baltachev commited on
Commit
9454a6f
1 Parent(s): 84c9828

Make the hf model compatible with transformers older versions

Browse files

Hello!
Yesterday we merged the model compatible with hf CLIP class. It works fine on the latest transformers version.

However, I found out that it can't be loaded on the older versions of transformers (specifically, 4.20.1), so I converted the open_clip model on transformers==4.20.1 and it works like a charm on version >= 4.19.

Decided not to check on older versions because there are quite outdated.

Files changed (2) hide show
  1. config.json +18 -12
  2. pytorch_model.bin +2 -2
config.json CHANGED
@@ -1,5 +1,4 @@
1
  {
2
- "_commit_hash": null,
3
  "architectures": [
4
  "CLIPModel"
5
  ],
@@ -13,13 +12,13 @@
13
  "architectures": null,
14
  "attention_dropout": 0.0,
15
  "bad_words_ids": null,
16
- "begin_suppress_tokens": null,
17
  "bos_token_id": 0,
18
  "chunk_size_feed_forward": 0,
19
  "cross_attention_hidden_size": null,
20
  "decoder_start_token_id": null,
21
  "diversity_penalty": 0.0,
22
  "do_sample": false,
 
23
  "early_stopping": false,
24
  "encoder_no_repeat_ngram_size": 0,
25
  "eos_token_id": 2,
@@ -60,17 +59,14 @@
60
  "pad_token_id": 1,
61
  "prefix": null,
62
  "problem_type": null,
63
- "projection_dim": 512,
64
  "pruned_heads": {},
65
  "remove_invalid_values": false,
66
  "repetition_penalty": 1.0,
67
  "return_dict": true,
68
  "return_dict_in_generate": false,
69
  "sep_token_id": null,
70
- "suppress_tokens": null,
71
  "task_specific_params": null,
72
  "temperature": 1.0,
73
- "tf_legacy_loss": false,
74
  "tie_encoder_decoder": false,
75
  "tie_word_embeddings": true,
76
  "tokenizer_class": null,
@@ -78,11 +74,17 @@
78
  "top_p": 1.0,
79
  "torch_dtype": null,
80
  "torchscript": false,
81
- "transformers_version": "4.29.1",
82
  "typical_p": 1.0,
83
  "use_bfloat16": false,
84
  "vocab_size": 49408
85
  },
 
 
 
 
 
 
86
  "torch_dtype": "float32",
87
  "transformers_version": null,
88
  "vision_config": {
@@ -91,13 +93,13 @@
91
  "architectures": null,
92
  "attention_dropout": 0.0,
93
  "bad_words_ids": null,
94
- "begin_suppress_tokens": null,
95
  "bos_token_id": null,
96
  "chunk_size_feed_forward": 0,
97
  "cross_attention_hidden_size": null,
98
  "decoder_start_token_id": null,
99
  "diversity_penalty": 0.0,
100
  "do_sample": false,
 
101
  "early_stopping": false,
102
  "encoder_no_repeat_ngram_size": 0,
103
  "eos_token_id": null,
@@ -130,7 +132,6 @@
130
  "num_attention_heads": 16,
131
  "num_beam_groups": 1,
132
  "num_beams": 1,
133
- "num_channels": 3,
134
  "num_hidden_layers": 24,
135
  "num_return_sequences": 1,
136
  "output_attentions": false,
@@ -140,17 +141,14 @@
140
  "patch_size": 14,
141
  "prefix": null,
142
  "problem_type": null,
143
- "projection_dim": 512,
144
  "pruned_heads": {},
145
  "remove_invalid_values": false,
146
  "repetition_penalty": 1.0,
147
  "return_dict": true,
148
  "return_dict_in_generate": false,
149
  "sep_token_id": null,
150
- "suppress_tokens": null,
151
  "task_specific_params": null,
152
  "temperature": 1.0,
153
- "tf_legacy_loss": false,
154
  "tie_encoder_decoder": false,
155
  "tie_word_embeddings": true,
156
  "tokenizer_class": null,
@@ -158,8 +156,16 @@
158
  "top_p": 1.0,
159
  "torch_dtype": null,
160
  "torchscript": false,
161
- "transformers_version": "4.29.1",
162
  "typical_p": 1.0,
163
  "use_bfloat16": false
 
