Spaces:
Running
on
Zero
Running
on
Zero
fancyfeast
commited on
Commit
•
9d2cfda
1
Parent(s):
fd8406c
Initial commit
Browse files- 9em124t2-499968/clip_model.pt +3 -0
- 9em124t2-499968/config.yaml +39 -0
- 9em124t2-499968/image_adapter.pt +3 -0
- 9em124t2-499968/text_model/README.md +202 -0
- 9em124t2-499968/text_model/adapter_config.json +29 -0
- 9em124t2-499968/text_model/adapter_model.safetensors +3 -0
- app.py +242 -0
- requirements.txt +6 -0
9em124t2-499968/clip_model.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7d7b0548d12fa649370896982c2af9d03d43285b782bd47639c96e6e0b29473c
|
3 |
+
size 1713067838
|
9em124t2-499968/config.yaml
ADDED
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
wandb_project: joy-caption-1
|
2 |
+
device_batch_size: 2
|
3 |
+
batch_size: 256
|
4 |
+
learning_rate: 0.0002
|
5 |
+
warmup_samples: 18000
|
6 |
+
max_samples: 500000
|
7 |
+
save_every: 50000
|
8 |
+
test_every: 50000
|
9 |
+
use_amp: true
|
10 |
+
grad_scaler: true
|
11 |
+
lr_scheduler_type: cosine
|
12 |
+
min_lr_ratio: 0.0
|
13 |
+
allow_tf32: true
|
14 |
+
seed: 69
|
15 |
+
num_workers: 8
|
16 |
+
optimizer_type: adamw
|
17 |
+
adam_beta1: 0.9
|
18 |
+
adam_beta2: 0.999
|
19 |
+
adam_eps: 1.0e-08
|
20 |
+
adam_weight_decay: 0.0
|
21 |
+
clip_grad_norm: 1.0
|
22 |
+
dataset: fancyfeast/joy-captioning-20240917a
|
23 |
+
clip_model: google/siglip-so400m-patch14-384
|
24 |
+
text_model: meta-llama/Meta-Llama-3.1-8B
|
25 |
+
resume: null
|
26 |
+
gradient_checkpointing: false
|
27 |
+
test_size: 2048
|
28 |
+
grad_scaler_init: 65536.0
|
29 |
+
max_caption_length: 257
|
30 |
+
num_image_tokens: 32
|
31 |
+
adapter_type: mlp
|
32 |
+
text_model_dtype: bfloat16
|
33 |
+
pre_test: false
|
34 |
+
train_image_model: true
|
35 |
+
image_model_lr: null
|
36 |
+
train_lora: true
|
37 |
+
lora_r: 64
|
38 |
+
lora_alpha: 16
|
39 |
+
lora_dropout: 0.1
|
9em124t2-499968/image_adapter.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e53c3bf8df745a3c19ae3c70dbf9bf23cfdc8f3fdb937000a4eafd2a36914661
|
3 |
+
size 86067714
|
9em124t2-499968/text_model/README.md
ADDED
@@ -0,0 +1,202 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
base_model: meta-llama/Meta-Llama-3.1-8B
|
3 |
+
library_name: peft
|
4 |
+
---
|
5 |
+
|
6 |
+
# Model Card for Model ID
|
7 |
+
|
8 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
9 |
+
|
10 |
+
|
11 |
+
|
12 |
+
## Model Details
|
13 |
+
|
14 |
+
### Model Description
|
15 |
+
|
16 |
+
<!-- Provide a longer summary of what this model is. -->
|
17 |
+
|
18 |
+
|
19 |
+
|
20 |
+
- **Developed by:** [More Information Needed]
|
21 |
+
- **Funded by [optional]:** [More Information Needed]
|
22 |
+
- **Shared by [optional]:** [More Information Needed]
|
23 |
+
- **Model type:** [More Information Needed]
|
24 |
+
- **Language(s) (NLP):** [More Information Needed]
|
25 |
+
- **License:** [More Information Needed]
|
26 |
+
- **Finetuned from model [optional]:** [More Information Needed]
|
27 |
+
|
28 |
+
### Model Sources [optional]
|
29 |
+
|
30 |
+
<!-- Provide the basic links for the model. -->
|
31 |
+
|
32 |
+
- **Repository:** [More Information Needed]
|
33 |
+
- **Paper [optional]:** [More Information Needed]
|
34 |
+
- **Demo [optional]:** [More Information Needed]
|
35 |
+
|
36 |
+
## Uses
|
37 |
+
|
38 |
+
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
39 |
+
|
40 |
+
### Direct Use
|
41 |
+
|
42 |
+
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
43 |
+
|
44 |
+
[More Information Needed]
|
45 |
+
|
46 |
+
### Downstream Use [optional]
|
47 |
+
|
48 |
+
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
|
49 |
+
|
50 |
+
[More Information Needed]
|
51 |
+
|
52 |
+
### Out-of-Scope Use
|
53 |
+
|
54 |
+
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
|
55 |
+
|
56 |
+
[More Information Needed]
|
57 |
+
|
58 |
+
## Bias, Risks, and Limitations
|
59 |
+
|
60 |
+
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
|
61 |
+
|
62 |
+
[More Information Needed]
|
63 |
+
|
64 |
+
### Recommendations
|
65 |
+
|
66 |
+
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
|
67 |
+
|
68 |
+
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
|
69 |
+
|
70 |
+
## How to Get Started with the Model
|
71 |
+
|
72 |
+
Use the code below to get started with the model.
|
73 |
+
|
74 |
+
[More Information Needed]
|
75 |
+
|
76 |
+
## Training Details
|
77 |
+
|
78 |
+
### Training Data
|
79 |
+
|
80 |
+
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
|
81 |
+
|
82 |
+
[More Information Needed]
|
83 |
+
|
84 |
+
### Training Procedure
|
85 |
+
|
86 |
+
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
|
87 |
+
|
88 |
+
#### Preprocessing [optional]
|
89 |
+
|
90 |
+
[More Information Needed]
|
91 |
+
|
92 |
+
|
93 |
+
#### Training Hyperparameters
|
94 |
+
|
95 |
+
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
|
96 |
+
|
97 |
+
#### Speeds, Sizes, Times [optional]
|
98 |
+
|
99 |
+
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
|
100 |
+
|
101 |
+
[More Information Needed]
|
102 |
+
|
103 |
+
## Evaluation
|
104 |
+
|
105 |
+
<!-- This section describes the evaluation protocols and provides the results. -->
|
106 |
+
|
107 |
+
### Testing Data, Factors & Metrics
|
108 |
+
|
109 |
+
#### Testing Data
|
110 |
+
|
111 |
+
<!-- This should link to a Dataset Card if possible. -->
|
112 |
+
|
113 |
+
[More Information Needed]
|
114 |
+
|
115 |
+
#### Factors
|
116 |
+
|
117 |
+
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
118 |
+
|
119 |
+
[More Information Needed]
|
120 |
+
|
121 |
+
#### Metrics
|
122 |
+
|
123 |
+
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
|
124 |
+
|
125 |
+
[More Information Needed]
|
126 |
+
|
127 |
+
### Results
|
128 |
+
|
129 |
+
[More Information Needed]
|
130 |
+
|
131 |
+
#### Summary
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
## Model Examination [optional]
|
136 |
+
|
137 |
+
<!-- Relevant interpretability work for the model goes here -->
|
138 |
+
|
139 |
+
[More Information Needed]
|
140 |
+
|
141 |
+
## Environmental Impact
|
142 |
+
|
143 |
