Piotr Zalewski commited on
Commit
4f31f18
1 Parent(s): 6d47f47

retrying 32k context

Browse files
Files changed (1) hide show
  1. config.json +3 -3
config.json CHANGED
@@ -15,7 +15,7 @@
15
  "hidden_size": 3584,
16
  "initializer_range": 0.02,
17
  "intermediate_size": 14336,
18
- "max_position_embeddings": 8192,
19
  "model_type": "gemma2",
20
  "num_attention_heads": 16,
21
  "num_hidden_layers": 42,
@@ -23,8 +23,8 @@
23
  "pad_token_id": 0,
24
  "query_pre_attn_scalar": 256,
25
  "rms_norm_eps": 1e-06,
26
- "rope_theta": 10000.0,
27
- "sliding_window": 4096,
28
  "sliding_window_size": 4096,
29
  "torch_dtype": "bfloat16",
30
  "transformers_version": "4.42.0.dev0",
 
15
  "hidden_size": 3584,
16
  "initializer_range": 0.02,
17
  "intermediate_size": 14336,
18
+ "max_position_embeddings": 32768,
19
  "model_type": "gemma2",
20
  "num_attention_heads": 16,
21
  "num_hidden_layers": 42,
 
23
  "pad_token_id": 0,
24
  "query_pre_attn_scalar": 256,
25
  "rms_norm_eps": 1e-06,
26
+ "rope_theta": 160000.0,
27
+ "sliding_window": 32768,
28
  "sliding_window_size": 4096,
29
  "torch_dtype": "bfloat16",
30
  "transformers_version": "4.42.0.dev0",