Hugging Face's logo

keras
/
llama3.1_guard_8b

Text Generation
KerasHub
Model card Files Community
Divyasreepat commited on
Commit
89264f9
·
verified ·
1 Parent(s): c477554

Upload folder using huggingface_hub

Browse files
Files changed (11) hide show
  1. README.md +26 -0
  2. assets/tokenizer/merges.txt +0 -0
  3. assets/tokenizer/vocabulary.json +0 -0
  4. config.json +23 -0
  5. metadata.json +9 -0
  6. model.weights.json +297 -0
  7. model_00000.weights.h5 +3 -0
  8. model_00001.weights.h5 +3 -0
  9. preprocessor.json +51 -0
  10. task.json +82 -0
  11. tokenizer.json +30 -0
README.md ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: keras-hub
3
+ pipeline_tag: text-generation
4
+ ---
5
+ This is a [`Llama3` model](https://keras.io/api/keras_hub/models/llama3) uploaded using the KerasHub library and can be used with JAX, TensorFlow, and PyTorch backends.
6
+ This model is related to a `CausalLM` task.
7
+
8
+ Model config:
9
+ * **name:** llama3_backbone
10
+ * **trainable:** True
11
+ * **vocabulary_size:** 128256
12
+ * **num_layers:** 32
13
+ * **num_query_heads:** 32
14
+ * **hidden_dim:** 4096
15
+ * **intermediate_dim:** 14336
16
+ * **rope_max_wavelength:** 500000.0
17
+ * **rope_position_scaling_factor:** 1.0
18
+ * **rope_frequency_adjustment_factor:** 8.0
19
+ * **rope_low_freq_factor:** 1.0
20
+ * **rope_high_freq_factor:** 4.0
21
+ * **rope_pretraining_sequence_length:** 8192
22
+ * **num_key_value_heads:** 8
23
+ * **layer_norm_epsilon:** 1e-06
24
+ * **dropout:** 0
25
+
26
+ This model card has been generated automatically and should be completed by the model author. See [Model Cards documentation](https://huggingface.co/docs/hub/model-cards) for more information.
assets/tokenizer/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
assets/tokenizer/vocabulary.json ADDED
The diff for this file is too large to render. See raw diff
 
config.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.llama3.llama3_backbone",
3
+ "class_name": "Llama3Backbone",
4
+ "config": {
5
+ "name": "llama3_backbone",
6
+ "trainable": true,
7
+ "vocabulary_size": 128256,
8
+ "num_layers": 32,
9
+ "num_query_heads": 32,
10
+ "hidden_dim": 4096,
11
+ "intermediate_dim": 14336,
12
+ "rope_max_wavelength": 500000.0,
13
+ "rope_position_scaling_factor": 1.0,
14
+ "rope_frequency_adjustment_factor": 8.0,
15
+ "rope_low_freq_factor": 1.0,
16
+ "rope_high_freq_factor": 4.0,
17
+ "rope_pretraining_sequence_length": 8192,
18
+ "num_key_value_heads": 8,
19
+ "layer_norm_epsilon": 1e-06,
20
+ "dropout": 0
21
+ },
22
+ "registered_name": "keras_hub>Llama3Backbone"
23
+ }
metadata.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "keras_version": "3.10.0",
3
+ "keras_hub_version": "0.21.1",
4
+ "parameter_count": 8030261248,
5
+ "date_saved": "2025-06-06@19:14:25",
6
+ "tasks": [
7
+ "LlamaCausalLM"
8
+ ]
9
+ }
model.weights.json ADDED
@@ -0,0 +1,297 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 16060522496.0
4
+ },
5
+ "weight_map": {
6
+ "/layers/reversible_embedding/vars": "model_00000.weights.h5",
7
+ "/layers/llama_transformer_decoder/_feedforward_gate_dense/vars": "model_00000.weights.h5",
8
+ "/layers/llama_transformer_decoder/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
9
+ "/layers/llama_transformer_decoder/_feedforward_layernorm/vars": "model_00000.weights.h5",
10
+ "/layers/llama_transformer_decoder/_feedforward_output_dense/vars": "model_00000.weights.h5",
11
+ "/layers/llama_transformer_decoder/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
12
+ "/layers/llama_transformer_decoder/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
13
+ "/layers/llama_transformer_decoder/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
14
+ "/layers/llama_transformer_decoder/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
15
+ "/layers/llama_transformer_decoder/_self_attention_layernorm/vars": "model_00000.weights.h5",
16
+ "/layers/llama_transformer_decoder_1/_feedforward_gate_dense/vars": "model_00000.weights.h5",
17
+ "/layers/llama_transformer_decoder_1/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
18
+ "/layers/llama_transformer_decoder_1/_feedforward_layernorm/vars": "model_00000.weights.h5",
19
+ "/layers/llama_transformer_decoder_1/_feedforward_output_dense/vars": "model_00000.weights.h5",
20
+ "/layers/llama_transformer_decoder_1/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
21
+ "/layers/llama_transformer_decoder_1/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
22
+ "/layers/llama_transformer_decoder_1/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
