Update tiny models for MiniCPMV4_7ForConditionalGeneration
Browse files- config.json +86 -0
- generation_config.json +10 -0
- model.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,86 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"MiniCPMV4_7ForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"downsample_mode": "16x",
|
| 6 |
+
"drop_vision_last_layer": false,
|
| 7 |
+
"dtype": "float32",
|
| 8 |
+
"image_end_id": 11,
|
| 9 |
+
"image_size": 32,
|
| 10 |
+
"image_start_id": 10,
|
| 11 |
+
"image_token_id": 100,
|
| 12 |
+
"insert_layer_id": 0,
|
| 13 |
+
"merge_kernel_size": [
|
| 14 |
+
2,
|
| 15 |
+
2
|
| 16 |
+
],
|
| 17 |
+
"merger_times": 1,
|
| 18 |
+
"model_type": "minicpmv4_7",
|
| 19 |
+
"newline_id": 14,
|
| 20 |
+
"patch_size": 8,
|
| 21 |
+
"slice_end_id": 13,
|
| 22 |
+
"slice_start_id": 12,
|
| 23 |
+
"text_config": {
|
| 24 |
+
"attention_bias": false,
|
| 25 |
+
"attention_dropout": 0.0,
|
| 26 |
+
"bos_token_id": 0,
|
| 27 |
+
"eos_token_id": 1,
|
| 28 |
+
"head_dim": 32,
|
| 29 |
+
"hidden_act": "silu",
|
| 30 |
+
"hidden_size": 64,
|
| 31 |
+
"initializer_range": 0.02,
|
| 32 |
+
"intermediate_size": 37,
|
| 33 |
+
"layer_types": [
|
| 34 |
+
"full_attention",
|
| 35 |
+
"linear_attention"
|
| 36 |
+
],
|
| 37 |
+
"linear_conv_kernel_dim": 2,
|
| 38 |
+
"linear_key_head_dim": 16,
|
| 39 |
+
"linear_num_key_heads": 4,
|
| 40 |
+
"linear_num_value_heads": 8,
|
| 41 |
+
"linear_value_head_dim": 16,
|
| 42 |
+
"max_position_embeddings": 512,
|
| 43 |
+
"model_type": "qwen3_5_text",
|
| 44 |
+
"num_attention_heads": 4,
|
| 45 |
+
"num_hidden_layers": 2,
|
| 46 |
+
"num_key_value_heads": 2,
|
| 47 |
+
"pad_token_id": 2,
|
| 48 |
+
"partial_rotary_factor": 0.25,
|
| 49 |
+
"rms_norm_eps": 1e-06,
|
| 50 |
+
"rope_parameters": {
|
| 51 |
+
"mrope_section": [
|
| 52 |
+
2,
|
| 53 |
+
1,
|
| 54 |
+
1
|
| 55 |
+
],
|
| 56 |
+
"partial_rotary_factor": 0.25,
|
| 57 |
+
"rope_theta": 10000,
|
| 58 |
+
"rope_type": "default",
|
| 59 |
+
"type": "default"
|
| 60 |
+
},
|
| 61 |
+
"tie_word_embeddings": true,
|
| 62 |
+
"use_cache": true,
|
| 63 |
+
"vocab_size": 256
|
| 64 |
+
},
|
| 65 |
+
"tie_word_embeddings": false,
|
| 66 |
+
"transformers_version": "5.19.0.dev0",
|
| 67 |
+
"video_token_id": 101,
|
| 68 |
+
"vision_config": {
|
| 69 |
+
"attention_dropout": 0.0,
|
| 70 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 71 |
+
"hidden_size": 64,
|
| 72 |
+
"image_size": 32,
|
| 73 |
+
"insert_layer_id": 0,
|
| 74 |
+
"intermediate_size": 128,
|
| 75 |
+
"layer_norm_eps": 1e-06,
|
| 76 |
+
"model_type": "minicpmv4_7_vision",
|
| 77 |
+
"num_attention_heads": 4,
|
| 78 |
+
"num_channels": 3,
|
| 79 |
+
"num_hidden_layers": 2,
|
| 80 |
+
"patch_size": 8,
|
| 81 |
+
"window_kernel_size": [
|
| 82 |
+
2,
|
| 83 |
+
2
|
| 84 |
+
]
|
| 85 |
+
}
|
| 86 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 0,
|
| 4 |
+
"eos_token_id": 1,
|
| 5 |
+
"output_attentions": false,
|
| 6 |
+
"output_hidden_states": false,
|
| 7 |
+
"pad_token_id": 2,
|
| 8 |
+
"transformers_version": "5.19.0.dev0",
|
| 9 |
+
"use_cache": true
|
| 10 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d4a758b59e76b6d4de94c69057cb9442c1d9c3c6dbd21dfa53d39ebb35ea3db2
|
| 3 |
+
size 1847544
|