GLM-4.6-FP8-dynamic

from transformers import AutoModelForCausalLM, AutoTokenizer

from llmcompressor import oneshot
from llmcompressor.modifiers.quantization import QuantizationModifier
from llmcompressor.utils import dispatch_for_generation

MODEL_ID = "zai-org/GLM-4.6"

# Load model.
model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID, torch_dtype="auto", trust_remote_code=True, device_map=None
)
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)

# Configure the quantization algorithm and scheme.
recipe = QuantizationModifier(
    targets="Linear",
    scheme="FP8_DYNAMIC",
    ignore = [
        "lm_head",
    ]
)

# Apply quantization.
# FP8_DYNAMIC uses data-free quantization, so no calibration dataset needed
oneshot(model=model, recipe=recipe, trust_remote_code_model=True)

# Save to disk in compressed-tensors format.
SAVE_DIR = "./" + MODEL_ID.rstrip("/").split("/")[-1] + "-FP8-dynamic"
model.save_pretrained(SAVE_DIR, save_compressed=True)
tokenizer.save_pretrained(SAVE_DIR)