Hugging Face's logo

mlx-community
/
Mamba-Codestral-7B-v0.1-4bit

MLX
Safetensors
Model card Files Community
1

Upload folder using huggingface_hub

#1
by Goekdeniz-Guelmez - opened
base: refs/heads/main
←
from: refs/pr/1
Discussion Files changed
+282879
-0
Files changed (8) hide show
  1. README.md +34 -0
  2. config.json +51 -0
  3. model.safetensors +3 -0
  4. model.safetensors.index.json +846 -0
  5. special_tokens_map.json +23 -0
  6. tokenizer.json +0 -0
  7. tokenizer.model +3 -0
  8. tokenizer_config.json +0 -0
README.md ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: mistralai/Mamba-Codestral-7B-v0.1
3
+ license: apache-2.0
4
+ tags:
5
+ - mlx
6
+ extra_gated_description: If you want to learn more about how we process your personal
7
+ data, please read our <a href="https://mistral.ai/terms/">Privacy Policy</a>.
8
+ ---
9
+
10
+ # mlx-community/Mamba-Codestral-7B-v0.1-4bit
11
+
12
+ The Model [mlx-community/Mamba-Codestral-7B-v0.1-4bit](https://huggingface.co/mlx-community/Mamba-Codestral-7B-v0.1-4bit) was converted to MLX format from [mistralai/Mamba-Codestral-7B-v0.1](https://huggingface.co/mistralai/Mamba-Codestral-7B-v0.1) using mlx-lm version **0.18.2**.
13
+
14
+ ## Use with mlx
15
+
16
+ ```bash
17
+ pip install mlx-lm
18
+ ```
19
+
20
+ ```python
21
+ from mlx_lm import load, generate
22
+
23
+ model, tokenizer = load("mlx-community/Mamba-Codestral-7B-v0.1-4bit")
24
+
25
+ prompt="hello"
26
+
27
+ if hasattr(tokenizer, "apply_chat_template") and tokenizer.chat_template is not None:
28
+ messages = [{"role": "user", "content": prompt}]
29
+ prompt = tokenizer.apply_chat_template(
30
+ messages, tokenize=False, add_generation_prompt=True
31
+ )
32
+
33
+ response = generate(model, tokenizer, prompt=prompt, verbose=True)
34
+ ```
config.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Mamba2ForCausalLM"
4
+ ],
5
+ "bos_token_id": 0,
6
+ "chunk_size": 256,
7
+ "conv_kernel": 4,
8
+ "eos_token_id": 0,
9
+ "expand": 2,
10
+ "head_dim": 64,
11
+ "hidden_act": "silu",
12
+ "hidden_size": 4096,
13
+ "initializer_range": 0.1,
14
+ "intermediate_size": 8192,
15
+ "layer_norm_epsilon": 1e-05,
16
+ "model_type": "mamba2",
17
+ "n_groups": 8,
18
+ "norm_before_gate": true,
19
+ "num_heads": 128,
20
+ "num_hidden_layers": 64,
21
+ "pad_token_id": 0,
22
+ "quantization": {
23
+ "group_size": 64,
24
+ "bits": 4
25
+ },
26
+ "quantization_config": {
27
+ "group_size": 64,
28
+ "bits": 4
29
+ },
30
+ "rescale_prenorm_residual": false,
31
+ "residual_in_fp32": true,
32
+ "rms_norm": true,
33
+ "state_size": 128,
34
+ "tie_word_embeddings": false,
35
+ "time_step_floor": 0.0001,
36
+ "time_step_init_scheme": "random",
37
+ "time_step_limit": [
38
+ 0.0,
39
+ Infinity
40
+ ],
41
+ "time_step_max": 0.1,
42
+ "time_step_min": 0.001,
43
+ "time_step_rank": 256,
44
+ "time_step_scale": 1.0,
45
+ "torch_dtype": "bfloat16",
46
+ "transformers_version": "4.44.0.dev0",
47
+ "use_bias": false,
48
+ "use_cache": true,
49
+ "use_conv_bias": true,
50
+ "vocab_size": 32768
51
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21a0f0ce8194f8ef5726f7da5ebfed3bc10217b1d23e8ee8d05179b52992cf22
3
+ size 4104014183
model.safetensors.index.json ADDED
@@ -0,0 +1,846 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 4103921664
4
+ },
5
+ "weight_map": {
6
+ "backbone.embeddings.biases": "model.safetensors",
7
+ "backbone.embeddings.scales": "model.safetensors",
8
+ "backbone.embeddings.weight": "model.safetensors",
9
+ "backbone.layers.0.mixer.A_log": "model.safetensors",
10
+ "backbone.layers.0.mixer.D": "model.safetensors",
11
+ "backbone.layers.0.mixer.conv1d.bias": "model.safetensors",
12
+ "backbone.layers.0.mixer.conv1d.weight": "model.safetensors",
13
+ "backbone.layers.0.mixer.dt_bias": "model.safetensors",
14
+ "backbone.layers.0.mixer.in_proj.biases": "model.safetensors",
15
+ "backbone.layers.0.mixer.in_proj.scales": "model.safetensors",
16
+ "backbone.layers.0.mixer.in_proj.weight": "model.safetensors",
17
+ "backbone.layers.0.mixer.norm.weight": "model.safetensors",
18
+ "backbone.layers.0.mixer.out_proj.biases": "model.safetensors",
19
+ "backbone.layers.0.mixer.out_proj.scales": "model.safetensors",
20
+ "backbone.layers.0.mixer.out_proj.weight": "model.safetensors",
21
+ "backbone.layers.0.norm.weight": "model.safetensors",
22
+ "backbone.layers.1.mixer.A_log": "model.safetensors",
23
+ "backbone.layers.1.mixer.D": "model.safetensors",
24
+ "backbone.layers.1.mixer.conv1d.bias": "model.safetensors",
25
+ "backbone.layers.1.mixer.conv1d.weight": "model.safetensors",
26
+ "backbone.layers.1.mixer.dt_bias": "model.safetensors",
27
+ "backbone.layers.1.mixer.in_proj.biases": "model.safetensors",
28
+ "backbone.layers.1.mixer.in_proj.scales": "model.safetensors",
29
+ "backbone.layers.1.mixer.in_proj.weight": "model.safetensors",
30
+ "backbone.layers.1.mixer.norm.weight": "model.safetensors",
31
+ "backbone.layers.1.mixer.out_proj.biases": "model.safetensors",
32
+ "backbone.layers.1.mixer.out_proj.scales": "model.safetensors",
33
+ "backbone.layers.1.mixer.out_proj.weight": "model.safetensors",
34
+ "backbone.layers.1.norm.weight": "model.safetensors",
35
+ "backbone.layers.10.mixer.A_log": "model.safetensors",
36
+ "backbone.layers.10.mixer.D": "model.safetensors",
37
+ "backbone.layers.10.mixer.conv1d.bias": "model.safetensors",
38
+ "backbone.layers.10.mixer.conv1d.weight": "model.safetensors",
39
+ "backbone.layers.10.mixer.dt_bias": "model.safetensors",
40
+ "backbone.layers.10.mixer.in_proj.biases": "model.safetensors",
41
+ "backbone.layers.10.mixer.in_proj.scales": "model.safetensors",
42
+ "backbone.layers.10.mixer.in_proj.weight": "model.safetensors",
43
+ "backbone.layers.10.mixer.norm.weight": "model.safetensors",
44
+ "backbone.layers.10.mixer.out_proj.biases": "model.safetensors",
45
+ "backbone.layers.10.mixer.out_proj.scales": "model.safetensors",
46
+ "backbone.layers.10.mixer.out_proj.weight": "model.safetensors",
47
+ "backbone.layers.10.norm.weight": "model.safetensors",
48
+ "backbone.layers.11.mixer.A_log": "model.safetensors",
49
+ "backbone.layers.11.mixer.D": "model.safetensors",
50
+ "backbone.layers.11.mixer.conv1d.bias": "model.safetensors",
51
+ "backbone.layers.11.mixer.conv1d.weight": "model.safetensors",
52
+ "backbone.layers.11.mixer.dt_bias": "model.safetensors",
53
+ "backbone.layers.11.mixer.in_proj.biases": "model.safetensors",
54
+ "backbone.layers.11.mixer.in_proj.scales": "model.safetensors",
55
+ "backbone.layers.11.mixer.in_proj.weight": "model.safetensors",
56
+ "backbone.layers.11.mixer.norm.weight": "model.safetensors",
57
+ "backbone.layers.11.mixer.out_proj.biases": "model.safetensors",
58
+ "backbone.layers.11.mixer.out_proj.scales": "model.safetensors",
59
+ "backbone.layers.11.mixer.out_proj.weight": "model.safetensors",
60
+ "backbone.layers.11.norm.weight": "model.safetensors",
61
+ "backbone.layers.12.mixer.A_log": "model.safetensors",
62
+ "backbone.layers.12.mixer.D": "model.safetensors",
63
+ "backbone.layers.12.mixer.conv1d.bias": "model.safetensors",
64
+ "backbone.layers.12.mixer.conv1d.weight": "model.safetensors",
65
+ "backbone.layers.12.mixer.dt_bias": "model.safetensors",
66
+ "backbone.layers.12.mixer.in_proj.biases": "model.safetensors",
67
+ "backbone.layers.12.mixer.in_proj.scales": "model.safetensors",
68
+ "backbone.layers.12.mixer.in_proj.weight": "model.safetensors",
69
+ "backbone.layers.12.mixer.norm.weight": "model.safetensors",
