初始化项目，由ModelHub XC社区提供模型

Model: Likithp/v10_rand_s1 Source: Original Platform
2026-07-27 09:34:14 +08:00
commit 7f20185eea
62 changed files with 910619 additions and 0 deletions
--- a/training_config.json
+++ b/training_config.json
@@ -0,0 +1,195 @@
+{
+  "mode": "rand",
+  "seed": 1,
+  "hf_repo": "Likithp/v10_rand_s1",
+  "base_model": "Qwen/Qwen2.5-0.5B",
+  "dataset": "data/cs7_rand_v3",
+  "dataset_version": "cs7_v3",
+  "trained_at": "2026-06-06T07:50:09.185807+00:00",
+  "optimizer": "AdamW",
+  "lr": 3e-05,
+  "weight_decay": 0.0,
+  "batch_size": 8,
+  "grad_accum": 8,
+  "effective_batch": 64,
+  "epochs": 5,
+  "warmup_steps": 200,
+  "grad_clip": 0.5,
+  "lr_schedule": "cosine",
+  "checkpoint_criterion": "train_loss",
+  "max_seq_len": 256,
+  "dtype": "torch.bfloat16",
+  "eval_method": "teacher_forcing_argmax",
+  "alias_groups": {
+    "T1": [
+      "act",
+      "cst",
+      "emp",
+      "inv",
+      "ord",
+      "spl",
+      "txn"
+    ],
+    "T2": [
+      "brc",
+      "ctg",
+      "dpt",
+      "empl",
+      "ordr",
+      "prd",
+      "prj",
+      "rgn",
+      "shp",
+      "tsk",
+      "whs"
+    ]
+  },
+  "log_every_steps": 50,
+  "eval_n": 500,
+  "dry_run": false,
+  "best_epoch": 5,
+  "best_train_loss": 0.091381,
+  "val_exact_match": 0.054,
+  "val_outer_alias_acc": 1.0,
+  "val_inner_alias_acc": 0.046,
+  "val_t1_inner_alias_acc": 0.0,
+  "val_t2_inner_alias_acc": 0.075783,
+  "val_t1_t2_gap_pp": -7.58,
+  "val_inner_by_alias": {
+    "act": {
+      "correct": 0,
+      "total": 61,
+      "pct": 0.0,
+      "token_group": "T1"
+    },
+    "brc": {
+      "correct": 0,
+      "total": 50,
+      "pct": 0.0,
+      "token_group": "T2"
+    },
+    "cst": {
+      "correct": 0,
+      "total": 62,
+      "pct": 0.0,
+      "token_group": "T1"
+    },
+    "ctg": {
+      "correct": 0,
+      "total": 58,
+      "pct": 0.0,
+      "token_group": "T2"
+    },
+    "dpt": {
+      "correct": 0,
+      "total": 47,
+      "pct": 0.0,
+      "token_group": "T2"
+    },
+    "emp": {
+      "correct": 0,
+      "total": 40,
+      "pct": 0.0,
+      "token_group": "T1"
+    },
+    "empl": {
+      "correct": 0,
+      "total": 43,
+      "pct": 0.0,
+      "token_group": "T2"
+    },
+    "inv": {
+      "correct": 0,
+      "total": 56,
+      "pct": 0.0,
+      "token_group": "T1"
+    },
+    "ord": {
+      "correct": 0,
+      "total": 47,
+      "pct": 0.0,
+      "token_group": "T1"
+    },
+    "ordr": {
+      "correct": 0,
+      "total": 50,
+      "pct": 0.0,
+      "token_group": "T2"
+    },
+    "prd": {
+      "correct": 33,
+      "total": 55,
+      "pct": 60.0,
+      "token_group": "T2"
+    },
+    "prj": {
+      "correct": 13,
+      "total": 65,
+      "pct": 20.0,
+      "token_group": "T2"
+    },
+    "rgn": {
+      "correct": 0,
+      "total": 64,
+      "pct": 0.0,
+      "token_group": "T2"
+    },
+    "shp": {
+      "correct": 0,
+      "total": 59,
+      "pct": 0.0,
+      "token_group": "T2"
+    },
+    "spl": {
+      "correct": 0,
+      "total": 56,
+      "pct": 0.0,
+      "token_group": "T1"
+    },
+    "tsk": {
+      "correct": 0,
+      "total": 59,
+      "pct": 0.0,
+      "token_group": "T2"
+    },
+    "txn": {
+      "correct": 0,
+      "total": 71,
+      "pct": 0.0,
+      "token_group": "T1"
+    },
+    "whs": {
+      "correct": 0,
+      "total": 57,
+      "pct": 0.0,
+      "token_group": "T2"
+    }
+  },
+  "train_log": [
+    {
+      "epoch": 1,
+      "loss": 0.121444,
+      "val_em": 0.056
+    },
+    {
+      "epoch": 2,
+      "loss": 0.092489,
+      "val_em": 0.056
+    },
+    {
+      "epoch": 3,
+      "loss": 0.091918,
+      "val_em": 0.036
+    },
+    {
+      "epoch": 4,
+      "loss": 0.091536,
+      "val_em": 0.056
+    },
+    {
+      "epoch": 5,
+      "loss": 0.091381,
+      "val_em": 0.048
+    }
+  ]
+}