23
+ "/layers/llama_transformer_decoder_1/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
24
+ "/layers/llama_transformer_decoder_1/_self_attention_layernorm/vars": "model_00000.weights.h5",
25
+ "/layers/llama_transformer_decoder_2/_feedforward_gate_dense/vars": "model_00000.weights.h5",
26
+ "/layers/llama_transformer_decoder_2/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
27
+ "/layers/llama_transformer_decoder_2/_feedforward_layernorm/vars": "model_00000.weights.h5",
28
+ "/layers/llama_transformer_decoder_2/_feedforward_output_dense/vars": "model_00000.weights.h5",
29
+ "/layers/llama_transformer_decoder_2/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
30
+ "/layers/llama_transformer_decoder_2/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
31
+ "/layers/llama_transformer_decoder_2/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
32
+ "/layers/llama_transformer_decoder_2/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
33
+ "/layers/llama_transformer_decoder_2/_self_attention_layernorm/vars": "model_00000.weights.h5",
34
+ "/layers/llama_transformer_decoder_3/_feedforward_gate_dense/vars": "model_00000.weights.h5",
35
+ "/layers/llama_transformer_decoder_3/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
36
+ "/layers/llama_transformer_decoder_3/_feedforward_layernorm/vars": "model_00000.weights.h5",
37
+ "/layers/llama_transformer_decoder_3/_feedforward_output_dense/vars": "model_00000.weights.h5",
38
+ "/layers/llama_transformer_decoder_3/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
39
+ "/layers/llama_transformer_decoder_3/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
40
+ "/layers/llama_transformer_decoder_3/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
41
+ "/layers/llama_transformer_decoder_3/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
42
+ "/layers/llama_transformer_decoder_3/_self_attention_layernorm/vars": "model_00000.weights.h5",
43
+ "/layers/llama_transformer_decoder_4/_feedforward_gate_dense/vars": "model_00000.weights.h5",
44
+ "/layers/llama_transformer_decoder_4/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
45
+ "/layers/llama_transformer_decoder_4/_feedforward_layernorm/vars": "model_00000.weights.h5",
46
+ "/layers/llama_transformer_decoder_4/_feedforward_output_dense/vars": "model_00000.weights.h5",
47
+ "/layers/llama_transformer_decoder_4/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
48
+ "/layers/llama_transformer_decoder_4/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
49
+ "/layers/llama_transformer_decoder_4/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
50
+ "/layers/llama_transformer_decoder_4/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
51
+ "/layers/llama_transformer_decoder_4/_self_attention_layernorm/vars": "model_00000.weights.h5",
52
+ "/layers/llama_transformer_decoder_5/_feedforward_gate_dense/vars": "model_00000.weights.h5",
53
+ "/layers/llama_transformer_decoder_5/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
54
+ "/layers/llama_transformer_decoder_5/_feedforward_layernorm/vars": "model_00000.weights.h5",
55
+ "/layers/llama_transformer_decoder_5/_feedforward_output_dense/vars": "model_00000.weights.h5",
56
+ "/layers/llama_transformer_decoder_5/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
57
+ "/layers/llama_transformer_decoder_5/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
58
+ "/layers/llama_transformer_decoder_5/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
59
+ "/layers/llama_transformer_decoder_5/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
60
+ "/layers/llama_transformer_decoder_5/_self_attention_layernorm/vars": "model_00000.weights.h5",
61
+ "/layers/llama_transformer_decoder_6/_feedforward_gate_dense/vars": "model_00000.weights.h5",
62
+ "/layers/llama_transformer_decoder_6/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
63
+ "/layers/llama_transformer_decoder_6/_feedforward_layernorm/vars": "model_00000.weights.h5",
64
+ "/layers/llama_transformer_decoder_6/_feedforward_output_dense/vars": "model_00000.weights.h5",
65
+ "/layers/llama_transformer_decoder_6/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
66
+ "/layers/llama_transformer_decoder_6/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
67
+ "/layers/llama_transformer_decoder_6/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
68
+ "/layers/llama_transformer_decoder_6/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
69
