amd / Qwen3.6-35B-A3B-w4a16-llmcompressor

Qwen3.6-35B-A3B-w4a16-llmcompressor-v0.12.0

Model Overview

• Model Architecture: Qwen3_5MoeForCausalLM
  • Input: Text
  • Output: Text
• Source Model: Qwen3.6-35B-A3B
• Supported Hardware: AMD EPYC (CPU inference)
• Preferred Operating System: Linux
• Inference Engine: vLLM v0.26.0
• Quantization Framework: LLM Compressor v0.12.0
• Quantization Method: 4-bit Weight-Only Quantization (W4A16)
• Compatible Stack:
  • ZenDNN v6.1.0
  • ZenTorch v2.11.0.3
  • PyTorch v2.11.0
  • LLM Compressor v0.12.0
  • vLLM v0.26.0

This is a quantized version of Qwen3.6-35B-A3B created by AMD using LLM Compressor (compressed-tensors) for ZenDNN-optimized CPU inference.

Quantization

The model was quantized from Qwen3.6-35B-A3B using LLM Compressor via the GPTQ algorithm. This reduces the model weights from 67.0 GiB to 18.1 GiB on disk (~73% reduction).

• Method: 4-bit Weight-Only Quantization (W4A16)
• Config: compressed-tensors, num_bits=4, type=int, symmetric=true, group_size=128
• Weights: INT4, symmetric, group-wise ( group_size=128 , actorder=static ), stored as pack-quantized
• Activations: BF16 (unquantized)
• Group Size: 128, auto-selected as the largest of 128/64/32 that divides every quantizable weight's column count
• Calibration: 128 examples from HuggingFaceH4/ultrachat_200k at a max sequence length of 2048
• Kept in BF16: the MoE router ( mlp.gate ), lm_head , embed_tokens , and the layer norms. The 256 routed experts, the shared expert, and the linear-attention projections are all quantized, which is why the reduction is close to the 75% ideal for INT4.

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from datasets import load_dataset

from llmcompressor import oneshot
from llmcompressor.modifiers.quantization.gptq import GPTQModifier

model_id = "Qwen/Qwen3.6-35B-A3B"
save_dir = "./Qwen3.6-35B-A3B-w4a16-llmcompressor-v0.12.0"

CALIB_SIZE = 128
MAX_SEQ_LENGTH = 2048

# Step 1: Load the BF16 model and tokenizer
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="cpu",
    dtype=torch.bfloat16,
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)

# Step 2: Load calibration data. GPTQ is data-driven: it needs real activations to
# build the per-layer Hessians used to compensate the rounding error.
ds = load_dataset("HuggingFaceH4/ultrachat_200k", split=f"train_sft[:{CALIB_SIZE}]")
ds = ds.map(
    lambda example: {"text": "\n".join(m["content"] for m in example["messages"] if m["content"])},
    remove_columns=ds.column_names,
)

# Step 3: Define the W4A16 GPTQ recipe. Routing layers are skipped: they are tiny,
# and a mis-routed token costs far more accuracy than 4-bit expert weights do.
# The W4A16 preset implies group_size=128, which is valid here because every
# quantized weight's column count (2048 hidden, 512 MoE intermediate) divides by 128.
recipe = GPTQModifier(
    targets="Linear",
    scheme="W4A16",
    ignore=[
        "lm_head",
        r"re:.*\.router$",
        r"re:.*\.router\..*",
        r"re:.*\.gate$",
        r"re:.*\.mlp\.gate$",
    ],
)

# Step 4: One-shot quantize with calibration and save in compressed-tensors format
oneshot(
    model=model,
    dataset=ds,
    recipe=recipe,
    max_seq_length=MAX_SEQ_LENGTH,
    processor=tokenizer,
)
model.save_pretrained(save_dir, save_compressed=True)
tokenizer.save_pretrained(save_dir)

# Smoke test
inputs = tokenizer("What are we having for dinner?", return_tensors="pt")
output = model.generate(**inputs, max_new_tokens=30)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Run end to end with the driver script:

numactl --physcpubind=0-95 python llm_compressor_quantize_and_run.py \
    --model_id Qwen/Qwen3.6-35B-A3B \
    --save_dir ./Qwen3.6-35B-A3B-w4a16-llmcompressor-v0.12.0 \
    --recipe gptq \
    --scheme W4A16 \
    --run

Quick Start

Use with vLLM

from vllm import LLM, SamplingParams

model = LLM(
    model="amd/Qwen3.6-35B-A3B-w4a16-llmcompressor-v0.12.0",
    dtype="bfloat16",
)

sampling_params = SamplingParams(temperature=0.7, max_tokens=256)
outputs = model.generate(["Hello, how are you?"], sampling_params)
print(outputs[0].outputs[0].text)

Requirements

torch==2.11.0
zentorch==2.11.0.3
vllm==0.26.0
llmcompressor==0.12.0

OpenMP Setup

For optimal performance, set LD_PRELOAD with libomp.so (LLVM OpenMP) or libiomp5.so (Intel OpenMP):

# Using LLVM OpenMP (llvmopenmp)
export LD_PRELOAD=$(find /path/to/env -name "libomp.so" | head -1)

# Or using Intel OpenMP (libiomp)
export LD_PRELOAD=$(find /path/to/env -name "libiomp5.so" | head -1)

Note: Set LD_PRELOAD before launching vLLM or any inference script.

Evaluation

The model was evaluated against the BF16 (unquantized) baseline on standard benchmarks using lm-evaluation-harness with the vLLM engine.

Evaluation results will be updated after benchmarking.

Evaluation Command

lm_eval \
    --model vllm \
    --model_args pretrained=amd/Qwen3.6-35B-A3B-w4a16-llmcompressor,tokenizer=Qwen/Qwen3.6-35B-A3B,dtype=bfloat16,max_model_len=4096,enable_thinking=False \
    --tasks gsm8k \
    --batch_size auto \
    --trust_remote_code \
    --num_fewshot 5 \
    --apply_chat_template \
    --log_samples \
    --gen_kwargs "max_gen_toks=2048" \
    --output_path .

Limitations

• Version Lock: This model is compatible with ZenDNN v6.1.0 / ZenTorch v2.11.0.3 / PyTorch v2.11.0. It may not load correctly on other versions.
• CPU Only: This model is optimized for AMD EPYC CPU inference via ZenDNN. It is not intended for GPU inference.
• Text-Only Checkpoint: The model was loaded through AutoModelForCausalLM during quantization, so the saved config is the text-only Qwen3_5MoeForCausalLM variant. The vision path of the source model is not carried over.
• Reasoning Model: Qwen3.6 emits thinking traces by default. Evaluate with enable_thinking=False (and a generous max_gen_toks ) so answer extraction stays comparable to the BF16 baseline.
• Accuracy Trade-off: 4-bit weight-only quantization is more aggressive than INT8 and trades some accuracy for a ~73% smaller memory footprint.

License

This model is distributed under the same license as the source model. See the LICENSE file for details.

Modifications copyright (c) 2026 Advanced Micro Devices, Inc. All rights reserved.