danielhanchen commited on
Commit
2b40dca
·
verified ·
1 Parent(s): ac5b1ff

Upload folder using huggingface_hub

Browse files
config.json CHANGED
@@ -10,7 +10,6 @@
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,
@@ -22,7 +21,6 @@
22
  128,
23
  32
24
  ],
25
- "sscp_conv_eps": 0.001,
26
  "sscp_conv_group_norm_eps": 0.001,
27
  "sscp_conv_kernel_size": [
28
  [
@@ -129,7 +127,6 @@
129
  "altup_active_idx": 0,
130
  "altup_coef_clip": 120.0,
131
  "altup_correct_scale": true,
132
- "altup_lr_multiplier": 1.0,
133
  "altup_num_inputs": 4,
134
  "attention_bias": false,
135
  "attention_dropout": 0.0,
@@ -210,7 +207,6 @@
210
  "num_hidden_layers": 30,
211
  "num_key_value_heads": 2,
212
  "num_kv_shared_layers": 10,
213
- "query_pre_attn_scalar": 256,
214
  "rms_norm_eps": 1e-06,
215
  "rope_local_base_freq": 10000.0,
216
  "rope_scaling": null,
@@ -222,11 +218,11 @@
222
  "vocab_size_per_layer_input": 262144
223
  },
224
  "torch_dtype": "bfloat16",
225
- "transformers_version": "4.53.0",
226
  "unsloth_fixed": true,
227
  "vision_config": {
228
  "architecture": "mobilenetv5_300m_enc",
229
- "do_pooling": true,
230
  "hidden_size": 2048,
231
  "initializer_range": 0.02,
232
  "label_names": [
 
10
  "conf_conv_kernel_size": 5,
11
  "conf_num_attention_heads": 8,
12
  "conf_num_hidden_layers": 12,
 
13
  "conf_reduction_factor": 4,
14
  "conf_residual_weight": 0.5,
15
  "gradient_clipping": 10000000000.0,
 
21
  128,
22
  32
23
  ],
 
24
  "sscp_conv_group_norm_eps": 0.001,
25
  "sscp_conv_kernel_size": [
26
  [
 
127
  "altup_active_idx": 0,
128
  "altup_coef_clip": 120.0,
129
  "altup_correct_scale": true,
 
130
  "altup_num_inputs": 4,
131
  "attention_bias": false,
132
  "attention_dropout": 0.0,
 
207
  "num_hidden_layers": 30,
208
  "num_key_value_heads": 2,
209
  "num_kv_shared_layers": 10,
 
210
  "rms_norm_eps": 1e-06,
211
  "rope_local_base_freq": 10000.0,
212
  "rope_scaling": null,
 
218
  "vocab_size_per_layer_input": 262144
219
  },
220
  "torch_dtype": "bfloat16",
221
+ "transformers_version": "4.53.1",
222
  "unsloth_fixed": true,
223
  "vision_config": {
224
  "architecture": "mobilenetv5_300m_enc",
225
+ "do_pooling": false,
226
  "hidden_size": 2048,
227
  "initializer_range": 0.02,
228
  "label_names": [
generation_config.json CHANGED
@@ -6,5 +6,5 @@
6
  "pad_token_id": 0,
7
  "top_k": 64,
8
  "top_p": 0.95,
9
- "transformers_version": "4.53.0"
10
  }
 
6
  "pad_token_id": 0,
7
  "top_k": 64,
8
  "top_p": 0.95,
9
+ "transformers_version": "4.53.1"
10
  }
model-00001-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:01a9c74c1d69b300df4b0816755a72b10aac6cb6239a2dd56f9fb3d8e3f28bbe
3
  size 2650152952
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dcb67227b2b0f92e9c47eed6cad816fcc226a6b0cf9cd8f6bd1e3b834419ef70
3
  size 2650152952
preprocessor_config.json CHANGED
@@ -41,7 +41,7 @@
41
  "processor_class": "Gemma3nProcessor",
42
  "resample": 2,
43
  "rescale_factor": 0.00392156862745098,
44
- "return_attention_mask": false,
45
  "return_tensors": null,
46
  "sampling_rate": 16000,
47
  "size": {
 
41
  "processor_class": "Gemma3nProcessor",
42
  "resample": 2,
43
  "rescale_factor": 0.00392156862745098,
44
+ "return_attention_mask": true,
45
  "return_tensors": null,
46
  "sampling_rate": 16000,
47
  "size": {