Upload recovered code-plan SFT LoRA

README.md

@@ -1,3 +1,94 @@
 ---
-license: openmdw-1.0
+library_name: peft
+license: apache-2.0
+base_model: Qwen/Qwen2.5-Coder-7B
+datasets:
+- keryszhan/agent-code-rl-artifacts
+pipeline_tag: text-generation
+language:
+- en
+- zh
+tags:
+- code
+- code-generation
+- llama-factory
+- lora
+- reasoning
 ---
+
+# Qwen2.5-Coder-7B Code-Plan SFT LoRA
+
+LoRA adapter trained on four-step planning followed by Python code generation.
+The adapter is a recovered artifact from the Agent Code RL project; it is the
+SFT starting point used before the later PRM/GRPO work.
+
+This is an adapter, not a standalone model. Load it with
+`Qwen/Qwen2.5-Coder-7B`.
+
+## Usage
+
+```python
+from peft import PeftModel
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+base_id = "Qwen/Qwen2.5-Coder-7B"
+adapter_id = "keryszhan/qwen2.5-coder-7b-code-plan-sft"
+
+tokenizer = AutoTokenizer.from_pretrained(base_id, trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained(
+    base_id,
+    device_map="auto",
+    torch_dtype="auto",
+    trust_remote_code=True,
+)
+model = PeftModel.from_pretrained(model, adapter_id)
+```
+
+The training examples use an instruction/response format. Responses contain
+four explicit reasoning steps and one Python code block; callers should preserve
+that protocol when constructing prompts.
+
+## Training
+
+- Base model: `Qwen/Qwen2.5-Coder-7B`
+- Method: LoRA SFT via LLaMA-Factory
+- Rank / alpha / dropout: 16 / 16 / 0.05
+- Target modules: attention and MLP projection layers
+- Learning rate: `2e-4`
+- Effective batch size: 16
+- Epochs: 3
+- Seed: 42
+- Final step: 177
+- Recorded training loss: 0.3711
+
+Recorded framework versions: PEFT 0.15.1, Transformers 4.51.3,
+PyTorch 2.5.1+cu121, Datasets 3.5.0, and Tokenizers 0.21.1.
+
+## Evaluation evidence
+
+Recovered project records report approximately 44.8% for the base model and
+61.29% for an SFT evaluation. A separate 122-item `prm_val` evaluation records
+70 passes (57.38%). These historical measurements were not reconstructed from
+scratch during repository cleanup, so treat them as project evidence rather
+than a standardized leaderboard result.
+
+The later GRPO evaluation files must not be attributed to this adapter: the
+final GRPO adapter is no longer available, and one candidate result has no
+checkpoint identity.
+
+## Limitations
+
+- Generates and may execute Python code; use an actual sandbox for untrusted output.
+- Trained on HumanEval/MBPP-derived tasks and should not be used to claim an
+  uncontaminated benchmark result on those task families.
+- The model may emit incorrect reasoning, insecure code, or non-terminating code.
+- Training-data redistribution is tracked separately from this model release.
+
+The related process dataset is maintained at
+[`keryszhan/agent-code-rl-artifacts`](https://huggingface.co/datasets/keryszhan/agent-code-rl-artifacts).
+
+## License
+
+The base model is Apache-2.0. This adapter is released under Apache-2.0 subject
+to the terms and attribution requirements of the base model and applicable
+training-data licenses.

adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-Coder-7B",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "k_proj",
+    "q_proj",
+    "v_proj",
+    "o_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:099806554d5b0080448a5e7f1d46faeeb7bed35703f05c2f734b98200fa2fce4
+size 161533192

