File size: 19,057 Bytes
764a0fb
 
 
0f20421
764a0fb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
---
base_model: nicholasKluge/TeenyTinyLlama-160m
co2_eq_emissions:
  emissions: 5600
  geographical_location: Germany
  hardware_used: NVIDIA A100-SXM4-40GB
  source: CodeCarbon
  training_type: pre-training
datasets:
- nicholasKluge/Pt-Corpus-Instruct
inference: true
language:
- pt
library_name: transformers
license: apache-2.0
metrics:
- perplexity
model-index:
- name: TeenyTinyLlama-160m
  results:
  - dataset:
      args:
        num_few_shot: 3
      name: ENEM Challenge (No Images)
      split: train
      type: eduagarcia/enem_challenge
    metrics:
    - name: accuracy
      type: acc
      value: 19.24
    source:
      name: Open Portuguese LLM Leaderboard
      url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=nicholasKluge/TeenyTinyLlama-160m
    task:
      name: Text Generation
      type: text-generation
  - dataset:
      args:
        num_few_shot: 3
      name: BLUEX (No Images)
      split: train
      type: eduagarcia-temp/BLUEX_without_images
    metrics:
    - name: accuracy
      type: acc
      value: 23.09
    source:
      name: Open Portuguese LLM Leaderboard
      url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=nicholasKluge/TeenyTinyLlama-160m
    task:
      name: Text Generation
      type: text-generation
  - dataset:
      args:
        num_few_shot: 3
      name: OAB Exams
      split: train
      type: eduagarcia/oab_exams
    metrics:
    - name: accuracy
      type: acc
      value: 22.37
    source:
      name: Open Portuguese LLM Leaderboard
      url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=nicholasKluge/TeenyTinyLlama-160m
    task:
      name: Text Generation
      type: text-generation
  - dataset:
      args:
        num_few_shot: 15
      name: Assin2 RTE
      split: test
      type: assin2
    metrics:
    - name: f1-macro
      type: f1_macro
      value: 53.97
    source:
      name: Open Portuguese LLM Leaderboard
      url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=nicholasKluge/TeenyTinyLlama-160m
    task:
      name: Text Generation
      type: text-generation
  - dataset:
      args:
        num_few_shot: 15
      name: Assin2 STS
      split: test
      type: eduagarcia/portuguese_benchmark
    metrics:
    - name: pearson
      type: pearson
      value: 0.24
    source:
      name: Open Portuguese LLM Leaderboard
      url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=nicholasKluge/TeenyTinyLlama-160m
    task:
      name: Text Generation
      type: text-generation
  - dataset:
      args:
        num_few_shot: 15
      name: FaQuAD NLI
      split: test
      type: ruanchaves/faquad-nli
    metrics:
    - name: f1-macro
      type: f1_macro
      value: 43.97
    source:
      name: Open Portuguese LLM Leaderboard
      url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=nicholasKluge/TeenyTinyLlama-160m
    task:
      name: Text Generation
      type: text-generation
  - dataset:
      args:
        num_few_shot: 25
      name: HateBR Binary
      split: test
      type: ruanchaves/hatebr
    metrics:
    - name: f1-macro
      type: f1_macro
      value: 36.92
    source:
      name: Open Portuguese LLM Leaderboard
      url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=nicholasKluge/TeenyTinyLlama-160m
    task:
      name: Text Generation
      type: text-generation
  - dataset:
      args:
        num_few_shot: 25
      name: PT Hate Speech Binary
      split: test
      type: hate_speech_portuguese
    metrics:
    - name: f1-macro
      type: f1_macro
      value: 42.63
    source:
      name: Open Portuguese LLM Leaderboard
      url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=nicholasKluge/TeenyTinyLlama-160m
    task:
      name: Text Generation
      type: text-generation
  - dataset:
      args:
        num_few_shot: 25
      name: tweetSentBR
      split: test
      type: eduagarcia-temp/tweetsentbr
    metrics:
    - name: f1-macro
      type: f1_macro
      value: 11.39
    source:
      name: Open Portuguese LLM Leaderboard
      url: https://huggingface.co/spaces/eduagarcia/open_pt_llm_leaderboard?query=nicholasKluge/TeenyTinyLlama-160m
    task:
      name: Text Generation
      type: text-generation
model_creator: nicholasKluge
model_name: TeenyTinyLlama-160m
pipeline_tag: text-generation
quantized_by: afrideva
tags:
- text-generation-inference
- gguf
- ggml
- quantized
widget:
- example_title: Exemplo
  text: 'A PUCRS é uma universidade '
- example_title: Exemplo
  text: A muitos anos atrás, em uma galáxia muito distante, vivia uma raça de
- example_title: Exemplo
  text: Em meio a um escândalo, a frente parlamentar pediu ao Senador Silva para
---