+ "/layers/llama_transformer_decoder_6/_self_attention_layernorm/vars": "model_00000.weights.h5",
70
+ "/layers/llama_transformer_decoder_7/_feedforward_gate_dense/vars": "model_00000.weights.h5",
71
+ "/layers/llama_transformer_decoder_7/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
72
+ "/layers/llama_transformer_decoder_7/_feedforward_layernorm/vars": "model_00000.weights.h5",
73
+ "/layers/llama_transformer_decoder_7/_feedforward_output_dense/vars": "model_00000.weights.h5",
74
+ "/layers/llama_transformer_decoder_7/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
75
+ "/layers/llama_transformer_decoder_7/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
76
+ "/layers/llama_transformer_decoder_7/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
77
+ "/layers/llama_transformer_decoder_7/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
78
+ "/layers/llama_transformer_decoder_7/_self_attention_layernorm/vars": "model_00000.weights.h5",
79
+ "/layers/llama_transformer_decoder_8/_feedforward_gate_dense/vars": "model_00000.weights.h5",
80
+ "/layers/llama_transformer_decoder_8/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
81
+ "/layers/llama_transformer_decoder_8/_feedforward_layernorm/vars": "model_00000.weights.h5",
82
+ "/layers/llama_transformer_decoder_8/_feedforward_output_dense/vars": "model_00000.weights.h5",
83
+ "/layers/llama_transformer_decoder_8/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
84
+ "/layers/llama_transformer_decoder_8/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
85
+ "/layers/llama_transformer_decoder_8/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
86
+ "/layers/llama_transformer_decoder_8/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
87
+ "/layers/llama_transformer_decoder_8/_self_attention_layernorm/vars": "model_00000.weights.h5",
88
+ "/layers/llama_transformer_decoder_9/_feedforward_gate_dense/vars": "model_00000.weights.h5",
89
+ "/layers/llama_transformer_decoder_9/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
90
+ "/layers/llama_transformer_decoder_9/_feedforward_layernorm/vars": "model_00000.weights.h5",
91
+ "/layers/llama_transformer_decoder_9/_feedforward_output_dense/vars": "model_00000.weights.h5",
92
+ "/layers/llama_transformer_decoder_9/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
93
+ "/layers/llama_transformer_decoder_9/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
94
+ "/layers/llama_transformer_decoder_9/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
95
+ "/layers/llama_transformer_decoder_9/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
96
+ "/layers/llama_transformer_decoder_9/_self_attention_layernorm/vars": "model_00000.weights.h5",
97
+ "/layers/llama_transformer_decoder_10/_feedforward_gate_dense/vars": "model_00000.weights.h5",
98
+ "/layers/llama_transformer_decoder_10/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
99
+ "/layers/llama_transformer_decoder_10/_feedforward_layernorm/vars": "model_00000.weights.h5",
100
+ "/layers/llama_transformer_decoder_10/_feedforward_output_dense/vars": "model_00000.weights.h5",
101
+ "/layers/llama_transformer_decoder_10/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
102
+ "/layers/llama_transformer_decoder_10/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
103
+ "/layers/llama_transformer_decoder_10/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
104
+ "/layers/llama_transformer_decoder_10/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
105
+ "/layers/llama_transformer_decoder_10/_self_attention_layernorm/vars": "model_00000.weights.h5",
106
+ "/layers/llama_transformer_decoder_11/_feedforward_gate_dense/vars": "model_00000.weights.h5",
107
+ "/layers/llama_transformer_decoder_11/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
108
+ "/layers/llama_transformer_decoder_11/_feedforward_layernorm/vars": "model_00000.weights.h5",
109
+ "/layers/llama_transformer_decoder_11/_feedforward_output_dense/vars": "model_00000.weights.h5",
110
+ "/layers/llama_transformer_decoder_11/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
111
+ "/layers/llama_transformer_decoder_11/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
112
+ "/layers/llama_transformer_decoder_11/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
113
+ "/layers/llama_transformer_decoder_11/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
114
+ "/layers/llama_transformer_decoder_11/_self_attention_layernorm/vars": "model_00000.weights.h5",
115
+ "/layers/llama_transformer_decoder_12/_feedforward_gate_dense/vars": "model_00000.weights.h5",
