GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
azherali
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
azherali
Model Tree
unsloth/Qwen3-8B
Input Modalities
Output Modalities
Supported Functionality
python
from unsloth import FastLanguageModelimport 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 reasoning_start = "<reasoning>"reasoning_end = "</reasoning>"solution_start = "<SOLUTION>"solution_end = "</SOLUTION>" system_prompt = f"""You are given a problem. Think about the problem and provide your working out.Place your reasoning between {reasoning_start} and {reasoning_end}. Then, provide your final solution between{solution_start} and {solution_end}. Always answer in Urdu.""" message = [ {"role": "system", "content": system_prompt}, { "role": "user", "content": "پانچ بچوں نے 20 چاکلیٹس برابر بانٹیں۔ ہر بچے کو کتنی چاکلیٹس ملیں گی؟" }]text = tokenizer.apply_chat_template( message, tokenize=False, add_generation_prompt=True, enable_thinking=False, # Must add for generation) # Inference Using Tranformer Libraryfrom 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),) # Inference Using VLLM(fast)from vllm import SamplingParamsdef generate_answer(problem): message = [ {"role": "system", "content": system_prompt}, {"role": "user", "content": problem}, ] text = tokenizer.apply_chat_template( message, add_generation_prompt = True, # Must add for generation tokenize = False, enable_thinking=False, ) sampling_params = SamplingParams( max_tokens=2048, temperature=0.6, top_p=0.95, top_k=20, ) output = model.fast_generate( text, sampling_params = sampling_params, lora_request =None, )[0].outputs[0].text return output generate_answer("پانچ بچوں نے 20 چاکلیٹس برابر بانٹیں۔ ہر بچے کو کتنی چاکلیٹس ملیں گی؟")This model was trained with SFT.
from unsloth import FastLanguageModelimport 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 reasoning_start = "<reasoning>"reasoning_end = "</reasoning>"solution_start = "<SOLUTION>"solution_end = "</SOLUTION>" system_prompt = f"""You are given a problem. Think about the problem and provide your working out.Place your reasoning between {reasoning_start} and {reasoning_end}. Then, provide your final solution between{solution_start} and {solution_end}. Always answer in Urdu.""" message = [ {"role": "system", "content": system_prompt}, { "role": "user", "content": "پانچ بچوں نے 20 چاکلیٹس برابر بانٹیں۔ ہر بچے کو کتنی چاکلیٹس ملیں گی؟" }]text = tokenizer.apply_chat_template( message, tokenize=False, add_generation_prompt=True, enable_thinking=False, # Must add for generation) # Inference Using Tranformer Libraryfrom 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),) # Inference Using VLLM(fast)from vllm import SamplingParamsdef generate_answer(problem): message = [ {"role": "system", "content": system_prompt}, {"role": "user", "content": problem}, ] text = tokenizer.apply_chat_template( message, add_generation_prompt = True, # Must add for generation tokenize = False, enable_thinking=False, ) sampling_params = SamplingParams( max_tokens=2048, temperature=0.6, top_p=0.95, top_k=20, ) output = model.fast_generate( text, sampling_params = sampling_params, lora_request =None, )[0].outputs[0].text return output generate_answer("پانچ بچوں نے 20 چاکلیٹس برابر بانٹیں۔ ہر بچے کو کتنی چاکلیٹس ملیں گی؟")