Text Generation
Transformers
Safetensors
English
Chinese
qwen3_5_moe
image-text-to-text
agent
apodex
arxiv:2608.23283
conversational
8-bit precision
modelopt_mixed
Instructions to use apodex/Apodex-1.1-mini-NVFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use apodex/Apodex-1.1-mini-NVFP4 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="apodex/Apodex-1.1-mini-NVFP4") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("apodex/Apodex-1.1-mini-NVFP4") model = AutoModelForMultimodalLM.from_pretrained("apodex/Apodex-1.1-mini-NVFP4", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use apodex/Apodex-1.1-mini-NVFP4 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "apodex/Apodex-1.1-mini-NVFP4" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "apodex/Apodex-1.1-mini-NVFP4", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/apodex/Apodex-1.1-mini-NVFP4
- SGLang
How to use apodex/Apodex-1.1-mini-NVFP4 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "apodex/Apodex-1.1-mini-NVFP4" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "apodex/Apodex-1.1-mini-NVFP4", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "apodex/Apodex-1.1-mini-NVFP4" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "apodex/Apodex-1.1-mini-NVFP4", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use apodex/Apodex-1.1-mini-NVFP4 with Docker Model Runner:
docker model run hf.co/apodex/Apodex-1.1-mini-NVFP4
| { | |
| "producer": { | |
| "name": "modelopt", | |
| "version": "0.44.0" | |
| }, | |
| "quantization": { | |
| "quant_algo": "MIXED_PRECISION", | |
| "kv_cache_quant_algo": "FP8", | |
| "quantized_layers": { | |
| "model.language_model.layers.0.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.0.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.0.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.0.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.0.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.0.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.0.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.1.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.1.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.1.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.1.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.1.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.1.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.1.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.2.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.2.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.2.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.2.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.2.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.2.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.2.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.3.self_attn.q_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.3.self_attn.k_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.3.self_attn.v_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.3.self_attn.o_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.3.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.3.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.3.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.3.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.4.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.4.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.4.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.4.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.4.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.4.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.4.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.5.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.5.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.5.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.5.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.5.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.5.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.5.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.6.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.6.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.6.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.6.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.6.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.6.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.6.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.7.self_attn.q_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.7.self_attn.k_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.7.self_attn.v_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.7.self_attn.o_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.7.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.7.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.7.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.7.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.8.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.8.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.8.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.8.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.8.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.8.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.8.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.9.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.9.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.9.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.9.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.9.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.9.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.9.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.10.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.10.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.10.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.10.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.10.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.10.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.10.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.11.self_attn.q_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.11.self_attn.k_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.11.self_attn.v_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.11.self_attn.o_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.11.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.11.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.11.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.11.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.12.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.12.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.12.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.12.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.12.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.12.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.12.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.13.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.13.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.13.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.13.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.13.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.13.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.13.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.14.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.14.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.14.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.14.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.14.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.14.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.14.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.15.self_attn.q_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.15.self_attn.k_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.15.self_attn.v_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.15.self_attn.o_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.15.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.15.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.15.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.15.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.16.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.16.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.16.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.16.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.16.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.16.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.16.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.17.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.17.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.17.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.17.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.17.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.17.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.17.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.18.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.18.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.18.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.18.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.18.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.18.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.18.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.19.self_attn.q_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.19.self_attn.k_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.19.self_attn.v_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.19.self_attn.o_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.19.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.19.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.19.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.19.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.20.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.20.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.20.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.20.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.20.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.20.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.20.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.21.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.21.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.21.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.21.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.21.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.21.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.21.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.22.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.22.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.22.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.22.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.22.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.22.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.22.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.23.self_attn.q_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.23.self_attn.k_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.23.self_attn.v_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.23.self_attn.o_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.23.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.23.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.23.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.23.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.24.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.24.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.24.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.24.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.24.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.24.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.24.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.25.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.25.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.25.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.25.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.25.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.25.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.25.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.26.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.26.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.26.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.26.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.26.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.26.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.26.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.27.self_attn.q_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.27.self_attn.k_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.27.self_attn.v_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.27.self_attn.o_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.27.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.27.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.27.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.27.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.28.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.28.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.28.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.28.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.28.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.28.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.28.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.29.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.29.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.29.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.29.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.29.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.29.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.29.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.30.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.30.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.30.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.30.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.30.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.30.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.30.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.31.self_attn.q_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.31.self_attn.k_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.31.self_attn.v_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.31.self_attn.o_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.31.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.31.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.31.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.31.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.32.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.32.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.32.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.32.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.32.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.32.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.32.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.33.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.33.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.33.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.33.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.33.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.33.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.33.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.34.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.34.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.34.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.34.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.34.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.34.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.34.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.35.self_attn.q_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.35.self_attn.k_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.35.self_attn.v_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.35.self_attn.o_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.35.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.35.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.35.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.35.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.36.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.36.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.36.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.36.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.36.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.36.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.36.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.37.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.37.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.37.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.37.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.37.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.37.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.37.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.38.linear_attn.out_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.38.linear_attn.in_proj_qkv": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.38.linear_attn.in_proj_z": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.38.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.38.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.38.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.38.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.39.self_attn.q_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.39.self_attn.k_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.39.self_attn.v_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.39.self_attn.o_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.39.mlp.experts": { | |
| "quant_algo": "NVFP4", | |
| "group_size": 16 | |
| }, | |
| "model.language_model.layers.39.mlp.shared_expert.gate_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.39.mlp.shared_expert.up_proj": { | |
| "quant_algo": "FP8" | |
| }, | |
| "model.language_model.layers.39.mlp.shared_expert.down_proj": { | |
| "quant_algo": "FP8" | |
| } | |
| }, | |
| "exclude_modules": [ | |
| "mtp*", | |
| "mtp.layers.0*" | |
| ] | |
| } | |
| } |