Add tiny-random PaliGemma LoRA fixture requiring key_mapping

README.md

@@ -0,0 +1,33 @@
+---
+library_name: transformers
+tags:
+- test-fixture
+- paligemma
+- colpali
+- peft
+- lora
+---
+
+# tomaarsen/tiny-random-paligemma-lora-key-mapping
+
+Tiny-random PaliGemma checkpoint bundling a LoRA adapter that **requires a `key_mapping` to load onto the
+underlying `PaliGemmaModel`**.
+
+It mirrors [`vidore/colpali`](https://huggingface.co/vidore/colpali) at tiny scale: the adapter's text weights
+are stored under the old `language_model.model.layers.*` layout, so loading them onto today's
+`PaliGemmaModel` (`language_model.layers.*`) needs:
+
+```python
+from transformers import PaliGemmaModel
+
+model = PaliGemmaModel.from_pretrained(
+    "tomaarsen/tiny-random-paligemma-lora-key-mapping",
+    key_mapping={r"language_model\.model\.": "language_model."},
+)
+```
+
+`PaliGemmaForConditionalGeneration` auto-bridges this (via the `llava` conversion) and does not need the
+mapping; the bare `PaliGemmaModel` does. Every `lora_A` weight is filled with `0.0234` and every
+`lora_B` weight with `0.0567`, so a test can assert the adapter was restored from the checkpoint.
+
+Generated by `scripts/create_tiny_paligemma_lora_key_mapping_fixture.py`.

adapter_config.json

@@ -0,0 +1,43 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": {
+    "base_model_class": "PaliGemmaForConditionalGeneration",
+    "parent_library": "transformers.models.paligemma.modeling_paligemma"
+  },
+  "base_model_name_or_path": "tomaarsen/tiny-random-paligemma-lora-key-mapping",
+  "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": false,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 8,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.2.dev0@76c37d4686cf7245a9a6998fd1e2619d62c9322d",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": ".*language_model.*\\.(q_proj|v_proj)$",
+  "target_parameters": null,
+  "task_type": null,
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fce525bfd50c235ff1c11b89ede25b0c1d047e7b9d637297b16492fa0b794e43
+size 7848

config.json

@@ -0,0 +1,57 @@
+{
+  "architectures": [
+    "PaliGemmaForConditionalGeneration"
+  ],
+  "dtype": "float32",
+  "hidden_size": 2048,
+  "image_token_index": 0,
+  "model_type": "paligemma",
+  "projection_dim": 32,
+  "text_config": {
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "bos_token_id": 2,
+    "eos_token_id": 1,
+    "head_dim": 8,
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_activation": "gelu_pytorch_tanh",
+    "hidden_size": 32,
+    "initializer_range": 0.02,
+    "intermediate_size": 37,
+    "max_position_embeddings": 512,
+    "model_type": "gemma",
+    "num_attention_heads": 4,
+    "num_hidden_layers": 2,
+    "num_image_tokens": 16,
+    "num_key_value_heads": 1,
+    "pad_token_id": 1,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "rope_theta": 10000.0,
+      "rope_type": "default"
+    },
+    "tie_word_embeddings": true,
+    "use_bidirectional_attention": true,
+    "use_cache": true,
+    "vocab_size": 99
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.13.0.dev0",
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 32,
+    "image_size": 20,
+    "intermediate_size": 37,
+    "layer_norm_eps": 1e-06,
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 4,
+    "num_channels": 3,
+    "num_hidden_layers": 2,
+    "num_image_tokens": 4,
+    "num_key_value_heads": 1,
+    "patch_size": 5,
+    "projection_dim": 32
+  },
+  "vocab_size": 257152
+}

generation_config.json

@@ -0,0 +1,10 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 2,
+  "eos_token_id": 1,
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "pad_token_id": 1,
+  "transformers_version": "5.13.0.dev0",
+  "use_cache": true
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e323085a13881dd707a7d21ebc1d6ff968970c743dcea0158fec6ff292b3b36
+size 167508