+
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
|
144 |
+
|
145 |
+
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
|
146 |
+
|
147 |
+
- **Hardware Type:** [More Information Needed]
|
148 |
+
- **Hours used:** [More Information Needed]
|
149 |
+
- **Cloud Provider:** [More Information Needed]
|
150 |
+
- **Compute Region:** [More Information Needed]
|
151 |
+
- **Carbon Emitted:** [More Information Needed]
|
152 |
+
|
153 |
+
## Technical Specifications [optional]
|
154 |
+
|
155 |
+
### Model Architecture and Objective
|
156 |
+
|
157 |
+
[More Information Needed]
|
158 |
+
|
159 |
+
### Compute Infrastructure
|
160 |
+
|
161 |
+
[More Information Needed]
|
162 |
+
|
163 |
+
#### Hardware
|
164 |
+
|
165 |
+
[More Information Needed]
|
166 |
+
|
167 |
+
#### Software
|
168 |
+
|
169 |
+
[More Information Needed]
|
170 |
+
|
171 |
+
## Citation [optional]
|
172 |
+
|
173 |
+
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
|
174 |
+
|
175 |
+
**BibTeX:**
|
176 |
+
|
177 |
+
[More Information Needed]
|
178 |
+
|
179 |
+
**APA:**
|
180 |
+
|
181 |
+
[More Information Needed]
|
182 |
+
|
183 |
+
## Glossary [optional]
|
184 |
+
|
185 |
+
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
|
186 |
+
|
187 |
+
[More Information Needed]
|
188 |
+
|
189 |
+
## More Information [optional]
|
190 |
+
|
191 |
+
[More Information Needed]
|
192 |
+
|
193 |
+
## Model Card Authors [optional]
|
194 |
+
|
195 |
+
[More Information Needed]
|
196 |
+
|
197 |
+
## Model Card Contact
|
198 |
+
|
199 |
+
[More Information Needed]
|
200 |
+
### Framework versions
|
201 |
+
|
202 |
+
- PEFT 0.12.0
|
9em124t2-499968/text_model/adapter_config.json
ADDED
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"alpha_pattern": {},
|
3 |
+
"auto_mapping": null,
|
4 |
+
"base_model_name_or_path": "meta-llama/Meta-Llama-3.1-8B",
|
5 |
+
"bias": "none",
|
6 |
+
"fan_in_fan_out": false,
|
7 |
+
"inference_mode": true,
|
8 |
+
"init_lora_weights": true,
|
9 |
+
"layer_replication": null,
|
10 |
+
"layers_pattern": null,
|
11 |
+
"layers_to_transform": null,
|
12 |
+
"loftq_config": {},
|
13 |
+
"lora_alpha": 16,
|
14 |
+
"lora_dropout": 0.1,
|
15 |
+
"megatron_config": null,
|
16 |
+
"megatron_core": "megatron.core",
|
17 |
+
"modules_to_save": null,
|
18 |
+
"peft_type": "LORA",
|
19 |
+
"r": 64,
|
20 |
+
"rank_pattern": {},
|
21 |
+
"revision": null,
|
22 |
+
"target_modules": [
|
23 |
+
"q_proj",
|
24 |
+
"v_proj"
|
25 |
+
],
|
26 |
+
"task_type": "CAUSAL_LM",
|
27 |
+
"use_dora": false,
|
28 |
+
"use_rslora": false
|
29 |
+
}
|
9em124t2-499968/text_model/adapter_model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b48221de174ab0db7b46b4833118c5c0a4c2bf0b51b77b4cc4ab04651bd06cca
|
3 |
+
size 109069176
|
app.py
ADDED
@@ -0,0 +1,242 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import spaces
|
2 |
+
import gradio as gr
|
3 |
+
from huggingface_hub import InferenceClient
|
4 |
+
from torch import nn
|
5 |
+
from transformers import AutoModel, AutoProcessor, AutoTokenizer, PreTrainedTokenizer, PreTrainedTokenizerFast, AutoModelForCausalLM
