You can deploy this model and start earning money today!

amd/
GLM-5.1-NVFP4

Name: GLM-5.1-NVFP4
Availability: InStock

Views

No views yet

Model Overview

Model Architecture: GLM-5.1
- Input: Text
- Output: Text
Supported Hardware Microarchitecture: AMD MI300/MI350/MI355 (emulation)
ROCm: 7.2.2
PyTorch: 2.10.0
Transformers: 5.2.0
Operating System(s): Linux
Inference Engine: vLLM
Model Optimizer: AMD-Quark (V0.12)
- Quantized layers: experts and shared_experts
- Weight quantization: NVFP4, Static
- Activation quantization: NVFP4, Dynamic
Calibration Dataset: Pile

This model was built with GLM-5.1 model by applying AMD-Quark for NVFP4 quantization.

Model Quantization

The model was quantized from zai-org/GLM-5.1 using AMD-Quark. The weights and activations are quantized to NVFP4.

Quantization scripts:

sudo sysctl -w vm.max_map_count=4194304
cd Quark/examples/torch/language_modeling/llm_ptq/
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
export MODEL_DIR=zai-org/GLM-5.1
export output_dir=amd/GLM-5.1-NVFP4
exclude_layers="*self_attn* *mlp.gate lm_head *mlp.gate_proj *mlp.up_proj *mlp.down_proj"
python3 quantize_quark.py --model_dir $MODEL_DIR \
                          --quant_scheme nvfp4 \
                          --num_calib_data 128 \
                          --exclude_layers $exclude_layers \
                          --model_export hf_format \
                          --output_dir $output_dir \
                          --multi_gpu balanced

Deployment

Use with vLLM

This model can be deployed efficiently using the vLLM backend.

Evaluation

The model was evaluated on GSM8K benchmarks.

Accuracy

Benchmark	GLM-5.1	GLM-5.1-NVFP4(this model)	Recovery
GSM8K (flexible-extract)	95.38	95.68	100.31%

Reproduction

The GSM8K result was obtained using the lm-evaluation-harness framework, based on the Docker image rocm/vllm-dev:nightly_main_20260603.

Install the lm-eval (Version: 0.4.12) in container first.

pip install lm-eval[api]

Launching server

export VLLM_ROCM_USE_AITER=1
export VLLM_ROCM_USE_AITER_FP8BMM=0
export VLLM_ROCM_USE_AITER_FP4BMM=0
HIP_VISIBLE_DEVICES=4,5,6,7 vllm serve amd/GLM-5.1-NVFP4 \
  -tp 4 \
  --block-size 1 \
  --trust-remote-code \
  --max-model-len 4096 \
  --port 8082

Evaluating model in a new terminal

lm_eval \
  --model local-completions \
  --model_args '{"model": "amd/GLM-5.1-NVFP4", "base_url": "http://localhost:8082/v1/completions", "num_concurrent": 32, "max_retries": 10, "max_gen_toks": 2048, "tokenizer_backend": null, "tokenized_requests": false}' \
  --tasks gsm8k \
  --batch_size auto \
  --num_fewshot 5 \
  --trust_remote_code

amd/GLM-5.1-NVFP4