Update tiny models for Gemma3nTextModel
Browse files- .gitattributes +1 -0
- config.json +63 -0
- model.safetensors +3 -0
- preprocessor_config.json +51 -0
- tokenizer.json +3 -0
- tokenizer_config.json +31 -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 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"activation_sparsity_pattern": [
|
| 3 |
+
0.0,
|
| 4 |
+
0.0,
|
| 5 |
+
0.0,
|
| 6 |
+
0.0
|
| 7 |
+
],
|
| 8 |
+
"altup_active_idx": 0,
|
| 9 |
+
"altup_coef_clip": 120.0,
|
| 10 |
+
"altup_correct_scale": true,
|
| 11 |
+
"altup_num_inputs": 2,
|
| 12 |
+
"architectures": [
|
| 13 |
+
"Gemma3nTextModel"
|
| 14 |
+
],
|
| 15 |
+
"attention_bias": false,
|
| 16 |
+
"attention_dropout": 0.0,
|
| 17 |
+
"bos_token_id": 2,
|
| 18 |
+
"dtype": "float32",
|
| 19 |
+
"eos_token_id": 1,
|
| 20 |
+
"final_logit_softcapping": 30.0,
|
| 21 |
+
"head_dim": 8,
|
| 22 |
+
"hidden_activation": "gelu_pytorch_tanh",
|
| 23 |
+
"hidden_size": 16,
|
| 24 |
+
"hidden_size_per_layer_input": 16,
|
| 25 |
+
"initializer_range": 0.02,
|
| 26 |
+
"intermediate_size": [
|
| 27 |
+
22,
|
| 28 |
+
22,
|
| 29 |
+
22,
|
| 30 |
+
22
|
| 31 |
+
],
|
| 32 |
+
"laurel_rank": 64,
|
| 33 |
+
"layer_types": [
|
| 34 |
+
"full_attention",
|
| 35 |
+
"sliding_attention",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"sliding_attention"
|
| 38 |
+
],
|
| 39 |
+
"max_position_embeddings": 512,
|
| 40 |
+
"model_type": "gemma3n_text",
|
| 41 |
+
"num_attention_heads": 2,
|
| 42 |
+
"num_hidden_layers": 4,
|
| 43 |
+
"num_key_value_heads": 2,
|
| 44 |
+
"num_kv_shared_layers": 2,
|
| 45 |
+
"pad_token_id": 0,
|
| 46 |
+
"rms_norm_eps": 1e-06,
|
| 47 |
+
"rope_parameters": {
|
| 48 |
+
"full_attention": {
|
| 49 |
+
"rope_theta": 1000000.0,
|
| 50 |
+
"rope_type": "default"
|
| 51 |
+
},
|
| 52 |
+
"sliding_attention": {
|
| 53 |
+
"rope_theta": 10000.0,
|
| 54 |
+
"rope_type": "default"
|
| 55 |
+
}
|
| 56 |
+
},
|
| 57 |
+
"sliding_window": 512,
|
| 58 |
+
"tie_word_embeddings": true,
|
| 59 |
+
"transformers_version": "5.16.0.dev0",
|
| 60 |
+
"use_cache": true,
|
| 61 |
+
"vocab_size": 262400,
|
| 62 |
+
"vocab_size_per_layer_input": 99
|
| 63 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fc260a38f84fb22c0e4a2bf418a6475d09d1588d21139e3b9c841005e7e924c0
|
| 3 |
+
size 16908296
|
preprocessor_config.json
ADDED
|
@@ -0,0 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"crop_size": null,
|
| 3 |
+
"data_format": "channels_first",
|
| 4 |
+
"default_to_square": false,
|
| 5 |
+
"device": null,
|
| 6 |
+
"disable_grouping": null,
|
| 7 |
+
"dither": 0.0,
|
| 8 |
+
"do_center_crop": null,
|
| 9 |
+
"do_convert_rgb": null,
