Hugging Face's logo

Spaces:
alphabell
/
Model-Observatory
Running

App Files Community
4
Model-Observatory / models.json
alphabell-research's picture
alphabell-research
Publish the measured Flint and Qwen explorer
638598e verified
Raw History Blame
6.95 kB
[
{
"id": "flint",
"title": "Recursive Flint",
"short_name": "Flint",
"family": "MoE transformer",
"quantization": "Q8_0",
"source_url": "https://huggingface.co/RecursiveSI/flint/blob/main/config.json",
"learned_sink": true,
"stats": [
[
"49",
"",
"layers"
],
[
"32",
"",
"query heads / layer"
],
[
"6",
"256",
"routed experts / token"
],
[
"2.5B",
"39.8B",
"active parameters"
]
],
"architecture": {
"architectures": [
"FlintForCausalLM"
],
"model_type": "flint",
"attention_bias": false,
"attention_dropout": 0,
"bos_token_id": 150000,
"decoder_sparse_step": 1,
"eos_token_id": [
150001,
150003,
150005,
150018
],
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 2048,
"initializer_range": 0.02,
"intermediate_size": 6144,
"layer_types": [
"full_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"full_attention"
],
"max_position_embeddings": 262144,
"mlp_only_layers": [
0,
1
],
"moe_intermediate_size": 512,
"norm_topk_prob": true,
"num_attention_heads": 32,
"num_experts": 256,
"num_experts_per_tok": 6,
"num_hidden_layers": 49,
"num_key_value_heads": 4,
"no_rope_layer_indices": [
0,
1,
5,
9,
13,
17,
21,
25,
29,
33,
37,
41,
45,
48
],
"output_router_logits": false,
"pad_token_id": 150001,
"rms_norm_eps": 1e-06,
"rope_theta": 10000,
"router_aux_loss_coef": 0.0005,
"router_score_function": "sigmoid",
"shared_expert_intermediate_size": 1024,
"sliding_window": 513,
"softmax_type": "learnable",
"tie_word_embeddings": false,
"torch_dtype": "bfloat16",
"use_cache": true,
"use_sliding_window": true,
"vocab_size": 150144,
"transformers_version": "5.12.1"
}
},
{
"id": "qwen",
"title": "Qwen3.8-27B",
"short_name": "Qwen",
"family": "Hybrid recurrent transformer",
"quantization": "Q8_0",
"source_url": "https://huggingface.co/Qwen/Qwen3.8-27B/blob/main/config.json",
"learned_sink": false,
"stats": [
[
"64",
"",
"layers"
],
[
"48",
"24",
"linear / full heads"
],
[
"17408",
"",
"dense FFN width"
],
[
"5120",
"",
"hidden width"
]
],
"architecture": {
"architectures": [
"Qwen3_5ForConditionalGeneration"
],
"model_type": "qwen3_5_text",
"num_hidden_layers": 64,
"hidden_size": 5120,
"intermediate_size": 17408,
"num_attention_heads": 24,
"num_key_value_heads": 4,
"head_dim": 256,
"attn_output_gate": true,
"full_attention_interval": 4,
"layer_types": [
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention",
"linear_attention",
"linear_attention",
"linear_attention",
"full_attention"
],
"linear_conv_kernel_dim": 4,
"linear_key_head_dim": 128,
"linear_value_head_dim": 128,
"linear_num_key_heads": 16,
"linear_num_value_heads": 48,
"mamba_ssm_dtype": "float32",
"rms_norm_eps": 1e-06,
"max_position_embeddings": 262144,
"vocab_size": 248320,
"partial_rotary_factor": 0.25,
"rope_parameters": {
"mrope_interleaved": true,
"mrope_section": [
11,
11,
10
],
"partial_rotary_factor": 0.25,
"rope_theta": 10000000,
"rope_type": "default"
},
"tie_word_embeddings": false,
"hidden_act": "silu",
"num_experts": 0,
"first_k_dense_replace": 64
}
}
]