Commit
·
ce9f014
1
Parent(s):
ff0745b
Training in progress, step 3334
Browse files- .gitignore +1 -0
- config.json +84 -0
- pytorch_model.bin +3 -0
- training_args.bin +3 -0
.gitignore
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
checkpoint-*/
|
config.json
ADDED
|
@@ -0,0 +1,84 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"act_dim": 6,
|
| 3 |
+
"action_tanh": true,
|
| 4 |
+
"activation_function": "relu",
|
| 5 |
+
"adv_act_dim": 6,
|
| 6 |
+
"architectures": [
|
| 7 |
+
"MultipartARDT"
|
| 8 |
+
],
|
| 9 |
+
"attn_pdrop": 0.1,
|
| 10 |
+
"bos_token_id": 50256,
|
| 11 |
+
"context_size": 20,
|
| 12 |
+
"embd_pdrop": 0.1,
|
| 13 |
+
"eos_token_id": 50256,
|
| 14 |
+
"flag": 0,
|
| 15 |
+
"hidden_size": 128,
|
| 16 |
+
"initializer_range": 0.02,
|
| 17 |
+
"lambda1": 0.1,
|
| 18 |
+
"lambda2": 1.0,
|
| 19 |
+
"layer_norm_epsilon": 1e-05,
|
| 20 |
+
"log_interval_steps": 100,
|
| 21 |
+
"max_ep_len": 999,
|
| 22 |
+
"max_ep_return": 2054.7955639899983,
|
| 23 |
+
"max_obs_len": 999,
|
| 24 |
+
"max_obs_return": 2054.7955639899983,
|
| 25 |
+
"min_ep_return": -31.996896582000012,
|
| 26 |
+
"min_obs_return": -31.996896582000012,
|
| 27 |
+
"model_type": "decision_transformer",
|
| 28 |
+
"n_head": 1,
|
| 29 |
+
"n_inner": null,
|
| 30 |
+
"n_layer": 3,
|
| 31 |
+
"n_positions": 1024,
|
| 32 |
+
"pr_act_dim": 6,
|
| 33 |
+
"reorder_and_upcast_attn": false,
|
| 34 |
+
"resid_pdrop": 0.1,
|
| 35 |
+
"returns_scale": 1000,
|
| 36 |
+
"rtg_shift": 31.996996582000012,
|
| 37 |
+
"scale_attn_by_inverse_layer_idx": false,
|
| 38 |
+
"scale_attn_weights": true,
|
| 39 |
+
"state_dim": 17,
|
| 40 |
+
"state_mean": [
|
| 41 |
+
-0.11173252361577966,
|
| 42 |
+
0.06332421416752207,
|
| 43 |
+
0.10642913798176151,
|
| 44 |
+
0.016696281865103354,
|
| 45 |
+
0.01292316632921915,
|
| 46 |
+
0.0252802808143255,
|
| 47 |
+
-0.163862748115139,
|
| 48 |
+
-0.2785617364192604,
|
| 49 |
+
1.800042231687943,
|
| 50 |
+
-0.023767931373515298,
|
| 51 |
+
0.0033658001950578565,
|
| 52 |
+
0.016833230392745077,
|
| 53 |
+
-0.05654895540758762,
|
| 54 |
+
0.057139100560966265,
|
| 55 |
+
-0.0686476457401574,
|
| 56 |
+
0.04198423127298879,
|
| 57 |
+
0.005969089117189445
|
| 58 |
+
],
|
| 59 |
+
"state_std": [
|
| 60 |
+
0.0985398737294268,
|
| 61 |
+
0.19583829223764262,
|
| 62 |
+
0.3315948858702887,
|
| 63 |
+
0.32217133624379024,
|
| 64 |
+
0.39319405952592607,
|
| 65 |
+
0.4463002698189581,
|
| 66 |
+
0.3517920731882343,
|
| 67 |
+
0.2777303946401783,
|
| 68 |
+
0.7745754765732835,
|
| 69 |
+
0.9800573190633699,
|
| 70 |
+
1.365188942289809,
|
| 71 |
+
4.87113384669823,
|
| 72 |
+
5.058535039406536,
|
| 73 |
+
6.395587223733579,
|
| 74 |
+
5.947895650994883,
|
| 75 |
+
6.4933930314513075,
|
| 76 |
+
5.279823051159204
|
| 77 |
+
],
|
| 78 |
+
"torch_dtype": "float32",
|
| 79 |
+
"total_train_steps": 10000,
|
| 80 |
+
"transformers_version": "4.31.0",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 1,
|
| 83 |
+
"warmup_steps": 1000
|
| 84 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:024db5141dbc2864b23d99dd7632d1c664ffa1631351ac232dd252cdae09f1c3
|
| 3 |
+
size 10397121
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f47eaf90ef47a228c231a0dadef32e2e7592805825beb12785ba04d9877c8627
|
| 3 |
+
size 4283
|