116
+ "/layers/llama_transformer_decoder_12/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
117
+ "/layers/llama_transformer_decoder_12/_feedforward_layernorm/vars": "model_00000.weights.h5",
118
+ "/layers/llama_transformer_decoder_12/_feedforward_output_dense/vars": "model_00000.weights.h5",
119
+ "/layers/llama_transformer_decoder_12/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
120
+ "/layers/llama_transformer_decoder_12/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
121
+ "/layers/llama_transformer_decoder_12/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
122
+ "/layers/llama_transformer_decoder_12/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
123
+ "/layers/llama_transformer_decoder_12/_self_attention_layernorm/vars": "model_00000.weights.h5",
124
+ "/layers/llama_transformer_decoder_13/_feedforward_gate_dense/vars": "model_00000.weights.h5",
125
+ "/layers/llama_transformer_decoder_13/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
126
+ "/layers/llama_transformer_decoder_13/_feedforward_layernorm/vars": "model_00000.weights.h5",
127
+ "/layers/llama_transformer_decoder_13/_feedforward_output_dense/vars": "model_00000.weights.h5",
128
+ "/layers/llama_transformer_decoder_13/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
129
+ "/layers/llama_transformer_decoder_13/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
130
+ "/layers/llama_transformer_decoder_13/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
131
+ "/layers/llama_transformer_decoder_13/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
132
+ "/layers/llama_transformer_decoder_13/_self_attention_layernorm/vars": "model_00000.weights.h5",
133
+ "/layers/llama_transformer_decoder_14/_feedforward_gate_dense/vars": "model_00000.weights.h5",
134
+ "/layers/llama_transformer_decoder_14/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
135
+ "/layers/llama_transformer_decoder_14/_feedforward_layernorm/vars": "model_00000.weights.h5",
136
+ "/layers/llama_transformer_decoder_14/_feedforward_output_dense/vars": "model_00000.weights.h5",
137
+ "/layers/llama_transformer_decoder_14/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
138
+ "/layers/llama_transformer_decoder_14/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
139
+ "/layers/llama_transformer_decoder_14/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
140
+ "/layers/llama_transformer_decoder_14/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
141
+ "/layers/llama_transformer_decoder_14/_self_attention_layernorm/vars": "model_00000.weights.h5",
142
+ "/layers/llama_transformer_decoder_15/_feedforward_gate_dense/vars": "model_00000.weights.h5",
143
+ "/layers/llama_transformer_decoder_15/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
144
+ "/layers/llama_transformer_decoder_15/_feedforward_layernorm/vars": "model_00000.weights.h5",
145
+ "/layers/llama_transformer_decoder_15/_feedforward_output_dense/vars": "model_00000.weights.h5",
146
+ "/layers/llama_transformer_decoder_15/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
147
+ "/layers/llama_transformer_decoder_15/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
148
+ "/layers/llama_transformer_decoder_15/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
149
+ "/layers/llama_transformer_decoder_15/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
150
+ "/layers/llama_transformer_decoder_15/_self_attention_layernorm/vars": "model_00000.weights.h5",
151
+ "/layers/llama_transformer_decoder_16/_feedforward_gate_dense/vars": "model_00000.weights.h5",
152
+ "/layers/llama_transformer_decoder_16/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
153
+ "/layers/llama_transformer_decoder_16/_feedforward_layernorm/vars": "model_00000.weights.h5",
154
+ "/layers/llama_transformer_decoder_16/_feedforward_output_dense/vars": "model_00000.weights.h5",
155
+ "/layers/llama_transformer_decoder_16/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
156
+ "/layers/llama_transformer_decoder_16/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
157
+ "/layers/llama_transformer_decoder_16/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
158
+ "/layers/llama_transformer_decoder_16/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
159
+ "/layers/llama_transformer_decoder_16/_self_attention_layernorm/vars": "model_00000.weights.h5",
160
+ "/layers/llama_transformer_decoder_17/_feedforward_gate_dense/vars": "model_00000.weights.h5",
161
+ "/layers/llama_transformer_decoder_17/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
