| { | |
| "geitje-7b": { | |
| "compute_dtype": "auto", | |
| "dutch_coverage": "fine-tuned", | |
| "model_name": "Rijgersberg/GEITje-7B", | |
| "model_type": "fine-tuned", | |
| "num_parameters": 7241732096, | |
| "quantization": "8-bit" | |
| }, | |
| "geitje-7b-chat": { | |
| "compute_dtype": "auto", | |
| "dutch_coverage": "fine-tuned", | |
| "model_name": "Rijgersberg/GEITje-7B-chat", | |
| "model_type": "instruction-tuned", | |
| "num_parameters": 7241732096, | |
| "quantization": "8-bit" | |
| }, | |
| "gpt-neo-1.3b-dutch": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "pretrained", | |
| "model_name": "yhavinga/gpt-neo-1.3B-dutch", | |
| "model_type": "pretrained", | |
| "num_parameters": 1315575808, | |
| "quantization": "8-bit" | |
| }, | |
| "gpt-neo-125m-dutch": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "pretrained", | |
| "model_name": "yhavinga/gpt-neo-125M-dutch", | |
| "model_type": "pretrained", | |
| "num_parameters": 125198592, | |
| "quantization": "8-bit" | |
| }, | |
| "gpt2-large-dutch": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "pretrained", | |
| "model_name": "yhavinga/gpt2-large-dutch", | |
| "model_type": "pretrained", | |
| "num_parameters": 774030080, | |
| "quantization": "8-bit" | |
| }, | |
| "gpt2-medium-dutch": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "pretrained", | |
| "model_name": "yhavinga/gpt2-medium-dutch", | |
| "model_type": "pretrained", | |
| "num_parameters": 354823168, | |
| "quantization": "8-bit" | |
| }, | |
| "llama-2-13b-chat-dutch": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "fine-tuned", | |
| "model_name": "BramVanroy/Llama-2-13b-chat-dutch", | |
| "model_type": "instruction-tuned", | |
| "num_parameters": 13015864320, | |
| "quantization": "8-bit" | |
| }, | |
| "llama-2-13b-chat-hf": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "none", | |
| "model_name": "meta-llama/Llama-2-13b-chat-hf", | |
| "model_type": "RL-tuned", | |
| "num_parameters": 13015864320, | |
| "quantization": "8-bit" | |
| }, | |
| "llama-2-13b-hf": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "none", | |
| "model_name": "meta-llama/Llama-2-13b-hf", | |
| "model_type": "pretrained", | |
| "num_parameters": 13015864320, | |
| "quantization": "8-bit" | |
| }, | |
| "llama-2-7b-chat-hf": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "none", | |
| "model_name": "meta-llama/Llama-2-7b-chat-hf", | |
| "model_type": "RL-tuned", | |
| "num_parameters": 6738415616, | |
| "quantization": "8-bit" | |
| }, | |
| "llama-2-7b-hf": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "none", | |
| "model_name": "meta-llama/Llama-2-7b-hf", | |
| "model_type": "pretrained", | |
| "num_parameters": 6738415616, | |
| "quantization": "8-bit" | |
| }, | |
| "llama2-13b-ft-mc4_nl_cleaned_tiny": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "fine-tuned", | |
| "model_name": "BramVanroy/llama2-13b-ft-mc4_nl_cleaned_tiny", | |
| "model_type": "fine-tuned", | |
| "num_parameters": 13015864320, | |
| "quantization": "8-bit" | |
| }, | |
| "mistral-7b-v0.1": { | |
| "compute_dtype": "auto", | |
| "dutch_coverage": "none", | |
| "model_name": "mistralai/Mistral-7B-v0.1", | |
| "model_type": "pretrained", | |
| "num_parameters": 7241732096, | |
| "quantization": "8-bit" | |
| }, | |
| "mixtral-8x7b-v0.1": { | |
| "compute_dtype": "auto", | |
| "dutch_coverage": "none", | |
| "model_name": "mistralai/Mixtral-8x7B-v0.1", | |
| "model_type": "pretrained", | |
| "num_parameters": 46702792704, | |
| "quantization": "8-bit" | |
| }, | |
| "neural-chat-7b-v3-1": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "none", | |
| "model_name": "Intel/neural-chat-7b-v3-1", | |
| "model_type": "RL-tuned", | |
| "num_parameters": 7241732096, | |
| "quantization": "8-bit" | |
| }, | |
| "orca-2-13b": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "none", | |
| "model_name": "microsoft/Orca-2-13b", | |
| "model_type": "instruction-tuned", | |
| "num_parameters": 13015895040, | |
| "quantization": "8-bit" | |
| }, | |
| "orca-2-7b": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "none", | |
| "model_name": "microsoft/Orca-2-7b", | |
| "model_type": "instruction-tuned", | |
| "num_parameters": 6738440192, | |
| "quantization": "8-bit" | |
| }, | |
| "phi-2": { | |
| "compute_dtype": "auto", | |
| "dutch_coverage": "none", | |
| "model_name": "microsoft/phi-2", | |
| "model_type": "instruction-tuned", | |
| "num_parameters": 2779683840, | |
| "quantization": "8-bit" | |
| }, | |
| "towerbase-7b-v0.1": { | |
| "compute_dtype": "auto", | |
| "dutch_coverage": "fine-tuned", | |
| "model_name": "Unbabel/TowerBase-7B-v0.1", | |
| "model_type": "fine-tuned", | |
| "num_parameters": 6738415616, | |
| "quantization": "8-bit" | |
| }, | |
| "towerinstruct-7b-v0.1": { | |
| "compute_dtype": "auto", | |
| "dutch_coverage": "fine-tuned", | |
| "model_name": "Unbabel/TowerInstruct-7B-v0.1", | |
| "model_type": "instruction-tuned", | |
| "num_parameters": 6738472960, | |
| "quantization": "8-bit" | |
| }, | |
| "yi-6b": { | |
| "compute_dtype": "auto", | |
| "dutch_coverage": "none", | |
| "model_name": "01-ai/Yi-6B", | |
| "model_type": "pretrained", | |
| "num_parameters": 6061035520, | |
| "quantization": "8-bit" | |
| }, | |
| "yi-6b-chat": { | |
| "compute_dtype": "auto", | |
| "dutch_coverage": "none", | |
| "model_name": "01-ai/Yi-6B-Chat", | |
| "model_type": "instruction-tuned", | |
| "num_parameters": 6061035520, | |
| "quantization": "8-bit" | |
| }, | |
| "zephyr-7b-beta": { | |
| "compute_dtype": "bfloat16", | |
| "dutch_coverage": "none", | |
| "model_name": "HuggingFaceH4/zephyr-7b-beta", | |
| "model_type": "RL-tuned", | |
| "num_parameters": 7241732096, | |
| "quantization": "8-bit" | |
| } | |
| } |