|
| 10 |
+
"do_normalize": false,
|
| 11 |
+
"do_rescale": true,
|
| 12 |
+
"do_resize": true,
|
| 13 |
+
"feature_extractor_type": "Gemma3nAudioFeatureExtractor",
|
| 14 |
+
"feature_size": 128,
|
| 15 |
+
"fft_length": 1024,
|
| 16 |
+
"fft_overdrive": true,
|
| 17 |
+
"frame_length": 512,
|
| 18 |
+
"hop_length": 160,
|
| 19 |
+
"image_mean": [
|
| 20 |
+
0.5,
|
| 21 |
+
0.5,
|
| 22 |
+
0.5
|
| 23 |
+
],
|
| 24 |
+
"image_processor_type": "SiglipImageProcessorFast",
|
| 25 |
+
"image_seq_length": 256,
|
| 26 |
+
"image_std": [
|
| 27 |
+
0.5,
|
| 28 |
+
0.5,
|
| 29 |
+
0.5
|
| 30 |
+
],
|
| 31 |
+
"input_data_format": null,
|
| 32 |
+
"input_scale_factor": 1.0,
|
| 33 |
+
"max_frequency": 7600.0,
|
| 34 |
+
"mel_floor": 1e-05,
|
| 35 |
+
"min_frequency": 125.0,
|
| 36 |
+
"padding_side": "right",
|
| 37 |
+
"padding_value": 0.0,
|
| 38 |
+
"per_bin_mean": null,
|
| 39 |
+
"per_bin_stddev": null,
|
| 40 |
+
"preemphasis": 0.97,
|
| 41 |
+
"preemphasis_htk_flavor": true,
|
| 42 |
+
"resample": 2,
|
| 43 |
+
"rescale_factor": 0.00392156862745098,
|
| 44 |
+
"return_attention_mask": true,
|
| 45 |
+
"return_tensors": null,
|
| 46 |
+
"sampling_rate": 16000,
|
| 47 |
+
"size": {
|
| 48 |
+
"height": 768,
|
| 49 |
+
"width": 768
|
| 50 |
+
}
|
| 51 |
+
}
|
tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6aa40b46d55d1fa5f1c0b218de3a695942c620858ba34aa94e2aa705efc6f071
|
| 3 |
+
size 33442552
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"audio_token": "<audio_soft_token>",
|
| 3 |
+
"backend": "tokenizers",
|
| 4 |
+
"boa_token": "<start_of_audio>",
|
| 5 |
+
"boi_token": "<start_of_image>",
|
| 6 |
+
"bos_token": "<bos>",
|
| 7 |
+
"clean_up_tokenization_spaces": false,
|
| 8 |
+
"eoa_token": "<end_of_audio>",
|
| 9 |
+
"eoi_token": "<end_of_image>",
|
| 10 |
+
"eos_token": "<eos>",
|
| 11 |
+
"image_token": "<image_soft_token>",
|
| 12 |
+
"is_local": true,
|
| 13 |
+
"local_files_only": false,
|
| 14 |
+
"mask_token": "<mask>",
|
| 15 |
+
"model_max_length": 512,
|
| 16 |
+
"model_specific_special_tokens": {
|
| 17 |
+
"audio_token": "<audio_soft_token>",
|
| 18 |
+
"boa_token": "<start_of_audio>",
|
| 19 |
+
"boi_token": "<start_of_image>",
|
| 20 |
+
"eoa_token": "<end_of_audio>",
|
| 21 |
+
"eoi_token": "<end_of_image>",
|
| 22 |
+
"image_token": "<image_soft_token>"
|
| 23 |
+
},
|
| 24 |
+
"pad_token": "<pad>",
|
| 25 |
+
"processor_class": "Gemma3nProcessor",
|
| 26 |
+
"sp_model_kwargs": null,
|
| 27 |
+
"spaces_between_special_tokens": false,
|
| 28 |
+
"tokenizer_class": "GemmaTokenizer",
|
| 29 |
+
"unk_token": "<unk>",
|
| 30 |
+
"use_default_system_prompt": false
|
| 31 |
+
}
|