162
+ "/layers/llama_transformer_decoder_17/_feedforward_layernorm/vars": "model_00000.weights.h5",
163
+ "/layers/llama_transformer_decoder_17/_feedforward_output_dense/vars": "model_00000.weights.h5",
164
+ "/layers/llama_transformer_decoder_17/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
165
+ "/layers/llama_transformer_decoder_17/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
166
+ "/layers/llama_transformer_decoder_17/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
167
+ "/layers/llama_transformer_decoder_17/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
168
+ "/layers/llama_transformer_decoder_17/_self_attention_layernorm/vars": "model_00000.weights.h5",
169
+ "/layers/llama_transformer_decoder_18/_feedforward_gate_dense/vars": "model_00000.weights.h5",
170
+ "/layers/llama_transformer_decoder_18/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
171
+ "/layers/llama_transformer_decoder_18/_feedforward_layernorm/vars": "model_00000.weights.h5",
172
+ "/layers/llama_transformer_decoder_18/_feedforward_output_dense/vars": "model_00000.weights.h5",
173
+ "/layers/llama_transformer_decoder_18/_self_attention_layer/_key_dense/vars": "model_00000.weights.h5",
174
+ "/layers/llama_transformer_decoder_18/_self_attention_layer/_output_dense/vars": "model_00000.weights.h5",
175
+ "/layers/llama_transformer_decoder_18/_self_attention_layer/_query_dense/vars": "model_00000.weights.h5",
176
+ "/layers/llama_transformer_decoder_18/_self_attention_layer/_value_dense/vars": "model_00000.weights.h5",
177
+ "/layers/llama_transformer_decoder_18/_self_attention_layernorm/vars": "model_00000.weights.h5",
178
+ "/layers/llama_transformer_decoder_19/_feedforward_gate_dense/vars": "model_00000.weights.h5",
179
+ "/layers/llama_transformer_decoder_19/_feedforward_intermediate_dense/vars": "model_00000.weights.h5",
180
+ "/layers/llama_transformer_decoder_19/_feedforward_layernorm/vars": "model_00000.weights.h5",
181
+ "/layers/llama_transformer_decoder_19/_feedforward_output_dense/vars": "model_00001.weights.h5",
182
+ "/layers/llama_transformer_decoder_19/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
183
+ "/layers/llama_transformer_decoder_19/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
184
+ "/layers/llama_transformer_decoder_19/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
185
+ "/layers/llama_transformer_decoder_19/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
186
+ "/layers/llama_transformer_decoder_19/_self_attention_layernorm/vars": "model_00001.weights.h5",
187
+ "/layers/llama_transformer_decoder_20/_feedforward_gate_dense/vars": "model_00001.weights.h5",
188
+ "/layers/llama_transformer_decoder_20/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
189
+ "/layers/llama_transformer_decoder_20/_feedforward_layernorm/vars": "model_00001.weights.h5",
190
+ "/layers/llama_transformer_decoder_20/_feedforward_output_dense/vars": "model_00001.weights.h5",
191
+ "/layers/llama_transformer_decoder_20/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
192
+ "/layers/llama_transformer_decoder_20/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
193
+ "/layers/llama_transformer_decoder_20/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
194
+ "/layers/llama_transformer_decoder_20/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
195
+ "/layers/llama_transformer_decoder_20/_self_attention_layernorm/vars": "model_00001.weights.h5",
196
+ "/layers/llama_transformer_decoder_21/_feedforward_gate_dense/vars": "model_00001.weights.h5",
197
+ "/layers/llama_transformer_decoder_21/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
198
+ "/layers/llama_transformer_decoder_21/_feedforward_layernorm/vars": "model_00001.weights.h5",
199
+ "/layers/llama_transformer_decoder_21/_feedforward_output_dense/vars": "model_00001.weights.h5",
200
+ "/layers/llama_transformer_decoder_21/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
201
+ "/layers/llama_transformer_decoder_21/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
202
+ "/layers/llama_transformer_decoder_21/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
203
+ "/layers/llama_transformer_decoder_21/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
204
+ "/layers/llama_transformer_decoder_21/_self_attention_layernorm/vars": "model_00001.weights.h5",
205
+ "/layers/llama_transformer_decoder_22/_feedforward_gate_dense/vars": "model_00001.weights.h5",
206
+ "/layers/llama_transformer_decoder_22/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
