初始化项目，由ModelHub XC社区提供模型

Model: Mathieu-Thomas-JOSSET/joke-finetome-model-gguf-phi4-20260112-081758
Source: Original Platform

reports/config_snapshot.json

@@ -0,0 +1,28 @@
+{
+  "MODEL_NAME": "unsloth/Phi-4-unsloth-bnb-4bit",
+  "CHAT_TEMPLATE": "phi-4",
+  "MAX_SEQ_LENGTH": 2048,
+  "LOAD_IN_4BIT": true,
+  "DATASET_NAME": "Mathieu-Thomas-JOSSET/michael_abab_conversations_infini_instruct.jsonl",
+  "DATASET_SPLIT": "train",
+  "PER_DEVICE_TRAIN_BATCH_SIZE": 2,
+  "GRADIENT_ACCUMULATION_STEPS": 4,
+  "WARMUP_STEPS": 10,
+  "MAX_STEPS": 2000,
+  "LEARNING_RATE": 9.95267419777795e-06,
+  "LR_AUTO_ENABLED": true,
+  "LR_AUTO_USE_N": "train",
+  "LR_AUTO_N_REF": 1436,
+  "LR_AUTO_BASE": 1e-05,
+  "LR_AUTO_MULT": 0.5,
+  "LR_AUTO_FINAL": 5e-06,
+  "WEIGHT_DECAY": 0.009206070410847844,
+  "LR_SCHEDULER_TYPE": "linear",
+  "SEED": 3407,
+  "PLOTLY_DARK_MODE": true,
+  "PLOTLY_BASE_COLOR": "#00CC96",
+  "PLOTLY_EMA_SPAN": 25,
+  "HF_REPO_ID_MERGED_RESOLVED": "Mathieu-Thomas-JOSSET/joke-20260112-081758",
+  "HF_REPO_ID_GGUF_RESOLVED": "Mathieu-Thomas-JOSSET/joke-finetome-model-gguf-phi4-20260112-081758",
+  "HF_ARTIFACTS_DIR_IN_REPO": "reports"
+}

reports/inference_sample.json

@@ -0,0 +1,10 @@
+{
+  "source": "dataset",
+  "index": 232,
+  "messages": [
+    {
+      "role": "user",
+      "content": "Dwight: \"Oh, man.\"\nMichael: \"How did we do it?\"\nDwight: \"I don’t … have no idea.\""
+    }
+  ]
+}

reports/run_manifest.json

@@ -0,0 +1,57 @@
+{
+  "started_at": "2026-01-12T08:52:02+01:00",
+  "repos": {
+    "merged": "Mathieu-Thomas-JOSSET/joke-20260112-081758",
+    "gguf": "Mathieu-Thomas-JOSSET/joke-finetome-model-gguf-phi4-20260112-081758"
+  },
+  "model_name": "unsloth/Phi-4-unsloth-bnb-4bit",
+  "dataset": {
+    "name": "Mathieu-Thomas-JOSSET/michael_abab_conversations_infini_instruct.jsonl",
+    "split": "train"
+  },
+  "training": {
+    "max_steps": 2000,
+    "learning_rate": 9.95267419777795e-06,
+    "per_device_train_batch_size": 2,
+    "gradient_accumulation_steps": 4,
+    "max_seq_length": 2048,
+    "seed": 3407,
+    "optimizer": "adamw_8bit",
+    "lr_scheduler_type": "linear"
+  },
+  "auto_lr": {
+    "enabled": true,
+    "use_n": "train",
+    "n_ref": 1436,
+    "base": 1e-05,
+    "mult": 0.5,
+    "final": 5e-06
+  },
+  "metrics": {
+    "train_runtime": 2220.807,
+    "train_samples_per_second": 5.944,
+    "train_steps_per_second": 0.743,
+    "total_flos": 4.35104765343744e+16,
+    "train_loss": 1.654274252106746,
+    "epoch": 3.3342618384401113
+  },
+  "best": {
+    "checkpoint": "/content/outputs/continue_r1_from_350_20260112_073729/checkpoint-100",
+    "metric": 2.2380564212799072,
+    "metric_name": "eval_loss"
+  },
+  "plotly": {
+    "html": "reports/training_loss_step.html",
+    "png": null
+  },
+  "inference_sample": {
+    "source": "dataset",
+    "index": 232,
+    "messages": [
+      {
+        "role": "user",
+        "content": "Dwight: \"Oh, man.\"\nMichael: \"How did we do it?\"\nDwight: \"I don’t … have no idea.\""
+      }
+    ]
+  }
+}