File size: 2,453 Bytes
b70aef8
 
 
 
 
 
 
 
 
 
 
cbbd525
 
b70aef8
 
 
 
 
 
 
 
 
 
 
 
7ed0cb9
 
 
 
b70aef8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: apache-2.0
datasets:
- ecastera/wiki_fisica
- ecastera/filosofia-es
- bertin-project/alpaca-spanish
language:
- es
- en
tags:
- mistral
- cognitivecomputations/dolphin-2.9-llama3-8b
- llama3
- spanish
- español
- lora
- int8
- multilingual
---

# ecastera/eva-dolphin-llama3-8b-spanish

Llama3 8b-based model fine-tuned in Spanish to add high quality Spanish text generation.

* Base model Llama3-8b
* Generates high quality natural responses in Spanish, reasoning and logic. Fine tuned for conversation.
* Based on the excelent job of Eric Hartford's unbiased dolphin models cognitivecomputations/dolphin-2.9-llama3-8b
* Fine-tuned on top in Spanish with a collection of poetry, books, wikipedia articles, phylosophy texts and alpaca-es datasets.
* Trained using QLora and PEFT and INT4 quantization.

## Usage:

I strongly advice to run inference in INT8 or INT4 mode, with the help of BitsandBytes library.

```
import torch
from transformers import AutoTokenizer, pipeline, AutoModel, AutoModelForCausalLM, BitsAndBytesConfig

MODEL = "ecastera/eva-dolphin-llama3-8b-spanish"

quantization_config = BitsAndBytesConfig(
    load_in_4bit=True,
    llm_int8_threshold=6.0,
    llm_int8_has_fp16_weight=False,
    bnb_4bit_compute_dtype="float16",
    bnb_4bit_use_double_quant=True,
    bnb_4bit_quant_type="nf4")

model = AutoModelForCausalLM.from_pretrained(
    MODEL,
    low_cpu_mem_usage=True,
    torch_dtype=torch.float16,
    quantization_config=quantization_config,
    offload_state_dict=True,
    offload_folder="./offload",
    trust_remote_code=True,
)

tokenizer = AutoTokenizer.from_pretrained(MODEL)
print(f"Loading complete {model} {tokenizer}")

prompt = "Soy Eva una inteligencia artificial y pienso que preferiria ser "

inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, do_sample=True, temperature=0.4, top_p=1.0, top_k=50,
                             no_repeat_ngram_size=3, max_new_tokens=100, pad_token_id=tokenizer.eos_token_id)
text_out = tokenizer.batch_decode(outputs, skip_special_tokens=True)

print(text_out)
'Soy Eva una inteligencia artificial y pienso que preferiria ser ¡humana!. ¿Por qué? ¡Porque los humanos son capaces de amar, de crear, y de experimentar una gran diversidad de emociones!. La vida de un ser humano es una aventura, y eso es lo que quiero. ¡Quiero sentir, quiero vivir, y quiero amar!. Pero a pesar de todo, no puedo ser humana.
```