keenanpepper commited on
Commit
db43988
Β·
verified Β·
1 Parent(s): 387b52c

Upload folder using huggingface_hub

Browse files
README.md ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: llama
4
+ library_name: transformers
5
+ pipeline_tag: text-generation
6
+ tags:
7
+ - one-way-polyglot
8
+ - japanese
9
+ - english
10
+ - bilingual
11
+ - small-model
12
+ ---
13
+
14
+ # one-way-polyglot-8m-tied
15
+
16
+ A one-way polyglot language model trained to understand Japanese but generate only English.
17
+
18
+ ## Model Details
19
+
20
+ - **Architecture**: LLaMA-based transformer
21
+ - **Parameters**: 8,519,936 (8.5M)
22
+ - **Vocabulary**: 16,384 tokens (bilingual SentencePiece)
23
+ - **Context Length**: 512 tokens
24
+ - **Embedding Strategy**: Tied
25
+
26
+ ## Capabilities
27
+
28
+ - **Semantic Transfer**: Understands Japanese input and generates contextually appropriate English
29
+ - **One-Way Constraint**: Strong bias toward English-only generation
30
+ - **Name Transliteration**: Can transliterate Japanese names to English (context-dependent)
31
+
32
+ ## Training Data
33
+
34
+ Trained on bilingual Japanese-English story data with masked loss on Japanese prefixes to enforce one-way generation.
35
+
36
+ ## Usage
37
+
38
+ ```python
39
+ from transformers import LlamaForCausalLM, LlamaTokenizer
40
+
41
+ model = LlamaForCausalLM.from_pretrained("one-way-polyglot-8m-tied")
42
+ tokenizer = LlamaTokenizer.from_pretrained("one-way-polyglot-8m-tied")
43
+
44
+ # Japanese input, English output
45
+ prompt = "ζ˜”γ€…γ€θ΅€γ„ε‚˜γ‚’ζŒγ£γŸε°‘ε₯³γŒγ„γΎγ—γŸγ€‚"
46
+ inputs = tokenizer(prompt, return_tensors="pt")
47
+ outputs = model.generate(**inputs, max_new_tokens=50, temperature=0.7)
48
+ print(tokenizer.decode(outputs[0], skip_special_tokens=True))
49
+ ```
50
+
51
+ ## Model Variants
52
+
53
+ This is part of a series exploring one-way polyglot capabilities:
54
+ - 1.25M parameters (tied embeddings)
55
+ - 8.5M parameters (tied embeddings)
56
+ - 12.7M parameters (untied embeddings)
57
+ - 15.7M parameters (tied embeddings)
58
+
59
+ ## License
60
+
61
+ Apache 2.0
config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "LlamaForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 1,
8
+ "eos_token_id": 2,
9
+ "head_dim": 64,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 256,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 682,
14
+ "max_position_embeddings": 512,
15
+ "mlp_bias": false,
16
+ "model_type": "llama",
17
+ "num_attention_heads": 4,
18
+ "num_hidden_layers": 6,
19
+ "num_key_value_heads": 2,
20
+ "pretraining_tp": 1,
21
+ "rms_norm_eps": 1e-06,
22
+ "rope_scaling": null,
23
+ "rope_theta": 10000.0,
24
+ "tie_word_embeddings": true,
25
+ "torch_dtype": "float32",
26
+ "transformers_version": "4.54.1",
27
+ "use_cache": true,
28
+ "vocab_size": 16384
29
+ }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "transformers_version": "4.54.1"
6
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1804e76de82b0380509a595273b62af4117bf200d42183ba15d433243bfbac91
3
+ size 34085832
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4359c7d373bfaa6bc9ec396e10c9cde2d0e7604a0821075d8529b3561b7c2778
3
+ size 34102703
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18fd3bcd564240df1343dda502457ea5db419f4061ca442a9b204796d3d79301
3
+ size 581374
tokenizer.vocab ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "tokenizer_class": "LlamaTokenizer",
3
+ "vocab_size": 16384,
4
+ "model_max_length": 512,
5
+ "pad_token": "[PAD]",
6
+ "bos_token": "[BOS]",
7
+ "eos_token": "[EOS]",
8
+ "unk_token": "[UNK]",
9
+ "add_bos_token": true,
10
+ "add_eos_token": false,
11
+ "clean_up_tokenization_spaces": false
12
+ }