File size: 3,680 Bytes
6c0108e
 
 
 
 
 
 
0b7962a
6c0108e
 
 
71549ea
6c0108e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
967d4b7
 
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
---
language:
- or
---

# Model Card for Model ID

[![License: CC BY-NC-SA 4.0](https://img.shields.io/badge/License-CC_BY--NC--SA_4.0-lightgrey.svg)](https://creativecommons.org/licenses/by-nc-sa/4.0/)

## Model description

odiagenAI-model-v0 is based on Llama-7b and finetuned with 52k Odia translated data from the open-source Stanford-Alpaca, resulting in good Odia instruction understanding and response generation capabilities. 

The code of Odia data generation and other detailed information can be found in our Github project repository: https://github.com/shantipriyap/OdiaGenAI.
This repo contains a low-rank adapter for LLaMA-7b fit on the Stanford Alpaca dataset.



## Training hyper-parameters
| Parameter | Value |
| ------ | ------ |
| Batch size | 128 |
| Learning rate | 3e-4 |
| Epochs | 2 |
|Cutoff length | 256 |
|Weight_decay | 0.001 |
|Warmup_rate | 0.1 |
|LR_scheduler | linear |
|Lora r | 16 |
|Lora target modules | (q_proj, k_proj, v_proj, o_proj) |




Model can be easily loaded with AutoModelForCausalLM.
``` python

import torch
from peft import PeftModel
import transformers

assert (
    "LlamaTokenizer" in transformers._import_structure["models.llama"]
), "LLaMA is now in HuggingFace's main branch.\nPlease reinstall it: pip uninstall transformers && pip install git+https://github.com/huggingface/transformers.git"
from transformers import LlamaTokenizer, LlamaForCausalLM, GenerationConfig

tokenizer = LlamaTokenizer.from_pretrained("decapoda-research/llama-7b-hf")

BASE_MODEL = "decapoda-research/llama-7b-hf"
LORA_WEIGHTS = "OdiaGenAI/odiagenAI-model-v0"


model = LlamaForCausalLM.from_pretrained(
        BASE_MODEL,
        load_in_8bit=False,
        torch_dtype=torch.float16,
        device_map="auto",
    )
model = PeftModel.from_pretrained(
        model, LORA_WEIGHTS, torch_dtype=torch.float16, force_download=True
    )

def generate_prompt(instruction, input=None):
    if input:
        return f"""ନିମ୍ନରେ ଏକ ନିର୍ଦ୍ଦେଶନାମା ଯାହାକି ଏକ କାର୍ଯ୍ୟକୁ ବର୍ଣ୍ଣନା କରେ, ଏକ ଇନପୁଟ୍ ସହିତ ଯୋଡି ଯାହା ପରବର୍ତ୍ତୀ ପ୍ରସଙ୍ଗ ପ୍ରଦାନ କରେ | ଏକ ପ୍ରତିକ୍ରିୟା ଲେଖନ୍ତୁ ଯାହା ଅନୁରୋଧକୁ ସଠିକ୍ ଭାବରେ ସମାପ୍ତ କରେ |
### ନିର୍ଦ୍ଦେଶ:
{instruction}
### ଇନପୁଟ୍:
{input}
### ପ୍ରତିକ୍ରିୟା:"""
    else:
        return f"""ନିମ୍ନରେ ଏକ ନିର୍ଦ୍ଦେଶ ଯାହାକି ଏକ କାର୍ଯ୍ୟକୁ ବର୍ଣ୍ଣନା କରେ | ଏକ ପ୍ରତିକ୍ରିୟା ଲେଖନ୍ତୁ ଯାହା ଅନୁରୋଧକୁ ସଠିକ୍ ଭାବରେ ସମାପ୍ତ କରେ |
### ନିର୍ଦ୍ଦେଶ:
{instruction}
### ପ୍ରତିକ୍ରିୟା:"""

prompt = generate_prompt(instruction, input)
inputs = tokenizer(prompt, return_tensors="pt")
input_ids = inputs["input_ids"].to(device)
generation_config = GenerationConfig(
    temperature=0.1,
    top_p=0.75,
    top_k=40,
    num_beams=4,
    **kwargs,
)
with torch.no_grad():
    generation_output = model.generate(
        input_ids=input_ids,
        generation_config=generation_config,
        return_dict_in_generate=True,
        output_scores=True,
        max_new_tokens=128,
    )
s = generation_output.sequences[0]
output = tokenizer.decode(s)
print(output.split("### Response:")[1].strip())


```


Instructions for running it can be found at https://github.com/shantipriyap/OdiaGenAI.