sayed0am/Nanonets-OCR2-3B-FP8-Dynamic

Creation Code

from transformers import AutoProcessor, Qwen2_5_VLForConditionalGeneration

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

MODEL_ID = "nanonets/Nanonets-OCR2-3B"

# Load model.
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(MODEL_ID, torch_dtype="auto")
processor = AutoProcessor.from_pretrained(MODEL_ID)

# Configure the quantization algorithm and scheme.
# In this case, we:
#   * quantize the weights to fp8 with per channel via ptq
#   * quantize the activations to fp8 with dynamic per token
recipe = QuantizationModifier(
    targets="Linear",
    scheme="FP8_DYNAMIC",
    ignore=["lm_head", "re:visual.*", "re:model.visual.*"],
)

# Apply quantization and save to disk in compressed-tensors format.
oneshot(model=model, recipe=recipe)