Files changed (1) hide show
  1. config.json +6 -0
config.json CHANGED
@@ -33,6 +33,12 @@
33
  "num_key_value_heads": 128,
34
  "num_nextn_predict_layers": 1,
35
  "pretraining_tp": 1,
 
 
 
 
 
 
36
  "q_lora_rank": 1536,
37
  "qk_nope_head_dim": 128,
38
  "qk_rope_head_dim": 64,
 
33
  "num_key_value_heads": 128,
34
  "num_nextn_predict_layers": 1,
35
  "pretraining_tp": 1,
36
+ "quantization_config": {
37
+ "activation_scheme": "dynamic",
38
+ "bits": 8,
39
+ "group_size": -1,
40
+ "quant_method": "w8a8_int8"
41
+ },
42
  "q_lora_rank": 1536,
43
  "qk_nope_head_dim": 128,
44
  "qk_rope_head_dim": 64,