RedHatAI/whisper-tiny-FP8-Dynamic

from vllm.assets.audio import AudioAsset
from vllm import LLM, SamplingParams

# prepare model
llm = LLM(
    model="neuralmagic/whisper-tiny-FP8-Dynamic",
    max_model_len=448,
    max_num_seqs=400,
    limit_mm_per_prompt={"audio": 1},
)

# prepare inputs
inputs = {  # Test explicit encoder/decoder prompt
    "encoder_prompt": {
        "prompt": "",
        "multi_modal_data": {
            "audio": AudioAsset("winning_call").audio_and_sample_rate,
        },
    },
    "decoder_prompt": "<|startoftranscript|>",
}

# generate response
print("========== SAMPLE GENERATION ==============")
outputs = llm.generate(inputs, SamplingParams(temperature=0.0, max_tokens=64))
print(f"PROMPT  : {outputs[0].prompt}")
print(f"RESPONSE: {outputs[0].outputs[0].text}")
print("==========================================")

python quantize.py \
    --model_path openai/whisper-tiny \
    --quant_path output_dir/whisper-tiny-FP8-Dynamic

import argparse
import torch
import os
from datasets import load_dataset
from transformers import WhisperProcessor
from llmcompressor import oneshot
from llmcompressor.modifiers.quantization import QuantizationModifier
from llmcompressor.transformers.tracing import TraceableWhisperForConditionalGeneration
from compressed_tensors.quantization import QuantizationType

# --- Args ---
parser = argparse.ArgumentParser()
parser.add_argument('--model_path', type=str, required=True)
parser.add_argument('--quant_path', type=str, required=True)
parser.add_argument('--observer', type=str, default="minmax")
args = parser.parse_args()

# --- Load Model ---
model = TraceableWhisperForConditionalGeneration.from_pretrained(
    args.model_path,
    device_map="auto",
    torch_dtype="auto",
)
model.config.forced_decoder_ids = None
processor = WhisperProcessor.from_pretrained(args.model_path)

# --- Recipe (FP8 Dynamic) ---
recipe = [
    QuantizationModifier(
        targets="Linear",
        scheme="FP8_DYNAMIC",
        sequential_targets=["WhisperEncoderLayer", "WhisperDecoderLayer"],
        ignore=["re:.*lm_head"],
    )
]

# --- Run oneshot ---
oneshot(
    model=model,
    recipe=recipe,
    trust_remote_code_model=True,
)

# --- Save ---
os.makedirs(args.quant_path, exist_ok=True)
model.save_pretrained(args.quant_path, save_compressed=True)
processor.save_pretrained(args.quant_path)

lmms-eval \
    --model=whisper_vllm \
    --model_args="pretrained=neuralmagic-ent/whisper-tiny-FP8-Dynamic" \
    --batch_size 64 \
    --output_path <output_file_path> \
    --tasks librispeech

lmms-eval \
    --model=whisper_vllm \
    --model_args="pretrained=neuralmagic-ent/whisper-tiny-FP8-Dynamic" \
    --batch_size 64 \
    --output_path <output_file_path> \
    --tasks fleurs