70
+ "backbone.layers.12.mixer.out_proj.biases": "model.safetensors",
71
+ "backbone.layers.12.mixer.out_proj.scales": "model.safetensors",
72
+ "backbone.layers.12.mixer.out_proj.weight": "model.safetensors",
73
+ "backbone.layers.12.norm.weight": "model.safetensors",
74
+ "backbone.layers.13.mixer.A_log": "model.safetensors",
75
+ "backbone.layers.13.mixer.D": "model.safetensors",
76
+ "backbone.layers.13.mixer.conv1d.bias": "model.safetensors",
77
+ "backbone.layers.13.mixer.conv1d.weight": "model.safetensors",
78
+ "backbone.layers.13.mixer.dt_bias": "model.safetensors",
79
+ "backbone.layers.13.mixer.in_proj.biases": "model.safetensors",
80
+ "backbone.layers.13.mixer.in_proj.scales": "model.safetensors",
81
+ "backbone.layers.13.mixer.in_proj.weight": "model.safetensors",
82
+ "backbone.layers.13.mixer.norm.weight": "model.safetensors",
83
+ "backbone.layers.13.mixer.out_proj.biases": "model.safetensors",
84
+ "backbone.layers.13.mixer.out_proj.scales": "model.safetensors",
85
+ "backbone.layers.13.mixer.out_proj.weight": "model.safetensors",
86
+ "backbone.layers.13.norm.weight": "model.safetensors",
87
+ "backbone.layers.14.mixer.A_log": "model.safetensors",
88
+ "backbone.layers.14.mixer.D": "model.safetensors",
89
+ "backbone.layers.14.mixer.conv1d.bias": "model.safetensors",
90
+ "backbone.layers.14.mixer.conv1d.weight": "model.safetensors",
91
+ "backbone.layers.14.mixer.dt_bias": "model.safetensors",
92
+ "backbone.layers.14.mixer.in_proj.biases": "model.safetensors",
93
+ "backbone.layers.14.mixer.in_proj.scales": "model.safetensors",
94
+ "backbone.layers.14.mixer.in_proj.weight": "model.safetensors",
95
+ "backbone.layers.14.mixer.norm.weight": "model.safetensors",
96
+ "backbone.layers.14.mixer.out_proj.biases": "model.safetensors",
97
+ "backbone.layers.14.mixer.out_proj.scales": "model.safetensors",
98
+ "backbone.layers.14.mixer.out_proj.weight": "model.safetensors",
99
+ "backbone.layers.14.norm.weight": "model.safetensors",
100
+ "backbone.layers.15.mixer.A_log": "model.safetensors",
101
+ "backbone.layers.15.mixer.D": "model.safetensors",
102
+ "backbone.layers.15.mixer.conv1d.bias": "model.safetensors",
103
+ "backbone.layers.15.mixer.conv1d.weight": "model.safetensors",
104
+ "backbone.layers.15.mixer.dt_bias": "model.safetensors",
105
+ "backbone.layers.15.mixer.in_proj.biases": "model.safetensors",
106
+ "backbone.layers.15.mixer.in_proj.scales": "model.safetensors",
107
+ "backbone.layers.15.mixer.in_proj.weight": "model.safetensors",
108
+ "backbone.layers.15.mixer.norm.weight": "model.safetensors",
109
+ "backbone.layers.15.mixer.out_proj.biases": "model.safetensors",
110
+ "backbone.layers.15.mixer.out_proj.scales": "model.safetensors",
111
+ "backbone.layers.15.mixer.out_proj.weight": "model.safetensors",
112
+ "backbone.layers.15.norm.weight": "model.safetensors",
113
+ "backbone.layers.16.mixer.A_log": "model.safetensors",
114
+ "backbone.layers.16.mixer.D": "model.safetensors",
115
+ "backbone.layers.16.mixer.conv1d.bias": "model.safetensors",
116
+ "backbone.layers.16.mixer.conv1d.weight": "model.safetensors",
117
+ "backbone.layers.16.mixer.dt_bias": "model.safetensors",
118
+ "backbone.layers.16.mixer.in_proj.biases": "model.safetensors",
119
+ "backbone.layers.16.mixer.in_proj.scales": "model.safetensors",
120
+ "backbone.layers.16.mixer.in_proj.weight": "model.safetensors",
121
+ "backbone.layers.16.mixer.norm.weight": "model.safetensors",
122
+ "backbone.layers.16.mixer.out_proj.biases": "model.safetensors",
123
+ "backbone.layers.16.mixer.out_proj.scales": "model.safetensors",
124
+ "backbone.layers.16.mixer.out_proj.weight": "model.safetensors",
125
+ "backbone.layers.16.norm.weight": "model.safetensors",
126
+ "backbone.layers.17.mixer.A_log": "model.safetensors",
127
+ "backbone.layers.17.mixer.D": "model.safetensors",
128
+ "backbone.layers.17.mixer.conv1d.bias": "model.safetensors",
129
+ "backbone.layers.17.mixer.conv1d.weight": "model.safetensors",
130
+ "backbone.layers.17.mixer.dt_bias": "model.safetensors",
131
+ "backbone.layers.17.mixer.in_proj.biases": "model.safetensors",
132
+ "backbone.layers.17.mixer.in_proj.scales": "model.safetensors",
133
+ "backbone.layers.17.mixer.in_proj.weight": "model.safetensors",
134
+ "backbone.layers.17.mixer.norm.weight": "model.safetensors",
135
+ "backbone.layers.17.mixer.out_proj.biases": "model.safetensors",
136
+ "backbone.layers.17.mixer.out_proj.scales": "model.safetensors",
137
+ "backbone.layers.17.mixer.out_proj.weight": "model.safetensors",
138
+ "backbone.layers.17.norm.weight": "model.safetensors",
139
+ "backbone.layers.18.mixer.A_log": "model.safetensors",
140
+ "backbone.layers.18.mixer.D": "model.safetensors",
141
+ "backbone.layers.18.mixer.conv1d.bias": "model.safetensors",
142
+ "backbone.layers.18.mixer.conv1d.weight": "model.safetensors",
143
+ "backbone.layers.18.mixer.dt_bias": "model.safetensors",
144
+ "backbone.layers.18.mixer.in_proj.biases": "model.safetensors",
145
+ "backbone.layers.18.mixer.in_proj.scales": "model.safetensors",
146
+ "backbone.layers.18.mixer.in_proj.weight": "model.safetensors",
147
+ "backbone.layers.18.mixer.norm.weight": "model.safetensors",
148
+ "backbone.layers.18.mixer.out_proj.biases": "model.safetensors",
149
+ "backbone.layers.18.mixer.out_proj.scales": "model.safetensors",
150
+ "backbone.layers.18.mixer.out_proj.weight": "model.safetensors",
151
+ "backbone.layers.18.norm.weight": "model.safetensors",
152
+ "backbone.layers.19.mixer.A_log": "model.safetensors",
153
+ "backbone.layers.19.mixer.D": "model.safetensors",
154
+ "backbone.layers.19.mixer.conv1d.bias": "model.safetensors",
155
+ "backbone.layers.19.mixer.conv1d.weight": "model.safetensors",
156
+ "backbone.layers.19.mixer.dt_bias": "model.safetensors",
157
+ "backbone.layers.19.mixer.in_proj.biases": "model.safetensors",
158
+ "backbone.layers.19.mixer.in_proj.scales": "model.safetensors",
159
+ "backbone.layers.19.mixer.in_proj.weight": "model.safetensors",
160
+ "backbone.layers.19.mixer.norm.weight": "model.safetensors",
161
+ "backbone.layers.19.mixer.out_proj.biases": "model.safetensors",
162
+ "backbone.layers.19.mixer.out_proj.scales": "model.safetensors",
163
+ "backbone.layers.19.mixer.out_proj.weight": "model.safetensors",
164
+ "backbone.layers.19.norm.weight": "model.safetensors",
165
+ "backbone.layers.2.mixer.A_log": "model.safetensors",
166
+ "backbone.layers.2.mixer.D": "model.safetensors",
167
+ "backbone.layers.2.mixer.conv1d.bias": "model.safetensors",
168
+ "backbone.layers.2.mixer.conv1d.weight": "model.safetensors",
169
+ "backbone.layers.2.mixer.dt_bias": "model.safetensors",
170
+ "backbone.layers.2.mixer.in_proj.biases": "model.safetensors",
171
+ "backbone.layers.2.mixer.in_proj.scales": "model.safetensors",
172
+ "backbone.layers.2.mixer.in_proj.weight": "model.safetensors",
173
+ "backbone.layers.2.mixer.norm.weight": "model.safetensors",
174
+ "backbone.layers.2.mixer.out_proj.biases": "model.safetensors",
175
+ "backbone.layers.2.mixer.out_proj.scales": "model.safetensors",
176
+ "backbone.layers.2.mixer.out_proj.weight": "model.safetensors",
177
+ "backbone.layers.2.norm.weight": "model.safetensors",
178
+ "backbone.layers.20.mixer.A_log": "model.safetensors",
179
+ "backbone.layers.20.mixer.D": "model.safetensors",
180
+ "backbone.layers.20.mixer.conv1d.bias": "model.safetensors",
181
