Upload folder using huggingface_hub

README.md

@@ -4,6 +4,7 @@ license: apache-2.0
 
 ```python
 import os
+import torch
 from transformers import (
     AutoConfig,
     AutoModelForImageTextToText,
@@ -12,28 +13,39 @@ from transformers import (
 )
 
 model_id = "mistralai/Mistral-Small-3.1-24B-Instruct-2503"
+torch.manual_seed(42)
 config = AutoConfig.from_pretrained(model_id)
 
+config.tie_word_embeddings = False
+config.text_config.tie_word_embeddings = False
+
 config.text_config.num_hidden_layers = 2
-config.text_config.hidden_size = 8
-config.text_config.intermediate_size = 64
-config.text_config.num_attention_heads = 8
-config.text_config.num_key_value_heads = 4
-config.text_config.head_dim = 32
+config.text_config.hidden_size = 64
+config.text_config.intermediate_size = 128
+config.text_config.num_attention_heads = 4
+config.text_config.num_key_value_heads = 2
+config.text_config.head_dim = 16
+config.text_config.max_position_embeddings = 512
 
 config.vision_config.num_hidden_layers = 2
-config.vision_config.hidden_size = 128
-config.vision_config.intermediate_size = 64
+config.vision_config.hidden_size = 64
+config.vision_config.intermediate_size = 128
 config.vision_config.num_attention_heads = 4
-config.vision_config.head_dim = 32
+config.vision_config.head_dim = 16
+config.vision_config.image_size = 56
+
+for subconfig in (config, config.text_config, config.vision_config):
+    subconfig.dtype = "float32"
+    subconfig.torch_dtype = "float32"
 
 model = AutoModelForImageTextToText.from_config(config)
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 processor = AutoProcessor.from_pretrained(model_id)
+processor.image_processor.size = {"longest_edge": 56}
 
 output_dir = "./tiny-random-mistral3"
 os.makedirs(output_dir, exist_ok=True)
-model.save_pretrained(output_dir)
+model.save_pretrained(output_dir, safe_serialization=True)
 tokenizer.save_pretrained(output_dir)
 processor.save_pretrained(output_dir)
-```
+```

config.json

@@ -2,7 +2,7 @@
   "architectures": [
     "Mistral3ForConditionalGeneration"
   ],
-  "dtype": "bfloat16",
+  "dtype": "float32",
   "image_token_index": 10,
   "model_type": "mistral3",
   "multimodal_projector_bias": false,
@@ -10,37 +10,50 @@
   "spatial_merge_size": 2,
   "text_config": {
     "attention_dropout": 0.0,
-    "head_dim": 32,
+    "bos_token_id": 1,
+    "dtype": "float32",
+    "eos_token_id": 2,
+    "head_dim": 16,
     "hidden_act": "silu",
-    "hidden_size": 8,
+    "hidden_size": 64,
     "initializer_range": 0.02,
-    "intermediate_size": 64,
-    "max_position_embeddings": 131072,
+    "intermediate_size": 128,
+    "max_position_embeddings": 512,
     "model_type": "mistral",
-    "num_attention_heads": 8,
+    "num_attention_heads": 4,
     "num_hidden_layers": 2,
-    "num_key_value_heads": 4,
+    "num_key_value_heads": 2,
+    "pad_token_id": null,
     "rms_norm_eps": 1e-05,
-    "rope_theta": 1000000000.0,
+    "rope_parameters": {
+      "rope_theta": 1000000000.0,
+      "rope_type": "default"
+    },
     "sliding_window": null,
+    "tie_word_embeddings": false,
     "use_cache": true,
     "vocab_size": 131072
   },
-  "transformers_version": "4.57.0",
+  "tie_word_embeddings": false,
+  "transformers_version": "5.17.0",
   "vision_config": {
     "attention_dropout": 0.0,
-    "head_dim": 32,
+    "dtype": "float32",
+    "head_dim": 16,
     "hidden_act": "silu",
-    "hidden_size": 128,
-    "image_size": 1540,
+    "hidden_size": 64,
+    "image_size": 56,
     "initializer_range": 0.02,
-    "intermediate_size": 64,
+    "intermediate_size": 128,
     "model_type": "pixtral",
     "num_attention_heads": 4,
     "num_channels": 3,
     "num_hidden_layers": 2,
     "patch_size": 14,
-    "rope_theta": 10000.0
+    "rope_parameters": {
+      "rope_theta": 10000.0,
+      "rope_type": "axial"
+    }
   },
   "vision_feature_layer": -1
 }

generation_config.json

@@ -2,5 +2,8 @@
   "_from_model_config": true,
   "bos_token_id": 1,
   "eos_token_id": 2,
-  "transformers_version": "4.57.0"
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "transformers_version": "5.17.0",
+  "use_cache": true
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:20dee5337f75f487dc2ee099b8b2dd43acc806e331fe8dd46340a8056b40d022
-size 4876304
+oid sha256:9c681b8749bf9ba8ef5123b4fbcbf9997bc250d272f0be81ae55feb06b278f15
+size 67988584

processor_config.json

@@ -1,6 +1,31 @@
 {
   "image_break_token": "[IMG_BREAK]",
   "image_end_token": "[IMG_END]",
+  "image_processor": {
+    "data_format": "channels_first",
+    "default_to_square": true,
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "image_mean": [
+      0.48145466,
+      0.4578275,
+      0.40821073
+    ],
+    "image_processor_type": "PixtralImageProcessor",
+    "image_std": [
+      0.26862954,
+      0.26130258,
+      0.27577711
+    ],
+    "patch_size": 14,
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "size": {
+      "longest_edge": 56
+    }
+  },
   "image_token": "[IMG]",
   "patch_size": 14,
   "processor_class": "PixtralProcessor",