Text Generation
Transformers
Safetensors
English
deepseek-v2
multi-head-latent-attention
mla
mixture-of-experts
Mixture of Experts
tinystories
tiny-model
validation
debug-model
Instructions to use shibatch/tinydeepseekv2-3m with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use shibatch/tinydeepseekv2-3m with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="shibatch/tinydeepseekv2-3m")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("shibatch/tinydeepseekv2-3m", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use shibatch/tinydeepseekv2-3m with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "shibatch/tinydeepseekv2-3m" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "shibatch/tinydeepseekv2-3m", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/shibatch/tinydeepseekv2-3m
- SGLang
How to use shibatch/tinydeepseekv2-3m with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "shibatch/tinydeepseekv2-3m" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "shibatch/tinydeepseekv2-3m", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "shibatch/tinydeepseekv2-3m" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "shibatch/tinydeepseekv2-3m", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use shibatch/tinydeepseekv2-3m with Docker Model Runner:
docker model run hf.co/shibatch/tinydeepseekv2-3m
File size: 5,544 Bytes
d093b93 | 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 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 | {
"model_type": "deepseek_v2_moe",
"hf_model_type": "deepseek_v2",
"parameter_count": 2928392,
"tokenizer_size": 1003,
"final_eval_loss": 1.4225675538182259,
"final_eval_ppl": 4.1477563679496425,
"final_train_lm_loss": 1.409641880750656,
"router_usage": {
"model.layers.1.mlp.gate": [
0.2502,
0.2487,
0.2522,
0.2488
],
"model.layers.2.mlp.gate": [
0.2538,
0.2481,
0.2487,
0.2495
],
"model.layers.3.mlp.gate": [
0.2512,
0.2454,
0.2519,
0.2514
],
"model.layers.4.mlp.gate": [
0.2465,
0.2535,
0.2542,
0.2458
]
},
"args": {
"train_files": [
"data/train-00000-of-00004-2d5a1467fff1081b.parquet",
"data/train-00001-of-00004-5852b56a2bd28fd9.parquet",
"data/train-00002-of-00004-a26307300439e943.parquet",
"data/train-00003-of-00004-d243063613e5a057.parquet"
],
"output_dir": "runs/deepseek_v2_3m_fix2_100p_1epoch",
"max_rows": 0,
"data_fraction": 1.0,
"vocab_size": 1024,
"base_vocab_size": 1000,
"min_frequency": 2,
"legacy_tokenizer": true,
"legacy_special_token_ids": true,
"deepseek_arch": "v2",
"hidden_size": 216,
"intermediate_size": 432,
"num_hidden_layers": 5,
"num_attention_heads": 8,
"num_key_value_heads": 8,
"qk_nope_head_dim": 16,
"qk_rope_head_dim": 16,
"v_head_dim": 32,
"q_lora_rank": -1,
"kv_lora_rank": 64,
"attention_bias": false,
"attention_dropout": 0.0,
"rms_norm_eps": 1e-06,
"initializer_range": 0.02,
"rope_theta": 10000,
"first_k_dense_replace": 1,
"n_routed_experts": 4,
"n_shared_experts": 1,
"num_experts_per_tok": 1,
"n_group": 0,
"topk_group": 0,
"topk_method": "greedy",
"norm_topk_prob": false,
"routed_scaling_factor": 4.0,
"moe_intermediate_size": 128,
"router_aux_loss_alpha": 0.01,
"official_router_init": true,
"max_position_embeddings": 2048,
"tie_word_embeddings": true,
"use_cache": false,
"mlp_bias": false,
"block_size": 256,
"batch_size": 16,
"max_steps": 152568,
"num_epochs": 1.0,
"learning_rate": 0.0003,
"weight_decay": 0.0,
"grad_clip": 1.0,
"warmup_steps": 2000,
"min_lr_ratio": 0.1,
"validation_fraction": 0.01,
"log_steps": 500,
"eval_steps": 10000,
"eval_batches": 32,
"save_steps": 10000,
"resume_from_checkpoint": "",
"seed": 1234,
"dtype": "float32",
"device": "cuda",
"num_workers": 0,
"generate_prompts": [
"Once upon",
"There was a little",
"One day"
],
"generate_new_tokens": 100,
"generation_temperature": 0.8,
"generation_top_p": 0.95,
"generation_top_k": 40,
"generation_repetition_penalty": 1.1,
"do_sample": true
},
"config": {
"transformers_version": "5.14.1",
"architectures": [
"DeepseekV2ForCausalLM"
],
"output_hidden_states": false,
"return_dict": true,
"dtype": "float32",
"chunk_size_feed_forward": 0,
"is_encoder_decoder": false,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"problem_type": null,
"vocab_size": 1024,
"hidden_size": 216,
"intermediate_size": 432,
"num_hidden_layers": 5,
"num_attention_heads": 8,
"num_key_value_heads": 8,
"hidden_act": "silu",
"max_position_embeddings": 2048,
"initializer_range": 0.02,
"rms_norm_eps": 1e-06,
"use_cache": false,
"pad_token_id": 1000,
"bos_token_id": 1000,
"eos_token_id": 1001,
"pretraining_tp": 1,
"tie_word_embeddings": true,
"rope_parameters": {
"rope_type": "default",
"rope_theta": 10000
},
"attention_bias": false,
"attention_dropout": 0.0,
"mlp_bias": false,
"head_dim": 16,
"first_k_dense_replace": 1,
"kv_lora_rank": 64,
"q_lora_rank": null,
"n_group": null,
"n_routed_experts": 4,
"n_shared_experts": 1,
"qk_nope_head_dim": 16,
"qk_rope_head_dim": 16,
"routed_scaling_factor": 4.0,
"topk_group": null,
"topk_method": "greedy",
"norm_topk_prob": false,
"v_head_dim": 32,
"num_experts_per_tok": 1,
"moe_intermediate_size": 128,
"_name_or_path": "",
"aux_loss_alpha": 0.01,
"seq_aux": true,
"model_type": "deepseek_v2",
"output_attentions": false
},
"generations": {
"Once upon": "Once upon a time, there was a little boy named Timmy. He loved to play in the sand and splash in the water. One day, he was playing with his toy car when it slipped on some rocks. His mom came and said, \"Oh no, I'm sorry! Let's bury something that is not ours so we can find out.\"\n\nTimmy didn't understand what \"oray\" meant, but",
"There was a little": "There was a little girl named Lily who loved to explore. One day, she saw a big bush with lots of yummy peaches on it. She wanted to eat one, but her mom said no.\n\nLily didn't listen and kept trying to eat her peach. She tried and tried but the bush wouldn't eat. Finally, she ate all the peaches and started to feel sick. Her mom told her not to eat",
"One day": "One day, the boy saw a palace in the park. He thought it was the most special palace he had ever seen! He wanted to play with it, so he called out to his friend, Alice.\n\nAlice asked the palace: \"Can I go in?\"\n\nHis friend replied: \"Yes, but you must be very kind and gentle. What do you think?\"\n\n"
}
} |