# TeenyTinyLlama-160m-GGUF

Quantized GGUF model files for [TeenyTinyLlama-160m](https://huggingface.co/nicholasKluge/TeenyTinyLlama-160m) from [nicholasKluge](https://huggingface.co/nicholasKluge)

## Original Model Card:

# TeenyTinyLlama-160m

<img src="./logo.png" alt="A little llama wearing a mushroom hat and a monocle." height="200">

## Model Summary

Large language models (LLMs) have significantly advanced natural language processing, but their progress has yet to be equal across languages. While most LLMs are trained in high-resource languages like English, multilingual models generally underperform monolingual ones. Additionally, aspects of their multilingual foundation sometimes restrict the byproducts they produce, like computational demands and licensing regimes. Hence, we developed the _TeenyTinyLlama_ pair: two compact models for Brazilian Portuguese text generation.

Read our preprint on [ArXiv](https://arxiv.org/abs/2401.16640).

## Details

- **Architecture:** a Transformer-based model pre-trained via causal language modeling
- **Size:** 162,417,408 parameters
- **Context length:** 2048 tokens
- **Dataset:** [Pt-Corpus Instruct](https://huggingface.co/datasets/nicholasKluge/Pt-Corpus-Instruct) (6.2B tokens)
- **Language:** Portuguese
- **Number of steps:** 458,000 
- **GPU:** 1 NVIDIA A100-SXM4-40GB
- **Training time**: ~ 36 hours
- **Emissions:** 5.6 KgCO2 (Germany)
- **Total energy consumption:** 15.5 kWh

This repository has the [source code](https://github.com/Nkluge-correa/TeenyTinyLlama) used to train this model. The main libraries used are:

- [Transformers](https://github.com/huggingface/transformers)
- [PyTorch](https://github.com/pytorch/pytorch)
- [Datasets](https://github.com/huggingface/datasets)
- [Tokenizers](https://github.com/huggingface/tokenizers)
- [Sentencepiece](https://github.com/google/sentencepiece)
- [Accelerate](https://github.com/huggingface/accelerate)
- [FlashAttention](https://github.com/Dao-AILab/flash-attention)
- [Codecarbon](https://github.com/mlco2/codecarbon)

## Intended Uses

The primary intended use of TeenyTinyLlama is to research the challenges related to developing language models for low-resource languages. Checkpoints saved during training are intended to provide a controlled setting for performing scientific experiments. You may also further fine-tune and adapt TeenyTinyLlama for deployment, as long as your use is following the Apache 2.0 license. If you decide to use pre-trained TeenyTinyLlama as a basis for your fine-tuned model, please conduct your own risk and bias assessment.

## Out-of-scope Use

TeenyTinyLlama is not intended for deployment. It is not a product and should not be used for human-facing interactions.

TeenyTinyLlama models are Brazilian Portuguese language only and are not suitable for translation or generating text in other languages.

TeenyTinyLlama has not been fine-tuned for downstream contexts in which language models are commonly deployed.

## Basic usage

Using the `pipeline`:

```python
from transformers import pipeline

generator = pipeline("text-generation", model="nicholasKluge/TeenyTinyLlama-160m")

completions  = generator("Astronomia é a ciência", num_return_sequences=2, max_new_tokens=100)

for comp in completions:
  print(f"🤖 {comp['generated_text']}")
```

Using the `AutoTokenizer` and `AutoModelForCausalLM`:

```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# Load model and the tokenizer
tokenizer = AutoTokenizer.from_pretrained("nicholasKluge/TeenyTinyLlama-160m", revision='main')
model = AutoModelForCausalLM.from_pretrained("nicholasKluge/TeenyTinyLlama-160m", revision='main')

# Pass the model to your device
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

model.eval()
model.to(device)

# Tokenize the inputs and pass them to the device
inputs = tokenizer("Astronomia é a ciência", return_tensors="pt").to(device)

# Generate some text
completions = model.generate(**inputs, num_return_sequences=2, max_new_tokens=100)

# Print the generated text
for i, completion in enumerate(completions):
    print(f'🤖 {tokenizer.decode(completion)}')
```

## Limitations

Like almost all other language models trained on large text datasets scraped from the web, the TTL pair exhibited behavior that does not make them an out-of-the-box solution to many real-world applications, especially those requiring factual, reliable, nontoxic text generation. Our models are all subject to the following:

- **Hallucinations:** This model can produce content that can be mistaken for truth but is, in fact, misleading or entirely false, i.e., hallucination.