+ "backbone.layers.20.mixer.conv1d.weight": "model.safetensors",
182
+ "backbone.layers.20.mixer.dt_bias": "model.safetensors",
183
+ "backbone.layers.20.mixer.in_proj.biases": "model.safetensors",
184
+ "backbone.layers.20.mixer.in_proj.scales": "model.safetensors",
185
+ "backbone.layers.20.mixer.in_proj.weight": "model.safetensors",
186
+ "backbone.layers.20.mixer.norm.weight": "model.safetensors",
187
+ "backbone.layers.20.mixer.out_proj.biases": "model.safetensors",
188
+ "backbone.layers.20.mixer.out_proj.scales": "model.safetensors",
189
+ "backbone.layers.20.mixer.out_proj.weight": "model.safetensors",
190
+ "backbone.layers.20.norm.weight": "model.safetensors",
191
+ "backbone.layers.21.mixer.A_log": "model.safetensors",
192
+ "backbone.layers.21.mixer.D": "model.safetensors",
193
+ "backbone.layers.21.mixer.conv1d.bias": "model.safetensors",
194
+ "backbone.layers.21.mixer.conv1d.weight": "model.safetensors",
195
+ "backbone.layers.21.mixer.dt_bias": "model.safetensors",
196
+ "backbone.layers.21.mixer.in_proj.biases": "model.safetensors",
197
+ "backbone.layers.21.mixer.in_proj.scales": "model.safetensors",
198
+ "backbone.layers.21.mixer.in_proj.weight": "model.safetensors",
199
+ "backbone.layers.21.mixer.norm.weight": "model.safetensors",
200
+ "backbone.layers.21.mixer.out_proj.biases": "model.safetensors",
201
+ "backbone.layers.21.mixer.out_proj.scales": "model.safetensors",
202
+ "backbone.layers.21.mixer.out_proj.weight": "model.safetensors",
203
+ "backbone.layers.21.norm.weight": "model.safetensors",
204
+ "backbone.layers.22.mixer.A_log": "model.safetensors",
205
+ "backbone.layers.22.mixer.D": "model.safetensors",
206
+ "backbone.layers.22.mixer.conv1d.bias": "model.safetensors",
207
+ "backbone.layers.22.mixer.conv1d.weight": "model.safetensors",
208
+ "backbone.layers.22.mixer.dt_bias": "model.safetensors",
209
+ "backbone.layers.22.mixer.in_proj.biases": "model.safetensors",
210
+ "backbone.layers.22.mixer.in_proj.scales": "model.safetensors",
211
+ "backbone.layers.22.mixer.in_proj.weight": "model.safetensors",
212
+ "backbone.layers.22.mixer.norm.weight": "model.safetensors",
213
+ "backbone.layers.22.mixer.out_proj.biases": "model.safetensors",
214
+ "backbone.layers.22.mixer.out_proj.scales": "model.safetensors",
215
+ "backbone.layers.22.mixer.out_proj.weight": "model.safetensors",
216
+ "backbone.layers.22.norm.weight": "model.safetensors",
217
+ "backbone.layers.23.mixer.A_log": "model.safetensors",
218
+ "backbone.layers.23.mixer.D": "model.safetensors",
219
+ "backbone.layers.23.mixer.conv1d.bias": "model.safetensors",
220
+ "backbone.layers.23.mixer.conv1d.weight": "model.safetensors",
221
+ "backbone.layers.23.mixer.dt_bias": "model.safetensors",
222
+ "backbone.layers.23.mixer.in_proj.biases": "model.safetensors",
223
+ "backbone.layers.23.mixer.in_proj.scales": "model.safetensors",
224
+ "backbone.layers.23.mixer.in_proj.weight": "model.safetensors",
225
+ "backbone.layers.23.mixer.norm.weight": "model.safetensors",
226
+ "backbone.layers.23.mixer.out_proj.biases": "model.safetensors",
227
+ "backbone.layers.23.mixer.out_proj.scales": "model.safetensors",
228
+ "backbone.layers.23.mixer.out_proj.weight": "model.safetensors",
229
+ "backbone.layers.23.norm.weight": "model.safetensors",
230
+ "backbone.layers.24.mixer.A_log": "model.safetensors",
231
+ "backbone.layers.24.mixer.D": "model.safetensors",
232
+ "backbone.layers.24.mixer.conv1d.bias": "model.safetensors",
233
+ "backbone.layers.24.mixer.conv1d.weight": "model.safetensors",
234
+ "backbone.layers.24.mixer.dt_bias": "model.safetensors",
235
+ "backbone.layers.24.mixer.in_proj.biases": "model.safetensors",
236
+ "backbone.layers.24.mixer.in_proj.scales": "model.safetensors",
237
+ "backbone.layers.24.mixer.in_proj.weight": "model.safetensors",
238
+ "backbone.layers.24.mixer.norm.weight": "model.safetensors",
239
+ "backbone.layers.24.mixer.out_proj.biases": "model.safetensors",
240
+ "backbone.layers.24.mixer.out_proj.scales": "model.safetensors",
241
+ "backbone.layers.24.mixer.out_proj.weight": "model.safetensors",
242
+ "backbone.layers.24.norm.weight": "model.safetensors",
243
+ "backbone.layers.25.mixer.A_log": "model.safetensors",
244
+ "backbone.layers.25.mixer.D": "model.safetensors",
245
+ "backbone.layers.25.mixer.conv1d.bias": "model.safetensors",
246
+ "backbone.layers.25.mixer.conv1d.weight": "model.safetensors",
247
+ "backbone.layers.25.mixer.dt_bias": "model.safetensors",
248
+ "backbone.layers.25.mixer.in_proj.biases": "model.safetensors",
249
+ "backbone.layers.25.mixer.in_proj.scales": "model.safetensors",
250
+ "backbone.layers.25.mixer.in_proj.weight": "model.safetensors",
251
+ "backbone.layers.25.mixer.norm.weight": "model.safetensors",
252
+ "backbone.layers.25.mixer.out_proj.biases": "model.safetensors",
253
+ "backbone.layers.25.mixer.out_proj.scales": "model.safetensors",
254
+ "backbone.layers.25.mixer.out_proj.weight": "model.safetensors",
255
+ "backbone.layers.25.norm.weight": "model.safetensors",
256
+ "backbone.layers.26.mixer.A_log": "model.safetensors",
257
+ "backbone.layers.26.mixer.D": "model.safetensors",
258
+ "backbone.layers.26.mixer.conv1d.bias": "model.safetensors",
259
+ "backbone.layers.26.mixer.conv1d.weight": "model.safetensors",
260
+ "backbone.layers.26.mixer.dt_bias": "model.safetensors",
261
+ "backbone.layers.26.mixer.in_proj.biases": "model.safetensors",
262
+ "backbone.layers.26.mixer.in_proj.scales": "model.safetensors",
263
+ "backbone.layers.26.mixer.in_proj.weight": "model.safetensors",
264
+ "backbone.layers.26.mixer.norm.weight": "model.safetensors",
265
+ "backbone.layers.26.mixer.out_proj.biases": "model.safetensors",
266
+ "backbone.layers.26.mixer.out_proj.scales": "model.safetensors",
267
+ "backbone.layers.26.mixer.out_proj.weight": "model.safetensors",
268
+ "backbone.layers.26.norm.weight": "model.safetensors",
269
+ "backbone.layers.27.mixer.A_log": "model.safetensors",
270
+ "backbone.layers.27.mixer.D": "model.safetensors",
271
+ "backbone.layers.27.mixer.conv1d.bias": "model.safetensors",
272
+ "backbone.layers.27.mixer.conv1d.weight": "model.safetensors",
273
+ "backbone.layers.27.mixer.dt_bias": "model.safetensors",
274
+ "backbone.layers.27.mixer.in_proj.biases": "model.safetensors",
275
+ "backbone.layers.27.mixer.in_proj.scales": "model.safetensors",
276
+ "backbone.layers.27.mixer.in_proj.weight": "model.safetensors",
277
+ "backbone.layers.27.mixer.norm.weight": "model.safetensors",
278
+ "backbone.layers.27.mixer.out_proj.biases": "model.safetensors",
279
+ "backbone.layers.27.mixer.out_proj.scales": "model.safetensors",
280
+ "backbone.layers.27.mixer.out_proj.weight": "model.safetensors",
281
+ "backbone.layers.27.norm.weight": "model.safetensors",
282
+ "backbone.layers.28.mixer.A_log": "model.safetensors",
283
+ "backbone.layers.28.mixer.D": "model.safetensors",
284
+ "backbone.layers.28.mixer.conv1d.bias": "model.safetensors",
285
+ "backbone.layers.28.mixer.conv1d.weight": "model.safetensors",
286
+ "backbone.layers.28.mixer.dt_bias": "model.safetensors",
287
+ "backbone.layers.28.mixer.in_proj.biases": "model.safetensors",
288
+ "backbone.layers.28.mixer.in_proj.scales": "model.safetensors",
289
+ "backbone.layers.28.mixer.in_proj.weight": "model.safetensors",
290
+ "backbone.layers.28.mixer.norm.weight": "model.safetensors",
291
+ "backbone.layers.28.mixer.out_proj.biases": "model.safetensors",
292
