ripple-preview:65b

32 17 minutes ago

65B dense model built for agentic coding, with up to 1M tokens of context.

tools thinking 65b
... /
json
c0bbf99d20a5 ยท 2.2kB
{
"architectures": [
"DFlash2DraftModel"
],
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 100257,
"dflash_config": {
"block_size": 16,
"conv_group_size": 16,
"conv_kernel_size": 4,
"mask_token_id": 100278,
"selector_rank": 256,
"selector_top_k": 32,
"target_layer_ids": [
1,
20,
39,
58,
77
]
},
"draft_vocab_size": 100352,
"dtype": "bfloat16",
"eos_token_id": [
100265,
100257
],
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 5120,
"initializer_range": 0.02,
"intermediate_size": 40960,
"is_causal": false,
"layer_types": [
"sliding_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention"
],
"max_position_embeddings": 131072,
"max_window_layers": 5,
"model_type": "qwen3",
"no_rope_layers": [
4
],
"num_attention_heads": 80,
"num_hidden_layers": 5,
"num_key_value_heads": 8,
"num_target_layers": 80,
"pad_token_id": 100277,
"rms_norm_eps": 1e-06,
"rope_parameters": {
"rope_theta": 10000.0,
"rope_type": "default"
},
"rope_theta": 10000.0,
"sliding_window": 4096,
"tie_word_embeddings": false,
"transformers_version": "5.12.1",
"use_cache": true,
"use_sliding_window": true,
"vocab_size": 100352,
"quantization_config": {
"config_groups": {
"group_0": {
"weights": {
"dynamic": false,
"num_bits": 4,
"type": "float",
"group_size": 16
},
"targets": [
"Linear"
]
}
},
"ignore": [
"candidate_selector*",
"layers.0.attention_conv*",
"layers.0.mlp_conv*",
"layers.1.attention_conv*",
"layers.1.mlp_conv*",
"layers.2.attention_conv*",
"layers.2.mlp_conv*",
"layers.3.attention_conv*",
"layers.3.mlp_conv*",
"layers.4.attention_conv*",
"layers.4.mlp_conv*"
],
"quant_algo": "W4A16_NVFP4",
"group_size": 16,
"producer": {
"name": "modelopt",
"version": "0.47.0rc1.dev283+gf1d68fd4d"
},
"quant_method": "modelopt"
}
}