- **Biases and Toxicity:** This model inherits the social and historical stereotypes from the data used to train it. Given these biases, the model can produce toxic content, i.e., harmful, offensive, or detrimental to individuals, groups, or communities.

- **Unreliable Code:** The model may produce incorrect code snippets and statements. These code generations should not be treated as suggestions or accurate solutions.

- **Language Limitations:** The model is primarily designed to understand standard Brazilian Portuguese. Other languages might challenge its comprehension, leading to potential misinterpretations or errors in response.

- **Repetition and Verbosity:** The model may get stuck on repetition loops (especially if the repetition penalty during generations is set to a meager value) or produce verbose responses unrelated to the prompt it was given.

Hence, even though our models are released with a permissive license, we urge users to perform their risk analysis on these models if intending to use them for real-world applications and also have humans moderating the outputs of these models in applications where they will interact with an audience, guaranteeing users are always aware they are interacting with a language model.

## Evaluations

During our training runs, both models showed consistent convergence. At no point did our evaluation curves show signs of overfitting or saturation. In the case of our 460m parameter model, we intentionally trained past the optimal point by approximately 75,000 steps to assess if there were any signs of saturation, but our evaluations consistently gave better results. We hypothesize that our models are under-trained but can improve if further trained to pass the Chinchilla optimal range.

| Processed Tokens | Perplexity | Energy Consumption (kWh)  | Emissions (KgCO2eq)  |
|------------------|------------|---------------------------|----------------------|
| 8.1M             | 20.49      | 9.40                      | 3.34                 |
| 1.6B             | 16.90      | 18.82                     | 6.70                 |
| 2.4B             | 15.43      | 28.59                     | 10.16                |
| 3.2B             | 14.64      | 38.20                     | 13.57                |
| 4.0B             | 14.08      | 48.04                     | 17.07                |
| 4.9B             | 13.61      | 57.74                     | 20.52                |
| 5.7B             | 13.25      | 67.32                     | 23.92                |
| 6.5B             | 12.87      | 76.84                     | 27.30                |
| 7.3B             | 12.57      | 86.40                     | 30.70                |
| 8.1B             | 12.27      | 96.19                     | 34.18                |
| 9.0B             | 11.96      | 106.06                    | 37.70                |
| 9.8B             | 11.77      | 115.69                    | 41.31                |

## Benchmarks

Evaluations on benchmarks were performed using the [Language Model Evaluation Harness](https://github.com/EleutherAI/lm-evaluation-harness) (by [EleutherAI](https://www.eleuther.ai/)). [Laiviet](https://github.com/laiviet/lm-evaluation-harness) translated the tasks from the LM-Evaluation-Harness we used. The results of models marked with an "*" were extracted from the [Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard).

|                  | **ARC**   | **HellaSwag** | **MMLU**  | **TruthfulQA** | **Average** |
|------------------|-----------|---------------|-----------|----------------|-------------|
| Pythia-410m      | 24.83*    | 41.29*        | 25.99*    | 40.95*         | 33.26       |
| **TTL-460m**     | 29.40     | 33.00         | 28.55     | 41.10          | 33.01       |
| Bloom-560m       | 24.74*    | 37.15*        | 24.22*    | 42.44*         | 32.13       |
| Xglm-564M        | 25.56     | 34.64*        | 25.18*    | 42.53          | 31.97       |
| OPT-350m         | 23.55*    | 36.73*        | 26.02*    | 40.83*         | 31.78       |
| **TTL-160m**     | 26.15     | 29.29         | 28.11     | 41.12          | 31.16       |
| Pythia-160m      | 24.06*    | 31.39*        | 24.86*    | 44.34*         | 31.16       |
| OPT-125m         | 22.87*    | 31.47*        | 26.02*    | 42.87*         | 30.80       |
| GPorTuguese-2    | 22.48     | 29.62         | 27.36     | 41.44          | 30.22       |
| Gpt2-small       | 21.48*    | 31.60*        | 25.79*    | 40.65*         | 29.97       |
| Multilingual GPT | 23.81     | 26.37*        | 25.17*    | 39.62          | 28.73       |

Evaluations on Brazilian Portuguese benchmarks were performed using a [Portuguese implementation of the EleutherAI LM Evaluation Harness](https://github.com/eduagarcia/lm-evaluation-harness-pt) (created by [Eduardo Garcia](https://github.com/eduagarcia/lm-evaluation-harness-pt)).

