mgoin / Mixtral-8x7B-Instruct-v0.1-FP8

vllm (pretrained=/home/mgoin/code/llm-compressor/examples/quantizing_moe/Mixtral-8x7B-Instruct-v0.1-FP8,tensor_parallel_size=2), gen_kwargs: (None), limit: None, num_fewshot: 5, batch_size: auto
|Tasks|Version|     Filter     |n-shot|  Metric   |   |Value |   |Stderr|
|-----|------:|----------------|-----:|-----------|---|-----:|---|-----:|
|gsm8k|      3|flexible-extract|     5|exact_match|↑  |0.6338|±  |0.0133|
|     |       |strict-match    |     5|exact_match|↑  |0.6293|±  |0.0133|

Creation

from typing import List

from transformers import AutoTokenizer

from llmcompressor.modifiers.quantization import QuantizationModifier
from llmcompressor.transformers import SparseAutoModelForCausalLM, oneshot
from llmcompressor.transformers.compression.helpers import calculate_offload_device_map

MODEL_ID = "mistralai/Mixtral-8x7B-Instruct-v0.1"
NUM_GPUS = 2

# Adjust based off number of desired GPUs
device_map = calculate_offload_device_map(
    MODEL_ID, reserve_for_hessians=True, num_gpus=NUM_GPUS, torch_dtype="auto"
)

model = SparseAutoModelForCausalLM.from_pretrained(
    MODEL_ID, device_map=device_map, torch_dtype="auto"
)
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)


# Dataset config parameters
DATASET_ID = "open_platypus"
MAX_SEQ_LENGTH = 2048
NUM_CALIBRATION_SAMPLES = 512

# Save location of quantized model
OUTPUT_DIR = f"{MODEL_ID.split('/')[-1]}-FP8"
SAVE_COMPRESSED = True

layers_to_ignore: List[str] = [
    "lm_head",
    "re:.*block_sparse_moe.gate",  # does not quantize well
]

recipe = QuantizationModifier(
    scheme="FP8", targets="Linear", ignore=layers_to_ignore
)


oneshot(
    model=model,
    tokenizer=tokenizer,
    dataset=DATASET_ID,
    recipe=recipe,
    max_seq_length=MAX_SEQ_LENGTH,
    num_calibration_samples=NUM_CALIBRATION_SAMPLES,
    save_compressed=SAVE_COMPRESSED,
    overwrite_output_dir=True,
    output_dir=OUTPUT_DIR,
)

# Confirm generations of the quantized model look sane.
print("========== SAMPLE GENERATION ==============")
input_ids = tokenizer("Hello my name is", return_tensors="pt").input_ids.to("cuda")
output = model.generate(input_ids, max_new_tokens=20)
print(tokenizer.decode(output[0]))
print("==========================================")