Hugging Face's logo

unsloth
/
gemma-3n-E4B-unsloth-bnb-4bit

Image-Text-to-Text
Transformers
Safetensors
English
gemma3n
gemma3
unsloth
gemma
google
4-bit precision
bitsandbytes
Model card Files
xet
Community
1
danielhanchen commited on
Commit
be510eb
·
verified ·
1 Parent(s): c3a5b29

Upload folder using huggingface_hub

Browse files
Files changed (13) hide show
  1. .gitattributes +1 -0
  2. config.json +222 -0
  3. generation_config.json +10 -0
  4. model-00001-of-00003.safetensors +3 -0
  5. model-00002-of-00003.safetensors +3 -0
  6. model-00003-of-00003.safetensors +3 -0
  7. model.safetensors.index.json +0 -0
  8. preprocessor_config.json +50 -0
  9. processor_config.json +5 -0
  10. special_tokens_map.json +36 -0
  11. tokenizer.json +3 -0
  12. tokenizer.model +3 -0
  13. tokenizer_config.json +0 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
config.json ADDED
@@ -0,0 +1,222 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Gemma3nForConditionalGeneration"
4
+ ],
5
+ "audio_config": {
6
+ "conf_attention_chunk_size": 12,
7
+ "conf_attention_context_left": 13,
8
+ "conf_attention_context_right": 0,
9
+ "conf_attention_logit_cap": 50.0,
10
+ "conf_conv_kernel_size": 5,
11
+ "conf_num_attention_heads": 8,
12
+ "conf_num_hidden_layers": 12,
13
+ "conf_positional_bias_size": 256,
14
+ "conf_reduction_factor": 4,
15
+ "conf_residual_weight": 0.5,
16
+ "gradient_clipping": 10000000000.0,
17
+ "hidden_size": 1536,
18
+ "input_feat_size": 128,
19
+ "model_type": "gemma3n_audio",
20
+ "rms_norm_eps": 1e-06,
21
+ "sscp_conv_channel_size": [
22
+ 128,
23
+ 32
24
+ ],
25
+ "sscp_conv_eps": 0.001,
26
+ "sscp_conv_kernel_size": [
27
+ [
28
+ 3,
29
+ 3
30
+ ],
31
+ [
32
+ 3,
33
+ 3
34
+ ]
35
+ ],
36
+ "sscp_conv_stride_size": [
37
+ [
38
+ 2,
39
+ 2
40
+ ],
41
+ [
42
+ 2,
43
+ 2
44
+ ]
45
+ ],
46
+ "torch_dtype": "bfloat16",
47
+ "vocab_offset": 262272,
48
+ "vocab_size": 128
49
+ },
50
+ "audio_soft_tokens_per_image": 188,
51
+ "audio_token_id": 262273,
52
+ "boa_token_id": 256000,
53
+ "boi_token_id": 255999,
54
+ "bos_token_id": 2,
55
+ "eoa_token_id": 262272,
56
+ "eoi_token_id": 262144,
57
+ "eos_token_id": 1,
58
+ "image_token_id": 262145,
59
+ "initializer_range": 0.02,
60
+ "model_type": "gemma3n",
61
+ "pad_token_id": 0,
62
+ "quantization_config": {
63
+ "_load_in_4bit": true,
64
+ "_load_in_8bit": false,
65
+ "bnb_4bit_compute_dtype": "bfloat16",
66
+ "bnb_4bit_quant_storage": "uint8",
67
+ "bnb_4bit_quant_type": "nf4",
68
+ "bnb_4bit_use_double_quant": true,
69
+ "llm_int8_enable_fp32_cpu_offload": false,
70
+ "llm_int8_has_fp16_weight": false,
71
+ "llm_int8_skip_modules": [
72
+ "embed_tokens",
73
+ "embedding",
74
+ "lm_head",
75
+ "multi_modal_projector",
76
+ "merger",
77
+ "modality_projection",
78
+ "correction_coefs",
79
+ "prediction_coefs",
80
+ "modality_router",
81
+ "embedding_projection",
82
+ "input_proj_linear",
83
+ "laurel",
84
+ "per_layer_input_gate",
85
+ "per_layer_projection",
86
+ "audio_tower",
87
+ "embed_vision",
88
+ "embed_audio"
89
+ ],
90
+ "llm_int8_threshold": 6.0,
91
+ "load_in_4bit": true,
92
+ "load_in_8bit": false,
93
+ "quant_method": "bitsandbytes"
