File size: 3,448 Bytes
7a741dc
 
 
 
785446d
7a741dc
 
 
4ed1f75
7a741dc
 
f22717d
41dc771
f22717d
 
 
7a741dc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6d37e44
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
---
license: apache-2.0
datasets:
- HuggingFaceH4/no_robots
base_model: mistralai/Mistral-7B-v0.1
language:
- en
pipeline_tag: text-generation
thumbnail: https://huggingface.co/mrm8488/mistral-7b-ft-h4-no_robots_instructions/resolve/main/mistralh4-removebg-preview.png?download=true
---

<div style="text-align:center;width:250px;height:250px;">
    <img src="https://huggingface.co/mrm8488/mistral-7b-ft-h4-no_robots_instructions/resolve/main/mistralh4-removebg-preview.png?download=true" alt="limstral logo"">
</div>
<br />

## Mistral 7B fine-tuned on H4/No Robots instructions
This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the [HuggingFaceH4/no_robots](https://huggingface.co/datasets/HuggingFaceH4/no_robots) dataset for instruction following downstream task.

## Training procedure

The model was loaded on **8 bits** and fine-tuned on the LIMA dataset using the **LoRA** PEFT technique with the `huggingface/peft` library and `trl/sft` for one epoch on 1 x A100 (40GB) GPU.

SFT Trainer params:
```
trainer = SFTTrainer(
    model=model,
    train_dataset=train_ds,
    eval_dataset=test_ds,
    peft_config=peft_config,
    dataset_text_field="text",
    max_seq_length=2048,
    tokenizer=tokenizer,
    args=training_arguments,
    packing=False
)
```

LoRA config:
```
config = LoraConfig(
        lora_alpha=16,
        lora_dropout=0.1,
        r=64,
        bias="none",
        task_type="CAUSAL_LM",
        target_modules = ['q_proj', 'k_proj', 'down_proj', 'v_proj', 'o_proj', 'gate_proj', 'up_proj']
    )
```

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 0.0002
- train_batch_size: 2
- eval_batch_size: 8
- seed: 66
- gradient_accumulation_steps: 64
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.03
- num_epochs: 2
- mixed_precision_training: Native AMP

### Training results

| Step | Training Loss | Validation Loss |
|------|---------------|-----------------|
| 10   | 1.796200      | 1.774305        |
| 20   | 1.769700      | 1.679720        |
| 30   | 1.626800      | 1.667754        |
| 40   | 1.663400      | 1.665188        |
| 50   | 1.565700      | 1.659000        |
| 60   | 1.660300      | 1.658270        |




### Usage
```py
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

repo_id = "mrm8488/mistral-7b-ft-h4-no_robots_instructions"

model = AutoModelForCausalLM.from_pretrained(repo_id, torch_dtype=torch.bfloat16)
tokenizer = AutoTokenizer.from_pretrained(repo_id)

gen = pipeline("text-generation", model=model, tokenizer=tokenizer, device=0)

instruction = "[INST] Write an email to say goodbye to me boss [\INST]"
res = gen(instruction, max_new_tokens=512, temperature=0.3, top_p=0.75, top_k=40, repetition_penalty=1.2, eos_token_id=2)
print(res[0]['generated_text'])
```

### Framework versions

- Transformers 4.35.0.dev0
- Pytorch 2.1.0+cu118
- Datasets 2.14.6
- Tokenizers 0.14.1

### Citation
```
@misc {manuel_romero_2023,
	author       = { {Manuel Romero} },
	title        = { mistral-7b-ft-h4-no_robots_instructions (Revision 785446d) },
	year         = 2023,
	url          = { https://huggingface.co/mrm8488/mistral-7b-ft-h4-no_robots_instructions },
	doi          = { 10.57967/hf/1426 },
	publisher    = { Hugging Face }
}
```