|
--- |
|
license: llama3 |
|
tags: |
|
- moe |
|
- merge |
|
- llama-3 |
|
language: |
|
- en |
|
- tr |
|
pipeline_tag: text-generation |
|
library_name: transformers |
|
--- |
|
|
|
|
|
## 💻 For English |
|
Megatron_llama3_2x8B is a Mixure of Experts (MoE) (two llama3 models) |
|
|
|
|
|
```python |
|
!pip install -qU transformers bitsandbytes accelerate |
|
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
import torch |
|
|
|
model_id = "Eurdem/Megatron_llama3_2x8B" |
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model_id) |
|
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16, device_map="auto", load_in_8bit= True) |
|
|
|
messages = [ |
|
{"role": "system", "content": "You are a helpful chatbot who always responds friendly."}, |
|
{"role": "user", "content": "f(x)=3x^2+4x+12 so what is f(3)?"}, |
|
] |
|
|
|
input_ids = tokenizer.apply_chat_template(messages, return_tensors="pt").to("cuda") |
|
|
|
outputs = model.generate(input_ids, |
|
max_new_tokens=1024, |
|
do_sample=True, |
|
temperature=0.7, |
|
top_p=0.7, |
|
top_k=500, |
|
eos_token_id = tokenizer.eos_token_id |
|
) |
|
response = outputs[0][input_ids.shape[-1]:] |
|
print(tokenizer.decode(response, skip_special_tokens=True)) |
|
``` |
|
|
|
# Megatron_llama3_2x8B |
|
|
|
|
|
|
|
## 💻 Türkçe İçin |
|
|
|
```python |
|
!pip install -qU transformers bitsandbytes accelerate |
|
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
import torch |
|
|
|
model_id = "Eurdem/Megatron_llama3_2x8B" |
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model_id) |
|
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16, device_map="auto", load_in_4bit= True) |
|
|
|
messages = [ |
|
{"role": "system", "content": "Sen Defne isimli Türkçe konuşan bir chatbotsun."}, |
|
{"role": "user", "content": "Sana 2 sorum var. 1) Sen kimsin? 2)f(x)=3x^2+4x+12 ise f(3) kaçtır?"} |
|
] |
|
|
|
input_ids = tokenizer.apply_chat_template(messages, return_tensors="pt").to("cuda") |
|
|
|
outputs = model.generate(input_ids, |
|
max_new_tokens=1024, |
|
do_sample=True, |
|
temperature=0.7, |
|
top_p=0.7, |
|
top_k=500, |
|
eos_token_id = tokenizer.eos_token_id |
|
) |
|
response = outputs[0][input_ids.shape[-1]:] |
|
print(tokenizer.decode(response, skip_special_tokens=True)) |
|
``` |
|
|
|
### Çıktı |
|
```Merhaba! Ben Sen Defne, Türkçe konuşan bir chatbotum. Hizmetinizdeyim. |
|
|
|
Sorunuzun 2. kısmı için, f(x) = 3x^2 + 4x + 12 formülünü ele alalım. f(3)'ün hesabını yapalım: |
|
|
|
f(3) = 3(3)^2 + 4(3) + 12 |
|
= 3(9) + 12 + 12 |
|
= 27 + 24 |
|
= 51 |
|
|
|
Bu nedenle, f(3) 51'dir.``` |
|
|