+ "backbone.layers.28.mixer.out_proj.scales": "model.safetensors",
293
+ "backbone.layers.28.mixer.out_proj.weight": "model.safetensors",
294
+ "backbone.layers.28.norm.weight": "model.safetensors",
295
+ "backbone.layers.29.mixer.A_log": "model.safetensors",
296
+ "backbone.layers.29.mixer.D": "model.safetensors",
297
+ "backbone.layers.29.mixer.conv1d.bias": "model.safetensors",
298
+ "backbone.layers.29.mixer.conv1d.weight": "model.safetensors",
299
+ "backbone.layers.29.mixer.dt_bias": "model.safetensors",
300
+ "backbone.layers.29.mixer.in_proj.biases": "model.safetensors",
301
+ "backbone.layers.29.mixer.in_proj.scales": "model.safetensors",
302
+ "backbone.layers.29.mixer.in_proj.weight": "model.safetensors",
303
+ "backbone.layers.29.mixer.norm.weight": "model.safetensors",
304
+ "backbone.layers.29.mixer.out_proj.biases": "model.safetensors",
305
+ "backbone.layers.29.mixer.out_proj.scales": "model.safetensors",
306
+ "backbone.layers.29.mixer.out_proj.weight": "model.safetensors",
307
+ "backbone.layers.29.norm.weight": "model.safetensors",
308
+ "backbone.layers.3.mixer.A_log": "model.safetensors",
309
+ "backbone.layers.3.mixer.D": "model.safetensors",
310
+ "backbone.layers.3.mixer.conv1d.bias": "model.safetensors",
311
+ "backbone.layers.3.mixer.conv1d.weight": "model.safetensors",
312
+ "backbone.layers.3.mixer.dt_bias": "model.safetensors",
313
+ "backbone.layers.3.mixer.in_proj.biases": "model.safetensors",
314
+ "backbone.layers.3.mixer.in_proj.scales": "model.safetensors",
315
+ "backbone.layers.3.mixer.in_proj.weight": "model.safetensors",
316
+ "backbone.layers.3.mixer.norm.weight": "model.safetensors",
317
+ "backbone.layers.3.mixer.out_proj.biases": "model.safetensors",
318
+ "backbone.layers.3.mixer.out_proj.scales": "model.safetensors",
319
+ "backbone.layers.3.mixer.out_proj.weight": "model.safetensors",
320
+ "backbone.layers.3.norm.weight": "model.safetensors",
321
+ "backbone.layers.30.mixer.A_log": "model.safetensors",
322
+ "backbone.layers.30.mixer.D": "model.safetensors",
323
+ "backbone.layers.30.mixer.conv1d.bias": "model.safetensors",
324
+ "backbone.layers.30.mixer.conv1d.weight": "model.safetensors",
325
+ "backbone.layers.30.mixer.dt_bias": "model.safetensors",
326
+ "backbone.layers.30.mixer.in_proj.biases": "model.safetensors",
327
+ "backbone.layers.30.mixer.in_proj.scales": "model.safetensors",
328
+ "backbone.layers.30.mixer.in_proj.weight": "model.safetensors",
329
+ "backbone.layers.30.mixer.norm.weight": "model.safetensors",
330
+ "backbone.layers.30.mixer.out_proj.biases": "model.safetensors",
331
+ "backbone.layers.30.mixer.out_proj.scales": "model.safetensors",
332
+ "backbone.layers.30.mixer.out_proj.weight": "model.safetensors",
333
+ "backbone.layers.30.norm.weight": "model.safetensors",
334
+ "backbone.layers.31.mixer.A_log": "model.safetensors",
335
+ "backbone.layers.31.mixer.D": "model.safetensors",
336
+ "backbone.layers.31.mixer.conv1d.bias": "model.safetensors",
337
+ "backbone.layers.31.mixer.conv1d.weight": "model.safetensors",
338
+ "backbone.layers.31.mixer.dt_bias": "model.safetensors",
339
+ "backbone.layers.31.mixer.in_proj.biases": "model.safetensors",
340
+ "backbone.layers.31.mixer.in_proj.scales": "model.safetensors",
341
+ "backbone.layers.31.mixer.in_proj.weight": "model.safetensors",
342
+ "backbone.layers.31.mixer.norm.weight": "model.safetensors",
343
+ "backbone.layers.31.mixer.out_proj.biases": "model.safetensors",
344
+ "backbone.layers.31.mixer.out_proj.scales": "model.safetensors",
345
+ "backbone.layers.31.mixer.out_proj.weight": "model.safetensors",
346
+ "backbone.layers.31.norm.weight": "model.safetensors",
347
+ "backbone.layers.32.mixer.A_log": "model.safetensors",
348
+ "backbone.layers.32.mixer.D": "model.safetensors",
349
+ "backbone.layers.32.mixer.conv1d.bias": "model.safetensors",
350
+ "backbone.layers.32.mixer.conv1d.weight": "model.safetensors",
351
+ "backbone.layers.32.mixer.dt_bias": "model.safetensors",
352
+ "backbone.layers.32.mixer.in_proj.biases": "model.safetensors",
353
+ "backbone.layers.32.mixer.in_proj.scales": "model.safetensors",
354
+ "backbone.layers.32.mixer.in_proj.weight": "model.safetensors",
355
+ "backbone.layers.32.mixer.norm.weight": "model.safetensors",
356
+ "backbone.layers.32.mixer.out_proj.biases": "model.safetensors",
357
+ "backbone.layers.32.mixer.out_proj.scales": "model.safetensors",
358
+ "backbone.layers.32.mixer.out_proj.weight": "model.safetensors",
359
+ "backbone.layers.32.norm.weight": "model.safetensors",
360
+ "backbone.layers.33.mixer.A_log": "model.safetensors",
361
+ "backbone.layers.33.mixer.D": "model.safetensors",
362
+ "backbone.layers.33.mixer.conv1d.bias": "model.safetensors",
363
+ "backbone.layers.33.mixer.conv1d.weight": "model.safetensors",
364
+ "backbone.layers.33.mixer.dt_bias": "model.safetensors",
365
+ "backbone.layers.33.mixer.in_proj.biases": "model.safetensors",
366
+ "backbone.layers.33.mixer.in_proj.scales": "model.safetensors",
367
+ "backbone.layers.33.mixer.in_proj.weight": "model.safetensors",
368
+ "backbone.layers.33.mixer.norm.weight": "model.safetensors",
369
+ "backbone.layers.33.mixer.out_proj.biases": "model.safetensors",
370
+ "backbone.layers.33.mixer.out_proj.scales": "model.safetensors",
371
+ "backbone.layers.33.mixer.out_proj.weight": "model.safetensors",
372
+ "backbone.layers.33.norm.weight": "model.safetensors",
373
+ "backbone.layers.34.mixer.A_log": "model.safetensors",
374
+ "backbone.layers.34.mixer.D": "model.safetensors",
375
+ "backbone.layers.34.mixer.conv1d.bias": "model.safetensors",
376
+ "backbone.layers.34.mixer.conv1d.weight": "model.safetensors",
377
+ "backbone.layers.34.mixer.dt_bias": "model.safetensors",
378
+ "backbone.layers.34.mixer.in_proj.biases": "model.safetensors",
379
+ "backbone.layers.34.mixer.in_proj.scales": "model.safetensors",
380
+ "backbone.layers.34.mixer.in_proj.weight": "model.safetensors",
381
+ "backbone.layers.34.mixer.norm.weight": "model.safetensors",
382
+ "backbone.layers.34.mixer.out_proj.biases": "model.safetensors",
383
+ "backbone.layers.34.mixer.out_proj.scales": "model.safetensors",
384
+ "backbone.layers.34.mixer.out_proj.weight": "model.safetensors",
385
+ "backbone.layers.34.norm.weight": "model.safetensors",
386
+ "backbone.layers.35.mixer.A_log": "model.safetensors",
387
+ "backbone.layers.35.mixer.D": "model.safetensors",
388
+ "backbone.layers.35.mixer.conv1d.bias": "model.safetensors",
389
+ "backbone.layers.35.mixer.conv1d.weight": "model.safetensors",
390
+ "backbone.layers.35.mixer.dt_bias": "model.safetensors",
391
+ "backbone.layers.35.mixer.in_proj.biases": "model.safetensors",
392
+ "backbone.layers.35.mixer.in_proj.scales": "model.safetensors",
393
+ "backbone.layers.35.mixer.in_proj.weight": "model.safetensors",
394
+ "backbone.layers.35.mixer.norm.weight": "model.safetensors",
395
+ "backbone.layers.35.mixer.out_proj.biases": "model.safetensors",
396
+ "backbone.layers.35.mixer.out_proj.scales": "model.safetensors",
397
+ "backbone.layers.35.mixer.out_proj.weight": "model.safetensors",
398
+ "backbone.layers.35.norm.weight": "model.safetensors",
399
+ "backbone.layers.36.mixer.A_log": "model.safetensors",
400
+ "backbone.layers.36.mixer.D": "model.safetensors",
401
+ "backbone.layers.36.mixer.conv1d.bias": "model.safetensors",
402
+ "backbone.layers.36.mixer.conv1d.weight": "model.safetensors",
403
+ "backbone.layers.36.mixer.dt_bias": "model.safetensors",
