Views
No views yet
atrost/nanochat-d24-matformer-h16-natural-sft/d24_matformer_h16_natural_sft at step 000485.matformerXL{"attn_alpha_init_value": 1.0, "dec_alpha_init_value": 1.0, "ffn_alpha_init_value": 1.0, "head_dim": 80, "matformer_ffn_hidden_dims": [5120], "matformer_head_counts": [16], "n_embd": 1280, "n_head": 16, "n_kv_head": 16, "n_layer": 24, "norm_type": "rmsnorm", "sequence_len": 2048, "vocab_size": 32768, "window_pattern": "L"}1from nanochat.hf_utils import stage_hf_checkpoint
2
3stage_hf_checkpoint(
4 repo_id="atrost/nanochat-d24-matformer-h16-natural-sft-xl",
5 path_in_repo="d24_matformer_h16_natural_sft_xl",
6 model_tag="d24_matformer_h16_natural_sft_xl",
7 step=485,
8 base_dir="/tmp/nanochat-eval",
9)NANOCHAT_BASE_DIR=/tmp/nanochat-eval python -m scripts.base_eval --model-tag d24_matformer_h16_natural_sft_xl --step 485