Image Generation Parameters
2025-02-022 turns8,381 charsgpt-4o-mini
Summary
Configuring DeepSeek Janus model parameters for image generation tasks
Messages
import os
import PIL.Image
import torch
import numpy as np
from transformers import AutoModelForCausalLM
from janus.models import MultiModalityCausalLM, VLChatProcessor
# specify the path to the model
model_path = "deepseek-ai/Janus-Pro-7B"
vl_chat_processor: VLChatProcessor = VLChatProcessor.from_pretrained(model_path)
tokenizer = vl_chat_processor.tokenizer
vl_gpt: MultiModalityCausalLM = AutoModelForCausalLM.from_pretrained(
model_path, trust_remote_code=True
)
vl_gpt = vl_gpt.to(torch.bfloat16).cuda().eval()
conversation = [
{
"role": "<|User|>",
"content": "A stunning princess from kabul in red, white traditional clothing, blue eyes, brown hair",
},
{"role": "<|Assistant|>", "content": ""},
]
sft_format = vl_chat_processor.apply_sft_template_for_multi_turn_prompts(
conversations=conversation,
sft_format=vl_chat_processor.sft_format,
system_prompt="",
)
prompt = sft_format + vl_chat_processor.image_start_tag
@torch.inference_mode()
def generate(
mmgpt: MultiModalityCausalLM,
vl_chat_processor: VLChatProcessor,
prompt: str,
temperature: float = 1,
parallel_size: int = 16,
cfg_weight: float = 5,
image_token_num_per_image: int = 576,
img_size: int = 384,
patch_size: int = 16,
):
input_ids = vl_chat_processor.tokenizer.encode(prompt)
input_ids = torch.LongTensor(input_ids)
tokens = torch.zeros((parallel_size*2, len(input_ids)), dtype=torch.int).cuda()
for i in range(parallel_size*2):
tokens[i, :] = input_ids
if i % 2 != 0:
tokens[i, 1:-1] = vl_chat_processor.pad_id
inputs_embeds = mmgpt.language_model.get_input_embeddings()(tokens)
generated_tokens = torch.zeros((parallel_size, image_token_num_per_image), dtype=torch.int).cuda()
for i in range(image_token_num_per_image):
outputs = mmgpt.language_model.model(inputs_embeds=inputs_embeds, use_cache=True, past_key_values=outputs.past_key_values if i != 0 else None)
hidden_states = outputs.last_hidden_state
logits = mmgpt.gen_head(hidden_states[:, -1, :])
logit_cond = logits[0::2, :]
logit_uncond = logits[1::2, :]
logits = logit_uncond + cfg_weight * (logit_cond-logit_uncond)
probs = torch.softmax(logits / temperature, dim=-1)
next_token = torch.multinomial(probs, num_samples=1)
generated_tokens[:, i] = next_token.squeeze(dim=-1)
next_token = torch.cat([next_token.unsqueeze(dim=1), next_token.unsqueeze(dim=1)], dim=1).view(-1)
img_embeds = mmgpt.prepare_gen_img_embeds(next_token)
inputs_embeds = img_embeds.unsqueeze(dim=1)
dec = mmgpt.gen_vision_model.decode_code(generated_tokens.to(dtype=torch.int), shape=[parallel_size, 8, img_size//patch_size, img_size//patch_size])
dec = dec.to(torch.float32).cpu().numpy().transpose(0, 2, 3, 1)
dec = np.clip((dec + 1) / 2 * 255, 0, 255)
visual_img = np.zeros((parallel_size, img_size, img_size, 3), dtype=np.uint8)
visual_img[:, :, :] = dec
os.makedirs('generated_samples', exist_ok=True)
for i in range(parallel_size):
save_path = os.path.join('generated_samples', "img_{}.jpg".format(i))
PIL.Image.fromarray(visual_img[i]).save(save_path)
generate(
vl_gpt,
vl_chat_processor,
prompt,
) TEMPERATURE = 0.7
CFG_WEIGHT = 7
IMAGE_TOKEN_NUM = 1024
IMG_SIZE = 512
PATCH_SIZE = 16 --- make this work with these parameters and make sure it just generates one image
To adjust the code to use the new parameters (`TEMPERATURE`, `CFG_WEIGHT`, `IMAGE_TOKEN_NUM`, `IMG_SIZE`, and `PATCH_SIZE`), and ensure it generates only one image, you can make the following changes to the code:
1. **Set the Parameters:**
You'll need to modify the function call to include the new values for `TEMPERATURE`, `CFG_WEIGHT`, `IMAGE_TOKEN_NUM`, `IMG_SIZE`, and `PATCH_SIZE`. The `parallel_size` should be set to 1 because you want to generate a single image.
2. **Update the `generate` function:**
We'll ensure that the values passed match your desired configuration.
Here's the adjusted code:
```python
import os
import PIL.Image
import torch
import numpy as np
from transformers import AutoModelForCausalLM
from janus.models import MultiModalityCausalLM, VLChatProcessor
# specify the path to the model
model_path = "deepseek-ai/Janus-Pro-7B"
vl_chat_processor: VLChatProcessor = VLChatProcessor.from_pretrained(model_path)
tokenizer = vl_chat_processor.tokenizer
vl_gpt: MultiModalityCausalLM = AutoModelForCausalLM.from_pretrained(
model_path, trust_remote_code=True
)
vl_gpt = vl_gpt.to(torch.bfloat16).cuda().eval()
conversation = [
{
"role": "<|User|>",
"content": "A stunning princess from kabul in red, white traditional clothing, blue eyes, brown hair",
},
{"role": "<|Assistant|>", "content": ""},
]
sft_format = vl_chat_processor.apply_sft_template_for_multi_turn_prompts(
conversations=conversation,
sft_format=vl_chat_processor.sft_format,
system_prompt="",
)
prompt = sft_format + vl_chat_processor.image_start_tag
@torch.inference_mode()
def generate(
mmgpt: MultiModalityCausalLM,
vl_chat_processor: VLChatProcessor,
prompt: str,
temperature: float = 1,
parallel_size: int = 1, # Generate one image
cfg_weight: float = 7, # Adjusted CFG weight
image_token_num_per_image: int = 1024, # Adjusted image token size
img_size: int = 512, # Adjusted image size
patch_size: int = 16, # Adjusted patch size
):
input_ids = vl_chat_processor.tokenizer.encode(prompt)
input_ids = torch.LongTensor(input_ids)
tokens = torch.zeros((parallel_size*2, len(input_ids)), dtype=torch.int).cuda()
for i in range(parallel_size*2):
tokens[i, :] = input_ids
if i % 2 != 0:
tokens[i, 1:-1] = vl_chat_processor.pad_id
inputs_embeds = mmgpt.language_model.get_input_embeddings()(tokens)
generated_tokens = torch.zeros((parallel_size, image_token_num_per_image), dtype=torch.int).cuda()
for i in range(image_token_num_per_image):
outputs = mmgpt.language_model.model(inputs_embeds=inputs_embeds, use_cache=True, past_key_values=outputs.past_key_values if i != 0 else None)
hidden_states = outputs.last_hidden_state
logits = mmgpt.gen_head(hidden_states[:, -1, :])
logit_cond = logits[0::2, :]
logit_uncond = logits[1::2, :]
logits = logit_uncond + cfg_weight * (logit_cond - logit_uncond)
probs = torch.softmax(logits / temperature, dim=-1)
next_token = torch.multinomial(probs, num_samples=1)
generated_tokens[:, i] = next_token.squeeze(dim=-1)
next_token = torch.cat([next_token.unsqueeze(dim=1), next_token.unsqueeze(dim=1)], dim=1).view(-1)
img_embeds = mmgpt.prepare_gen_img_embeds(next_token)
inputs_embeds = img_embeds.unsqueeze(dim=1)
dec = mmgpt.gen_vision_model.decode_code(generated_tokens.to(dtype=torch.int), shape=[parallel_size, 8, img_size//patch_size, img_size//patch_size])
dec = dec.to(torch.float32).cpu().numpy().transpose(0, 2, 3, 1)
dec = np.clip((dec + 1) / 2 * 255, 0, 255)
visual_img = np.zeros((parallel_size, img_size, img_size, 3), dtype=np.uint8)
visual_img[:, :, :] = dec
os.makedirs('generated_samples', exist_ok=True)
for i in range(parallel_size):
save_path = os.path.join('generated_samples', "img_{}.jpg".format(i))
PIL.Image.fromarray(visual_img[i]).save(save_path)
# Call generate function with updated parameters
generate(
vl_gpt,
vl_chat_processor,
prompt,
temperature=0.7, # Temperature
cfg_weight=7, # CFG Weight
image_token_num_per_image=1024, # Image token number
img_size=512, # Image size
patch_size=16 # Patch size
)
```
### Key Changes:
1. **Parallel Size:**
The `parallel_size` is set to `1` to generate only one image.
2. **Temperature, CFG Weight, Image Token Size, Image Size, Patch Size:**
The function call is updated to reflect the new values you've provided.
3. **Generated Image:**
The image will be saved in the `generated_samples` folder, where you can inspect the output.
This code should now generate a single image based on your updated parameters. Let me know if you need further adjustments!
gpt-4o-mini