404
+ "backbone.layers.36.mixer.in_proj.biases": "model.safetensors",
405
+ "backbone.layers.36.mixer.in_proj.scales": "model.safetensors",
406
+ "backbone.layers.36.mixer.in_proj.weight": "model.safetensors",
407
+ "backbone.layers.36.mixer.norm.weight": "model.safetensors",
408
+ "backbone.layers.36.mixer.out_proj.biases": "model.safetensors",
409
+ "backbone.layers.36.mixer.out_proj.scales": "model.safetensors",
410
+ "backbone.layers.36.mixer.out_proj.weight": "model.safetensors",
411
+ "backbone.layers.36.norm.weight": "model.safetensors",
412
+ "backbone.layers.37.mixer.A_log": "model.safetensors",
413
+ "backbone.layers.37.mixer.D": "model.safetensors",
414
+ "backbone.layers.37.mixer.conv1d.bias": "model.safetensors",
415
+ "backbone.layers.37.mixer.conv1d.weight": "model.safetensors",
416
+ "backbone.layers.37.mixer.dt_bias": "model.safetensors",
417
+ "backbone.layers.37.mixer.in_proj.biases": "model.safetensors",
418
+ "backbone.layers.37.mixer.in_proj.scales": "model.safetensors",
419
+ "backbone.layers.37.mixer.in_proj.weight": "model.safetensors",
420
+ "backbone.layers.37.mixer.norm.weight": "model.safetensors",
421
+ "backbone.layers.37.mixer.out_proj.biases": "model.safetensors",
422
+ "backbone.layers.37.mixer.out_proj.scales": "model.safetensors",
423
+ "backbone.layers.37.mixer.out_proj.weight": "model.safetensors",
424
+ "backbone.layers.37.norm.weight": "model.safetensors",
425
+ "backbone.layers.38.mixer.A_log": "model.safetensors",
426
+ "backbone.layers.38.mixer.D": "model.safetensors",
427
+ "backbone.layers.38.mixer.conv1d.bias": "model.safetensors",
428
+ "backbone.layers.38.mixer.conv1d.weight": "model.safetensors",
429
+ "backbone.layers.38.mixer.dt_bias": "model.safetensors",
430
+ "backbone.layers.38.mixer.in_proj.biases": "model.safetensors",
431
+ "backbone.layers.38.mixer.in_proj.scales": "model.safetensors",
432
+ "backbone.layers.38.mixer.in_proj.weight": "model.safetensors",
433
+ "backbone.layers.38.mixer.norm.weight": "model.safetensors",
434
+ "backbone.layers.38.mixer.out_proj.biases": "model.safetensors",
435
+ "backbone.layers.38.mixer.out_proj.scales": "model.safetensors",
436
+ "backbone.layers.38.mixer.out_proj.weight": "model.safetensors",
437
+ "backbone.layers.38.norm.weight": "model.safetensors",
438
+ "backbone.layers.39.mixer.A_log": "model.safetensors",
439
+ "backbone.layers.39.mixer.D": "model.safetensors",
440
+ "backbone.layers.39.mixer.conv1d.bias": "model.safetensors",
441
+ "backbone.layers.39.mixer.conv1d.weight": "model.safetensors",
442
+ "backbone.layers.39.mixer.dt_bias": "model.safetensors",
443
+ "backbone.layers.39.mixer.in_proj.biases": "model.safetensors",
444
+ "backbone.layers.39.mixer.in_proj.scales": "model.safetensors",
445
+ "backbone.layers.39.mixer.in_proj.weight": "model.safetensors",
446
+ "backbone.layers.39.mixer.norm.weight": "model.safetensors",
447
+ "backbone.layers.39.mixer.out_proj.biases": "model.safetensors",
448
+ "backbone.layers.39.mixer.out_proj.scales": "model.safetensors",
449
+ "backbone.layers.39.mixer.out_proj.weight": "model.safetensors",
450
+ "backbone.layers.39.norm.weight": "model.safetensors",
451
+ "backbone.layers.4.mixer.A_log": "model.safetensors",
452
+ "backbone.layers.4.mixer.D": "model.safetensors",
453
+ "backbone.layers.4.mixer.conv1d.bias": "model.safetensors",
454
+ "backbone.layers.4.mixer.conv1d.weight": "model.safetensors",
455
+ "backbone.layers.4.mixer.dt_bias": "model.safetensors",
456
+ "backbone.layers.4.mixer.in_proj.biases": "model.safetensors",
457
+ "backbone.layers.4.mixer.in_proj.scales": "model.safetensors",
458
+ "backbone.layers.4.mixer.in_proj.weight": "model.safetensors",
459
+ "backbone.layers.4.mixer.norm.weight": "model.safetensors",
460
+ "backbone.layers.4.mixer.out_proj.biases": "model.safetensors",
461
+ "backbone.layers.4.mixer.out_proj.scales": "model.safetensors",
462
+ "backbone.layers.4.mixer.out_proj.weight": "model.safetensors",
463
+ "backbone.layers.4.norm.weight": "model.safetensors",
464
+ "backbone.layers.40.mixer.A_log": "model.safetensors",
465
+ "backbone.layers.40.mixer.D": "model.safetensors",
466
+ "backbone.layers.40.mixer.conv1d.bias": "model.safetensors",
467
+ "backbone.layers.40.mixer.conv1d.weight": "model.safetensors",
468
+ "backbone.layers.40.mixer.dt_bias": "model.safetensors",
469
+ "backbone.layers.40.mixer.in_proj.biases": "model.safetensors",
470
+ "backbone.layers.40.mixer.in_proj.scales": "model.safetensors",
471
+ "backbone.layers.40.mixer.in_proj.weight": "model.safetensors",
472
+ "backbone.layers.40.mixer.norm.weight": "model.safetensors",
473
+ "backbone.layers.40.mixer.out_proj.biases": "model.safetensors",
474
+ "backbone.layers.40.mixer.out_proj.scales": "model.safetensors",
475
+ "backbone.layers.40.mixer.out_proj.weight": "model.safetensors",
476
+ "backbone.layers.40.norm.weight": "model.safetensors",
477
+ "backbone.layers.41.mixer.A_log": "model.safetensors",
478
+ "backbone.layers.41.mixer.D": "model.safetensors",
479
+ "backbone.layers.41.mixer.conv1d.bias": "model.safetensors",
480
+ "backbone.layers.41.mixer.conv1d.weight": "model.safetensors",
481
+ "backbone.layers.41.mixer.dt_bias": "model.safetensors",
482
+ "backbone.layers.41.mixer.in_proj.biases": "model.safetensors",
483
+ "backbone.layers.41.mixer.in_proj.scales": "model.safetensors",
484
+ "backbone.layers.41.mixer.in_proj.weight": "model.safetensors",
485
+ "backbone.layers.41.mixer.norm.weight": "model.safetensors",
486
+ "backbone.layers.41.mixer.out_proj.biases": "model.safetensors",
487
+ "backbone.layers.41.mixer.out_proj.scales": "model.safetensors",
488
+ "backbone.layers.41.mixer.out_proj.weight": "model.safetensors",
489
+ "backbone.layers.41.norm.weight": "model.safetensors",
490
+ "backbone.layers.42.mixer.A_log": "model.safetensors",
491
+ "backbone.layers.42.mixer.D": "model.safetensors",
492
+ "backbone.layers.42.mixer.conv1d.bias": "model.safetensors",
493
+ "backbone.layers.42.mixer.conv1d.weight": "model.safetensors",
494
+ "backbone.layers.42.mixer.dt_bias": "model.safetensors",
495
+ "backbone.layers.42.mixer.in_proj.biases": "model.safetensors",
496
+ "backbone.layers.42.mixer.in_proj.scales": "model.safetensors",
497
+ "backbone.layers.42.mixer.in_proj.weight": "model.safetensors",
498
+ "backbone.layers.42.mixer.norm.weight": "model.safetensors",
499
+ "backbone.layers.42.mixer.out_proj.biases": "model.safetensors",
500
+ "backbone.layers.42.mixer.out_proj.scales": "model.safetensors",
501
+ "backbone.layers.42.mixer.out_proj.weight": "model.safetensors",
502
+ "backbone.layers.42.norm.weight": "model.safetensors",
503
+ "backbone.layers.43.mixer.A_log": "model.safetensors",
504
+ "backbone.layers.43.mixer.D": "model.safetensors",
505
+ "backbone.layers.43.mixer.conv1d.bias": "model.safetensors",
506
+ "backbone.layers.43.mixer.conv1d.weight": "model.safetensors",
507
+ "backbone.layers.43.mixer.dt_bias": "model.safetensors",
508
+ "backbone.layers.43.mixer.in_proj.biases": "model.safetensors",
509
+ "backbone.layers.43.mixer.in_proj.scales": "model.safetensors",
510
+ "backbone.layers.43.mixer.in_proj.weight": "model.safetensors",
511
+ "backbone.layers.43.mixer.norm.weight": "model.safetensors",
512
+ "backbone.layers.43.mixer.out_proj.biases": "model.safetensors",
513
+ "backbone.layers.43.mixer.out_proj.scales": "model.safetensors",
514
+ "backbone.layers.43.mixer.out_proj.weight": "model.safetensors",
