initial_queue.jsonl

// 1- base models <=7B
{"model": "TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "meta-llama/Llama-2-7b-hf", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "mistralai/Mistral-7B-v0.1", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "huggyllama/llama-7b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "openlm-research/open_llama_3b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "openlm-research/open_llama_3b_v2", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "openlm-research/open_llama_7b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "openlm-research/open_llama_7b_v2", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
// 2 - portuguese models
{"model": "maritaca-ai/sabia-7b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "dominguesm/canarim-7b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "22h/open-cabrita3b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "recogna-nlp/bode-7b-alpaca-pt-br", "base_model": "meta-llama/Llama-2-7b-chat-hf", "revision": "main", "precision": "float16", "weight_type": "Adapter", "model_type": "💬 : chat models (RLHF, DPO, IFT, ...)"}
{"model": "recogna-nlp/bode-13b-alpaca-pt-br", "base_model": "meta-llama/Llama-2-13b-chat-hf", "revision": "main", "precision": "float16", "weight_type": "Adapter", "model_type": "💬 : chat models (RLHF, DPO, IFT, ...)"}
{"model": "22h/cabrita_7b_pt_850000", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "22h/cabrita-lora-v0-1", "base_model": "huggyllama/llama-7b", "revision": "main", "precision": "float16", "weight_type": "Adapter", "model_type": "🔶 : fine-tuned"}
{"model": "pierreguillou/gpt2-small-portuguese", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "pucpr/gpt2-bio-pt", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "unicamp-dl/ptt5-small-portuguese-vocab", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "unicamp-dl/ptt5-base-portuguese-vocab", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "unicamp-dl/ptt5-large-portuguese-vocab", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "unicamp-dl/ptt5-small-t5-vocab", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "unicamp-dl/ptt5-base-t5-vocab", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "unicamp-dl/ptt5-large-t5-vocab", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "josu/gpt-neo-pt-br", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "josu/gpt-neo-pt-1.3B", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "monilouise/opt125M_portuguese", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "HeyLucasLeao/gpt-neo-small-portuguese", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
// 3 - Larger base models <= 13B
{"model": "meta-llama/Llama-2-13b-hf", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "huggyllama/llama-13b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "openlm-research/open_llama_13b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "upstage/SOLAR-10.7B-v1.0", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
// other must-have <=7B
{"model": "01-ai/Yi-6B", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "Unbabel/TowerBase-7B-v0.1", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🆎 : language adapted models (FP, FT, ...)"}
{"model": "tiiuae/falcon-7b", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "bigscience/bloom-560m", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "bigscience/bloom-1b7", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "bigscience/bloom-3b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "bigscience/bloom-7b1", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "gpt2", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "stabilityai/stablelm-2-1_6b", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "stabilityai/stablelm-3b-4e1t", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "t5-small", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "t5-base", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "t5-large", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/mt5-small", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/mt5-base", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/mt5-large", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
// Larger base models >13B
{"model": "mistralai/Mixtral-8x7B-v0.1", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "huggyllama/llama-30b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "01-ai/Yi-34B", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "meta-llama/Llama-2-70b-hf", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "huggyllama/llama-65b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
// others
{"model": "internlm/internlm2-7b", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🔶 : fine-tuned"}
{"model": "internlm/internlm2-base-7b", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "internlm/internlm-7b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "internlm/internlm2-20b", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🔶 : fine-tuned"}
{"model": "internlm/internlm2-base-20b", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "internlm/internlm-20b", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "Qwen/Qwen-1_8B", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "Qwen/Qwen-7B", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "Qwen/Qwen-14B", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "Qwen/Qwen-72B", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "DAMO-NLP-MT/polylm-1.7b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "DAMO-NLP-MT/polylm-13b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "Deci/DeciLM-6b", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🔶 : fine-tuned"}
{"model": "Deci/DeciLM-7B", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🔶 : fine-tuned"}
{"model": "EleutherAI/pythia-14m", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-70m-deduped", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-160m-deduped", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-410m-deduped", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-1b-deduped", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-2.8b-deduped", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-6.9b-deduped", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-12b-deduped", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "facebook/opt-125m", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "facebook/opt-350m", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "facebook/opt-1.3b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "facebook/opt-2.7b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "facebook/opt-6.7b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "facebook/opt-13b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "facebook/opt-30b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "facebook/opt-66b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/gpt-neo-125m", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/gpt-neo-1.3B", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/gpt-neo-2.7B", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/gpt-j-6b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/gpt-neox-20b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "stabilityai/stablelm-base-alpha-3b-v2", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "stabilityai/stablelm-base-alpha-7b-v2", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "stabilityai/stablelm-base-alpha-3b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "stabilityai/stablelm-base-alpha-7b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "tiiuae/falcon-40b", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "openai-community/openai-gpt", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "openai-community/gpt2-medium", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "openai-community/gpt2-large", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "openai-community/gpt2-xl", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "microsoft/phi-1", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "microsoft/phi-1_5", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "microsoft/phi-2", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "mosaicml/mpt-7b", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "01-ai/Yi-6B-200K", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "01-ai/Yi-34B-200K", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/t5-v1_1-base", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/t5-v1_1-small", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/t5-v1_1-large", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/t5-v1_1-xl", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/t5-v1_1-xxl", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/mt5-xl", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/mt5-xxl", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/umt5-small", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/umt5-base", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/umt5-xl", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "google/umt5-xxl", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "AdaptLLM/law-LLM", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🔶 : fine-tuned"}
{"model": "AdaptLLM/medicine-LLM", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🔶 : fine-tuned"}
{"model": "AdaptLLM/finance-LLM", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🔶 : fine-tuned"}
{"model": "AdaptLLM/law-LLM-13B", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🔶 : fine-tuned"}
{"model": "AdaptLLM/medicine-LLM-13B", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🔶 : fine-tuned"}
{"model": "AdaptLLM/finance-LLM-13B", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🔶 : fine-tuned"}
{"model": "cerebras/Cerebras-GPT-111M", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "cerebras/Cerebras-GPT-256M", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "cerebras/Cerebras-GPT-590M", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "cerebras/Cerebras-GPT-1.3B", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "cerebras/Cerebras-GPT-2.7B", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "cerebras/Cerebras-GPT-6.7B", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "cerebras/Cerebras-GPT-13B", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "cerebras/btlm-3b-8k-base", "base_model": "", "revision": "main", "precision": "bfloat16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "ai-forever/mGPT-13B", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "ai-forever/mGPT", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-70m", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-160m", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-410m", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-1b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-2.8b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-6.9b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}
{"model": "EleutherAI/pythia-12b", "base_model": "", "revision": "main", "precision": "float16", "weight_type": "Original", "model_type": "🟢 : pretrained"}