Upload 9 files

README.md

@@ -1,3 +1,90 @@
----
-license: cc-by-4.0
----
+**License**: CC‑BY‑4.0  
+**Datasets**:  
+- `clarin-pl/poquad`  
+
+**Language(s)**: Polish (pl)  
+**Library**: 🤗 Transformers  
+
+### Model Overview  
+
+- **Model name**: `radlab/polish-qa-v2-bnb`  
+- **Developer**: [radlab.dev](https://radlab.dev)  
+- **Model type**: Extractive Question‑Answering (QA)  
+- **Base model**: `sdadas/polish-roberta-large-v2` (fine‑tuned for QA)  
+- **Quantization**: 8‑bit inference‑only quantization via **bitsandbytes** (`load_in_8bit=True`, double‑quantization enabled, `qa_outputs` excluded from quantization)  
+- **Maximum context size**: 512 tokens  
+
+### Intended Use  
+
+This model is designed for **extractive QA** on Polish text. Given a question and a context passage,
+it returns the most relevant span of the context as the answer.
+
+### Limitations  
+
+- The model works best with contexts up to 512 tokens. Longer passages should be truncated or split.  
+- 8‑bit quantization reduces memory usage and inference latency but may introduce a slight drop in accuracy 
+compared with the full‑precision model.  
+- Only suitable for inference; it cannot be further fine‑tuned while kept in 8‑bit mode.
+
+### How to Use  
+
+```python
+from transformers import pipeline
+
+model_path = "radlab/polish-qa-v2-bnb"
+
+qa = pipeline(
+    "question-answering",
+    model=model_path,
+)
+
+question = "Co będzie w budowanym obiekcie?"
+context = """Pozwolenie na budowę zostało wydane w marcu. Pierwsze prace przygotowawcze
+na terenie przy ul. Wojska Polskiego już się rozpoczęły.
+Działkę ogrodzono, pojawił się również monitoring, a także kontenery
+dla pracowników budowy. Na ten moment nie jest znana lista sklepów,
+które pojawią się w nowym pasażu handlowym."""
+
+result = qa(
+    question=question,
+    context=context.replace("\n", " ")
+)
+
+print(result)
+```
+
+
+**Sample output**
+
+```json
+{
+  "score": 0.32568359375,
+  "start": 259,
+  "end": 268,
+  "answer": "sklepów,"
+}
+```
+
+
+### Technical Details  
+
+- **Quantization strategy**: `BitsAndBytesStrategy` (8‑bit, double‑quant, `qa_outputs` excluded).  
+- **Loading code (for reference)**  
+
+```python
+from transformers import AutoConfig, BitsAndBytesConfig, AutoModelForQuestionAnswering
+
+config = AutoConfig.from_pretrained(original_path)
+bnb_cfg = BitsAndBytesConfig(
+    load_in_8bit=True,
+    bnb_8bit_use_double_quant=True,
+    bnb_8bit_excluded_modules=["qa_outputs"],
+)
+
+model = AutoModelForQuestionAnswering.from_pretrained(
+    original_path,
+    config=config,
+    quantization_config=bnb_cfg,
+    device_map="auto",
+)
+```

added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<mask>": 128000
+}

config.json

@@ -0,0 +1,41 @@
+{
+  "architectures": [
+    "RobertaForQuestionAnswering"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "dtype": "float16",
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 514,
+  "model_type": "roberta",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "quantization_config": {
+    "_load_in_4bit": false,
+    "_load_in_8bit": true,
+    "bnb_4bit_compute_dtype": "float32",
+    "bnb_4bit_quant_storage": "uint8",
+    "bnb_4bit_quant_type": "fp4",
+    "bnb_4bit_use_double_quant": false,
+    "llm_int8_enable_fp32_cpu_offload": false,
+    "llm_int8_has_fp16_weight": false,
+    "llm_int8_skip_modules": null,
+    "llm_int8_threshold": 6.0,
+    "load_in_4bit": false,
+    "load_in_8bit": true,
+    "quant_method": "bitsandbytes"
+  },
+  "transformers_version": "4.56.2",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 128001
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85e32f37e1a7650d21a081cc66686d580e092a6208791aba7cd2fed4bfb897e3
+size 566804116

special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "<mask>",
+    "lstrip": true,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -0,0 +1,57 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "128000": {
+      "content": "<mask>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "errors": "replace",
+  "mask_token": "<mask>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "tokenizer_class": "RobertaTokenizer",
+  "trim_offsets": true,
+  "unk_token": "<unk>"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2eb8379a2ddcc06c081f2ee53c529e9bab3509adbf924e8246673d4ac50f3040
+size 4027