|                | **ASSIN2 RTE** | **ASSIN2 STS** | **BLUEX** | **ENEM** | **FAQUAD NLI** | **HateBR** | **OAB Exams** | **Average** |
|----------------|----------------|----------------|-----------|----------|----------------|------------|---------------|-------------|
| Qwen-1.8B      | 64.83          | 19.53          | 26.15     | 30.23    | 43.97          | 33.33      | 27.20         | 35.03       |
| TinyLlama-1.1B | 58.93          | 13.57          | 22.81     | 22.25    | 43.97          | 36.92      | 23.64         | 31.72       |
| **TTL-460m**   | 53.93          | 12.66          | 22.81     | 19.87    | 49.01          | 33.59      | 27.06         | 31.27       |
| XGLM-564m      | 49.61          | 22.91          | 19.61     | 19.38    | 43.97          | 33.99      | 23.42         | 30.41       |
| Bloom-1b7      | 53.60          | 4.81           | 21.42     | 18.96    | 43.97          | 34.89      | 23.05         | 28.67       |
| **TTL-160m**   | 53.36          | 2.58           | 21.84     | 18.75    | 43.97          | 36.88      | 22.60         | 28.56       |
| OPT-125m       | 39.77          | 2.00           | 21.84     | 17.42    | 43.97          | 47.04      | 22.78         | 27.83       |
| Pythia-160     | 33.33          | 12.81          | 16.13     | 16.66    | 50.36          | 41.09      | 22.82         | 27.60       |
| OLMo-1b        | 34.12          | 9.28           | 18.92     | 20.29    | 43.97          | 41.33      | 22.96         | 27.26       |
| Bloom-560m     | 33.33          | 8.48           | 18.92     | 19.03    | 43.97          | 37.07      | 23.05         | 26.26       |
| Pythia-410m    | 33.33          | 4.80           | 19.47     | 19.45    | 43.97          | 33.33      | 23.01         | 25.33       |
| OPT-350m       | 33.33          | 3.65           | 20.72     | 17.35    | 44.71          | 33.33      | 23.01         | 25.15       |
| GPT-2 small    | 33.26          | 0.00           | 10.43     | 11.20    | 43.52          | 33.68      | 13.12         | 20.74       |
| GPorTuguese    | 33.33          | 3.85           | 14.74     | 3.01     | 28.81          | 33.33      | 21.23         | 19.75       |
| Samba-1.1B     | 33.33          | 1.30           | 8.07      | 10.22    | 17.72          | 35.79      | 15.03         | 17.35       |

## Fine-Tuning Comparisons

To further evaluate the downstream capabilities of our models, we decided to employ a basic fine-tuning procedure for our TTL pair on a subset of tasks from the Poeta benchmark. We apply the same procedure for comparison purposes on both [BERTimbau](https://huggingface.co/neuralmind/bert-base-portuguese-cased) models, given that they are also LLM trained from scratch in Brazilian Portuguese and have a similar size range to our models. We used these comparisons to assess if our pre-training runs produced LLM capable of producing good results ("good" here means "close to BERTimbau") when utilized for downstream applications.

| Models          | IMDB      | FaQuAD-NLI | HateBr    | Assin2    | AgNews    | Average |
|-----------------|-----------|------------|-----------|-----------|-----------|---------|
| BERTimbau-large | **93.58** | 92.26      | 91.57     | **88.97** | 94.11     | 92.10   |
| BERTimbau-small | 92.22     | **93.07**  | 91.28     | 87.45     | 94.19     | 91.64   |
| **TTL-460m**    | 91.64     | 91.18      | **92.28** | 86.43     | **94.42** | 91.19   |
| **TTL-160m**    | 91.14     | 90.00      | 90.71     | 85.78     | 94.05     | 90.34   |

All the shown results are the higher accuracy scores achieved on the respective task test sets after fine-tuning the models on the training sets. All fine-tuning runs used the same hyperparameters, and the code implementation can be found in the [model cards](https://huggingface.co/nicholasKluge/TeenyTinyLlama-460m-HateBR) of our fine-tuned models.

## Cite as 🤗

```latex

@misc{correa24ttllama,
  title = {TeenyTinyLlama: open-source tiny language models trained in Brazilian Portuguese},
  author = {Corr{\^e}a, Nicholas Kluge and Falk, Sophia and Fatimah, Shiza and Sen, Aniket and De Oliveira, Nythamar},
  journal={arXiv preprint arXiv:2401.16640},
  year={2024}
}

```

## Funding

This repository was built as part of the RAIES ([Rede de Inteligência Artificial Ética e Segura](https://www.raies.org/)) initiative, a project supported by FAPERGS - ([Fundação de Amparo à Pesquisa do Estado do Rio Grande do Sul](https://fapergs.rs.gov.br/inicial)), Brazil.

## License

TeenyTinyLlama-160m is licensed under the Apache License, Version 2.0. See the [LICENSE](LICENSE) file for more details.