Goekdeniz-Guelmez
commited on
Commit
•
8af4d0f
1
Parent(s):
b966dba
Upload config.json
Browse files- config.json +43 -0
config.json
ADDED
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"FalconMambaForCausalLM"
|
4 |
+
],
|
5 |
+
"bos_token_id": 0,
|
6 |
+
"conv_kernel": 4,
|
7 |
+
"eos_token_id": 0,
|
8 |
+
"expand": 2,
|
9 |
+
"hidden_act": "silu",
|
10 |
+
"hidden_size": 4096,
|
11 |
+
"initializer_range": 0.1,
|
12 |
+
"intermediate_size": 8192,
|
13 |
+
"layer_norm_epsilon": 1e-05,
|
14 |
+
"mixer_rms_eps": 1e-05,
|
15 |
+
"model_type": "falcon_mamba",
|
16 |
+
"num_hidden_layers": 64,
|
17 |
+
"pad_token_id": 0,
|
18 |
+
"quantization": {
|
19 |
+
"group_size": 64,
|
20 |
+
"bits": 4
|
21 |
+
},
|
22 |
+
"quantization_config": {
|
23 |
+
"group_size": 64,
|
24 |
+
"bits": 4
|
25 |
+
},
|
26 |
+
"rescale_prenorm_residual": false,
|
27 |
+
"residual_in_fp32": true,
|
28 |
+
"state_size": 16,
|
29 |
+
"tie_word_embeddings": false,
|
30 |
+
"time_step_floor": 0.0001,
|
31 |
+
"time_step_init_scheme": "random",
|
32 |
+
"time_step_max": 0.1,
|
33 |
+
"time_step_min": 0.001,
|
34 |
+
"time_step_rank": 256,
|
35 |
+
"time_step_scale": 1.0,
|
36 |
+
"torch_dtype": "bfloat16",
|
37 |
+
"transformers_version": "4.47.0",
|
38 |
+
"use_bias": false,
|
39 |
+
"use_cache": true,
|
40 |
+
"use_conv_bias": true,
|
41 |
+
"use_mambapy": false,
|
42 |
+
"vocab_size": 65024
|
43 |
+
}
|