|
6 |
+
from pathlib import Path
|
7 |
+
import torch
|
8 |
+
import torch.amp.autocast_mode
|
9 |
+
from PIL import Image
|
10 |
+
import os
|
11 |
+
import torchvision.transforms.functional as TVF
|
12 |
+
|
13 |
+
|
14 |
+
CLIP_PATH = "google/siglip-so400m-patch14-384"
|
15 |
+
MODEL_PATH = "meta-llama/Meta-Llama-3.1-8B"
|
16 |
+
CHECKPOINT_PATH = Path("9em124t2-499968")
|
17 |
+
TITLE = "<h1><center>JoyCaption Alpha One (2024-09-20a)</center></h1>"
|
18 |
+
CAPTION_TYPE_MAP = {
|
19 |
+
("descriptive", "formal", False, False): ["Write a descriptive caption for this image in a formal tone."],
|
20 |
+
("descriptive", "formal", False, True): ["Write a descriptive caption for this image in a formal tone within {word_count} words."],
|
21 |
+
("descriptive", "formal", True, False): ["Write a {length} descriptive caption for this image in a formal tone."],
|
22 |
+
("descriptive", "informal", False, False): ["Write a descriptive caption for this image in a casual tone."],
|
23 |
+
("descriptive", "informal", False, True): ["Write a descriptive caption for this image in a casual tone within {word_count} words."],
|
24 |
+
("descriptive", "informal", True, False): ["Write a {length} descriptive caption for this image in a casual tone."],
|
25 |
+
|
26 |
+
("training_prompt", "formal", False, False): ["Write a stable diffusion prompt for this image."],
|
27 |
+
("training_prompt", "formal", False, True): ["Write a stable diffusion prompt for this image within {word_count} words."],
|
28 |
+
("training_prompt", "formal", True, False): ["Write a {length} stable diffusion prompt for this image."],
|
29 |
+
|
30 |
+
("rng-tags", "formal", False, False): ["Write a list of Booru tags for this image."],
|
31 |
+
("rng-tags", "formal", False, True): ["Write a list of Booru tags for this image within {word_count} words."],
|
32 |
+
("rng-tags", "formal", True, False): ["Write a {length} list of Booru tags for this image."],
|
33 |
+
}
|
34 |
+
|
35 |
+
HF_TOKEN = os.environ.get("HF_TOKEN", None)
|
36 |
+
|
37 |
+
|
38 |
+
class ImageAdapter(nn.Module):
|
39 |
+
def __init__(self, input_features: int, output_features: int, ln1: bool, pos_emb: bool, num_image_tokens: int, deep_extract: bool):
|
40 |
+
super().__init__()
|
41 |
+
self.deep_extract = deep_extract
|
42 |
+
|
43 |
+
if self.deep_extract:
|
44 |
+
input_features = input_features * 5
|
45 |
+
|
46 |
+
self.linear1 = nn.Linear(input_features, output_features)
|
47 |
+
self.activation = nn.GELU()
|
48 |
+
self.linear2 = nn.Linear(output_features, output_features)
|
49 |
+
self.ln1 = nn.Identity() if not ln1 else nn.LayerNorm(input_features)
|
50 |
+
self.pos_emb = None if not pos_emb else nn.Parameter(torch.zeros(num_image_tokens, input_features))
|
51 |
+
|
52 |
+
# Mode token
|
53 |
+
#self.mode_token = nn.Embedding(n_modes, output_features)
|
54 |
+
#self.mode_token.weight.data.normal_(mean=0.0, std=0.02) # Matches HF's implementation of llama3
|
55 |
+
|
56 |
+
