Uploading config.json
Browse files- config.json +0 -4
config.json
CHANGED
|
@@ -14,7 +14,6 @@
|
|
| 14 |
"eos_token_id": 163585,
|
| 15 |
"ep_size": 1,
|
| 16 |
"first_k_dense_replace": 1,
|
| 17 |
-
"head_dim": 64,
|
| 18 |
"hidden_act": "silu",
|
| 19 |
"hidden_size": 7168,
|
| 20 |
"initializer_range": 0.02,
|
|
@@ -35,7 +34,6 @@
|
|
| 35 |
"num_nextn_predict_layers": 0,
|
| 36 |
"pretraining_tp": 1,
|
| 37 |
"q_lora_rank": 1536,
|
| 38 |
-
"qk_head_dim": 192,
|
| 39 |
"qk_nope_head_dim": 128,
|
| 40 |
"qk_rope_head_dim": 64,
|
| 41 |
"quantization_config": {
|
|
@@ -54,7 +52,6 @@
|
|
| 54 |
"quant_method": "bitsandbytes"
|
| 55 |
},
|
| 56 |
"rms_norm_eps": 1e-06,
|
| 57 |
-
"rope_interleave": true,
|
| 58 |
"rope_scaling": {
|
| 59 |
"beta_fast": 1.0,
|
| 60 |
"beta_slow": 1.0,
|
|
@@ -62,7 +59,6 @@
|
|
| 62 |
"mscale": 1.0,
|
| 63 |
"mscale_all_dim": 1.0,
|
| 64 |
"original_max_position_embeddings": 4096,
|
| 65 |
-
"rope_type": "yarn",
|
| 66 |
"type": "yarn"
|
| 67 |
},
|
| 68 |
"rope_theta": 50000.0,
|
|
|
|
| 14 |
"eos_token_id": 163585,
|
| 15 |
"ep_size": 1,
|
| 16 |
"first_k_dense_replace": 1,
|
|
|
|
| 17 |
"hidden_act": "silu",
|
| 18 |
"hidden_size": 7168,
|
| 19 |
"initializer_range": 0.02,
|
|
|
|
| 34 |
"num_nextn_predict_layers": 0,
|
| 35 |
"pretraining_tp": 1,
|
| 36 |
"q_lora_rank": 1536,
|
|
|
|
| 37 |
"qk_nope_head_dim": 128,
|
| 38 |
"qk_rope_head_dim": 64,
|
| 39 |
"quantization_config": {
|
|
|
|
| 52 |
"quant_method": "bitsandbytes"
|
| 53 |
},
|
| 54 |
"rms_norm_eps": 1e-06,
|
|
|
|
| 55 |
"rope_scaling": {
|
| 56 |
"beta_fast": 1.0,
|
| 57 |
"beta_slow": 1.0,
|
|
|
|
| 59 |
"mscale": 1.0,
|
| 60 |
"mscale_all_dim": 1.0,
|
| 61 |
"original_max_position_embeddings": 4096,
|
|
|
|
| 62 |
"type": "yarn"
|
| 63 |
},
|
| 64 |
"rope_theta": 50000.0,
|