Troviku-1.1 / config.json
Trouter-Library's picture
Create config.json
06a7092 verified
raw
history blame
4.43 kB
{
"model_name": "Troviku-1.1",
"model_type": "CausalLM",
"organization": "OpenTrouter",
"version": "1.1.0",
"release_date": "2025-01-15",
"architecture": {
"model_type": "transformer",
"hidden_size": 4096,
"num_hidden_layers": 32,
"num_attention_heads": 32,
"num_key_value_heads": 8,
"intermediate_size": 14336,
"hidden_act": "silu",
"max_position_embeddings": 8192,
"rope_theta": 10000.0,
"rope_scaling": null,
"attention_bias": false,
"attention_dropout": 0.0,
"mlp_bias": false
},
"tokenizer": {
"tokenizer_type": "BPE",
"vocab_size": 32768,
"bos_token": "<|begin_of_text|>",
"eos_token": "<|end_of_text|>",
"pad_token": "<|pad|>",
"unk_token": "<|unknown|>",
"special_tokens": [
"<|begin_of_text|>",
"<|end_of_text|>",
"<|pad|>",
"<|unknown|>",
"<|code_start|>",
"<|code_end|>",
"<|comment_start|>",
"<|comment_end|>",
"<|python|>",
"<|javascript|>",
"<|java|>",
"<|cpp|>",
"<|rust|>",
"<|go|>",
"<|typescript|>"
]
},
"training": {
"dataset_size_tokens": 500000000000,
"training_steps": 1000000,
"batch_size": 4096,
"learning_rate": 3e-4,
"warmup_steps": 2000,
"weight_decay": 0.1,
"gradient_clipping": 1.0,
"optimizer": "AdamW",
"mixed_precision": "bf16"
},
"inference": {
"default_temperature": 0.7,
"default_top_p": 0.95,
"default_top_k": 50,
"default_max_tokens": 2048,
"repetition_penalty": 1.1,
"frequency_penalty": 0.0,
"presence_penalty": 0.0
},
"capabilities": {
"languages": [
"python",
"javascript",
"typescript",
"java",
"c",
"cpp",
"csharp",
"rust",
"go",
"ruby",
"php",
"swift",
"kotlin",
"scala",
"r",
"sql",
"html",
"css",
"bash",
"powershell",
"lua",
"perl",
"haskell",
"julia",
"matlab"
],
"tasks": [
"code_generation",
"code_completion",
"code_explanation",
"code_review",
"bug_detection",
"test_generation",
"documentation_generation",
"code_translation",
"refactoring",
"algorithm_design"
],
"specializations": [
"algorithms",
"data_structures",
"web_development",
"api_design",
"database_queries",
"devops_scripting",
"data_science",
"machine_learning_code"
]
},
"benchmarks": {
"humaneval": {
"score": 0.72,
"date": "2025-01-10",
"notes": "Pass@1 score on HumanEval benchmark"
},
"mbpp": {
"score": 0.68,
"date": "2025-01-10",
"notes": "Pass@1 score on MBPP benchmark"
},
"code_contests": {
"score": 0.45,
"date": "2025-01-10",
"notes": "Problems solved at first attempt"
}
},
"safety": {
"content_filtering": true,
"malicious_code_detection": true,
"license_compliance": true,
"pii_redaction": true,
"toxicity_filtering": true
},
"deployment": {
"supported_frameworks": [
"transformers",
"vllm",
"tensorrt-llm",
"ctranslate2"
],
"quantization_support": [
"int8",
"int4",
"fp16",
"bf16"
],
"hardware_requirements": {
"minimum_vram_gb": 16,
"recommended_vram_gb": 24,
"minimum_ram_gb": 32,
"recommended_ram_gb": 64
}
},
"api": {
"endpoint": "https://api.opentrouter.ai/v1/chat/completions",
"model_identifier": "OpenTrouter/Troviku-1.1",
"rate_limits": {
"requests_per_minute": 60,
"tokens_per_minute": 90000
},
"pricing": {
"input_tokens_per_million": 0.50,
"output_tokens_per_million": 1.50,
"currency": "USD"
}
},
"license": {
"type": "OpenTrouter Model License",
"commercial_use": true,
"attribution_required": true,
"modifications_allowed": false,
"redistribution_allowed": false
},
"metadata": {
"authors": [
"OpenTrouter Research Team"
],
"contact": "[email protected]",
"repository": "https://github.com/OpenTrouter/Troviku-1.1",
"documentation": "https://docs.opentrouter.ai/troviku",
"paper": "https://arxiv.org/abs/placeholder",
"demo": "https://demo.opentrouter.ai/troviku"
}
}