add VaseVLM-7B-rl adapter

model/7B-rl/README.md

@@ -0,0 +1,40 @@
+---
+base_model: Qwen/Qwen2.5-VL-7B-Instruct
+library_name: peft
+pipeline_tag: image-text-to-text
+tags:
+- lora
+- vasevlm
+- grpo
+- ancient-greek-pottery
+---
+
+# VaseVLM-7B-rl
+
+GRPO (RL) LoRA adapter for ancient Greek vase captioning (VaseVLM).
+
+## ⚠️ IMPORTANT — how to use correctly
+
+This RL adapter was trained **on top of the SFT model**, NOT on the raw base model.
+Applying it directly on `Qwen/Qwen2.5-VL-7B-Instruct` will NOT reproduce the results.
+You must stack it on the SFT model:
+
+**Step 1 — build the SFT model** = base + SFT adapter (merged)
+**Step 2 — apply this RL adapter** on the SFT model.
+
+```python
+from transformers import AutoModelForImageTextToText, AutoProcessor
+from peft import PeftModel
+
+# base
+base = AutoModelForImageTextToText.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct", torch_dtype="bfloat16", device_map="auto")
+
+# Step 1: apply SFT adapter and merge -> SFT model
+sft = PeftModel.from_pretrained(base, "AIGeeksGroup/VaseVLM", subfolder="model/7B-sft")
+sft = sft.merge_and_unload()
+
+# Step 2: apply this RL adapter on the SFT model
+model = PeftModel.from_pretrained(sft, "AIGeeksGroup/VaseVLM", subfolder="model/7B-rl")
+
+processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct")
+```

model/7B-rl/adapter_config.json

@@ -0,0 +1,38 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen2.5-VL-7B-Instruct",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "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": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.1,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": [],
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "^(model\\.language_model(?=\\.).*\\.(k_proj|q_proj|down_proj|o_proj|gate_proj|v_proj|up_proj))$",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

model/7B-rl/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75a474569b103b8ffc4b09f747e60b1eb261de135b3cf3dca9f540f2683eb6c6
+size 80797976

model/7B-rl/additional_config.json

@@ -0,0 +1 @@
+{"lora_dtype": null, "lorap_lr_ratio": null, "lorap_emb_lr": 1e-06}