Fill-Mask
Transformers
Safetensors
pakmosaic
pakistan
multilingual
encoder
research-preview
sota-target
custom_code
Instructions to use ProximaAI/PakMosaic-Small with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use ProximaAI/PakMosaic-Small with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("fill-mask", model="ProximaAI/PakMosaic-Small", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForMaskedLM model = AutoModelForMaskedLM.from_pretrained("ProximaAI/PakMosaic-Small", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Add Transformers AutoModelForMaskedLM remote code
Browse files- config.json +33 -75
config.json
CHANGED
|
@@ -1,75 +1,33 @@
|
|
| 1 |
-
{
|
| 2 |
-
"
|
| 3 |
-
"
|
| 4 |
-
"
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
"
|
| 10 |
-
"
|
| 11 |
-
"
|
| 12 |
-
"
|
| 13 |
-
"
|
| 14 |
-
"
|
| 15 |
-
"
|
| 16 |
-
"
|
| 17 |
-
"
|
| 18 |
-
"
|
| 19 |
-
"
|
| 20 |
-
"
|
| 21 |
-
"
|
| 22 |
-
"
|
| 23 |
-
"
|
| 24 |
-
"
|
| 25 |
-
"
|
| 26 |
-
"
|
| 27 |
-
"
|
| 28 |
-
"
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
"languages": [
|
| 35 |
-
"pus_Arab",
|
| 36 |
-
"urd_Arab",
|
| 37 |
-
"pnb_Arab",
|
| 38 |
-
"skr_Arab",
|
| 39 |
-
"snd_Arab",
|
| 40 |
-
"kas_Arab",
|
| 41 |
-
"guj_Gujr"
|
| 42 |
-
]
|
| 43 |
-
},
|
| 44 |
-
"masking": {
|
| 45 |
-
"kind": "fixed",
|
| 46 |
-
"mask_prob": 0.15,
|
| 47 |
-
"scheduled_peak": 0.3,
|
| 48 |
-
"replace_prob": 0.8,
|
| 49 |
-
"random_prob": 0.1,
|
| 50 |
-
"keep_prob": 0.1
|
| 51 |
-
},
|
| 52 |
-
"train": {
|
| 53 |
-
"max_seq_len": 512,
|
| 54 |
-
"smoke_seq_len": 128,
|
| 55 |
-
"token_budget": 2000000000,
|
| 56 |
-
"micro_batch_size": 2,
|
| 57 |
-
"grad_accum": 8,
|
| 58 |
-
"learning_rate": 0.0003,
|
| 59 |
-
"weight_decay": 0.01,
|
| 60 |
-
"warmup_steps": 50,
|
| 61 |
-
"max_grad_norm": 1.0,
|
| 62 |
-
"seed": 42,
|
| 63 |
-
"log_every": 10,
|
| 64 |
-
"checkpoint_every": 20000,
|
| 65 |
-
"keep_last_checkpoints": 2,
|
| 66 |
-
"require_expensive_confirm": true
|
| 67 |
-
},
|
| 68 |
-
"algebraic_parameters": {
|
| 69 |
-
"embeddings": 16384000,
|
| 70 |
-
"layers": 50343936,
|
| 71 |
-
"mlm_head": 263168,
|
| 72 |
-
"total": 66991616,
|
| 73 |
-
"trainable": 66991616
|
| 74 |
-
}
|
| 75 |
-
}
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": ["PakMosaicForMaskedLM"],
|
| 3 |
+
"model_type": "pakmosaic",
|
| 4 |
+
"auto_map": {
|
| 5 |
+
"AutoConfig": "configuration_pakmosaic.PakMosaicConfig",
|
| 6 |
+
"AutoModel": "modeling_pakmosaic.PakMosaicModel",
|
| 7 |
+
"AutoModelForMaskedLM": "modeling_pakmosaic.PakMosaicForMaskedLM"
|
| 8 |
+
},
|
| 9 |
+
"vocab_size": 32000,
|
| 10 |
+
"hidden_size": 512,
|
| 11 |
+
"num_hidden_layers": 12,
|
| 12 |
+
"num_attention_heads": 8,
|
| 13 |
+
"intermediate_size": 2048,
|
| 14 |
+
"max_position_embeddings": 512,
|
| 15 |
+
"rope_theta": 10000.0,
|
| 16 |
+
"attention_dropout": 0.0,
|
| 17 |
+
"hidden_dropout": 0.0,
|
| 18 |
+
"ffn_activation": "geglu",
|
| 19 |
+
"attention_kind": "full",
|
| 20 |
+
"local_window": 128,
|
| 21 |
+
"global_every": 3,
|
| 22 |
+
"use_bias": false,
|
| 23 |
+
"tie_word_embeddings": true,
|
| 24 |
+
"pad_token_id": 0,
|
| 25 |
+
"bos_token_id": 1,
|
| 26 |
+
"eos_token_id": 2,
|
| 27 |
+
"unk_token_id": 3,
|
| 28 |
+
"mask_token_id": 4,
|
| 29 |
+
"layer_norm_eps": 1e-05,
|
| 30 |
+
"initializer_range": 0.02,
|
| 31 |
+
"script_signal": "none",
|
| 32 |
+
"tokenizer_class": "PreTrainedTokenizerFast"
|
| 33 |
+
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|