen/AutoAWQ/releases/download/v0.2.0/autoawq-0.2.0+cu118-cp310-cp310-linux_x86_64.whl
84
+
85
+ ### Hardware
86
+ - Cloud provided: runpod.io