Safetensors
E1
sjain-profluent commited on
Commit
b06fe2d
·
verified ·
1 Parent(s): f61b414

Upload 2 files

Browse files
Files changed (2) hide show
  1. config.json +4 -3
  2. model.safetensors +1 -1
config.json CHANGED
@@ -4,6 +4,7 @@
4
  ],
5
  "bos_token_id": 1,
6
  "clip_qkv": 8,
 
7
  "eos_token_id": 2,
8
  "gated_mlp": true,
9
  "global_attention_every_n_layers": 3,
@@ -15,7 +16,7 @@
15
  "max_num_positions_global": 65536,
16
  "max_num_positions_within_seq": 8192,
17
  "max_num_sequences": 512,
18
- "model_type": "e1",
19
  "no_ffn_gradient_checkpointing": true,
20
  "num_attention_heads": 16,
21
  "num_hidden_layers": 20,
@@ -25,7 +26,7 @@
25
  "rope_theta_global": 500000.0,
26
  "rope_theta_within_seq": 10000.0,
27
  "tie_word_embeddings": false,
28
- "torch_dtype": "bfloat16",
29
- "transformers_version": "4.48.3",
30
  "vocab_size": 34
31
  }
 
4
  ],
5
  "bos_token_id": 1,
6
  "clip_qkv": 8,
7
+ "dtype": "bfloat16",
8
  "eos_token_id": 2,
9
  "gated_mlp": true,
10
  "global_attention_every_n_layers": 3,
 
16
  "max_num_positions_global": 65536,
17
  "max_num_positions_within_seq": 8192,
18
  "max_num_sequences": 512,
19
+ "model_type": "E1",
20
  "no_ffn_gradient_checkpointing": true,
21
  "num_attention_heads": 16,
22
  "num_hidden_layers": 20,
 
26
  "rope_theta_global": 500000.0,
27
  "rope_theta_within_seq": 10000.0,
28
  "tie_word_embeddings": false,
29
+ "transformers_version": "4.56.2",
30
+ "use_cache": true,
31
  "vocab_size": 34
32
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:59e20eecb66d8470fbd78b3c190b1045242541ae990f1e0e57991d83a5a5932a
3
  size 548657620
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31e09a2542f45b04e6ce4adafb3b657f21e2d56d12bf68fd2266b1576a80bc9b
3
  size 548657620