Update README.md
Browse files
README.md
CHANGED
@@ -1,6 +1,7 @@
|
|
1 |
---
|
2 |
language:
|
3 |
- en
|
|
|
4 |
license: apache-2.0
|
5 |
tags:
|
6 |
- text-generation-inference
|
@@ -11,6 +12,47 @@ tags:
|
|
11 |
base_model: unsloth/gemma-7b-bnb-4bit
|
12 |
---
|
13 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
# Uploaded model
|
15 |
|
16 |
- **Developed by:** AhmedBou
|
@@ -19,4 +61,4 @@ base_model: unsloth/gemma-7b-bnb-4bit
|
|
19 |
|
20 |
This gemma model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
|
21 |
|
22 |
-
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|
|
|
1 |
---
|
2 |
language:
|
3 |
- en
|
4 |
+
- ar
|
5 |
license: apache-2.0
|
6 |
tags:
|
7 |
- text-generation-inference
|
|
|
12 |
base_model: unsloth/gemma-7b-bnb-4bit
|
13 |
---
|
14 |
|
15 |
+
## Inference code:
|
16 |
+
Use this python code for inference
|
17 |
+
|
18 |
+
````python
|
19 |
+
# Installs Unsloth, Xformers (Flash Attention) and all other packages!
|
20 |
+
!pip install "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
|
21 |
+
!pip install --no-deps xformers trl peft accelerate bitsandbytes
|
22 |
+
|
23 |
+
from unsloth import FastLanguageModel
|
24 |
+
|
25 |
+
max_seq_length = 2048
|
26 |
+
dtype = None
|
27 |
+
load_in_4bit = True
|
28 |
+
|
29 |
+
model, tokenizer = FastLanguageModel.from_pretrained(
|
30 |
+
model_name = "AhmedBou/Gemma-7b-EngText-ArabicSummary",
|
31 |
+
max_seq_length = max_seq_length,
|
32 |
+
dtype = dtype,
|
33 |
+
load_in_4bit = load_in_4bit,
|
34 |
+
)
|
35 |
+
FastLanguageModel.for_inference(model)
|
36 |
+
|
37 |
+
input = """
|
38 |
+
past a news article here
|
39 |
+
"""
|
40 |
+
|
41 |
+
FastLanguageModel.for_inference(model) # Enable native 2x faster inference
|
42 |
+
inputs = tokenizer(
|
43 |
+
[
|
44 |
+
alpaca_prompt.format(
|
45 |
+
input, # input
|
46 |
+
"", # output - leave this blank for generation!
|
47 |
+
)
|
48 |
+
], return_tensors = "pt").to("cuda")
|
49 |
+
|
50 |
+
outputs = model.generate(**inputs, max_new_tokens = 64, use_cache = True)
|
51 |
+
tokenizer.batch_decode(outputs)
|
52 |
+
|
53 |
+
|
54 |
+
````
|
55 |
+
|
56 |
# Uploaded model
|
57 |
|
58 |
- **Developed by:** AhmedBou
|
|
|
61 |
|
62 |
This gemma model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
|
63 |
|
64 |
+
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|