Add ZipNN support

README.md

@@ -7,6 +7,42 @@ tags:
 - multimodal
 library_name: transformers
 ---
+# Disclaimer and Requirements
+
+This model is a clone of [**Qwen/Qwen2-VL-7B-Instruct**](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct) compressed using ZipNN. Compressed losslessly to 67% its original size, ZipNN saved ~6GB in storage and potentially ~2PB in data transfer **monthly**.
+
+### Requirement
+
+In order to use the model, ZipNN is necessary:
+```bash
+pip install zipnn
+```
+
+### Use This Model
+```python
+# Load model directly
+from transformers import AutoProcessor, AutoModelForSeq2SeqLM
+from zipnn import zipnn_hf
+
+zipnn_hf()
+
+processor = AutoProcessor.from_pretrained("royleibov/Qwen2-VL-7B-Instruct-ZipNN-Compressed")
+model = AutoModelForSeq2SeqLM.from_pretrained("royleibov/Qwen2-VL-7B-Instruct-ZipNN-Compressed")
+```
+### ZipNN
+ZipNN also allows you to seemlessly save local disk space in your cache after the model is downloaded.
+
+To compress the cached model, simply run:
+```bash
+python zipnn_compress_path.py safetensors --model royleibov/Qwen2-VL-7B-Instruct-ZipNN-Compressed --hf_cache
+```
+
+The model will be decompressed automatically and safely as long as `zipnn_hf()` is added at the top of the file like in the [example above](#use-this-model).
+
+To decompress manualy, simply run:
+```bash
+python zipnn_decompress_path.py --model royleibov/Qwen2-VL-7B-Instruct-ZipNN-Compressed --hf_cache
+```
 
 # Qwen2-VL-7B-Instruct
 
@@ -103,27 +139,30 @@ Here we show a code snippet to show you how to use the chat model with `transfor
 ```python
 from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor
 from qwen_vl_utils import process_vision_info
+from zipnn import zipnn_hf
+
+zipnn_hf()
 
 # default: Load the model on the available device(s)
 model = Qwen2VLForConditionalGeneration.from_pretrained(
-    "Qwen/Qwen2-VL-7B-Instruct", torch_dtype="auto", device_map="auto"
+    "royleibov/Qwen2-VL-7B-Instruct-ZipNN-Compressed", torch_dtype="auto", device_map="auto"
 )
 
 # We recommend enabling flash_attention_2 for better acceleration and memory saving, especially in multi-image and video scenarios.
 # model = Qwen2VLForConditionalGeneration.from_pretrained(
-#     "Qwen/Qwen2-VL-7B-Instruct",
+#     "royleibov/Qwen2-VL-7B-Instruct-ZipNN-Compressed",
 #     torch_dtype=torch.bfloat16,
 #     attn_implementation="flash_attention_2",
 #     device_map="auto",
 # )
 
 # default processer
-processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")
+processor = AutoProcessor.from_pretrained("royleibov/Qwen2-VL-7B-Instruct-ZipNN-Compressed")
 
 # The default range for the number of visual tokens per image in the model is 4-16384. You can set min_pixels and max_pixels according to your needs, such as a token count range of 256-1280, to balance speed and memory usage.
 # min_pixels = 256*28*28
 # max_pixels = 1280*28*28
-# processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct", min_pixels=min_pixels, max_pixels=max_pixels)
+# processor = AutoProcessor.from_pretrained("royleibov/Qwen2-VL-7B-Instruct-ZipNN-Compressed", min_pixels=min_pixels, max_pixels=max_pixels)
 
 messages = [
     {
@@ -172,12 +211,15 @@ import torch
 from torchvision import io
 from typing import Dict
 from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor
+from zipnn import zipnn_hf
+
+zipnn_hf()
 
 # Load the model in half-precision on the available device(s)
 model = Qwen2VLForConditionalGeneration.from_pretrained(
-    "Qwen/Qwen2-VL-7B-Instruct", torch_dtype="auto", device_map="auto"
+    "royleibov/Qwen2-VL-7B-Instruct-ZipNN-Compressed", torch_dtype="auto", device_map="auto"
 )
-processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct")
+processor = AutoProcessor.from_pretrained("royleibov/Qwen2-VL-7B-Instruct-ZipNN-Compressed")
 
 # Image
 url = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg"
@@ -418,7 +460,7 @@ The model supports a wide range of resolution inputs. By default, it uses the na
 min_pixels = 256 * 28 * 28
 max_pixels = 1280 * 28 * 28
 processor = AutoProcessor.from_pretrained(
-    "Qwen/Qwen2-VL-7B-Instruct", min_pixels=min_pixels, max_pixels=max_pixels
+    "royleibov/Qwen2-VL-7B-Instruct-ZipNN-Compressed", min_pixels=min_pixels, max_pixels=max_pixels
 )
 ```