# Other tokens (<|image_start|>, <|image_end|>, <|eot_id|>)
|
57 |
+
self.other_tokens = nn.Embedding(3, output_features)
|
58 |
+
self.other_tokens.weight.data.normal_(mean=0.0, std=0.02) # Matches HF's implementation of llama3
|
59 |
+
|
60 |
+
def forward(self, vision_outputs: torch.Tensor):
|
61 |
+
if self.deep_extract:
|
62 |
+
x = torch.concat((
|
63 |
+
vision_outputs[-2],
|
64 |
+
vision_outputs[3],
|
65 |
+
vision_outputs[7],
|
66 |
+
vision_outputs[13],
|
67 |
+
vision_outputs[20],
|
68 |
+
), dim=-1)
|
69 |
+
assert len(x.shape) == 3, f"Expected 3, got {len(x.shape)}" # batch, tokens, features
|
70 |
+
assert x.shape[-1] == vision_outputs[-2].shape[-1] * 5, f"Expected {vision_outputs[-2].shape[-1] * 5}, got {x.shape[-1]}"
|
71 |
+
else:
|
72 |
+
x = vision_outputs[-2]
|
73 |
+
|
74 |
+
x = self.ln1(x)
|
75 |
+
|
76 |
+
if self.pos_emb is not None:
|
77 |
+
assert x.shape[-2:] == self.pos_emb.shape, f"Expected {self.pos_emb.shape}, got {x.shape[-2:]}"
|
78 |
+
x = x + self.pos_emb
|
79 |
+
|
80 |
+
x = self.linear1(x)
|
81 |
+
x = self.activation(x)
|
82 |
+
x = self.linear2(x)
|
83 |
+
|
84 |
+
# Mode token
|
85 |
+
#mode_token = self.mode_token(mode)
|
86 |
+
#assert mode_token.shape == (x.shape[0], mode_token.shape[1], x.shape[2]), f"Expected {(x.shape[0], 1, x.shape[2])}, got {mode_token.shape}"
|
87 |
+
#x = torch.cat((x, mode_token), dim=1)
|
88 |
+
|
89 |
+
# <|image_start|>, IMAGE, <|image_end|>
|
90 |
+
other_tokens = self.other_tokens(torch.tensor([0, 1], device=self.other_tokens.weight.device).expand(x.shape[0], -1))
|
91 |
+
assert other_tokens.shape == (x.shape[0], 2, x.shape[2]), f"Expected {(x.shape[0], 2, x.shape[2])}, got {other_tokens.shape}"
|
92 |
+
x = torch.cat((other_tokens[:, 0:1], x, other_tokens[:, 1:2]), dim=1)
|
93 |
+
|
94 |
+
return x
|
95 |
+
|
96 |
+
def get_eot_embedding(self):
|
97 |
+
return self.other_tokens(torch.tensor([2], device=self.other_tokens.weight.device)).squeeze(0)
|
98 |
+
|
99 |
+
|
100 |
+
|
101 |
+
# Load CLIP
|
102 |
+
print("Loading CLIP")
|
103 |
+
clip_processor = AutoProcessor.from_pretrained(CLIP_PATH)
|
104 |
+
clip_model = AutoModel.from_pretrained(CLIP_PATH)
|
105 |
+
clip_model = clip_model.vision_model
|
106 |
+
|
107 |
+
if (CHECKPOINT_PATH / "clip_model.pt").exists():
|
108 |
+
print("Loading VLM's custom vision model")
|
109 |
+
checkpoint = torch.load(CHECKPOINT_PATH / "clip_model.pt", map_location='cpu')
|
110 |
+
checkpoint = {k.replace("_orig_mod.module.", ""): v for k, v in checkpoint.items()}
|
111 |
+
clip_model.load_state_dict(checkpoint)
|
112 |
+
del checkpoint
|
113 |
+
|
114 |
+
clip_model.eval()
|
115 |
+
clip_model.requires_grad_(False)
|
116 |
+
clip_model.to("cuda")
|
117 |
+
|
118 |
+
|
119 |
+
# Tokenizer
|
120 |
+
print("Loading tokenizer")
|
121 |
+
tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH, use_fast=False)
|
122 |
+
assert isinstance(tokenizer, PreTrainedTokenizer) or isinstance(tokenizer, PreTrainedTokenizerFast), f"Tokenizer is of type {type(tokenizer)}"
|
123 |
+
|
124 |
+
