init
Browse files- README.md +17 -0
- config.json +39 -0
- mm_projector.bin +3 -0
- trainer_state.json +0 -0
README.md
ADDED
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
inference: false
|
3 |
+
license: apache-2.0
|
4 |
+
---
|
5 |
+
|
6 |
+
# Model Card
|
7 |
+
|
8 |
+
Bunny is a family of lightweight multimodal models.
|
9 |
+
|
10 |
+
Bunny-pretrain-llama3-8b-siglip is the pretrained weights for [Bunny-Llama-3-8B-V](https://huggingface.co/BAAI/Bunny-Llama-3-8B-V), which leverages Llama-3-8B-Instruct as the language model backbone and SigLIP as the vision encoder.
|
11 |
+
It is pretrained on LAION-2M.
|
12 |
+
|
13 |
+
More details about this model can be found in [GitHub](https://github.com/BAAI-DCAI/Bunny).
|
14 |
+
|
15 |
+
# License
|
16 |
+
This project utilizes certain datasets and checkpoints that are subject to their respective original licenses. Users must comply with all terms and conditions of these original licenses.
|
17 |
+
The content of this project itself is licensed under the Apache license 2.0.
|
config.json
ADDED
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "bunny-pretrain-llama3-8b-siglip",
|
3 |
+
"architectures": [
|
4 |
+
"LlamaForCausalLM"
|
5 |
+
],
|
6 |
+
"attention_bias": false,
|
7 |
+
"attention_dropout": 0.0,
|
8 |
+
"bos_token_id": 128000,
|
9 |
+
"eos_token_id": 128001,
|
10 |
+
"freeze_mm_mlp_adapter": false,
|
11 |
+
"hidden_act": "silu",
|
12 |
+
"hidden_size": 4096,
|
13 |
+
"image_aspect_ratio": "square",
|
14 |
+
"initializer_range": 0.02,
|
15 |
+
"intermediate_size": 14336,
|
16 |
+
"max_position_embeddings": 8192,
|
17 |
+
"mm_hidden_size": 1152,
|
18 |
+
"mm_projector_lr": null,
|
19 |
+
"mm_projector_type": "mlp2x_gelu",
|
20 |
+
"mm_vision_tower": "google/siglip-so400m-patch14-384",
|
21 |
+
"model_type": "bunny-llama",
|
22 |
+
"num_attention_heads": 32,
|
23 |
+
"num_hidden_layers": 32,
|
24 |
+
"num_key_value_heads": 8,
|
25 |
+
"pretraining_tp": 1,
|
26 |
+
"rms_norm_eps": 1e-05,
|
27 |
+
"rope_scaling": null,
|
28 |
+
"rope_theta": 500000.0,
|
29 |
+
"tie_word_embeddings": false,
|
30 |
+
"tokenizer_model_max_length": 2048,
|
31 |
+
"tokenizer_padding_side": "right",
|
32 |
+
"torch_dtype": "bfloat16",
|
33 |
+
"transformers_version": "4.40.0",
|
34 |
+
"tune_mm_mlp_adapter": true,
|
35 |
+
"unfreeze_vision_tower": false,
|
36 |
+
"use_cache": true,
|
37 |
+
"use_mm_proj": true,
|
38 |
+
"vocab_size": 128256
|
39 |
+
}
|
mm_projector.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a0108fa6fb0f15d9ce0678e029d55ab09f571813d7b5db3593665c5570b6f2bc
|
3 |
+
size 43010104
|
trainer_state.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|