fancyfeast
commited on
Commit
•
de9952a
1
Parent(s):
d17e136
Update README.md example, processor works now, vLLM instructions
Browse files
README.md
CHANGED
@@ -33,10 +33,8 @@ Example usage:
|
|
33 |
|
34 |
```
|
35 |
import torch
|
36 |
-
import torch.amp
|
37 |
-
import torchvision.transforms.functional as TVF
|
38 |
from PIL import Image
|
39 |
-
from transformers import
|
40 |
|
41 |
|
42 |
IMAGE_PATH = "image.jpg"
|
@@ -47,27 +45,14 @@ MODEL_NAME = "fancyfeast/llama-joycaption-alpha-two-hf-llava"
|
|
47 |
# Load JoyCaption
|
48 |
# bfloat16 is the native dtype of the LLM used in JoyCaption (Llama 3.1)
|
49 |
# device_map=0 loads the model into the first GPU
|
50 |
-
|
51 |
llava_model = LlavaForConditionalGeneration.from_pretrained(MODEL_NAME, torch_dtype="bfloat16", device_map=0)
|
52 |
llava_model.eval()
|
53 |
|
54 |
with torch.no_grad():
|
55 |
-
# Load
|
56 |
-
# Normally you would use the Processor here, but the image module's processor
|
57 |
-
# has some buggy behavior and a simple resize in Pillow yields higher quality results
|
58 |
image = Image.open(IMAGE_PATH)
|
59 |
|
60 |
-
if image.size != (384, 384):
|
61 |
-
image = image.resize((384, 384), Image.LANCZOS)
|
62 |
-
|
63 |
-
image = image.convert("RGB")
|
64 |
-
pixel_values = TVF.pil_to_tensor(image)
|
65 |
-
|
66 |
-
# Normalize the image
|
67 |
-
pixel_values = pixel_values / 255.0
|
68 |
-
pixel_values = TVF.normalize(pixel_values, [0.5], [0.5])
|
69 |
-
pixel_values = pixel_values.to(torch.bfloat16).unsqueeze(0)
|
70 |
-
|
71 |
# Build the conversation
|
72 |
convo = [
|
73 |
{
|
@@ -81,30 +66,44 @@ with torch.no_grad():
|
|
81 |
]
|
82 |
|
83 |
# Format the conversation
|
84 |
-
|
85 |
-
|
86 |
-
#
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
|
|
|
|
|
|
102 |
|
103 |
# Trim off the prompt
|
104 |
-
generate_ids = generate_ids[input_ids.shape[1]:]
|
105 |
|
106 |
# Decode the caption
|
107 |
-
caption = tokenizer.decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)
|
108 |
caption = caption.strip()
|
109 |
print(caption)
|
110 |
-
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
```
|
35 |
import torch
|
|
|
|
|
36 |
from PIL import Image
|
37 |
+
from transformers import AutoProcessor, LlavaForConditionalGeneration
|
38 |
|
39 |
|
40 |
IMAGE_PATH = "image.jpg"
|
|
|
45 |
# Load JoyCaption
|
46 |
# bfloat16 is the native dtype of the LLM used in JoyCaption (Llama 3.1)
|
47 |
# device_map=0 loads the model into the first GPU
|
48 |
+
processor = AutoProcessor.from_pretrained(MODEL_NAME)
|
49 |
llava_model = LlavaForConditionalGeneration.from_pretrained(MODEL_NAME, torch_dtype="bfloat16", device_map=0)
|
50 |
llava_model.eval()
|
51 |
|
52 |
with torch.no_grad():
|
53 |
+
# Load image
|
|
|
|
|
54 |
image = Image.open(IMAGE_PATH)
|
55 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
56 |
# Build the conversation
|
57 |
convo = [
|
58 |
{
|
|
|
66 |
]
|
67 |
|
68 |
# Format the conversation
|
69 |
+
# WARNING: HF's handling of chat's on Llava models is very fragile. This specific combination of processor.apply_chat_template(), and processor() works
|
70 |
+
# but if using other combinations always inspect the final input_ids to ensure they are correct. Often times you will end up with multiple <bos> tokens
|
71 |
+
# if not careful, which can make the model perform poorly.
|
72 |
+
convo_string = processor.apply_chat_template(convo, tokenize = False, add_generation_prompt = True)
|
73 |
+
assert isinstance(convo_string, str)
|
74 |
+
|
75 |
+
# Process the inputs
|
76 |
+
inputs = processor(text=[convo_string], images=[image], return_tensors="pt").to('cuda')
|
77 |
+
inputs['pixel_values'] = inputs['pixel_values'].to(torch.bfloat16)
|
78 |
+
|
79 |
+
# Generate the captions
|
80 |
+
generate_ids = llava_model.generate(
|
81 |
+
**inputs,
|
82 |
+
max_new_tokens=300,
|
83 |
+
do_sample=True,
|
84 |
+
suppress_tokens=None,
|
85 |
+
use_cache=True,
|
86 |
+
temperature=0.6,
|
87 |
+
top_k=None,
|
88 |
+
top_p=0.9,
|
89 |
+
)[0]
|
90 |
|
91 |
# Trim off the prompt
|
92 |
+
generate_ids = generate_ids[inputs['input_ids'].shape[1]:]
|
93 |
|
94 |
# Decode the caption
|
95 |
+
caption = processor.tokenizer.decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)
|
96 |
caption = caption.strip()
|
97 |
print(caption)
|
98 |
+
```
|
99 |
+
|
100 |
+
|
101 |
+
## vLLM
|
102 |
+
|
103 |
+
vLLM provides the highest performance inference for JoyCaption, and an OpenAI compatible API so JoyCaption can be used like any other VLMs. Example usage:
|
104 |
+
|
105 |
+
```
|
106 |
+
vllm serve fancyfeast/llama-joycaption-alpha-two-hf-llava --max-model-len 4096 --enable-prefix-caching
|
107 |
+
```
|
108 |
+
|
109 |
+
VLMs are a bit finicky on vLLM, and vLLM is memory hungry, so you may have to adjust settings for your particular environment, such as forcing eager mode, adjusting max-model-len, adjusting gpu_memory_utilization, etc.
|