# LLM
|
125 |
+
print("Loading LLM")
|
126 |
+
if (CHECKPOINT_PATH / "text_model").exists:
|
127 |
+
print("Loading VLM's custom text model")
|
128 |
+
text_model = AutoModelForCausalLM.from_pretrained(CHECKPOINT_PATH / "text_model", device_map=0, torch_dtype=torch.bfloat16)
|
129 |
+
else:
|
130 |
+
text_model = AutoModelForCausalLM.from_pretrained(MODEL_PATH, device_map="auto", torch_dtype=torch.bfloat16)
|
131 |
+
|
132 |
+
text_model.eval()
|
133 |
+
|
134 |
+
# Image Adapter
|
135 |
+
print("Loading image adapter")
|
136 |
+
image_adapter = ImageAdapter(clip_model.config.hidden_size, text_model.config.hidden_size, False, False, 38, False)
|
137 |
+
image_adapter.load_state_dict(torch.load(CHECKPOINT_PATH / "image_adapter.pt", map_location="cpu"))
|
138 |
+
image_adapter.eval()
|
139 |
+
image_adapter.to("cuda")
|
140 |
+
|
141 |
+
|
142 |
+
@spaces.GPU()
|
143 |
+
@torch.no_grad()
|
144 |
+
def stream_chat(input_image: Image.Image, caption_type: str, caption_tone: str, caption_length: str | int) -> str:
|
145 |
+
torch.cuda.empty_cache()
|
146 |
+
|
147 |
+
length = None if caption_length == "any" else caption_length
|
148 |
+
prompt_key = (caption_type, caption_tone, isinstance(length, str), isinstance(length, int))
|
149 |
+
if prompt_key not in CAPTION_TYPE_MAP:
|
150 |
+
raise ValueError(f"Invalid caption type: {prompt_key}")
|
151 |
+
|
152 |
+
prompt_str = CAPTION_TYPE_MAP[prompt_key][0]
|
153 |
+
|
154 |
+
# Preprocess image
|
155 |
+
#image = clip_processor(images=input_image, return_tensors='pt').pixel_values
|
156 |
+
image = input_image.resize((384, 384), Image.LANCZOS)
|
157 |
+
pixel_values = TVF.pil_to_tensor(image).unsqueeze(0) / 255.0
|
158 |
+
pixel_values = TVF.normalize(pixel_values, [0.5], [0.5])
|
159 |
+
pixel_values = pixel_values.to('cuda')
|
160 |
+
|
161 |
+
# Tokenize the prompt
|
162 |
+
prompt = tokenizer.encode(prompt_str, return_tensors='pt', padding=False, truncation=False, add_special_tokens=False)
|
163 |
+
|
164 |
+
# Embed image
|
165 |
+
with torch.amp.autocast_mode.autocast('cuda', enabled=True):
|
166 |
+
vision_outputs = clip_model(pixel_values=pixel_values, output_hidden_states=True)
|
167 |
+
image_features = vision_outputs.hidden_states
|
168 |
+
embedded_images = image_adapter(image_features)
|
169 |
+
embedded_images = embedded_images.to('cuda')
|
170 |
+
|
171 |
+
# Embed prompt
|
172 |
+
prompt_embeds = text_model.model.embed_tokens(prompt.to('cuda'))
|
173 |
+
assert prompt_embeds.shape == (1, prompt.shape[1], text_model.config.hidden_size), f"Prompt shape is {prompt_embeds.shape}, expected {(1, prompt.shape[1], text_model.config.hidden_size)}"
|
174 |
+
embedded_bos = text_model.model.embed_tokens(torch.tensor([[tokenizer.bos_token_id]], device=text_model.device, dtype=torch.int64))
|
175 |
+
eot_embed = image_adapter.get_eot_embedding().unsqueeze(0).to(dtype=text_model.dtype)
|
176 |
+
|
177 |
+
# Construct prompts
|
178 |
+
inputs_embeds = torch.cat([
|
179 |
+
embedded_bos.expand(embedded_images.shape[0], -1, -1),
|
180 |
+