515
+ "backbone.layers.43.norm.weight": "model.safetensors",
516
+ "backbone.layers.44.mixer.A_log": "model.safetensors",
517
+ "backbone.layers.44.mixer.D": "model.safetensors",
518
+ "backbone.layers.44.mixer.conv1d.bias": "model.safetensors",
519
+ "backbone.layers.44.mixer.conv1d.weight": "model.safetensors",
520
+ "backbone.layers.44.mixer.dt_bias": "model.safetensors",
521
+ "backbone.layers.44.mixer.in_proj.biases": "model.safetensors",
522
+ "backbone.layers.44.mixer.in_proj.scales": "model.safetensors",
523
+ "backbone.layers.44.mixer.in_proj.weight": "model.safetensors",
524
+ "backbone.layers.44.mixer.norm.weight": "model.safetensors",
525
+ "backbone.layers.44.mixer.out_proj.biases": "model.safetensors",
526
+ "backbone.layers.44.mixer.out_proj.scales": "model.safetensors",
527
+ "backbone.layers.44.mixer.out_proj.weight": "model.safetensors",
528
+ "backbone.layers.44.norm.weight": "model.safetensors",
529
+ "backbone.layers.45.mixer.A_log": "model.safetensors",
530
+ "backbone.layers.45.mixer.D": "model.safetensors",
531
+ "backbone.layers.45.mixer.conv1d.bias": "model.safetensors",
532
+ "backbone.layers.45.mixer.conv1d.weight": "model.safetensors",
533
+ "backbone.layers.45.mixer.dt_bias": "model.safetensors",
534
+ "backbone.layers.45.mixer.in_proj.biases": "model.safetensors",
535
+ "backbone.layers.45.mixer.in_proj.scales": "model.safetensors",
536
+ "backbone.layers.45.mixer.in_proj.weight": "model.safetensors",
537
+ "backbone.layers.45.mixer.norm.weight": "model.safetensors",
538
+ "backbone.layers.45.mixer.out_proj.biases": "model.safetensors",
539
+ "backbone.layers.45.mixer.out_proj.scales": "model.safetensors",
540
+ "backbone.layers.45.mixer.out_proj.weight": "model.safetensors",
541
+ "backbone.layers.45.norm.weight": "model.safetensors",
542
+ "backbone.layers.46.mixer.A_log": "model.safetensors",
543
+ "backbone.layers.46.mixer.D": "model.safetensors",
544
+ "backbone.layers.46.mixer.conv1d.bias": "model.safetensors",
545
+ "backbone.layers.46.mixer.conv1d.weight": "model.safetensors",
546
+ "backbone.layers.46.mixer.dt_bias": "model.safetensors",
547
+ "backbone.layers.46.mixer.in_proj.biases": "model.safetensors",
548
+ "backbone.layers.46.mixer.in_proj.scales": "model.safetensors",
549
+ "backbone.layers.46.mixer.in_proj.weight": "model.safetensors",
550
+ "backbone.layers.46.mixer.norm.weight": "model.safetensors",
551
+ "backbone.layers.46.mixer.out_proj.biases": "model.safetensors",
552
+ "backbone.layers.46.mixer.out_proj.scales": "model.safetensors",
553
+ "backbone.layers.46.mixer.out_proj.weight": "model.safetensors",
554
+ "backbone.layers.46.norm.weight": "model.safetensors",
555
+ "backbone.layers.47.mixer.A_log": "model.safetensors",
556
+ "backbone.layers.47.mixer.D": "model.safetensors",
557
+ "backbone.layers.47.mixer.conv1d.bias": "model.safetensors",
558
+ "backbone.layers.47.mixer.conv1d.weight": "model.safetensors",
559
+ "backbone.layers.47.mixer.dt_bias": "model.safetensors",
560
+ "backbone.layers.47.mixer.in_proj.biases": "model.safetensors",
561
+ "backbone.layers.47.mixer.in_proj.scales": "model.safetensors",
562
+ "backbone.layers.47.mixer.in_proj.weight": "model.safetensors",
563
+ "backbone.layers.47.mixer.norm.weight": "model.safetensors",
564
+ "backbone.layers.47.mixer.out_proj.biases": "model.safetensors",
565
+ "backbone.layers.47.mixer.out_proj.scales": "model.safetensors",
566
+ "backbone.layers.47.mixer.out_proj.weight": "model.safetensors",
567
+ "backbone.layers.47.norm.weight": "model.safetensors",
568
+ "backbone.layers.48.mixer.A_log": "model.safetensors",
569
+ "backbone.layers.48.mixer.D": "model.safetensors",
570
+ "backbone.layers.48.mixer.conv1d.bias": "model.safetensors",
571
+ "backbone.layers.48.mixer.conv1d.weight": "model.safetensors",
572
+ "backbone.layers.48.mixer.dt_bias": "model.safetensors",
573
+ "backbone.layers.48.mixer.in_proj.biases": "model.safetensors",
574
+ "backbone.layers.48.mixer.in_proj.scales": "model.safetensors",
575
+ "backbone.layers.48.mixer.in_proj.weight": "model.safetensors",
576
+ "backbone.layers.48.mixer.norm.weight": "model.safetensors",
577
+ "backbone.layers.48.mixer.out_proj.biases": "model.safetensors",
578
+ "backbone.layers.48.mixer.out_proj.scales": "model.safetensors",
579
+ "backbone.layers.48.mixer.out_proj.weight": "model.safetensors",
580
+ "backbone.layers.48.norm.weight": "model.safetensors",
581
+ "backbone.layers.49.mixer.A_log": "model.safetensors",
582
+ "backbone.layers.49.mixer.D": "model.safetensors",
583
+ "backbone.layers.49.mixer.conv1d.bias": "model.safetensors",
584
+ "backbone.layers.49.mixer.conv1d.weight": "model.safetensors",
585
+ "backbone.layers.49.mixer.dt_bias": "model.safetensors",
586
+ "backbone.layers.49.mixer.in_proj.biases": "model.safetensors",
587
+ "backbone.layers.49.mixer.in_proj.scales": "model.safetensors",
588
+ "backbone.layers.49.mixer.in_proj.weight": "model.safetensors",
589
+ "backbone.layers.49.mixer.norm.weight": "model.safetensors",
590
+ "backbone.layers.49.mixer.out_proj.biases": "model.safetensors",
591
+ "backbone.layers.49.mixer.out_proj.scales": "model.safetensors",
592
+ "backbone.layers.49.mixer.out_proj.weight": "model.safetensors",
593
+ "backbone.layers.49.norm.weight": "model.safetensors",
594
+ "backbone.layers.5.mixer.A_log": "model.safetensors",
595
+ "backbone.layers.5.mixer.D": "model.safetensors",
596
+ "backbone.layers.5.mixer.conv1d.bias": "model.safetensors",
597
+ "backbone.layers.5.mixer.conv1d.weight": "model.safetensors",
598
+ "backbone.layers.5.mixer.dt_bias": "model.safetensors",
599
+ "backbone.layers.5.mixer.in_proj.biases": "model.safetensors",
600
+ "backbone.layers.5.mixer.in_proj.scales": "model.safetensors",
601
+ "backbone.layers.5.mixer.in_proj.weight": "model.safetensors",
602
+ "backbone.layers.5.mixer.norm.weight": "model.safetensors",
603
+ "backbone.layers.5.mixer.out_proj.biases": "model.safetensors",
604
+ "backbone.layers.5.mixer.out_proj.scales": "model.safetensors",
605
+ "backbone.layers.5.mixer.out_proj.weight": "model.safetensors",
606
+ "backbone.layers.5.norm.weight": "model.safetensors",
607
+ "backbone.layers.50.mixer.A_log": "model.safetensors",
608
+ "backbone.layers.50.mixer.D": "model.safetensors",
609
+ "backbone.layers.50.mixer.conv1d.bias": "model.safetensors",
610
+ "backbone.layers.50.mixer.conv1d.weight": "model.safetensors",
611
+ "backbone.layers.50.mixer.dt_bias": "model.safetensors",
612
+ "backbone.layers.50.mixer.in_proj.biases": "model.safetensors",
613
+ "backbone.layers.50.mixer.in_proj.scales": "model.safetensors",
614
+ "backbone.layers.50.mixer.in_proj.weight": "model.safetensors",
615
+ "backbone.layers.50.mixer.norm.weight": "model.safetensors",
616
+ "backbone.layers.50.mixer.out_proj.biases": "model.safetensors",
617
+ "backbone.layers.50.mixer.out_proj.scales": "model.safetensors",
618
+ "backbone.layers.50.mixer.out_proj.weight": "model.safetensors",
619
+ "backbone.layers.50.norm.weight": "model.safetensors",
620
+ "backbone.layers.51.mixer.A_log": "model.safetensors",
621
+ "backbone.layers.51.mixer.D": "model.safetensors",
622
+ "backbone.layers.51.mixer.conv1d.bias": "model.safetensors",
623
+ "backbone.layers.51.mixer.conv1d.weight": "model.safetensors",
624
+ "backbone.layers.51.mixer.dt_bias": "model.safetensors",
625
+ "backbone.layers.51.mixer.in_proj.biases": "model.safetensors",
626