207
+ "/layers/llama_transformer_decoder_22/_feedforward_layernorm/vars": "model_00001.weights.h5",
208
+ "/layers/llama_transformer_decoder_22/_feedforward_output_dense/vars": "model_00001.weights.h5",
209
+ "/layers/llama_transformer_decoder_22/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
210
+ "/layers/llama_transformer_decoder_22/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
211
+ "/layers/llama_transformer_decoder_22/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
212
+ "/layers/llama_transformer_decoder_22/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
213
+ "/layers/llama_transformer_decoder_22/_self_attention_layernorm/vars": "model_00001.weights.h5",
214
+ "/layers/llama_transformer_decoder_23/_feedforward_gate_dense/vars": "model_00001.weights.h5",
215
+ "/layers/llama_transformer_decoder_23/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
216
+ "/layers/llama_transformer_decoder_23/_feedforward_layernorm/vars": "model_00001.weights.h5",
217
+ "/layers/llama_transformer_decoder_23/_feedforward_output_dense/vars": "model_00001.weights.h5",
218
+ "/layers/llama_transformer_decoder_23/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
219
+ "/layers/llama_transformer_decoder_23/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
220
+ "/layers/llama_transformer_decoder_23/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
221
+ "/layers/llama_transformer_decoder_23/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
222
+ "/layers/llama_transformer_decoder_23/_self_attention_layernorm/vars": "model_00001.weights.h5",
223
+ "/layers/llama_transformer_decoder_24/_feedforward_gate_dense/vars": "model_00001.weights.h5",
224
+ "/layers/llama_transformer_decoder_24/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
225
+ "/layers/llama_transformer_decoder_24/_feedforward_layernorm/vars": "model_00001.weights.h5",
226
+ "/layers/llama_transformer_decoder_24/_feedforward_output_dense/vars": "model_00001.weights.h5",
227
+ "/layers/llama_transformer_decoder_24/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
228
+ "/layers/llama_transformer_decoder_24/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
229
+ "/layers/llama_transformer_decoder_24/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
230
+ "/layers/llama_transformer_decoder_24/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
231
+ "/layers/llama_transformer_decoder_24/_self_attention_layernorm/vars": "model_00001.weights.h5",
232
+ "/layers/llama_transformer_decoder_25/_feedforward_gate_dense/vars": "model_00001.weights.h5",
233
+ "/layers/llama_transformer_decoder_25/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
234
+ "/layers/llama_transformer_decoder_25/_feedforward_layernorm/vars": "model_00001.weights.h5",
235
+ "/layers/llama_transformer_decoder_25/_feedforward_output_dense/vars": "model_00001.weights.h5",
236
+ "/layers/llama_transformer_decoder_25/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
237
+ "/layers/llama_transformer_decoder_25/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
238
+ "/layers/llama_transformer_decoder_25/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
239
+ "/layers/llama_transformer_decoder_25/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
240
+ "/layers/llama_transformer_decoder_25/_self_attention_layernorm/vars": "model_00001.weights.h5",
241
+ "/layers/llama_transformer_decoder_26/_feedforward_gate_dense/vars": "model_00001.weights.h5",
242
+ "/layers/llama_transformer_decoder_26/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
243
+ "/layers/llama_transformer_decoder_26/_feedforward_layernorm/vars": "model_00001.weights.h5",
244
+ "/layers/llama_transformer_decoder_26/_feedforward_output_dense/vars": "model_00001.weights.h5",
245
+ "/layers/llama_transformer_decoder_26/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
246
+ "/layers/llama_transformer_decoder_26/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
247
+ "/layers/llama_transformer_decoder_26/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
248
+ "/layers/llama_transformer_decoder_26/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
249
+ "/layers/llama_transformer_decoder_26/_self_attention_layernorm/vars": "model_00001.weights.h5",
250
+ "/layers/llama_transformer_decoder_27/_feedforward_gate_dense/vars": "model_00001.weights.h5",
251
+ "/layers/llama_transformer_decoder_27/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
252
+ "/layers/llama_transformer_decoder_27/_feedforward_layernorm/vars": "model_00001.weights.h5",
