Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,69 @@
+---
+language:
+  - en
+license: apache-2.0
+library_name: transformers
+tags:
+  - mixture-of-experts
+  - MoE
+  - coding
+  - 4-bit
+  - nf4
+  - bitsandbytes
+  - quantized
+  - LFM2
+  - Qwen
+base_model:
+  - Akahsizrr/fuse-1-Lite
+  - LiquidAI/LFM2.5-2.6B
+  - Qwen/Qwen3.6-35B-A3B
+pipeline_tag: text-generation
+inference: false
+---
+
+# fuse-1 Lite — 4-bit Quantized (NF4)
+
+> **4-bit NF4 quantized version of [fuse-1 Lite](https://huggingface.co/Akahsizrr/fuse-1-Lite). 3.36 GB VRAM — runs on T4, RTX 3060, and consumer GPUs.**
+
+This is the bitsandbytes 4-bit quantized version of fuse-1 Lite, using NF4 (NormalFloat 4-bit) quantization with double quantization for maximum compression.
+
+## Usage
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+import torch
+
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_compute_dtype=torch.bfloat16,
+    bnb_4bit_use_double_quant=True,
+)
+
+model = AutoModelForCausalLM.from_pretrained(
+    "Akahsizrr/fuse-1-Lite-4bit",
+    quantization_config=bnb_config,
+    device_map="auto",
+    trust_remote_code=True,
+)
+tokenizer = AutoTokenizer.from_pretrained("Akahsizrr/fuse-1-Lite-4bit")
+
+messages = [{"role": "user", "content": "Write a Python function to check if a number is prime."}]
+text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+inputs = tokenizer(text, return_tensors="pt").to(model.device)
+
+with torch.no_grad():
+    outputs = model.generate(**inputs, max_new_tokens=512, do_sample=True, temperature=0.1)
+
+print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))
+```
+
+## VRAM Requirements
+
+| Precision | VRAM | GPU |
+|-----------|------|-----|
+| **4-bit (this model)** | **3.36 GB** | T4, RTX 3060, M2 Pro |
+| 8-bit | 6.00 GB | T4, L4, RTX 3060 |
+| bfloat16 | ~12 GB | L4, A10G, RTX 4090 |
+
+See the [main model card](https://huggingface.co/Akahsizrr/fuse-1-Lite) for full architecture details, training info, and technical report.