+ "backbone.layers.51.mixer.in_proj.scales": "model.safetensors",
627
+ "backbone.layers.51.mixer.in_proj.weight": "model.safetensors",
628
+ "backbone.layers.51.mixer.norm.weight": "model.safetensors",
629
+ "backbone.layers.51.mixer.out_proj.biases": "model.safetensors",
630
+ "backbone.layers.51.mixer.out_proj.scales": "model.safetensors",
631
+ "backbone.layers.51.mixer.out_proj.weight": "model.safetensors",
632
+ "backbone.layers.51.norm.weight": "model.safetensors",
633
+ "backbone.layers.52.mixer.A_log": "model.safetensors",
634
+ "backbone.layers.52.mixer.D": "model.safetensors",
635
+ "backbone.layers.52.mixer.conv1d.bias": "model.safetensors",
636
+ "backbone.layers.52.mixer.conv1d.weight": "model.safetensors",
637
+ "backbone.layers.52.mixer.dt_bias": "model.safetensors",
638
+ "backbone.layers.52.mixer.in_proj.biases": "model.safetensors",
639
+ "backbone.layers.52.mixer.in_proj.scales": "model.safetensors",
640
+ "backbone.layers.52.mixer.in_proj.weight": "model.safetensors",
641
+ "backbone.layers.52.mixer.norm.weight": "model.safetensors",
642
+ "backbone.layers.52.mixer.out_proj.biases": "model.safetensors",
643
+ "backbone.layers.52.mixer.out_proj.scales": "model.safetensors",
644
+ "backbone.layers.52.mixer.out_proj.weight": "model.safetensors",
645
+ "backbone.layers.52.norm.weight": "model.safetensors",
646
+ "backbone.layers.53.mixer.A_log": "model.safetensors",
647
+ "backbone.layers.53.mixer.D": "model.safetensors",
648
+ "backbone.layers.53.mixer.conv1d.bias": "model.safetensors",
649
+ "backbone.layers.53.mixer.conv1d.weight": "model.safetensors",
650
+ "backbone.layers.53.mixer.dt_bias": "model.safetensors",
651
+ "backbone.layers.53.mixer.in_proj.biases": "model.safetensors",
652
+ "backbone.layers.53.mixer.in_proj.scales": "model.safetensors",
653
+ "backbone.layers.53.mixer.in_proj.weight": "model.safetensors",
654
+ "backbone.layers.53.mixer.norm.weight": "model.safetensors",
655
+ "backbone.layers.53.mixer.out_proj.biases": "model.safetensors",
656
+ "backbone.layers.53.mixer.out_proj.scales": "model.safetensors",
657
+ "backbone.layers.53.mixer.out_proj.weight": "model.safetensors",
658
+ "backbone.layers.53.norm.weight": "model.safetensors",
659
+ "backbone.layers.54.mixer.A_log": "model.safetensors",
660
+ "backbone.layers.54.mixer.D": "model.safetensors",
661
+ "backbone.layers.54.mixer.conv1d.bias": "model.safetensors",
662
+ "backbone.layers.54.mixer.conv1d.weight": "model.safetensors",
663
+ "backbone.layers.54.mixer.dt_bias": "model.safetensors",
664
+ "backbone.layers.54.mixer.in_proj.biases": "model.safetensors",
665
+ "backbone.layers.54.mixer.in_proj.scales": "model.safetensors",
666
+ "backbone.layers.54.mixer.in_proj.weight": "model.safetensors",
667
+ "backbone.layers.54.mixer.norm.weight": "model.safetensors",
668
+ "backbone.layers.54.mixer.out_proj.biases": "model.safetensors",
669
+ "backbone.layers.54.mixer.out_proj.scales": "model.safetensors",
670
+ "backbone.layers.54.mixer.out_proj.weight": "model.safetensors",
671
+ "backbone.layers.54.norm.weight": "model.safetensors",
672
+ "backbone.layers.55.mixer.A_log": "model.safetensors",
673
+ "backbone.layers.55.mixer.D": "model.safetensors",
674
+ "backbone.layers.55.mixer.conv1d.bias": "model.safetensors",
675
+ "backbone.layers.55.mixer.conv1d.weight": "model.safetensors",
676
+ "backbone.layers.55.mixer.dt_bias": "model.safetensors",
677
+ "backbone.layers.55.mixer.in_proj.biases": "model.safetensors",
678
+ "backbone.layers.55.mixer.in_proj.scales": "model.safetensors",
679
+ "backbone.layers.55.mixer.in_proj.weight": "model.safetensors",
680
+ "backbone.layers.55.mixer.norm.weight": "model.safetensors",
681
+ "backbone.layers.55.mixer.out_proj.biases": "model.safetensors",
682
+ "backbone.layers.55.mixer.out_proj.scales": "model.safetensors",
683
+ "backbone.layers.55.mixer.out_proj.weight": "model.safetensors",
684
+ "backbone.layers.55.norm.weight": "model.safetensors",
685
+ "backbone.layers.56.mixer.A_log": "model.safetensors",
686
+ "backbone.layers.56.mixer.D": "model.safetensors",
687
+ "backbone.layers.56.mixer.conv1d.bias": "model.safetensors",
688
+ "backbone.layers.56.mixer.conv1d.weight": "model.safetensors",
689
+ "backbone.layers.56.mixer.dt_bias": "model.safetensors",
690
+ "backbone.layers.56.mixer.in_proj.biases": "model.safetensors",
691
+ "backbone.layers.56.mixer.in_proj.scales": "model.safetensors",
692
+ "backbone.layers.56.mixer.in_proj.weight": "model.safetensors",
693
+ "backbone.layers.56.mixer.norm.weight": "model.safetensors",
694
+ "backbone.layers.56.mixer.out_proj.biases": "model.safetensors",
695
+ "backbone.layers.56.mixer.out_proj.scales": "model.safetensors",
696
+ "backbone.layers.56.mixer.out_proj.weight": "model.safetensors",
697
+ "backbone.layers.56.norm.weight": "model.safetensors",
698
+ "backbone.layers.57.mixer.A_log": "model.safetensors",
699
+ "backbone.layers.57.mixer.D": "model.safetensors",
700
+ "backbone.layers.57.mixer.conv1d.bias": "model.safetensors",
701
+ "backbone.layers.57.mixer.conv1d.weight": "model.safetensors",
702
+ "backbone.layers.57.mixer.dt_bias": "model.safetensors",
703
+ "backbone.layers.57.mixer.in_proj.biases": "model.safetensors",
704
+ "backbone.layers.57.mixer.in_proj.scales": "model.safetensors",
705
+ "backbone.layers.57.mixer.in_proj.weight": "model.safetensors",
706
+ "backbone.layers.57.mixer.norm.weight": "model.safetensors",
707
+ "backbone.layers.57.mixer.out_proj.biases": "model.safetensors",
708
+ "backbone.layers.57.mixer.out_proj.scales": "model.safetensors",
709
+ "backbone.layers.57.mixer.out_proj.weight": "model.safetensors",
710
+ "backbone.layers.57.norm.weight": "model.safetensors",
711
+ "backbone.layers.58.mixer.A_log": "model.safetensors",
712
+ "backbone.layers.58.mixer.D": "model.safetensors",
713
+ "backbone.layers.58.mixer.conv1d.bias": "model.safetensors",
714
+ "backbone.layers.58.mixer.conv1d.weight": "model.safetensors",
715
+ "backbone.layers.58.mixer.dt_bias": "model.safetensors",
716
+ "backbone.layers.58.mixer.in_proj.biases": "model.safetensors",
717
+ "backbone.layers.58.mixer.in_proj.scales": "model.safetensors",
718
+ "backbone.layers.58.mixer.in_proj.weight": "model.safetensors",
719
+ "backbone.layers.58.mixer.norm.weight": "model.safetensors",
720
+ "backbone.layers.58.mixer.out_proj.biases": "model.safetensors",
721
+ "backbone.layers.58.mixer.out_proj.scales": "model.safetensors",
722
+ "backbone.layers.58.mixer.out_proj.weight": "model.safetensors",
723
+ "backbone.layers.58.norm.weight": "model.safetensors",
724
+ "backbone.layers.59.mixer.A_log": "model.safetensors",
725
+ "backbone.layers.59.mixer.D": "model.safetensors",
726
+ "backbone.layers.59.mixer.conv1d.bias": "model.safetensors",
727
+ "backbone.layers.59.mixer.conv1d.weight": "model.safetensors",
728
+ "backbone.layers.59.mixer.dt_bias": "model.safetensors",
729
+ "backbone.layers.59.mixer.in_proj.biases": "model.safetensors",
730
+ "backbone.layers.59.mixer.in_proj.scales": "model.safetensors",
731
+ "backbone.layers.59.mixer.in_proj.weight": "model.safetensors",
732
+ "backbone.layers.59.mixer.norm.weight": "model.safetensors",
733
+ "backbone.layers.59.mixer.out_proj.biases": "model.safetensors",
734
+ "backbone.layers.59.mixer.out_proj.scales": "model.safetensors",
735
+ "backbone.layers.59.mixer.out_proj.weight": "model.safetensors",
736
+ "backbone.layers.59.norm.weight": "model.safetensors",
737