 
 
 
 
 
 
 
164
  }
165
  }
 
1
  {
 
2
  "architectures": [
3
  "CLIPModel"
4
  ],
 
12
  "architectures": null,
13
  "attention_dropout": 0.0,
14
  "bad_words_ids": null,
 
15
  "bos_token_id": 0,
16
  "chunk_size_feed_forward": 0,
17
  "cross_attention_hidden_size": null,
18
  "decoder_start_token_id": null,
19
  "diversity_penalty": 0.0,
20
  "do_sample": false,
21
+ "dropout": 0.0,
22
  "early_stopping": false,
23
  "encoder_no_repeat_ngram_size": 0,
24
  "eos_token_id": 2,
 
59
  "pad_token_id": 1,
60
  "prefix": null,
61
  "problem_type": null,
 
62
  "pruned_heads": {},
63
  "remove_invalid_values": false,
64
  "repetition_penalty": 1.0,
65
  "return_dict": true,
66
  "return_dict_in_generate": false,
67
  "sep_token_id": null,
 
68
  "task_specific_params": null,
69
  "temperature": 1.0,
 
70
  "tie_encoder_decoder": false,
71
  "tie_word_embeddings": true,
72
  "tokenizer_class": null,
 
74
  "top_p": 1.0,
75
  "torch_dtype": null,
76
  "torchscript": false,
77
+ "transformers_version": "4.20.1",
78
  "typical_p": 1.0,
79
  "use_bfloat16": false,
80
  "vocab_size": 49408
81
  },
82
+ "text_config_dict": {
83
+ "hidden_act": "gelu",
84
+ "hidden_size": 768,
85
+ "intermediate_size": 3072,
86
+ "num_attention_heads": 12
87
+ },
88
  "torch_dtype": "float32",
89
  "transformers_version": null,
90
  "vision_config": {
 
93
  "architectures": null,
94
  "attention_dropout": 0.0,
95
  "bad_words_ids": null,
 
96
  "bos_token_id": null,
97
  "chunk_size_feed_forward": 0,
98
  "cross_attention_hidden_size": null,
99
  "decoder_start_token_id": null,
100
  "diversity_penalty": 0.0,
101
  "do_sample": false,
102
+ "dropout": 0.0,
103
  "early_stopping": false,
104
  "encoder_no_repeat_ngram_size": 0,
105
  "eos_token_id": null,
 
132
  "num_attention_heads": 16,
133
  "num_beam_groups": 1,
134
  "num_beams": 1,
 
135
  "num_hidden_layers": 24,
136
  "num_return_sequences": 1,
137
  "output_attentions": false,
 
141
  "patch_size": 14,
142
  "prefix": null,
143
  "problem_type": null,
 
144
  "pruned_heads": {},
145
  "remove_invalid_values": false,
146
  "repetition_penalty": 1.0,
147
  "return_dict": true,
148
  "return_dict_in_generate": false,
149
  "sep_token_id": null,
 
150
  "task_specific_params": null,
151
  "temperature": 1.0,
 
152
  "tie_encoder_decoder": false,
153
  "tie_word_embeddings": true,
154
  "tokenizer_class": null,
 
156
  "top_p": 1.0,
157
  "torch_dtype": null,
158
  "torchscript": false,
159
+ "transformers_version": "4.20.1",
160
  "typical_p": 1.0,
161
  "use_bfloat16": false
162
+ },
163
+ "vision_config_dict": {
164
+ "hidden_act": "gelu",
165
+ "hidden_size": 1024,
166
+ "intermediate_size": 4096,
167
+ "num_attention_heads": 16,
168
+ "num_hidden_layers": 24,
169
+ "patch_size": 14
170
  }
171
  }
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:cbae707ef72c3d2748f3c9482e9ca8645d8bbd915d59a86e4f11107c9b341068
3
- size 1710660257
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff37869878b61d3a1f2a82dbbd093927ea658f7477e1a347c0ba384591238904
3
+ size 1710664269