初始化项目，由ModelHub XC社区提供模型

Model: nikitastheo/babylm-ara-ell-sequential_interleaved Source: Original Platform
2026-07-20 04:10:13 +08:00
commit eaede82b41
9 changed files with 220705 additions and 0 deletions
--- a/training_metadata.json
+++ b/training_metadata.json
@@ -0,0 +1,67 @@
+{
+  "input_config": {
+    "model_type": "gpt2",
+    "architectures": [
+      "GPT2LMHeadModel"
+    ],
+    "activation_function": "gelu",
+    "attn_pdrop": 0.1,
+    "embd_pdrop": 0.1,
+    "resid_pdrop": 0.1,
+    "initializer_range": 0.02,
+    "layer_norm_epsilon": 1e-05,
+    "n_embd": 768,
+    "n_inner": 3072,
+    "n_head": 12,
+    "n_layer": 12,
+    "vocab_size": 1,
+    "n_ctx": 512,
+    "n_positions": 512
+  },
+  "num_steps_switch": 7140,
+  "save_steps": [
+    1,
+    2,
+    4,
+    8,
+    16,
+    32,
+    64,
+    128,
+    256,
+    512,
+    1000,
+    2000,
+    3000,
+    4000,
+    5000,
+    6000,
+    7000,
+    7140,
+    7141,
+    7142,
+    7144,
+    7148,
+    7156,
+    7172,
+    7204,
+    7268,
+    7396,
+    7652,
+    8000,
+    9000,
+    10000,
+    11000,
+    12000,
+    13000,
+    14000,
+    15000,
+    16000,
+    17000,
+    18000,
+    19000,
+    20000,
+    21000,
+    21390
+  ]
+}