add configs to repo

#4
by radames - opened
Files changed (2) hide show
  1. phi-1_5.json +27 -0
  2. puffin-phi-v2.json +37 -0
phi-1_5.json ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "phi-1.5-half",
3
+ "activation_function": "gelu",
4
+ "architectures": [
5
+ "MixFormerSequentialForCausalLM"
6
+ ],
7
+ "auto_map": {
8
+ "AutoConfig": "configuration_mixformer_sequential.MixFormerSequentialConfig",
9
+ "AutoModelForCausalLM": "modeling_mixformer_sequential.MixFormerSequentialForCausalLM"
10
+ },
11
+ "embd_pdrop": 0.0,
12
+ "initializer_range": 0.02,
13
+ "layer_norm_epsilon": 1e-05,
14
+ "model_type": "mixformer-sequential",
15
+ "n_embd": 2048,
16
+ "n_head": 32,
17
+ "n_inner": null,
18
+ "n_layer": 24,
19
+ "n_positions": 2048,
20
+ "resid_pdrop": 0.0,
21
+ "rotary_dim": 32,
22
+ "tie_word_embeddings": false,
23
+ "torch_dtype": "float16",
24
+ "transformers_version": "4.32.1",
25
+ "vocab_size": 51200,
26
+ "pad_vocab_size_multiple": 64
27
+ }
puffin-phi-v2.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "microsoft/phi-1_5",
3
+ "activation_function": "gelu",
4
+ "architecture": {
5
+ "block_cls": "parallel",
6
+ "mixer": {},
7
+ "mlp": {
8
+ "mlp_cls": "mlp"
9
+ }
10
+ },
11
+ "architectures": [
12
+ "MixFormerSequentialForCausalLM"
13
+ ],
14
+ "auto_map": {
15
+ "AutoConfig": "microsoft/phi-1_5--configuration_mixformer_sequential.MixFormerSequentialConfig",
16
+ "AutoModelForCausalLM": "microsoft/phi-1_5--modeling_mixformer_sequential.MixFormerSequentialForCausalLM"
17
+ },
18
+ "embd_layer": "default",
19
+ "embd_pdrop": 0.0,
20
+ "initializer_range": 0.02,
21
+ "layer_norm_epsilon": 1e-05,
22
+ "model_type": "mixformer-sequential",
23
+ "n_embd": 2048,
24
+ "n_head": 32,
25
+ "n_inner": null,
26
+ "n_layer": 24,
27
+ "n_positions": 2048,
28
+ "phyagi_version": "0.0.4.dev",
29
+ "resid_pdrop": 0.0,
30
+ "rotary_dim": 32,
31
+ "tie_word_embeddings": false,
32
+ "torch_dtype": "bfloat16",
33
+ "transformers_version": "4.34.0.dev0",
34
+ "use_cache": false,
35
+ "vocab_size": 50304,
36
+ "pad_vocab_size_multiple": 64
37
+ }