+ "backbone.layers.6.mixer.A_log": "model.safetensors",
738
+ "backbone.layers.6.mixer.D": "model.safetensors",
739
+ "backbone.layers.6.mixer.conv1d.bias": "model.safetensors",
740
+ "backbone.layers.6.mixer.conv1d.weight": "model.safetensors",
741
+ "backbone.layers.6.mixer.dt_bias": "model.safetensors",
742
+ "backbone.layers.6.mixer.in_proj.biases": "model.safetensors",
743
+ "backbone.layers.6.mixer.in_proj.scales": "model.safetensors",
744
+ "backbone.layers.6.mixer.in_proj.weight": "model.safetensors",
745
+ "backbone.layers.6.mixer.norm.weight": "model.safetensors",
746
+ "backbone.layers.6.mixer.out_proj.biases": "model.safetensors",
747
+ "backbone.layers.6.mixer.out_proj.scales": "model.safetensors",
748
+ "backbone.layers.6.mixer.out_proj.weight": "model.safetensors",
749
+ "backbone.layers.6.norm.weight": "model.safetensors",
750
+ "backbone.layers.60.mixer.A_log": "model.safetensors",
751
+ "backbone.layers.60.mixer.D": "model.safetensors",
752
+ "backbone.layers.60.mixer.conv1d.bias": "model.safetensors",
753
+ "backbone.layers.60.mixer.conv1d.weight": "model.safetensors",
754
+ "backbone.layers.60.mixer.dt_bias": "model.safetensors",
755
+ "backbone.layers.60.mixer.in_proj.biases": "model.safetensors",
756
+ "backbone.layers.60.mixer.in_proj.scales": "model.safetensors",
757
+ "backbone.layers.60.mixer.in_proj.weight": "model.safetensors",
758
+ "backbone.layers.60.mixer.norm.weight": "model.safetensors",
759
+ "backbone.layers.60.mixer.out_proj.biases": "model.safetensors",
760
+ "backbone.layers.60.mixer.out_proj.scales": "model.safetensors",
761
+ "backbone.layers.60.mixer.out_proj.weight": "model.safetensors",
762
+ "backbone.layers.60.norm.weight": "model.safetensors",
763
+ "backbone.layers.61.mixer.A_log": "model.safetensors",
764
+ "backbone.layers.61.mixer.D": "model.safetensors",
765
+ "backbone.layers.61.mixer.conv1d.bias": "model.safetensors",
766
+ "backbone.layers.61.mixer.conv1d.weight": "model.safetensors",
767
+ "backbone.layers.61.mixer.dt_bias": "model.safetensors",
768
+ "backbone.layers.61.mixer.in_proj.biases": "model.safetensors",
769
+ "backbone.layers.61.mixer.in_proj.scales": "model.safetensors",
770
+ "backbone.layers.61.mixer.in_proj.weight": "model.safetensors",
771
+ "backbone.layers.61.mixer.norm.weight": "model.safetensors",
772
+ "backbone.layers.61.mixer.out_proj.biases": "model.safetensors",
773
+ "backbone.layers.61.mixer.out_proj.scales": "model.safetensors",
774
+ "backbone.layers.61.mixer.out_proj.weight": "model.safetensors",
775
+ "backbone.layers.61.norm.weight": "model.safetensors",
776
+ "backbone.layers.62.mixer.A_log": "model.safetensors",
777
+ "backbone.layers.62.mixer.D": "model.safetensors",
778
+ "backbone.layers.62.mixer.conv1d.bias": "model.safetensors",
779
+ "backbone.layers.62.mixer.conv1d.weight": "model.safetensors",
780
+ "backbone.layers.62.mixer.dt_bias": "model.safetensors",
781
+ "backbone.layers.62.mixer.in_proj.biases": "model.safetensors",
782
+ "backbone.layers.62.mixer.in_proj.scales": "model.safetensors",
783
+ "backbone.layers.62.mixer.in_proj.weight": "model.safetensors",
784
+ "backbone.layers.62.mixer.norm.weight": "model.safetensors",
785
+ "backbone.layers.62.mixer.out_proj.biases": "model.safetensors",
786
+ "backbone.layers.62.mixer.out_proj.scales": "model.safetensors",
787
+ "backbone.layers.62.mixer.out_proj.weight": "model.safetensors",
788
+ "backbone.layers.62.norm.weight": "model.safetensors",
789
+ "backbone.layers.63.mixer.A_log": "model.safetensors",
790
+ "backbone.layers.63.mixer.D": "model.safetensors",
791
+ "backbone.layers.63.mixer.conv1d.bias": "model.safetensors",
792
+ "backbone.layers.63.mixer.conv1d.weight": "model.safetensors",
793
+ "backbone.layers.63.mixer.dt_bias": "model.safetensors",
794
+ "backbone.layers.63.mixer.in_proj.biases": "model.safetensors",
795
+ "backbone.layers.63.mixer.in_proj.scales": "model.safetensors",
796
+ "backbone.layers.63.mixer.in_proj.weight": "model.safetensors",
797
+ "backbone.layers.63.mixer.norm.weight": "model.safetensors",
798
+ "backbone.layers.63.mixer.out_proj.biases": "model.safetensors",
799
+ "backbone.layers.63.mixer.out_proj.scales": "model.safetensors",
800
+ "backbone.layers.63.mixer.out_proj.weight": "model.safetensors",
801
+ "backbone.layers.63.norm.weight": "model.safetensors",
802
+ "backbone.layers.7.mixer.A_log": "model.safetensors",
803
+ "backbone.layers.7.mixer.D": "model.safetensors",
804
+ "backbone.layers.7.mixer.conv1d.bias": "model.safetensors",
805
+ "backbone.layers.7.mixer.conv1d.weight": "model.safetensors",
806
+ "backbone.layers.7.mixer.dt_bias": "model.safetensors",
807
+ "backbone.layers.7.mixer.in_proj.biases": "model.safetensors",
808
+ "backbone.layers.7.mixer.in_proj.scales": "model.safetensors",
809
+ "backbone.layers.7.mixer.in_proj.weight": "model.safetensors",
810
+ "backbone.layers.7.mixer.norm.weight": "model.safetensors",
811
+ "backbone.layers.7.mixer.out_proj.biases": "model.safetensors",
812
+ "backbone.layers.7.mixer.out_proj.scales": "model.safetensors",
813
+ "backbone.layers.7.mixer.out_proj.weight": "model.safetensors",
814
+ "backbone.layers.7.norm.weight": "model.safetensors",
815
+ "backbone.layers.8.mixer.A_log": "model.safetensors",
816
+ "backbone.layers.8.mixer.D": "model.safetensors",
817
+ "backbone.layers.8.mixer.conv1d.bias": "model.safetensors",
818
+ "backbone.layers.8.mixer.conv1d.weight": "model.safetensors",
819
+ "backbone.layers.8.mixer.dt_bias": "model.safetensors",
820
+ "backbone.layers.8.mixer.in_proj.biases": "model.safetensors",
821
+ "backbone.layers.8.mixer.in_proj.scales": "model.safetensors",
822
+ "backbone.layers.8.mixer.in_proj.weight": "model.safetensors",
823
+ "backbone.layers.8.mixer.norm.weight": "model.safetensors",
824
+ "backbone.layers.8.mixer.out_proj.biases": "model.safetensors",
825
+ "backbone.layers.8.mixer.out_proj.scales": "model.safetensors",
826
+ "backbone.layers.8.mixer.out_proj.weight": "model.safetensors",
827
+ "backbone.layers.8.norm.weight": "model.safetensors",
828
+ "backbone.layers.9.mixer.A_log": "model.safetensors",
829
+ "backbone.layers.9.mixer.D": "model.safetensors",
830
+ "backbone.layers.9.mixer.conv1d.bias": "model.safetensors",
831
+ "backbone.layers.9.mixer.conv1d.weight": "model.safetensors",
832
+ "backbone.layers.9.mixer.dt_bias": "model.safetensors",
833
+ "backbone.layers.9.mixer.in_proj.biases": "model.safetensors",
834
+ "backbone.layers.9.mixer.in_proj.scales": "model.safetensors",
835
+ "backbone.layers.9.mixer.in_proj.weight": "model.safetensors",
836
+ "backbone.layers.9.mixer.norm.weight": "model.safetensors",
837
+ "backbone.layers.9.mixer.out_proj.biases": "model.safetensors",
838
+ "backbone.layers.9.mixer.out_proj.scales": "model.safetensors",
839
+ "backbone.layers.9.mixer.out_proj.weight": "model.safetensors",
840
+ "backbone.layers.9.norm.weight": "model.safetensors",
841
+ "backbone.norm_f.weight": "model.safetensors",
842
+ "lm_head.biases": "model.safetensors",
843
+ "lm_head.scales": "model.safetensors",
844
+ "lm_head.weight": "model.safetensors"
845
+ }
846
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "unk_token": {
17
+ "content": "<unk>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ }
23
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59f95e28944c062244741268596badc900df86c7f5ded05088d2da22a7379e06
3
+ size 587583
tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff