95 lines
2.4 KiB
Markdown
95 lines
2.4 KiB
Markdown
---
|
|
base_model:
|
|
- unsloth/Qwen3-8B
|
|
- azherali/Riazi-8B-CPT
|
|
tags:
|
|
- text-generation-inference
|
|
- transformers
|
|
- unsloth
|
|
- qwen3
|
|
- urdu
|
|
- urdu-reasoning
|
|
license: apache-2.0
|
|
language:
|
|
- ur
|
|
widget:
|
|
- messages:
|
|
- role: user
|
|
content: "2 + 3 کتنا ہوتا ہے؟"
|
|
|
|
- messages:
|
|
- role: user
|
|
content: "اگر آپ کے پاس 10 سیب ہیں اور 3 کھا لیتے ہیں، تو کتنے بچیں گے؟"
|
|
|
|
- messages:
|
|
- role: user
|
|
content: "ایک باپ کی عمر اپنے بیٹے کی عمر سے تین گنا ہے۔ 10 سال بعد باپ کی عمر 50 سال ہوگی۔ بیٹے کی موجودہ عمر کیا ہے؟"
|
|
|
|
- messages:
|
|
- role: user
|
|
content: "اگر ایک گاڑی 60 کلومیٹر فی گھنٹہ کی رفتار سے 2 گھنٹے چلے، تو کتنی دوری طے کرے گی؟"
|
|
---
|
|
|
|
|
|
|
|
|
|
## Quick start
|
|
|
|
```python
|
|
from unsloth import FastLanguageModel
|
|
import torch
|
|
|
|
max_seq_length = 2048 # Choose any! We auto support RoPE Scaling internally!
|
|
dtype = (
|
|
None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
|
|
)
|
|
load_in_4bit = False # Use 4bit quantization to reduce memory usage. Can be False.
|
|
load_in_8bit = False # Use 8bit quantization to reduce memory usage. Can be False.
|
|
|
|
model, tokenizer = FastLanguageModel.from_pretrained(
|
|
model_name="azherali/Riazi-8B", # Choose ANY
|
|
max_seq_length=max_seq_length,
|
|
dtype=dtype,
|
|
load_in_4bit=load_in_4bit,
|
|
load_in_8bit=load_in_8bit,
|
|
# token = "YOUR_HF_TOKEN", # HF Token for gated models
|
|
)
|
|
FastLanguageModel.for_inference(model) # Enable native 2x faster inference
|
|
|
|
messages = [
|
|
{
|
|
"role": "user",
|
|
"content": "پانچ بچوں نے 20 چاکلیٹس برابر بانٹیں۔ ہر بچے کو کتنی چاکلیٹس ملیں گی؟",
|
|
}
|
|
]
|
|
text = tokenizer.apply_chat_template(
|
|
messages,
|
|
tokenize=False,
|
|
add_generation_prompt=True, # Must add for generation
|
|
)
|
|
|
|
from transformers import TextStreamer
|
|
|
|
_ = model.generate(
|
|
**tokenizer(text, return_tensors="pt").to("cuda"),
|
|
temperature=0.6,
|
|
top_p=0.95,
|
|
top_k=20, # For non thinking
|
|
streamer=TextStreamer(tokenizer, skip_prompt=True),
|
|
)
|
|
```
|
|
|
|
## Training procedure
|
|
|
|
|
|
|
|
|
|
This model was trained with SFT.
|
|
|
|
### Framework versions
|
|
|
|
- TRL: 0.22.2
|
|
- Transformers: 4.56.2
|
|
- Pytorch: 2.12.0+rocm7.2
|
|
- Datasets: 4.3.0
|
|
- Tokenizers: 0.22.2 |