Upload folder using huggingface_hub
Browse files- config.json +65 -0
- model.pt +3 -0
config.json
ADDED
|
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "eqlm-medium-370m",
|
| 3 |
+
"hf_config": {
|
| 4 |
+
"org": "SandyResearch",
|
| 5 |
+
"name": "eqlm-medium-370m"
|
| 6 |
+
},
|
| 7 |
+
"block_size": 2048,
|
| 8 |
+
"n_embd": 1280,
|
| 9 |
+
"intermediate_size": 5120,
|
| 10 |
+
"num_attention_heads": 10,
|
| 11 |
+
"num_key_value_heads": 10,
|
| 12 |
+
"vocab_size": 32768,
|
| 13 |
+
"padding_multiple": 64,
|
| 14 |
+
"padded_vocab_size": 32768,
|
| 15 |
+
"rope_settings": {
|
| 16 |
+
"use_rope": true,
|
| 17 |
+
"rope_condense_ratio": 1,
|
| 18 |
+
"rope_base": 50000
|
| 19 |
+
},
|
| 20 |
+
"use_abacus": false,
|
| 21 |
+
"randomize_positions_from": null,
|
| 22 |
+
"block_class_name": "TransformerPreNormBlock",
|
| 23 |
+
"norm_class_name": "RMSNorm",
|
| 24 |
+
"attn_impl": "flash",
|
| 25 |
+
"norm_eps": 1e-05,
|
| 26 |
+
"mlp_class_name": "BaseMLP",
|
| 27 |
+
"nonlin_name": "ReLU2",
|
| 28 |
+
"bias": false,
|
| 29 |
+
"qk_bias": false,
|
| 30 |
+
"init_strategy": "scaled-zero",
|
| 31 |
+
"init_orthogonal": true,
|
| 32 |
+
"skip_initialization": false,
|
| 33 |
+
"mup_model_scaling_factor": 1,
|
| 34 |
+
"use_fused_head": "pytorch",
|
| 35 |
+
"debias_attention": false,
|
| 36 |
+
"center_attention": false,
|
| 37 |
+
"clip_qkv": null,
|
| 38 |
+
"qk_norm": true,
|
| 39 |
+
"logit_softcap": null,
|
| 40 |
+
"causal": true,
|
| 41 |
+
"activation_checkpoint_impl": "per-block",
|
| 42 |
+
"simple_ops": false,
|
| 43 |
+
"strategy": "ddp",
|
| 44 |
+
"n_backbone_layers": 15,
|
| 45 |
+
"n_fp_blocks": 2,
|
| 46 |
+
"tie_embeddings": true,
|
| 47 |
+
"solver": "anderson",
|
| 48 |
+
"max_iter": 64,
|
| 49 |
+
"min_iter": 8,
|
| 50 |
+
"tol": 0.0002,
|
| 51 |
+
"anderson_m": 5,
|
| 52 |
+
"anderson_beta": 1.0,
|
| 53 |
+
"backward_type": "onestep",
|
| 54 |
+
"backward_max_iter": 64,
|
| 55 |
+
"backward_min_iter": 6,
|
| 56 |
+
"backward_tol": 0.0002,
|
| 57 |
+
"adjoint_grad_clip": null,
|
| 58 |
+
"layer_scale_init": 0.75,
|
| 59 |
+
"gamma_max": 1.0,
|
| 60 |
+
"fp_lr_scale": 0.4,
|
| 61 |
+
"fp_wd": 0.1,
|
| 62 |
+
"recurrent_embedding_dimension": 1280,
|
| 63 |
+
"model_class_name": "EQLM",
|
| 64 |
+
"_class_name": "EQLMConfig"
|
| 65 |
+
}
|
model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:baf74b7473fa869c770230cddccc2e6b3ae493851971465f731a8d6496e87bef
|
| 3 |
+
size 1506002359
|