253
+ "/layers/llama_transformer_decoder_27/_feedforward_output_dense/vars": "model_00001.weights.h5",
254
+ "/layers/llama_transformer_decoder_27/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
255
+ "/layers/llama_transformer_decoder_27/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
256
+ "/layers/llama_transformer_decoder_27/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
257
+ "/layers/llama_transformer_decoder_27/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
258
+ "/layers/llama_transformer_decoder_27/_self_attention_layernorm/vars": "model_00001.weights.h5",
259
+ "/layers/llama_transformer_decoder_28/_feedforward_gate_dense/vars": "model_00001.weights.h5",
260
+ "/layers/llama_transformer_decoder_28/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
261
+ "/layers/llama_transformer_decoder_28/_feedforward_layernorm/vars": "model_00001.weights.h5",
262
+ "/layers/llama_transformer_decoder_28/_feedforward_output_dense/vars": "model_00001.weights.h5",
263
+ "/layers/llama_transformer_decoder_28/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
264
+ "/layers/llama_transformer_decoder_28/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
265
+ "/layers/llama_transformer_decoder_28/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
266
+ "/layers/llama_transformer_decoder_28/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
267
+ "/layers/llama_transformer_decoder_28/_self_attention_layernorm/vars": "model_00001.weights.h5",
268
+ "/layers/llama_transformer_decoder_29/_feedforward_gate_dense/vars": "model_00001.weights.h5",
269
+ "/layers/llama_transformer_decoder_29/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
270
+ "/layers/llama_transformer_decoder_29/_feedforward_layernorm/vars": "model_00001.weights.h5",
271
+ "/layers/llama_transformer_decoder_29/_feedforward_output_dense/vars": "model_00001.weights.h5",
272
+ "/layers/llama_transformer_decoder_29/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
273
+ "/layers/llama_transformer_decoder_29/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
274
+ "/layers/llama_transformer_decoder_29/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
275
+ "/layers/llama_transformer_decoder_29/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
276
+ "/layers/llama_transformer_decoder_29/_self_attention_layernorm/vars": "model_00001.weights.h5",
277
+ "/layers/llama_transformer_decoder_30/_feedforward_gate_dense/vars": "model_00001.weights.h5",
278
+ "/layers/llama_transformer_decoder_30/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
279
+ "/layers/llama_transformer_decoder_30/_feedforward_layernorm/vars": "model_00001.weights.h5",
280
+ "/layers/llama_transformer_decoder_30/_feedforward_output_dense/vars": "model_00001.weights.h5",
281
+ "/layers/llama_transformer_decoder_30/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
282
+ "/layers/llama_transformer_decoder_30/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
283
+ "/layers/llama_transformer_decoder_30/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
284
+ "/layers/llama_transformer_decoder_30/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
285
+ "/layers/llama_transformer_decoder_30/_self_attention_layernorm/vars": "model_00001.weights.h5",
286
+ "/layers/llama_transformer_decoder_31/_feedforward_gate_dense/vars": "model_00001.weights.h5",
287
+ "/layers/llama_transformer_decoder_31/_feedforward_intermediate_dense/vars": "model_00001.weights.h5",
288
+ "/layers/llama_transformer_decoder_31/_feedforward_layernorm/vars": "model_00001.weights.h5",
289
+ "/layers/llama_transformer_decoder_31/_feedforward_output_dense/vars": "model_00001.weights.h5",
290
+ "/layers/llama_transformer_decoder_31/_self_attention_layer/_key_dense/vars": "model_00001.weights.h5",
291
+ "/layers/llama_transformer_decoder_31/_self_attention_layer/_output_dense/vars": "model_00001.weights.h5",
292
+ "/layers/llama_transformer_decoder_31/_self_attention_layer/_query_dense/vars": "model_00001.weights.h5",
293
+ "/layers/llama_transformer_decoder_31/_self_attention_layer/_value_dense/vars": "model_00001.weights.h5",
294
+ "/layers/llama_transformer_decoder_31/_self_attention_layernorm/vars": "model_00001.weights.h5",
295
+ "/layers/llama_layer_norm/vars": "model_00001.weights.h5"
296
+ }
297
+ }
model_00000.weights.h5 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba5ce07c0c09eb1056cd129d45ece7ece037ee2038805122fc4e481164760289
3
+ size 10625162328