94
+ },
95
+ "text_config": {
96
+ "activation_sparsity_pattern": [
97
+ 0.95,
98
+ 0.95,
99
+ 0.95,
100
+ 0.95,
101
+ 0.95,
102
+ 0.95,
103
+ 0.95,
104
+ 0.95,
105
+ 0.95,
106
+ 0.95,
107
+ 0.0,
108
+ 0.0,
109
+ 0.0,
110
+ 0.0,
111
+ 0.0,
112
+ 0.0,
113
+ 0.0,
114
+ 0.0,
115
+ 0.0,
116
+ 0.0,
117
+ 0.0,
118
+ 0.0,
119
+ 0.0,
120
+ 0.0,
121
+ 0.0,
122
+ 0.0,
123
+ 0.0,
124
+ 0.0,
125
+ 0.0,
126
+ 0.0,
127
+ 0.0,
128
+ 0.0,
129
+ 0.0,
130
+ 0.0,
131
+ 0.0
132
+ ],
133
+ "altup_active_idx": 0,
134
+ "altup_coef_clip": 120.0,
135
+ "altup_correct_scale": true,
136
+ "altup_lr_multiplier": 1.0,
137
+ "altup_num_inputs": 4,
138
+ "attention_bias": false,
139
+ "attention_dropout": 0.0,
140
+ "final_logit_softcapping": 30.0,
141
+ "head_dim": 256,
142
+ "hidden_activation": "gelu_pytorch_tanh",
143
+ "hidden_size": 2048,
144
+ "hidden_size_per_layer_input": 256,
145
+ "initializer_range": 0.02,
146
+ "intermediate_size": 16384,
147
+ "laurel_rank": 64,
148
+ "layer_types": [
149
+ "sliding_attention",
150
+ "sliding_attention",
151
+ "sliding_attention",
152
+ "sliding_attention",
153
+ "full_attention",
154
+ "sliding_attention",
155
+ "sliding_attention",
156
+ "sliding_attention",
157
+ "sliding_attention",
158
+ "full_attention",
159
+ "sliding_attention",
160
+ "sliding_attention",
161
+ "sliding_attention",
162
+ "sliding_attention",
163
+ "full_attention",
164
+ "sliding_attention",
165
+ "sliding_attention",
166
+ "sliding_attention",
167
+ "sliding_attention",
168
+ "full_attention",
169
+ "sliding_attention",
170
+ "sliding_attention",
171
+ "sliding_attention",
172
+ "sliding_attention",
173
+ "full_attention",
174
+ "sliding_attention",
175
+ "sliding_attention",
176
+ "sliding_attention",
177
+ "sliding_attention",
178
+ "full_attention",
179
+ "sliding_attention",
180
+ "sliding_attention",
181
+ "sliding_attention",
182
+ "sliding_attention",
183
+ "full_attention"
184
+ ],
185
+ "max_position_embeddings": 32768,
186
+ "model_type": "gemma3n_text",
187
+ "num_attention_heads": 8,
188
+ "num_hidden_layers": 35,
189
+ "num_key_value_heads": 2,
190
+ "num_kv_shared_layers": 15,
191
+ "query_pre_attn_scalar": 256,
192
+ "rms_norm_eps": 1e-06,
193
+ "rope_local_base_freq": 10000.0,
194
+ "rope_scaling": null,
195
+ "rope_theta": 1000000.0,
196
+ "sliding_window": 512,
197
+ "torch_dtype": "bfloat16",
198
+ "use_cache": true,
199
+ "vocab_size": 262400,
200
+ "vocab_size_per_layer_input": 262144
201
+ },
202
+ "torch_dtype": "bfloat16",
203
+ "transformers_version": "4.53.0.dev0",
204
+ "unsloth_fixed": true,
205
+ "vision_config": {
206
+ "architecture": "mobilenetv5_300m_enc",
207
+ "do_pooling": true,
208
+ "hidden_size": 2048,
209
+ "initializer_range": 0.02,
210
+ "label_names": [
211
+ "LABEL_0",
212
+ "LABEL_1"
213
+ ],
214
+ "model_type": "gemma3n_vision",
215
+ "num_classes": 2,
216
+ "rms_norm_eps": 1e-06,
217
+ "torch_dtype": "bfloat16",
218
+ "vocab_offset": 262144,
219
+ "vocab_size": 128
220
+ },
221
+ "vision_soft_tokens_per_image": 256
222
+ }
generation_config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 2,
3
+ "cache_implementation": "hybrid",
4
+ "do_sample": true,
5
