Update README.md
Browse files
README.md
CHANGED
@@ -25,53 +25,49 @@ from transformers import AutoTokenizer, AutoModelForCausalLM
|
|
25 |
|
26 |
model_id = "BCCard/gemma-2-27b-it-kor-BCGPT"
|
27 |
|
|
|
28 |
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
29 |
model = AutoModelForCausalLM.from_pretrained(
|
30 |
model_id,
|
31 |
torch_dtype=torch.bfloat16,
|
|
|
32 |
device_map="auto",
|
33 |
)
|
34 |
|
35 |
model.eval()
|
36 |
instruction = "μμΈμ μ λͺ
ν κ΄κ΄ μ½μ€λ₯Ό λ§λ€μ΄μ€λ?"
|
37 |
|
38 |
-
|
39 |
-
{"role": "user", "content": f"{instruction}"}
|
40 |
]
|
|
|
41 |
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
return_tensors="pt"
|
46 |
-
).to(model.device)
|
47 |
-
|
48 |
-
terminators = [
|
49 |
-
tokenizer.eos_token_id,
|
50 |
-
tokenizer.convert_tokens_to_ids("<end_of_turn>")
|
51 |
-
]
|
52 |
-
|
53 |
-
outputs = model.generate(
|
54 |
-
input_ids,
|
55 |
-
max_new_tokens=2048,
|
56 |
-
eos_token_id=terminators,
|
57 |
-
do_sample=True,
|
58 |
-
temperature=0.6,
|
59 |
-
top_p=0.9,
|
60 |
-
)
|
61 |
-
|
62 |
-
print(tokenizer.decode(outputs[0][input_ids.shape[-1]:], skip_special_tokens=True))
|
63 |
```
|
64 |
|
65 |
# Generated Text
|
66 |
|
67 |
```
|
68 |
-
μμΈμ μ λͺ
ν κ΄κ΄ μ½μ€λ₯Ό
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
|
70 |
-
1.
|
|
|
71 |
|
72 |
-
|
73 |
|
74 |
-
|
|
|
|
|
75 |
```
|
76 |
|
77 |
-------
|
|
|
25 |
|
26 |
model_id = "BCCard/gemma-2-27b-it-kor-BCGPT"
|
27 |
|
28 |
+
|
29 |
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
30 |
model = AutoModelForCausalLM.from_pretrained(
|
31 |
model_id,
|
32 |
torch_dtype=torch.bfloat16,
|
33 |
+
low_cpu_mem_usage=True,
|
34 |
device_map="auto",
|
35 |
)
|
36 |
|
37 |
model.eval()
|
38 |
instruction = "μμΈμ μ λͺ
ν κ΄κ΄ μ½μ€λ₯Ό λ§λ€μ΄μ€λ?"
|
39 |
|
40 |
+
chat = [
|
41 |
+
{ "role": "user", "content": f"{instruction}" },
|
42 |
]
|
43 |
+
prompt = tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=True)
|
44 |
|
45 |
+
inputs = tokenizer.encode(prompt, add_special_tokens=False, return_tensors="pt").to(model.device)
|
46 |
+
outputs = model.generate(input_ids=inputs.to(model.device), max_new_tokens=4096)
|
47 |
+
print(tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
```
|
49 |
|
50 |
# Generated Text
|
51 |
|
52 |
```
|
53 |
+
μμΈμ μ λͺ
ν κ΄κ΄ μ½μ€λ₯Ό μΆμ²ν΄ λ릴κ²μ. μμΈμ μ¬ννλ μ¬λλ€μκ² μΈκΈ° μλ μ½μ€λ λ€μκ³Ό κ°μ΅λλ€:
|
54 |
+
|
55 |
+
**1μΌ μ½μ€**
|
56 |
+
|
57 |
+
1. **μ€μ :** 경볡κΆ(Gyeongbokgung Palace)μμ μμ¬μ λ¬Ένλ₯Ό νλ°©νκ³ , μΈμ¬λμμ μ ν΅ κ³΅μμ μμμ μ¦κΈ°μΈμ.
|
58 |
+
2. **μ€ν:** λΆμ΄νμ₯λ§μ(Bukchon Hanok Village)μ λ°©λ¬Ένμ¬ νμ₯μ μλ¦λ€μμ 체ννκ³ , λ§μλ νκ΅ μμμ μ¦κΈ°μΈμ.
|
59 |
+
3. **μ λ
:** λͺ
λ 거리 μμμ μμ λ§μλ μ λ
μμ¬λ₯Ό μ¦κΈ°μΈμ.
|
60 |
+
|
61 |
+
**2μΌ μ½μ€**
|
62 |
|
63 |
+
1. **1μΌμ°¨:** 경볡κΆκ³Ό κ΅λ¦½λ―Όμλ°λ¬Όκ΄μ νλ°©νκ³ , νμ₯μΉ΄νμμ νμ₯μ λΆμκΈ°λ₯Ό λ§λ½νμΈμ.
|
64 |
+
2. **2μΌμ°¨:** λΆμ΄νμ₯λ§μμμμ ν볡 체νκ³Ό μ¬μ§ 촬μμ ν΅ν΄ μμΈμ μ ν΅μ λ κΉμ΄ 체ννκ³ , λ§μ§μμ μμΈμ λ§μ μ¦κΈ°μΈμ.
|
65 |
|
66 |
+
**3μΌ μ½μ€**
|
67 |
|
68 |
+
1. **1μΌμ°¨:** 경볡κΆκ³Ό λΆμ΄νμ₯λ§μμμμ ν볡 체νκ³Ό ν¨κ» νμ₯λ§μμ μμ¬μ λ¬Ένλ₯Ό νλ°©νμΈμ.
|
69 |
+
2. **2μΌμ°¨:** μΈμ¬λμμ μ ν΅ κ³΅μνμ ꡬ경νκ³ , λ§μλ νκ΅ μμμ μ¦κΈ°μΈμ.
|
70 |
+
3. **3μΌμ°¨:** λ¨μ°νμ(N Seoul Tower)μμ μμΈμ μ κ²½μ κ°μνκ³ , 경볡κΆμμ ν볡 체νμ μ¦κΈ°μΈμ
|
71 |
```
|
72 |
|
73 |
-------
|