Commit
•
ef0fbeb
1
Parent(s):
440adc1
Delete config.json
Browse files- config.json +0 -33
config.json
DELETED
|
@@ -1,33 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"_name_or_path": "dominguesm/mambarim-110m",
|
| 3 |
-
"architectures": [
|
| 4 |
-
"MambaForCausalLM"
|
| 5 |
-
],
|
| 6 |
-
"bos_token_id": 1,
|
| 7 |
-
"conv_kernel": 4,
|
| 8 |
-
"eos_token_id": 2,
|
| 9 |
-
"expand": 2,
|
| 10 |
-
"hidden_act": "silu",
|
| 11 |
-
"hidden_size": 768,
|
| 12 |
-
"initializer_range": 0.1,
|
| 13 |
-
"intermediate_size": 1536,
|
| 14 |
-
"layer_norm_epsilon": 1e-05,
|
| 15 |
-
"model_type": "mamba",
|
| 16 |
-
"num_hidden_layers": 12,
|
| 17 |
-
"pad_token_id": 3,
|
| 18 |
-
"rescale_prenorm_residual": false,
|
| 19 |
-
"residual_in_fp32": true,
|
| 20 |
-
"state_size": 16,
|
| 21 |
-
"time_step_floor": 0.0001,
|
| 22 |
-
"time_step_init_scheme": "random",
|
| 23 |
-
"time_step_max": 0.1,
|
| 24 |
-
"time_step_min": 0.001,
|
| 25 |
-
"time_step_rank": 48,
|
| 26 |
-
"time_step_scale": 1.0,
|
| 27 |
-
"torch_dtype": "float16",
|
| 28 |
-
"transformers_version": "4.40.1",
|
| 29 |
-
"use_bias": false,
|
| 30 |
-
"use_cache": true,
|
| 31 |
-
"use_conv_bias": true,
|
| 32 |
-
"vocab_size": 32002
|
| 33 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|