Hugging Face's logo

mlfoundations
/
dclm-7b-it

Transformers
Safetensors
openlm
Inference Endpoints
Model card Files Community
4

Update config.json

#3
by jmercat - opened
base: refs/heads/main
←
from: refs/pr/3
Discussion Files changed
+25
-27
Files changed (1) hide show
  1. config.json +25 -27
config.json CHANGED
@@ -4,33 +4,31 @@
4
  ],
5
  "model_type": "openlm",
6
  "params": null,
7
- "params_args_dict": {
8
- "apply_qk_norm": true,
9
- "attn_activation": null,
10
- "attn_name": "auto",
11
- "attn_seq_scalar": null,
12
- "attn_seq_scalar_alpha": null,
13
- "dim": 4096,
14
- "ffn_type": "swiglu",
15
- "model": "open_lm_7b",
16
- "model_norm": "gain_only_lp_layer_norm",
17
- "moe_capacity_factor": 1.25,
18
- "moe_expert_model_parallelism": false,
19
- "moe_freq": 0,
20
- "moe_loss_weight": 0.1,
21
- "moe_num_experts": null,
22
- "moe_top_k": 2,
23
- "moe_weight_parallelism": false,
24
- "n_heads": 32,
25
- "n_layers": 32,
26
- "norm_eps": 1e-05,
27
- "positional_embedding_type": "rotary",
28
- "post_embed_norm": false,
29
- "qk_norm": true,
30
- "seq_len": 2048,
31
- "vocab_size": 50432,
32
- "weight_tying": false
33
- },
34
  "torch_dtype": "float32",
35
  "transformers_version": "4.41.2"
36
  }
 
4
  ],
5
  "model_type": "openlm",
6
  "params": null,
7
+ "apply_qk_norm": true,
8
+ "attn_activation": null,
9
+ "attn_name": "auto",
10
+ "attn_seq_scalar": null,
11
+ "attn_seq_scalar_alpha": null,
12
+ "dim": 4096,
13
+ "ffn_type": "swiglu",
14
+ "model": "open_lm_7b",
15
+ "model_norm": "gain_only_lp_layer_norm",
16
+ "moe_capacity_factor": 1.25,
17
+ "moe_expert_model_parallelism": false,
18
+ "moe_freq": 0,
19
+ "moe_loss_weight": 0.1,
20
+ "moe_num_experts": null,
21
+ "moe_top_k": 2,
22
+ "moe_weight_parallelism": false,
23
+ "n_heads": 32,
24
+ "n_layers": 32,
25
+ "norm_eps": 1e-05,
26
+ "positional_embedding_type": "rotary",
27
+ "post_embed_norm": false,
28
+ "qk_norm": true,
29
+ "seq_len": 2048,
30
+ "vocab_size": 50432,
31
+ "weight_tying": false,
 
 
32
  "torch_dtype": "float32",
33
  "transformers_version": "4.41.2"
34
  }