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 2.962025316455696,
+    "total_flos": 1.1532158740468531e+17,
+    "train_loss": 0.37109487339601677,
+    "train_runtime": 1389.7366,
+    "train_samples_per_second": 2.046,
+    "train_steps_per_second": 0.127
+}

trainer_log.jsonl

@@ -0,0 +1,18 @@
+{"current_steps": 10, "total_steps": 177, "loss": 0.634, "lr": 0.00019984815164333163, "epoch": 0.16877637130801687, "percentage": 5.65, "elapsed_time": "0:01:36", "remaining_time": "0:26:56"}
+{"current_steps": 20, "total_steps": 177, "loss": 0.488, "lr": 0.0001971614350559814, "epoch": 0.33755274261603374, "percentage": 11.3, "elapsed_time": "0:03:09", "remaining_time": "0:24:50"}
+{"current_steps": 30, "total_steps": 177, "loss": 0.4406, "lr": 0.00019120447901834706, "epoch": 0.5063291139240507, "percentage": 16.95, "elapsed_time": "0:04:30", "remaining_time": "0:22:07"}
+{"current_steps": 40, "total_steps": 177, "loss": 0.4385, "lr": 0.0001821777815225245, "epoch": 0.6751054852320675, "percentage": 22.6, "elapsed_time": "0:05:50", "remaining_time": "0:19:58"}
+{"current_steps": 50, "total_steps": 177, "loss": 0.4207, "lr": 0.00017038516128259115, "epoch": 0.8438818565400844, "percentage": 28.25, "elapsed_time": "0:07:10", "remaining_time": "0:18:12"}
+{"current_steps": 60, "total_steps": 177, "loss": 0.407, "lr": 0.00015622353186727544, "epoch": 1.0, "percentage": 33.9, "elapsed_time": "0:08:20", "remaining_time": "0:16:15"}
+{"current_steps": 70, "total_steps": 177, "loss": 0.3725, "lr": 0.00014016954246529696, "epoch": 1.1687763713080168, "percentage": 39.55, "elapsed_time": "0:09:38", "remaining_time": "0:14:44"}
+{"current_steps": 80, "total_steps": 177, "loss": 0.3632, "lr": 0.00012276353492572935, "epoch": 1.3375527426160336, "percentage": 45.2, "elapsed_time": "0:10:56", "remaining_time": "0:13:16"}
+{"current_steps": 90, "total_steps": 177, "loss": 0.3574, "lr": 0.00010459135704399718, "epoch": 1.5063291139240507, "percentage": 50.85, "elapsed_time": "0:12:14", "remaining_time": "0:11:49"}
+{"current_steps": 100, "total_steps": 177, "loss": 0.3303, "lr": 8.626464421815919e-05, "epoch": 1.6751054852320675, "percentage": 56.5, "elapsed_time": "0:13:29", "remaining_time": "0:10:23"}
+{"current_steps": 110, "total_steps": 177, "loss": 0.3385, "lr": 6.840023315140475e-05, "epoch": 1.8438818565400843, "percentage": 62.15, "elapsed_time": "0:14:45", "remaining_time": "0:08:59"}
+{"current_steps": 120, "total_steps": 177, "loss": 0.3286, "lr": 5.159940049010015e-05, "epoch": 2.0, "percentage": 67.8, "elapsed_time": "0:15:55", "remaining_time": "0:07:33"}
+{"current_steps": 130, "total_steps": 177, "loss": 0.2901, "lr": 3.642762517900322e-05, "epoch": 2.168776371308017, "percentage": 73.45, "elapsed_time": "0:17:11", "remaining_time": "0:06:12"}
+{"current_steps": 140, "total_steps": 177, "loss": 0.2916, "lr": 2.339555568810221e-05, "epoch": 2.3375527426160336, "percentage": 79.1, "elapsed_time": "0:18:27", "remaining_time": "0:04:52"}
+{"current_steps": 150, "total_steps": 177, "loss": 0.2865, "lr": 1.294182271221377e-05, "epoch": 2.5063291139240507, "percentage": 84.75, "elapsed_time": "0:19:43", "remaining_time": "0:03:32"}
+{"current_steps": 160, "total_steps": 177, "loss": 0.2854, "lr": 5.418275829936537e-06, "epoch": 2.6751054852320673, "percentage": 90.4, "elapsed_time": "0:21:01", "remaining_time": "0:02:14"}
+{"current_steps": 170, "total_steps": 177, "loss": 0.2898, "lr": 1.0781410234342094e-06, "epoch": 2.8438818565400843, "percentage": 96.05, "elapsed_time": "0:22:15", "remaining_time": "0:00:54"}
+{"current_steps": 177, "total_steps": 177, "epoch": 2.962025316455696, "percentage": 100.0, "elapsed_time": "0:23:09", "remaining_time": "0:00:00"}

