GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
g4me
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
g4me
Model Tree
Qwen/Qwen3-4B-Base
Input Modalities
Output Modalities
Supported Functionality
python
from transformers import AutoModelForCausalLM, AutoTokenizerimport torch model_id = "g4me/QWiki-4B-Base-LR1e5"tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained( model_id, torch_dtype=torch.bfloat16, device_map="auto",) prompt = "Era uma vez, em um reino distante,"inputs = tokenizer(prompt, return_tensors="pt").to(model.device)outputs = model.generate( **inputs, max_new_tokens=128, do_sample=True, temperature=0.7, top_p=0.9,)print(tokenizer.decode(outputs[0], skip_special_tokens=True))| Parameter | Value |
|---|---|
| Base model | Qwen/Qwen3-4B-Base |
| Training data | g4me/wikipedia-pt-br-corpus-2026-trainval |
| Sequence length | 4096 |
| Micro batch size | 16 |
| Global batch size | 512 |
| Max steps | 2174 |
| Peak learning rate | 1.0e-5 |
| Min learning rate | 1.0e-6 |
| LR scheduler | Cosine |
| Optimizer | AdamW (β1=0.9, β2=0.95, wd=0.1) |
| Precision | BF16 + FP8 hybrid (Transformer Engine, E4M3 fwd / E5M2 bwd) |
| Tensor parallelism | 4 |
| Pipeline parallelism | 1 |
| Sequence parallelism | true |
| Activation recompute | granularity=full + uniform + 1 (auto for Full-param) |
nvcr.io/nvidia/nemo:26.02)Inherits the license of the base model: see Qwen/Qwen3-4B-Base.
from transformers import AutoModelForCausalLM, AutoTokenizerimport torch model_id = "g4me/QWiki-4B-Base-LR1e5"tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained( model_id, torch_dtype=torch.bfloat16, device_map="auto",) prompt = "Era uma vez, em um reino distante,"inputs = tokenizer(prompt, return_tensors="pt").to(model.device)outputs = model.generate( **inputs, max_new_tokens=128, do_sample=True, temperature=0.7, top_p=0.9,)print(tokenizer.decode(outputs[0], skip_special_tokens=True))