ImportError: cannot import name 'GenerationMixin' from 'transformers.generation' (/usr/local/lib/python3.11/dist-packages/transformers/generation/__init__.py)

[omanandswami2005]

# 1. Install transformers 4.51.1 FIRST (critical!)
!pip install transformers==4.51.1 -q

# 2. Install torch 2.6.0
!pip install torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 --index-url https://download.pytorch.org/whl/cu118 -q

# 3. Install vLLM nightly (which now includes DeepSeek-OCR support as of 2025/10/23)
!pip install --pre vllm --extra-index-url https://wheels.vllm.ai/nightly -q

# 4. Pillow
!pip install pillow -q

# Optional: flash-attn (skip if fails)
try:
    !pip install flash-attn==2.7.3 --no-build-isolation -q
except:
    print("flash-attn skipped")

from vllm import LLM, SamplingParams
from vllm.model_executor.models.deepseek_ocr import NGramPerReqLogitsProcessor
from PIL import Image

# Load model (this may take 2-5 mins on Kaggle due to download + GPU setup)
llm = LLM(
    model="deepseek-ai/DeepSeek-OCR",
    enable_prefix_caching=False,
    mm_processor_cache_gb=0,
    logits_processors=[NGramPerReqLogitsProcessor],
    max_model_len=8192,  # required for long OCR outputs
    trust_remote_code=True,
    dtype="bfloat16",    # reduces VRAM usage; Kaggle T4 supports it
)

# Load your image
image = Image.open("/kaggle/working/test_image.jpg").convert("RGB")

# Set prompt (choose one based on your need)
prompt = "<image>\n<|grounding|>Convert the document to markdown."

model_input = [{
    "prompt": prompt,
    "multi_modal_data": {"image": image}
}]

sampling_params = SamplingParams(
    temperature=0.0,        # deterministic output
    max_tokens=8192,
    extra_args=dict(
        ngram_size=30,
        window_size=90,
        whitelist_token_ids={128821, 128822},  # <td>, </td> tokens
    ),
    skip_special_tokens=False,
)

# Generate
outputs = llm.generate(model_input, sampling_params)

# Print result
print("\n" + "="*50 + "\nOCR OUTPUT:\n" + "="*50)
print(outputs[0].outputs[0].text)

did as same described in documentation, still not working ? was testing in kaggle btw