trainer_state.json

@@ -0,0 +1,162 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.962025316455696,
+  "eval_steps": 500,
+  "global_step": 177,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.16877637130801687,
+      "grad_norm": 0.1649426966905594,
+      "learning_rate": 0.00019984815164333163,
+      "loss": 0.634,
+      "step": 10
+    },
+    {
+      "epoch": 0.33755274261603374,
+      "grad_norm": 0.09811161458492279,
+      "learning_rate": 0.0001971614350559814,
+      "loss": 0.488,
+      "step": 20
+    },
+    {
+      "epoch": 0.5063291139240507,
+      "grad_norm": 0.12041426450014114,
+      "learning_rate": 0.00019120447901834706,
+      "loss": 0.4406,
+      "step": 30
+    },
+    {
+      "epoch": 0.6751054852320675,
+      "grad_norm": 0.11681783199310303,
+      "learning_rate": 0.0001821777815225245,
+      "loss": 0.4385,
+      "step": 40
+    },
+    {
+      "epoch": 0.8438818565400844,
+      "grad_norm": 0.12966671586036682,
+      "learning_rate": 0.00017038516128259115,
+      "loss": 0.4207,
+      "step": 50
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.2353854924440384,
+      "learning_rate": 0.00015622353186727544,
+      "loss": 0.407,
+      "step": 60
+    },
+    {
+      "epoch": 1.1687763713080168,
+      "grad_norm": 0.172561913728714,
+      "learning_rate": 0.00014016954246529696,
+      "loss": 0.3725,
+      "step": 70
+    },
+    {
+      "epoch": 1.3375527426160336,
+      "grad_norm": 0.162893146276474,
+      "learning_rate": 0.00012276353492572935,
+      "loss": 0.3632,
+      "step": 80
+    },
+    {
+      "epoch": 1.5063291139240507,
+      "grad_norm": 0.1781071275472641,
+      "learning_rate": 0.00010459135704399718,
+      "loss": 0.3574,
+      "step": 90
+    },
+    {
+      "epoch": 1.6751054852320675,
+      "grad_norm": 0.2111092209815979,
+      "learning_rate": 8.626464421815919e-05,
+      "loss": 0.3303,
+      "step": 100
+    },
+    {
+      "epoch": 1.8438818565400843,
+      "grad_norm": 0.20615039765834808,
+      "learning_rate": 6.840023315140475e-05,
+      "loss": 0.3385,
+      "step": 110
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.49355044960975647,
+      "learning_rate": 5.159940049010015e-05,
+      "loss": 0.3286,
+      "step": 120
+    },
+    {
+      "epoch": 2.168776371308017,
+      "grad_norm": 0.22129110991954803,
+      "learning_rate": 3.642762517900322e-05,
+      "loss": 0.2901,
+      "step": 130
+    },
+    {
+      "epoch": 2.3375527426160336,
+      "grad_norm": 0.22937293350696564,
+      "learning_rate": 2.339555568810221e-05,
+      "loss": 0.2916,
+      "step": 140
+    },
+    {
+      "epoch": 2.5063291139240507,
+      "grad_norm": 0.2439713031053543,
+      "learning_rate": 1.294182271221377e-05,
+      "loss": 0.2865,
+      "step": 150
+    },
+    {
+      "epoch": 2.6751054852320673,
+      "grad_norm": 0.23686474561691284,
+      "learning_rate": 5.418275829936537e-06,
+      "loss": 0.2854,
+      "step": 160
+    },
+    {
+      "epoch": 2.8438818565400843,
+      "grad_norm": 0.2566355764865875,
+      "learning_rate": 1.0781410234342094e-06,
+      "loss": 0.2898,
+      "step": 170
+    },
+    {
+      "epoch": 2.962025316455696,
+      "step": 177,
+      "total_flos": 1.1532158740468531e+17,
+      "train_loss": 0.37109487339601677,
+      "train_runtime": 1389.7366,
+      "train_samples_per_second": 2.046,
+      "train_steps_per_second": 0.127
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 177,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 200,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1532158740468531e+17,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}