Upload Idefics3ForConditionalGeneration
Browse files- config.json +18 -4
- generation_config.json +7 -0
- model.safetensors +3 -0
config.json
CHANGED
@@ -1,5 +1,4 @@
|
|
1 |
{
|
2 |
-
"_name_or_path": "alikirec/SmolVLM-256M-Instruct-read-function-graph-v0",
|
3 |
"architectures": [
|
4 |
"Idefics3ForConditionalGeneration"
|
5 |
],
|
@@ -13,12 +12,16 @@
|
|
13 |
"architectures": [
|
14 |
"VLlama3ForCausalLM"
|
15 |
],
|
|
|
|
|
16 |
"head_dim": 64,
|
|
|
17 |
"hidden_size": 576,
|
18 |
"initializer_range": 0.041666666666666664,
|
19 |
"intermediate_size": 1536,
|
20 |
"is_llama_config": true,
|
21 |
"max_position_embeddings": 8192,
|
|
|
22 |
"model_type": "llama",
|
23 |
"neftune_noise_alpha": 0.0,
|
24 |
"num_attention_heads": 9,
|
@@ -99,17 +102,20 @@
|
|
99 |
"use_bfloat16": false
|
100 |
},
|
101 |
"pixel_shuffle_factor": 4,
|
|
|
102 |
"qk_layer_norms": false,
|
103 |
"rms_norm_eps": 1e-05,
|
104 |
"rope_interleaved": false,
|
|
|
105 |
"rope_theta": 100000,
|
106 |
-
"torch_dtype": "
|
107 |
"transformers.js_config": {
|
108 |
"kv_cache_dtype": {
|
109 |
"fp16": "float16",
|
110 |
"q4f16": "float16"
|
111 |
}
|
112 |
},
|
|
|
113 |
"use_resampler": false,
|
114 |
"vocab_size": 49280
|
115 |
},
|
@@ -121,22 +127,30 @@
|
|
121 |
"q4f16": "float16"
|
122 |
}
|
123 |
},
|
124 |
-
"transformers_version": "4.
|
125 |
"use_cache": true,
|
126 |
"vision_config": {
|
|
|
|
|
127 |
"hidden_size": 768,
|
128 |
"image_size": 512,
|
|
|
|
|
|
|
129 |
"max_image_size": {
|
130 |
"longest_edge": 512
|
131 |
},
|
132 |
"model_type": "idefics3_vision",
|
133 |
"num_attention_heads": 12,
|
|
|
|
|
134 |
"patch_size": 16,
|
135 |
"size": {
|
136 |
"longest_edge": 2048
|
137 |
},
|
138 |
"tie_word_embeddings": false,
|
|
|
139 |
"use_base_siglip": true
|
140 |
},
|
141 |
"vocab_size": 49280
|
142 |
-
}
|
|
|
1 |
{
|
|
|
2 |
"architectures": [
|
3 |
"Idefics3ForConditionalGeneration"
|
4 |
],
|
|
|
12 |
"architectures": [
|
13 |
"VLlama3ForCausalLM"
|
14 |
],
|
15 |
+
"attention_bias": false,
|
16 |
+
"attention_dropout": 0.0,
|
17 |
"head_dim": 64,
|
18 |
+
"hidden_act": "silu",
|
19 |
"hidden_size": 576,
|
20 |
"initializer_range": 0.041666666666666664,
|
21 |
"intermediate_size": 1536,
|
22 |
"is_llama_config": true,
|
23 |
"max_position_embeddings": 8192,
|
24 |
+
"mlp_bias": false,
|
25 |
"model_type": "llama",
|
26 |
"neftune_noise_alpha": 0.0,
|
27 |
"num_attention_heads": 9,
|
|
|
102 |
"use_bfloat16": false
|
103 |
},
|
104 |
"pixel_shuffle_factor": 4,
|
105 |
+
"pretraining_tp": 1,
|
106 |
"qk_layer_norms": false,
|
107 |
"rms_norm_eps": 1e-05,
|
108 |
"rope_interleaved": false,
|
109 |
+
"rope_scaling": null,
|
110 |
"rope_theta": 100000,
|
111 |
+
"torch_dtype": "float32",
|
112 |
"transformers.js_config": {
|
113 |
"kv_cache_dtype": {
|
114 |
"fp16": "float16",
|
115 |
"q4f16": "float16"
|
116 |
}
|
117 |
},
|
118 |
+
"use_cache": true,
|
119 |
"use_resampler": false,
|
120 |
"vocab_size": 49280
|
121 |
},
|
|
|
127 |
"q4f16": "float16"
|
128 |
}
|
129 |
},
|
130 |
+
"transformers_version": "4.51.3",
|
131 |
"use_cache": true,
|
132 |
"vision_config": {
|
133 |
+
"attention_dropout": 0.0,
|
134 |
+
"hidden_act": "gelu_pytorch_tanh",
|
135 |
"hidden_size": 768,
|
136 |
"image_size": 512,
|
137 |
+
"initializer_range": 0.02,
|
138 |
+
"intermediate_size": 3072,
|
139 |
+
"layer_norm_eps": 1e-06,
|
140 |
"max_image_size": {
|
141 |
"longest_edge": 512
|
142 |
},
|
143 |
"model_type": "idefics3_vision",
|
144 |
"num_attention_heads": 12,
|
145 |
+
"num_channels": 3,
|
146 |
+
"num_hidden_layers": 12,
|
147 |
"patch_size": 16,
|
148 |
"size": {
|
149 |
"longest_edge": 2048
|
150 |
},
|
151 |
"tie_word_embeddings": false,
|
152 |
+
"torch_dtype": "float32",
|
153 |
"use_base_siglip": true
|
154 |
},
|
155 |
"vocab_size": 49280
|
156 |
+
}
|
generation_config.json
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_from_model_config": true,
|
3 |
+
"bos_token_id": 0,
|
4 |
+
"eos_token_id": 49279,
|
5 |
+
"pad_token_id": 2,
|
6 |
+
"transformers_version": "4.51.3"
|
7 |
+
}
|
model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6c6fd68ac6074a39734d86e7907f5bd8ff981f048845adb7a166334a96d9e2b7
|
3 |
+
size 1025998224
|