Upload shared sparse checkpoint
Browse files- config.json +45 -0
- generation_config.json +6 -0
- model.safetensors +3 -0
- shared_checkpoint_metadata.json +10 -0
config.json
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SparseLlamaForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"bos_token_id": 0,
|
| 8 |
+
"dtype": "bfloat16",
|
| 9 |
+
"eos_token_id": 0,
|
| 10 |
+
"head_dim": 64,
|
| 11 |
+
"hidden_act": "silu",
|
| 12 |
+
"hidden_size": 2048,
|
| 13 |
+
"initializer_range": 0.02,
|
| 14 |
+
"intermediate_size": 5632,
|
| 15 |
+
"is_llama_config": true,
|
| 16 |
+
"max_position_embeddings": 2048,
|
| 17 |
+
"mlp_bias": false,
|
| 18 |
+
"model_type": "llama_sparse_relu",
|
| 19 |
+
"num_attention_heads": 32,
|
| 20 |
+
"num_hidden_layers": 38,
|
| 21 |
+
"num_key_value_heads": 32,
|
| 22 |
+
"pretraining_tp": 1,
|
| 23 |
+
"rms_norm_eps": 1e-05,
|
| 24 |
+
"rope_interleaved": false,
|
| 25 |
+
"rope_scaling": null,
|
| 26 |
+
"rope_theta": 100000,
|
| 27 |
+
"sparsity_baseline_run": false,
|
| 28 |
+
"sparsity_do_track": true,
|
| 29 |
+
"sparsity_end_warmup_step": null,
|
| 30 |
+
"sparsity_gated_mlp": true,
|
| 31 |
+
"sparsity_l1_coeff": 2e-05,
|
| 32 |
+
"sparsity_max_seq_len": 2048,
|
| 33 |
+
"sparsity_max_seq_length": 2048,
|
| 34 |
+
"sparsity_mlp_execution_logic": "training",
|
| 35 |
+
"sparsity_new_intermediate_size": 5632,
|
| 36 |
+
"sparsity_per_device_train_batch_size": 8,
|
| 37 |
+
"sparsity_reinitialize": true,
|
| 38 |
+
"sparsity_start_warmup_step": null,
|
| 39 |
+
"sparsity_tokens_until_dead": null,
|
| 40 |
+
"sparsity_train_batch_size": 512,
|
| 41 |
+
"tie_word_embeddings": true,
|
| 42 |
+
"transformers_version": "4.56.2",
|
| 43 |
+
"use_cache": true,
|
| 44 |
+
"vocab_size": 49152
|
| 45 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 0,
|
| 4 |
+
"eos_token_id": 0,
|
| 5 |
+
"transformers_version": "4.56.2"
|
| 6 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4f9f35e6ad5e528e69c82fe0bd0ef40b695322f0465db0481a9894441644ecba
|
| 3 |
+
size 4307905552
|
shared_checkpoint_metadata.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"checkpoint_format": "sparse_checkpoint",
|
| 3 |
+
"config_set": "2em5_2B",
|
| 4 |
+
"export_name": "SparseLM2B",
|
| 5 |
+
"model_kind": "2B",
|
| 6 |
+
"normalized_config_set": "2em5",
|
| 7 |
+
"source_run_dir": "results/fineweb_edu_dedup_45B/smollm2Bcustom/smollm2Bcustom_Gated_2048ctx_lr_0.001_l12e-05/2026.01.12093654",
|
| 8 |
+
"source_state_dict_file": "results/fineweb_edu_dedup_45B/smollm2Bcustom/smollm2Bcustom_Gated_2048ctx_lr_0.001_l12e-05/2026.01.12093654/model.safetensors",
|
| 9 |
+
"weight_file": "model.safetensors"
|
| 10 |
+
}
|