yujiepan / apriel-1.5-tiny-random

This tiny model is intended for debugging. It is randomly initialized using the configuration adapted from ServiceNow-AI/Apriel-1.5-15b-Thinker .

Example usage:

import re
import requests
import torch
from PIL import Image
from transformers import AutoProcessor, AutoModelForImageTextToText

# Load model
model_id = "yujiepan/apriel-1.5-tiny-random"
model = AutoModelForImageTextToText.from_pretrained(
    model_id,
    dtype=torch.bfloat16,
    device_map="auto"
)
processor = AutoProcessor.from_pretrained(model_id)
url = "https://picsum.photos/id/237/200/300"
image = Image.open(requests.get(url, stream=True).raw).convert("RGB")

chat = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Which animal is this?"},
            {"type": "image"},
        ],
    }
]
prompt = processor.apply_chat_template(chat, add_generation_prompt=True, tokenize=False)
inputs = processor(text=prompt, images=[image], return_tensors="pt").to(model.device)
inputs.pop("token_type_ids", None)
inputs['pixel_values'] = inputs['pixel_values'].to(model.dtype)

with torch.no_grad():
    output_ids = model.generate(**inputs, max_new_tokens=128, do_sample=True, temperature=0.6)

generated_ids = output_ids[:, inputs['input_ids'].shape[1]:]
output = processor.decode(generated_ids[0], skip_special_tokens=False)
print("Image Response:", output)

Codes to create this repo:

import json
from pathlib import Path

import accelerate
import torch
from huggingface_hub import file_exists, hf_hub_download
from transformers import (
    AutoConfig,
    AutoModelForCausalLM,
    AutoProcessor,
    GenerationConfig,
    AutoModelForImageTextToText,
    set_seed,
)

source_model_id = "ServiceNow-AI/Apriel-1.5-15b-Thinker"
save_folder = "/tmp/yujiepan/apriel-1.5-tiny-random"

processor = AutoProcessor.from_pretrained(source_model_id, trust_remote_code=True)
processor.save_pretrained(save_folder)

with open(hf_hub_download(source_model_id, filename='config.json', repo_type='model'), 'r', encoding='utf-8') as f:
    config_json = json.load(f)

config_json['text_config'].update({
    'head_dim': 32,
    'hidden_size': 8,
    'intermediate_size': 64,
    'num_hidden_layers': 2,
    'num_attention_heads': 8,
    'num_key_value_heads': 4,
})
config_json['vision_config'].update(
    {
        'head_dim': 32,
        'intermediate_size': 256,
        'hidden_size': 32 * 4,
        'num_attention_heads': 4,
        'num_hidden_layers': 2,
    }
)
with open(f"{save_folder}/config.json", "w", encoding='utf-8') as f:
    json.dump(config_json, f, indent=2)

config = AutoConfig.from_pretrained(
    save_folder,
    trust_remote_code=True,
)
print(config)
torch.set_default_dtype(torch.bfloat16)
model = AutoModelForImageTextToText.from_config(config, trust_remote_code=True).to(torch.bfloat16)
torch.set_default_dtype(torch.float32)
if file_exists(filename="generation_config.json", repo_id=source_model_id, repo_type='model'):
    model.generation_config = GenerationConfig.from_pretrained(
        source_model_id, trust_remote_code=True,
    )
    model.generation_config.do_sample = True
    print(model.generation_config)
model = model.cpu()
with torch.no_grad():
    for name, p in sorted(model.named_parameters()):
        torch.nn.init.normal_(p, 0, 0.1)
        print(name, p.shape)
model.save_pretrained(save_folder)

Printing the model:

LlavaForConditionalGeneration(
  (model): LlavaModel(
    (vision_tower): PixtralVisionModel(
      (patch_conv): Conv2d(3, 128, kernel_size=(16, 16), stride=(16, 16), bias=False)
      (ln_pre): PixtralRMSNorm((128,), eps=1e-05)
      (transformer): PixtralTransformer(
        (layers): ModuleList(
          (0-1): 2 x PixtralAttentionLayer(
            (attention_norm): PixtralRMSNorm((128,), eps=1e-05)
            (feed_forward): PixtralMLP(
              (gate_proj): Linear(in_features=128, out_features=256, bias=False)
              (up_proj): Linear(in_features=128, out_features=256, bias=False)
              (down_proj): Linear(in_features=256, out_features=128, bias=False)
              (act_fn): SiLU()
            )
            (attention): PixtralAttention(
              (k_proj): Linear(in_features=128, out_features=128, bias=False)
              (v_proj): Linear(in_features=128, out_features=128, bias=False)
              (q_proj): Linear(in_features=128, out_features=128, bias=False)
              (o_proj): Linear(in_features=128, out_features=128, bias=False)
            )
            (ffn_norm): PixtralRMSNorm((128,), eps=1e-05)
          )
        )
      )
      (patch_positional_embedding): PixtralRotaryEmbedding()
    )
    (multi_modal_projector): LlavaMultiModalProjector(
      (linear_1): Linear(in_features=128, out_features=8, bias=True)
      (act): GELUActivation()
      (linear_2): Linear(in_features=8, out_features=8, bias=True)
    )
    (language_model): MistralModel(
      (embed_tokens): Embedding(131072, 8)
      (layers): ModuleList(
        (0-1): 2 x MistralDecoderLayer(
          (self_attn): MistralAttention(
            (q_proj): Linear(in_features=8, out_features=256, bias=False)
            (k_proj): Linear(in_features=8, out_features=128, bias=False)
            (v_proj): Linear(in_features=8, out_features=128, bias=False)
            (o_proj): Linear(in_features=256, out_features=8, bias=False)
          )
          (mlp): MistralMLP(
            (gate_proj): Linear(in_features=8, out_features=64, bias=False)
            (up_proj): Linear(in_features=8, out_features=64, bias=False)
            (down_proj): Linear(in_features=64, out_features=8, bias=False)
            (act_fn): SiLU()
          )
          (input_layernorm): MistralRMSNorm((8,), eps=1e-05)
          (post_attention_layernorm): MistralRMSNorm((8,), eps=1e-05)
        )
      )
      (norm): MistralRMSNorm((8,), eps=1e-05)
      (rotary_emb): MistralRotaryEmbedding()
    )
  )
  (lm_head): Linear(in_features=8, out_features=131072, bias=False)
)