BoyaWu10 commited on
Commit
d700265
1 Parent(s): 673c0e3

Upload checkpoints

Browse files
Files changed (4) hide show
  1. README.md +14 -0
  2. config.json +48 -0
  3. mm_projector.bin +3 -0
  4. trainer_state.json +0 -0
README.md CHANGED
@@ -1,3 +1,17 @@
1
  ---
 
2
  license: apache-2.0
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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-minicpm-siglip is the pretrained weights for [bunny-minicpm-siglip-lora](https://huggingface.co/BoyaWu10/bunny-minicpm-siglip-lora), which leverages MiniCPM 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 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "bunny-pretrain-minicpm-siglip",
3
+ "architectures": [
4
+ "MiniCPMForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "auto_map": {
9
+ "AutoConfig": "configuration_minicpm.MiniCPMConfig",
10
+ "AutoModel": "modeling_minicpm.MiniCPMModel",
11
+ "AutoModelForCausalLM": "modeling_minicpm.MiniCPMForCausalLM",
12
+ "AutoModelForSeq2SeqLM": "modeling_minicpm.MiniCPMForCausalLM",
13
+ "AutoModelForSequenceClassification": "modeling_minicpm.MiniCPMForSequenceClassification"
14
+ },
15
+ "bos_token_id": 1,
16
+ "dim_model_base": 256,
17
+ "eos_token_id": 2,
18
+ "freeze_mm_mlp_adapter": false,
19
+ "hidden_act": "silu",
20
+ "hidden_size": 2304,
21
+ "image_aspect_ratio": "square",
22
+ "initializer_range": 0.1,
23
+ "intermediate_size": 5760,
24
+ "max_position_embeddings": 2048,
25
+ "mm_hidden_size": 1152,
26
+ "mm_projector_lr": null,
27
+ "mm_projector_type": "mlp2x_gelu",
28
+ "mm_vision_tower": "google/siglip-so400m-patch14-384",
29
+ "model_type": "bunny-minicpm",
30
+ "num_attention_heads": 36,
31
+ "num_hidden_layers": 40,
32
+ "num_key_value_heads": 36,
33
+ "pretraining_tp": 1,
34
+ "rms_norm_eps": 1e-05,
35
+ "rope_scaling": null,
36
+ "rope_theta": 10000.0,
37
+ "scale_depth": 1.4,
38
+ "scale_emb": 12,
39
+ "tokenizer_model_max_length": 2048,
40
+ "tokenizer_padding_side": "right",
41
+ "torch_dtype": "bfloat16",
42
+ "transformers_version": "4.38.2",
43
+ "tune_mm_mlp_adapter": true,
44
+ "unfreeze_vision_tower": false,
45
+ "use_cache": true,
46
+ "use_mm_proj": true,
47
+ "vocab_size": 122753
48
+ }
mm_projector.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83c64e2540daa9d75978b80b962544a7eefc82dfef9845e247b9f379ac9a9022
3
+ size 15936568
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff