Upload models/grpo_qwen06

models/grpo_qwen06/llm/adapter_config.json

@@ -32,13 +32,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "o_proj",
     "gate_proj",
+    "o_proj",
+    "up_proj",
+    "k_proj",
     "q_proj",
     "v_proj",
-    "down_proj",
-    "up_proj",
-    "k_proj"
+    "down_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

models/grpo_qwen06/llm/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6671a7c4cafe0d82f2a00f206f2ffab8037c1650d2229a4e0edb4e238181be53
+oid sha256:22e8bab57934dc231578270c1dbb1fad79b939595d8b9d0e4f18536d35941368
 size 80792456

models/grpo_qwen06/llm/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e3c7ba2b6d9654f5b6469870a1f687a8814cf7dcf1d423cbe08285b363a49d30
+oid sha256:5c02e8b0cfb206cfea5b653dc19d0490a566d3455973461b6b629e919ee42106
 size 161811835

models/grpo_qwen06/llm_best_0/adapter_config.json

@@ -32,13 +32,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "o_proj",
     "gate_proj",
+    "o_proj",
+    "up_proj",
+    "k_proj",
     "q_proj",
     "v_proj",
-    "down_proj",
-    "up_proj",
-    "k_proj"
+    "down_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

models/grpo_qwen06/llm_best_0/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6671a7c4cafe0d82f2a00f206f2ffab8037c1650d2229a4e0edb4e238181be53
+oid sha256:22e8bab57934dc231578270c1dbb1fad79b939595d8b9d0e4f18536d35941368
 size 80792456

models/grpo_qwen06/llm_best_0/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e3c7ba2b6d9654f5b6469870a1f687a8814cf7dcf1d423cbe08285b363a49d30
+oid sha256:5c02e8b0cfb206cfea5b653dc19d0490a566d3455973461b6b629e919ee42106
 size 161811835

models/grpo_qwen06/logs/eval_summary_5.json

@@ -1,53 +1,53 @@
 {
   "num_tokens": {
     "count": 64.0,
-    "mean": 101.078125,
-    "std": 117.161848,
-    "min": 4.0,
-    "25%": 26.0,
-    "50%": 42.0,
-    "75%": 144.75,
+    "mean": 190.59375,
+    "std": 102.343139,
+    "min": 73.0,
+    "25%": 110.5,
+    "50%": 140.0,
+    "75%": 285.25,
     "max": 384.0
   },
   "total_reward": {
     "count": 64.0,
-    "mean": 0.439597,
-    "std": 0.363853,
-    "min": -0.48348,
-    "25%": 0.174672,
-    "50%": 0.431622,
-    "75%": 0.780427,
-    "max": 0.983939
+    "mean": 2.355064,
+    "std": 0.74563,
+    "min": -0.47753,
+    "25%": 2.261295,
+    "50%": 2.535875,
+    "75%": 2.767111,
+    "max": 3.0
   },
   "think_format_reward": {
     "count": 64.0,
-    "mean": 0.0,
-    "std": 0.0,
+    "mean": 0.953125,
+    "std": 0.213042,
     "min": 0.0,
-    "25%": 0.0,
-    "50%": 0.0,
-    "75%": 0.0,
-    "max": 0.0
+    "25%": 1.0,
+    "50%": 1.0,
+    "75%": 1.0,
+    "max": 1.0
   },
   "output_datatype_reward": {
     "count": 64.0,
-    "mean": 0.0,
-    "std": 0.0,
-    "min": 0.0,
-    "25%": 0.0,
-    "50%": 0.0,
-    "75%": 0.0,
-    "max": 0.0
+    "mean": 0.460938,
+    "std": 0.16195,
+    "min": -0.5,
+    "25%": 0.5,
+    "50%": 0.5,
+    "75%": 0.5,
+    "max": 0.5
   },
   "output_schema_reward": {
     "count": 64.0,
-    "mean": 0.0,
-    "std": 0.0,
-    "min": 0.0,
-    "25%": 0.0,
-    "50%": 0.0,
-    "75%": 0.0,
-    "max": 0.0
+    "mean": 0.460938,
+    "std": 0.16195,
+    "min": -0.5,
+    "25%": 0.5,
+    "50%": 0.5,
+    "75%": 0.5,
+    "max": 0.5
   },
   "doom_loop_reward": {
     "count": 64.0,
@@ -61,18 +61,18 @@
   },
   "neuraltxt_reward": {
     "count": 64.0,
-    "mean": 0.486472,
-    "std": 0.296417,
-    "min": 0.01652,
-    "25%": 0.177485,
-    "50%": 0.453497,
-    "75%": 0.780427,
-    "max": 0.983939
+    "mean": 0.519127,
+    "std": 0.282372,
+    "min": 0.02247,
+    "25%": 0.261295,
+    "50%": 0.535875,
+    "75%": 0.767111,
+    "max": 1.0
   },
   "length_penalty_reward": {
     "count": 64.0,
-    "mean": -0.09375,
-    "std": 0.293785,
+    "mean": -0.078125,
+    "std": 0.27049,
     "min": -1.0,
     "25%": 0.0,
     "50%": 0.0,

models/grpo_qwen06/logs/train_metrics.jsonl

@@ -1 +1 @@
-{"training_steps": 240, "num_experiences": 320, "mean_loss": 0.008602914253666919, "mean_reward": 0.27183399023488164, "std_reward": 0.43991675177765277, "inference_score": 0.43959712982177734, "reward_breakdown": {"think_format_reward": {"mean": 0.0, "std": 0.0, "min": 0.0, "max": 0.0}, "output_datatype_reward": {"mean": 0.0, "std": 0.0, "min": 0.0, "max": 0.0}, "output_schema_reward": {"mean": 0.0, "std": 0.0, "min": 0.0, "max": 0.0}, "doom_loop_reward": {"mean": -0.015625, "std": 0.12401959300041199, "min": -1.0, "max": 0.0}, "neuraltxt_reward": {"mean": 0.44370901584625244, "std": 0.25785374641418457, "min": 0.0, "max": 0.9061836004257202}, "length_penalty_reward": {"mean": -0.3125, "std": 0.4635123908519745, "min": -1.0, "max": 0.0}}, "datatype_match_rate": 0.0, "schema_match_rate": 0.0, "think_format_reward_mean": 0.0, "output_datatype_reward_mean": 0.0, "output_schema_reward_mean": 0.0, "doom_loop_reward_mean": -0.015625, "neuraltxt_reward_mean": 0.44370899081695825, "length_penalty_reward_mean": -0.3125}
+{"training_steps": 240, "num_experiences": 320, "mean_loss": 0.01011337067787584, "mean_reward": 1.4421978439204395, "std_reward": 1.2547577476989022, "inference_score": 2.3550639152526855, "reward_breakdown": {"think_format_reward": {"mean": 0.5625, "std": 0.49607837200164795, "min": 0.0, "max": 1.0}, "output_datatype_reward": {"mean": 0.265625, "std": 0.26470428705215454, "min": -0.5, "max": 0.5}, "output_schema_reward": {"mean": 0.265625, "std": 0.26470428705215454, "min": -0.5, "max": 0.5}, "doom_loop_reward": {"mean": 0.0, "std": 0.0, "min": 0.0, "max": 0.0}, "neuraltxt_reward": {"mean": 0.46563535928726196, "std": 0.25898757576942444, "min": 0.06963597238063812, "max": 0.9661086797714233}, "length_penalty_reward": {"mean": -0.234375, "std": 0.42360755801200867, "min": -1.0, "max": 0.0}}, "datatype_match_rate": 0.546875, "schema_match_rate": 0.546875, "think_format_reward_mean": 0.5625, "output_datatype_reward_mean": 0.265625, "output_schema_reward_mean": 0.265625, "doom_loop_reward_mean": 0.0, "neuraltxt_reward_mean": 0.4656353541649878, "length_penalty_reward_mean": -0.234375}

models/grpo_qwen06/logs/train_metrics.txt

@@ -1 +1 @@
-training_steps=240 num_experiences=320 mean_loss=0.008602914253666919 mean_reward=0.27183399023488164 std_reward=0.43991675177765277 inference_score=0.43959712982177734 reward_breakdown={'think_format_reward': {'mean': 0.0, 'std': 0.0, 'min': 0.0, 'max': 0.0}, 'output_datatype_reward': {'mean': 0.0, 'std': 0.0, 'min': 0.0, 'max': 0.0}, 'output_schema_reward': {'mean': 0.0, 'std': 0.0, 'min': 0.0, 'max': 0.0}, 'doom_loop_reward': {'mean': -0.015625, 'std': 0.12401959300041199, 'min': -1.0, 'max': 0.0}, 'neuraltxt_reward': {'mean': 0.44370901584625244, 'std': 0.25785374641418457, 'min': 0.0, 'max': 0.9061836004257202}, 'length_penalty_reward': {'mean': -0.3125, 'std': 0.4635123908519745, 'min': -1.0, 'max': 0.0}} datatype_match_rate=0.0 schema_match_rate=0.0 think_format_reward_mean=0.0 output_datatype_reward_mean=0.0 output_schema_reward_mean=0.0 doom_loop_reward_mean=-0.015625 neuraltxt_reward_mean=0.44370899081695825 length_penalty_reward_mean=-0.3125
+training_steps=240 num_experiences=320 mean_loss=0.01011337067787584 mean_reward=1.4421978439204395 std_reward=1.2547577476989022 inference_score=2.3550639152526855 reward_breakdown={'think_format_reward': {'mean': 0.5625, 'std': 0.49607837200164795, 'min': 0.0, 'max': 1.0}, 'output_datatype_reward': {'mean': 0.265625, 'std': 0.26470428705215454, 'min': -0.5, 'max': 0.5}, 'output_schema_reward': {'mean': 0.265625, 'std': 0.26470428705215454, 'min': -0.5, 'max': 0.5}, 'doom_loop_reward': {'mean': 0.0, 'std': 0.0, 'min': 0.0, 'max': 0.0}, 'neuraltxt_reward': {'mean': 0.46563535928726196, 'std': 0.25898757576942444, 'min': 0.06963597238063812, 'max': 0.9661086797714233}, 'length_penalty_reward': {'mean': -0.234375, 'std': 0.42360755801200867, 'min': -1.0, 'max': 0.0}} datatype_match_rate=0.546875 schema_match_rate=0.546875 think_format_reward_mean=0.5625 output_datatype_reward_mean=0.265625 output_schema_reward_mean=0.265625 doom_loop_reward_mean=0.0 neuraltxt_reward_mean=0.4656353541649878 length_penalty_reward_mean=-0.234375