embedded_images.to(dtype=embedded_bos.dtype),
|
181 |
+
prompt_embeds.expand(embedded_images.shape[0], -1, -1),
|
182 |
+
eot_embed.expand(embedded_images.shape[0], -1, -1),
|
183 |
+
], dim=1)
|
184 |
+
|
185 |
+
input_ids = torch.cat([
|
186 |
+
torch.tensor([[tokenizer.bos_token_id]], dtype=torch.long),
|
187 |
+
torch.zeros((1, embedded_images.shape[1]), dtype=torch.long),
|
188 |
+
prompt,
|
189 |
+
torch.tensor([[tokenizer.convert_tokens_to_ids("<|eot_id|>")]], dtype=torch.long),
|
190 |
+
], dim=1).to('cuda')
|
191 |
+
attention_mask = torch.ones_like(input_ids)
|
192 |
+
|
193 |
+
#generate_ids = text_model.generate(input_ids, inputs_embeds=inputs_embeds, attention_mask=attention_mask, max_new_tokens=300, do_sample=False, suppress_tokens=None)
|
194 |
+
#generate_ids = text_model.generate(input_ids, inputs_embeds=inputs_embeds, attention_mask=attention_mask, max_new_tokens=300, do_sample=True, top_k=10, temperature=0.5, suppress_tokens=None)
|
195 |
+
generate_ids = text_model.generate(input_ids, inputs_embeds=inputs_embeds, attention_mask=attention_mask, max_new_tokens=300, do_sample=True, suppress_tokens=None) # Uses the default which is temp=0.6, top_p=0.9
|
196 |
+
|
197 |
+
# Trim off the prompt
|
198 |
+
generate_ids = generate_ids[:, input_ids.shape[1]:]
|
199 |
+
if generate_ids[0][-1] == tokenizer.eos_token_id or generate_ids[0][-1] == tokenizer.convert_tokens_to_ids("<|eot_id|>"):
|
200 |
+
generate_ids = generate_ids[:, :-1]
|
201 |
+
|
202 |
+
caption = tokenizer.batch_decode(generate_ids, skip_special_tokens=False, clean_up_tokenization_spaces=False)[0]
|
203 |
+
|
204 |
+
return caption.strip()
|
205 |
+
|
206 |
+
|
207 |
+
with gr.Blocks() as demo:
|
208 |
+
gr.HTML(TITLE)
|
209 |
+
|
210 |
+
with gr.Row():
|
211 |
+
with gr.Column():
|
212 |
+
input_image = gr.Image(type="pil", label="Input Image")
|
213 |
+
|
214 |
+
caption_type = gr.Dropdown(
|
215 |
+
choices=["descriptive", "training_prompt", "rng-tags"],
|
216 |
+
label="Caption Type",
|
217 |
+
value="descriptive",
|
218 |
+
)
|
219 |
+
|
220 |
+
caption_tone = gr.Dropdown(
|
221 |
+
choices=["formal", "informal"],
|
222 |
+
label="Caption Tone",
|
223 |
+
value="formal",
|
224 |
+
)
|
225 |
+
|
226 |
+
caption_length = gr.Dropdown(
|
227 |
+
choices=["any", "very short", "short", "medium-length", "long", "very long"] +
|
228 |
+
[str(i) for i in range(20, 261, 10)],
|
229 |
+
label="Caption Length",
|
230 |
+
value="any",
|
231 |
+
)
|
232 |
+
|
233 |
+
run_button = gr.Button("Caption")
|
234 |
+
|
235 |
+
with gr.Column():
|
236 |
+
output_caption = gr.Textbox(label="Caption")
|
237 |
+
|
238 |
+
run_button.click(fn=stream_chat, inputs=[input_image, caption_type, caption_tone, caption_length], outputs=[output_caption])
|
239 |
+
|
240 |
+
|
241 |
+
if __name__ == "__main__":
|
242 |
+
demo.launch()
|
requirements.txt
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
huggingface_hub==0.23.4
|
2 |
+
accelerate
|
3 |
+
torch
|
4 |
+
transformers==4.44.0
|
5 |
+
sentencepiece
|
6 |
+
peft==0.12.0
|