model_00001.weights.h5 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d4b90ac674a46df912adc82592d189197b8a14512351a5c7b492104e2a895f9
3
+ size 5436473240
preprocessor.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.llama3.llama3_causal_lm_preprocessor",
3
+ "class_name": "Llama3CausalLMPreprocessor",
4
+ "config": {
5
+ "name": "llama3_causal_lm_preprocessor_1",
6
+ "trainable": true,
7
+ "dtype": {
8
+ "module": "keras",
9
+ "class_name": "DTypePolicy",
10
+ "config": {
11
+ "name": "float32"
12
+ },
13
+ "registered_name": null
14
+ },
15
+ "tokenizer": {
16
+ "module": "keras_hub.src.models.llama3.llama3_tokenizer",
17
+ "class_name": "Llama3Tokenizer",
18
+ "config": {
19
+ "name": "llama3_tokenizer_1",
20
+ "trainable": true,
21
+ "dtype": {
22
+ "module": "keras",
23
+ "class_name": "DTypePolicy",
24
+ "config": {
25
+ "name": "int32"
26
+ },
27
+ "registered_name": null
28
+ },
29
+ "config_file": "tokenizer.json",
30
+ "sequence_length": null,
31
+ "add_prefix_space": false,
32
+ "unsplittable_tokens": [
33
+ "<|eom_id|>",
34
+ "<|python_tag|>",
35
+ "<|finetune_right_pad_id|>",
36
+ "<|begin_of_text|>",
37
+ "<|start_header_id|>",
38
+ "<|end_of_text|>",
39
+ "<|eot_id|>",
40
+ "<|end_header_id|>"
41
+ ]
42
+ },
43
+ "registered_name": "keras_hub>Llama3Tokenizer"
44
+ },
45
+ "config_file": "preprocessor.json",
46
+ "sequence_length": 1024,
47
+ "add_start_token": true,
48
+ "add_end_token": true
49
+ },
50
+ "registered_name": "keras_hub>Llama3CausalLMPreprocessor"
51
+ }
task.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.llama3.llama3_causal_lm",
3
+ "class_name": "Llama3CausalLM",
4
+ "config": {
5
+ "backbone": {
6
+ "module": "keras_hub.src.models.llama3.llama3_backbone",
7
+ "class_name": "Llama3Backbone",
8
+ "config": {
9
+ "name": "llama3_backbone",
10
+ "trainable": true,
11
+ "vocabulary_size": 128256,
12
+ "num_layers": 32,
13
+ "num_query_heads": 32,
14
+ "hidden_dim": 4096,
15
+ "intermediate_dim": 14336,
16
+ "rope_max_wavelength": 500000.0,
17
+ "rope_position_scaling_factor": 1.0,
18
+ "rope_frequency_adjustment_factor": 8.0,
19
+ "rope_low_freq_factor": 1.0,
20
+ "rope_high_freq_factor": 4.0,
21
+ "rope_pretraining_sequence_length": 8192,
22
+ "num_key_value_heads": 8,
23
+ "layer_norm_epsilon": 1e-06,
24
+ "dropout": 0
25
+ },
26
+ "registered_name": "keras_hub>Llama3Backbone"
27
+ },
28
+ "preprocessor": {
29
+ "module": "keras_hub.src.models.llama3.llama3_causal_lm_preprocessor",
30
+ "class_name": "Llama3CausalLMPreprocessor",
31
+ "config": {
32
+ "name": "llama3_causal_lm_preprocessor_1",
33
+ "trainable": true,
34
+ "dtype": {
35
+ "module": "keras",
36
+ "class_name": "DTypePolicy",
37
+ "config": {
38
+ "name": "float32"
39
+ },
40
+ "registered_name": null
41
+ },
42
+ "tokenizer": {
43
+ "module": "keras_hub.src.models.llama3.llama3_tokenizer",
44
+ "class_name": "Llama3Tokenizer",
45
+ "config": {
46
+ "name": "llama3_tokenizer_1",
47
+ "trainable": true,
48
+ "dtype": {
49
+ "module": "keras",
50
+ "class_name": "DTypePolicy",
51
+ "config": {
52
+ "name": "int32"
53
+ },
54
+ "registered_name": null
55
+ },
56
+ "config_file": "tokenizer.json",
57
+ "sequence_length": null,
58
+ "add_prefix_space": false,
59
+ "unsplittable_tokens": [
60
+ "<|eom_id|>",
61
+ "<|python_tag|>",
62
+ "<|finetune_right_pad_id|>",
63
+ "<|begin_of_text|>",
64
+ "<|start_header_id|>",
65
+ "<|end_of_text|>",
66
+ "<|eot_id|>",
67
+ "<|end_header_id|>"
68
+ ]
69
+ },
70
+ "registered_name": "keras_hub>Llama3Tokenizer"
71
+ },
72
+ "config_file": "preprocessor.json",
73
+ "sequence_length": 1024,
74
+ "add_start_token": true,
75
+ "add_end_token": true
76
+ },
77
+ "registered_name": "keras_hub>Llama3CausalLMPreprocessor"
78
+ },
79
+ "name": "llama3_causal_lm"
80
+ },
81
+ "registered_name": "keras_hub>Llama3CausalLM"
82
+ }
tokenizer.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "module": "keras_hub.src.models.llama3.llama3_tokenizer",
3
+ "class_name": "Llama3Tokenizer",
4
+ "config": {
5
+ "name": "llama3_tokenizer_1",
6
+ "trainable": true,
7
+ "dtype": {
8
+ "module": "keras",
9
+ "class_name": "DTypePolicy",
10
+ "config": {
11
+ "name": "int32"
12
+ },
13
+ "registered_name": null
14
+ },
15
+ "config_file": "tokenizer.json",
16
+ "sequence_length": null,
17
+ "add_prefix_space": false,
18
+ "unsplittable_tokens": [
19
+ "<|eom_id|>",
20
+ "<|python_tag|>",
21
+ "<|finetune_right_pad_id|>",
22
+ "<|begin_of_text|>",
23
+ "<|start_header_id|>",
24
+ "<|end_of_text|>",
25
+ "<|eot_id|>",
26
+ "<|end_header_id|>"
27
+ ]
28
+ },
29
+ "registered_name": "keras_hub>Llama3Tokenizer"
30
+ }