+ "eos_token_id": 1,
6
+ "pad_token_id": 0,
7
+ "top_k": 64,
8
+ "top_p": 0.95,
9
+ "transformers_version": "4.53.0.dev0"
10
+ }
model-00001-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16c89123fa97a11ba7b2c7e5f20d3c190262e0ed4c3f67d51f49ebe46af6f0f6
3
+ size 3764063764
model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4388f4421ff1e6ac422773f523859e369677bd8114bb82652618ee2392ad38c
3
+ size 4997088663
model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ac3f6647ce2f7b24a770caa89249e284ffb1ec77ebaad3c97581c3836ad6ad0
3
+ size 1101342016
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
preprocessor_config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "crop_size": null,
3
+ "data_format": "channels_first",
4
+ "default_to_square": false,
5
+ "device": null,
6
+ "dither": 0.0,
7
+ "do_center_crop": null,
8
+ "do_convert_rgb": null,
9
+ "do_normalize": false,
10
+ "do_rescale": true,
11
+ "do_resize": true,
12
+ "feature_size": 128,
13
+ "fft_length": 1024,
14
+ "fft_overdrive": true,
15
+ "frame_length": 512,
16
+ "hop_length": 160,
17
+ "image_mean": [
18
+ 0.5,
19
+ 0.5,
20
+ 0.5
21
+ ],
22
+ "image_processor_type": "SiglipImageProcessorFast",
23
+ "image_seq_length": 256,
24
+ "image_std": [
25
+ 0.5,
26
+ 0.5,
27
+ 0.5
28
+ ],
29
+ "input_data_format": null,
30
+ "input_scale_factor": 1.0,
31
+ "max_frequency": 7600.0,
32
+ "mel_floor": 1e-05,
33
+ "min_frequency": 125.0,
34
+ "padding_side": "left",
35
+ "padding_value": 0.0,
36
+ "per_bin_mean": null,
37
+ "per_bin_stddev": null,
38
+ "preemphasis": 0.97,
39
+ "preemphasis_htk_flavor": true,
40
+ "processor_class": "Gemma3nProcessor",
41
+ "resample": 2,
42
+ "rescale_factor": 0.00392156862745098,
43
+ "return_attention_mask": false,
44
+ "return_tensors": null,
45
+ "sampling_rate": 16000,
46
+ "size": {
47
+ "height": 768,
48
+ "width": 768
49
+ }
50
+ }
processor_config.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "audio_seq_length": 188,
3
+ "image_seq_length": 256,
4
+ "processor_class": "Gemma3nProcessor"
5
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "audio_token": "<audio_soft_token>",
3
+ "boa_token": "<start_of_audio>",
4
+ "boi_token": "<start_of_image>",
5
+ "bos_token": {
6
+ "content": "<bos>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false
11
+ },
12
+ "eoa_token": "<end_of_audio>",
13
+ "eoi_token": "<end_of_image>",
14
+ "eos_token": {
15
+ "content": "<eos>",
16
+ "lstrip": false,
17
+ "normalized": false,
18
+ "rstrip": false,
19
+ "single_word": false
20
+ },
21
+ "image_token": "<image_soft_token>",
22
+ "pad_token": {
23
+ "content": "<pad>",
24
+ "lstrip": false,
25
+ "normalized": false,
26
+ "rstrip": false,
27
+ "single_word": false
28
+ },
29
+ "unk_token": {
30
+ "content": "<unk>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false
35
+ }
36
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b6c35ee648c07754b44cd9e371c75d4caa05c4504910b7ad29b1847ee9d8ba5d
3
+ size 33442553
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ea5f0cc48abfbfc04d14562270a32e02149a3e7035f368cc5a462786f4a59961
3
+ size 4696020
tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff