Add files using upload-large-folder tool
Browse files- Keen-qwen4b-20b/README.md +16 -0
- Keen-qwen4b-20b/checkpoint-11202/README.md +40 -0
- Keen-qwen4b-20b/checkpoint-11202/config.json +83 -0
- Keen-qwen4b-20b/checkpoint-11202/generation_config.json +9 -0
- Keen-qwen4b-20b/checkpoint-11202/tokenizer_config.json +47 -0
- Keen-qwen4b-20b/checkpoint-13449/README.md +40 -0
- Keen-qwen4b-20b/checkpoint-13449/config.json +83 -0
- Keen-qwen4b-20b/checkpoint-13449/generation_config.json +9 -0
- Keen-qwen4b-20b/checkpoint-13449/tokenizer_config.json +47 -0
- Keen-qwen4b-20b/checkpoint-15698/README.md +40 -0
- Keen-qwen4b-20b/checkpoint-15698/config.json +83 -0
- Keen-qwen4b-20b/checkpoint-15698/generation_config.json +9 -0
- Keen-qwen4b-20b/checkpoint-15698/tokenizer_config.json +47 -0
- Keen-qwen4b-20b/checkpoint-17953/README.md +40 -0
- Keen-qwen4b-20b/checkpoint-17953/config.json +83 -0
- Keen-qwen4b-20b/checkpoint-17953/generation_config.json +9 -0
- Keen-qwen4b-20b/checkpoint-17953/tokenizer_config.json +47 -0
- Keen-qwen4b-20b/checkpoint-20218/README.md +40 -0
- Keen-qwen4b-20b/checkpoint-20218/config.json +83 -0
- Keen-qwen4b-20b/checkpoint-20218/generation_config.json +9 -0
- Keen-qwen4b-20b/checkpoint-20218/tokenizer_config.json +47 -0
- Keen-qwen4b-20b/checkpoint-22478/README.md +40 -0
- Keen-qwen4b-20b/checkpoint-22478/config.json +83 -0
- Keen-qwen4b-20b/checkpoint-22478/generation_config.json +9 -0
- Keen-qwen4b-20b/checkpoint-22478/tokenizer_config.json +47 -0
- Keen-qwen4b-20b/checkpoint-2253/README.md +40 -0
- Keen-qwen4b-20b/checkpoint-2253/config.json +83 -0
- Keen-qwen4b-20b/checkpoint-2253/generation_config.json +9 -0
- Keen-qwen4b-20b/checkpoint-2253/tokenizer_config.json +47 -0
- Keen-qwen4b-20b/checkpoint-4453/README.md +40 -0
- Keen-qwen4b-20b/checkpoint-4453/config.json +83 -0
- Keen-qwen4b-20b/checkpoint-4453/generation_config.json +9 -0
- Keen-qwen4b-20b/checkpoint-4453/tokenizer_config.json +47 -0
- Keen-qwen4b-20b/checkpoint-6701/README.md +40 -0
- Keen-qwen4b-20b/checkpoint-6701/config.json +83 -0
- Keen-qwen4b-20b/checkpoint-6701/generation_config.json +9 -0
- Keen-qwen4b-20b/checkpoint-6701/tokenizer_config.json +47 -0
- Keen-qwen4b-20b/checkpoint-8956/README.md +40 -0
- Keen-qwen4b-20b/checkpoint-8956/config.json +83 -0
- Keen-qwen4b-20b/checkpoint-8956/generation_config.json +9 -0
- Keen-qwen4b-20b/checkpoint-8956/tokenizer_config.json +47 -0
- README.md +32 -0
- Vanilla-qwen4b-20b/README.md +16 -0
- Vanilla-qwen4b-20b/checkpoint-1910/README.md +40 -0
- Vanilla-qwen4b-20b/checkpoint-9543/README.md +40 -0
- Vanilla-qwen4b-20b/checkpoint-9543/config.json +83 -0
- Vanilla-qwen4b-20b/checkpoint-9543/generation_config.json +9 -0
- Vanilla-qwen4b-20b/checkpoint-9543/tokenizer_config.json +47 -0
- checkpoints.json +202 -0
- manifest.json +626 -0
Keen-qwen4b-20b/README.md
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Keen-qwen4b-20b
|
| 2 |
+
|
| 3 |
+
Ten budget-boundary checkpoints, approximately every 2B body tokens from 2B through 20B. The additional final checkpoint at approximately 20.156B tokens is intentionally excluded.
|
| 4 |
+
|
| 5 |
+
| Checkpoint | Step | Body tokens | Status |
|
| 6 |
+
|---|---:|---:|---|
|
| 7 |
+
| [checkpoint-2253](checkpoint-2253) | 2,253 | 2,000,696,711 | Intermediate |
|
| 8 |
+
| [checkpoint-4453](checkpoint-4453) | 4,453 | 4,000,757,038 | Intermediate |
|
| 9 |
+
| [checkpoint-6701](checkpoint-6701) | 6,701 | 6,000,082,648 | Intermediate |
|
| 10 |
+
| [checkpoint-8956](checkpoint-8956) | 8,956 | 8,000,600,622 | Intermediate |
|
| 11 |
+
| [checkpoint-11202](checkpoint-11202) | 11,202 | 10,000,294,657 | Intermediate |
|
| 12 |
+
| [checkpoint-13449](checkpoint-13449) | 13,449 | 12,000,429,703 | Intermediate |
|
| 13 |
+
| [checkpoint-15698](checkpoint-15698) | 15,698 | 14,000,083,680 | Intermediate |
|
| 14 |
+
| [checkpoint-17953](checkpoint-17953) | 17,953 | 16,000,387,372 | Intermediate |
|
| 15 |
+
| [checkpoint-20218](checkpoint-20218) | 20,218 | 18,000,030,679 | Intermediate |
|
| 16 |
+
| [checkpoint-22478](checkpoint-22478) | 22,478 | 20,000,557,588 | Intermediate |
|
Keen-qwen4b-20b/checkpoint-11202/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Keen-qwen4b-20b: checkpoint-11202
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 11202
|
| 10 |
+
- Cumulative body tokens: 10,000,294,657
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Keen conditions passage reconstruction on `<fact>entity list</fact>`; the prefix participates in attention but is excluded from the training loss.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Keen-qwen4b-20b/checkpoint-11202"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("<fact>Paris, France</fact>\nParis", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Keen-qwen4b-20b/checkpoint-11202/config.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"attn_output_gate": true,
|
| 8 |
+
"bos_token_id": null,
|
| 9 |
+
"dtype": "bfloat16",
|
| 10 |
+
"eos_token_id": 248044,
|
| 11 |
+
"full_attention_interval": 4,
|
| 12 |
+
"head_dim": 256,
|
| 13 |
+
"hidden_act": "silu",
|
| 14 |
+
"hidden_size": 2560,
|
| 15 |
+
"initializer_range": 0.02,
|
| 16 |
+
"intermediate_size": 9216,
|
| 17 |
+
"layer_types": [
|
| 18 |
+
"linear_attention",
|
| 19 |
+
"linear_attention",
|
| 20 |
+
"linear_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"linear_attention",
|
| 23 |
+
"linear_attention",
|
| 24 |
+
"linear_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"linear_attention",
|
| 27 |
+
"linear_attention",
|
| 28 |
+
"linear_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"linear_attention",
|
| 31 |
+
"linear_attention",
|
| 32 |
+
"linear_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"linear_attention",
|
| 35 |
+
"linear_attention",
|
| 36 |
+
"linear_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"linear_attention",
|
| 39 |
+
"linear_attention",
|
| 40 |
+
"linear_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"linear_attention",
|
| 43 |
+
"linear_attention",
|
| 44 |
+
"linear_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"linear_attention",
|
| 47 |
+
"linear_attention",
|
| 48 |
+
"linear_attention",
|
| 49 |
+
"full_attention"
|
| 50 |
+
],
|
| 51 |
+
"linear_conv_kernel_dim": 4,
|
| 52 |
+
"linear_key_head_dim": 128,
|
| 53 |
+
"linear_num_key_heads": 16,
|
| 54 |
+
"linear_num_value_heads": 32,
|
| 55 |
+
"linear_value_head_dim": 128,
|
| 56 |
+
"mamba_ssm_dtype": "float32",
|
| 57 |
+
"max_position_embeddings": 262144,
|
| 58 |
+
"mlp_only_layers": [],
|
| 59 |
+
"model_type": "qwen3_5_text",
|
| 60 |
+
"mtp_num_hidden_layers": 1,
|
| 61 |
+
"mtp_use_dedicated_embeddings": false,
|
| 62 |
+
"num_attention_heads": 16,
|
| 63 |
+
"num_hidden_layers": 32,
|
| 64 |
+
"num_key_value_heads": 4,
|
| 65 |
+
"pad_token_id": 248044,
|
| 66 |
+
"partial_rotary_factor": 0.25,
|
| 67 |
+
"rms_norm_eps": 1e-06,
|
| 68 |
+
"rope_parameters": {
|
| 69 |
+
"mrope_interleaved": true,
|
| 70 |
+
"mrope_section": [
|
| 71 |
+
11,
|
| 72 |
+
11,
|
| 73 |
+
10
|
| 74 |
+
],
|
| 75 |
+
"partial_rotary_factor": 0.25,
|
| 76 |
+
"rope_theta": 10000000,
|
| 77 |
+
"rope_type": "default"
|
| 78 |
+
},
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"transformers_version": "5.6.2",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 248320
|
| 83 |
+
}
|
Keen-qwen4b-20b/checkpoint-11202/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 248044,
|
| 4 |
+
"output_attentions": false,
|
| 5 |
+
"output_hidden_states": false,
|
| 6 |
+
"pad_token_id": 248044,
|
| 7 |
+
"transformers_version": "5.6.2",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
Keen-qwen4b-20b/checkpoint-11202/tokenizer_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"audio_bos_token": "<|audio_start|>",
|
| 4 |
+
"audio_eos_token": "<|audio_end|>",
|
| 5 |
+
"audio_token": "<|audio_pad|>",
|
| 6 |
+
"backend": "tokenizers",
|
| 7 |
+
"bos_token": null,
|
| 8 |
+
"clean_up_tokenization_spaces": false,
|
| 9 |
+
"eos_token": "<|endoftext|>",
|
| 10 |
+
"errors": "replace",
|
| 11 |
+
"extra_special_tokens": [
|
| 12 |
+
"<|im_start|>",
|
| 13 |
+
"<|im_end|>",
|
| 14 |
+
"<|object_ref_start|>",
|
| 15 |
+
"<|object_ref_end|>",
|
| 16 |
+
"<|box_start|>",
|
| 17 |
+
"<|box_end|>",
|
| 18 |
+
"<|quad_start|>",
|
| 19 |
+
"<|quad_end|>",
|
| 20 |
+
"<|vision_start|>",
|
| 21 |
+
"<|vision_end|>",
|
| 22 |
+
"<|vision_pad|>",
|
| 23 |
+
"<|image_pad|>",
|
| 24 |
+
"<|video_pad|>",
|
| 25 |
+
"<fact>",
|
| 26 |
+
"</fact>"
|
| 27 |
+
],
|
| 28 |
+
"image_token": "<|image_pad|>",
|
| 29 |
+
"model_max_length": 32768,
|
| 30 |
+
"model_specific_special_tokens": {
|
| 31 |
+
"audio_bos_token": "<|audio_start|>",
|
| 32 |
+
"audio_eos_token": "<|audio_end|>",
|
| 33 |
+
"audio_token": "<|audio_pad|>",
|
| 34 |
+
"image_token": "<|image_pad|>",
|
| 35 |
+
"video_token": "<|video_pad|>",
|
| 36 |
+
"vision_bos_token": "<|vision_start|>",
|
| 37 |
+
"vision_eos_token": "<|vision_end|>"
|
| 38 |
+
},
|
| 39 |
+
"pad_token": "<|endoftext|>",
|
| 40 |
+
"pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
| 41 |
+
"split_special_tokens": false,
|
| 42 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 43 |
+
"unk_token": null,
|
| 44 |
+
"video_token": "<|video_pad|>",
|
| 45 |
+
"vision_bos_token": "<|vision_start|>",
|
| 46 |
+
"vision_eos_token": "<|vision_end|>"
|
| 47 |
+
}
|
Keen-qwen4b-20b/checkpoint-13449/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Keen-qwen4b-20b: checkpoint-13449
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 13449
|
| 10 |
+
- Cumulative body tokens: 12,000,429,703
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Keen conditions passage reconstruction on `<fact>entity list</fact>`; the prefix participates in attention but is excluded from the training loss.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Keen-qwen4b-20b/checkpoint-13449"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("<fact>Paris, France</fact>\nParis", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Keen-qwen4b-20b/checkpoint-13449/config.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"attn_output_gate": true,
|
| 8 |
+
"bos_token_id": null,
|
| 9 |
+
"dtype": "bfloat16",
|
| 10 |
+
"eos_token_id": 248044,
|
| 11 |
+
"full_attention_interval": 4,
|
| 12 |
+
"head_dim": 256,
|
| 13 |
+
"hidden_act": "silu",
|
| 14 |
+
"hidden_size": 2560,
|
| 15 |
+
"initializer_range": 0.02,
|
| 16 |
+
"intermediate_size": 9216,
|
| 17 |
+
"layer_types": [
|
| 18 |
+
"linear_attention",
|
| 19 |
+
"linear_attention",
|
| 20 |
+
"linear_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"linear_attention",
|
| 23 |
+
"linear_attention",
|
| 24 |
+
"linear_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"linear_attention",
|
| 27 |
+
"linear_attention",
|
| 28 |
+
"linear_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"linear_attention",
|
| 31 |
+
"linear_attention",
|
| 32 |
+
"linear_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"linear_attention",
|
| 35 |
+
"linear_attention",
|
| 36 |
+
"linear_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"linear_attention",
|
| 39 |
+
"linear_attention",
|
| 40 |
+
"linear_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"linear_attention",
|
| 43 |
+
"linear_attention",
|
| 44 |
+
"linear_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"linear_attention",
|
| 47 |
+
"linear_attention",
|
| 48 |
+
"linear_attention",
|
| 49 |
+
"full_attention"
|
| 50 |
+
],
|
| 51 |
+
"linear_conv_kernel_dim": 4,
|
| 52 |
+
"linear_key_head_dim": 128,
|
| 53 |
+
"linear_num_key_heads": 16,
|
| 54 |
+
"linear_num_value_heads": 32,
|
| 55 |
+
"linear_value_head_dim": 128,
|
| 56 |
+
"mamba_ssm_dtype": "float32",
|
| 57 |
+
"max_position_embeddings": 262144,
|
| 58 |
+
"mlp_only_layers": [],
|
| 59 |
+
"model_type": "qwen3_5_text",
|
| 60 |
+
"mtp_num_hidden_layers": 1,
|
| 61 |
+
"mtp_use_dedicated_embeddings": false,
|
| 62 |
+
"num_attention_heads": 16,
|
| 63 |
+
"num_hidden_layers": 32,
|
| 64 |
+
"num_key_value_heads": 4,
|
| 65 |
+
"pad_token_id": 248044,
|
| 66 |
+
"partial_rotary_factor": 0.25,
|
| 67 |
+
"rms_norm_eps": 1e-06,
|
| 68 |
+
"rope_parameters": {
|
| 69 |
+
"mrope_interleaved": true,
|
| 70 |
+
"mrope_section": [
|
| 71 |
+
11,
|
| 72 |
+
11,
|
| 73 |
+
10
|
| 74 |
+
],
|
| 75 |
+
"partial_rotary_factor": 0.25,
|
| 76 |
+
"rope_theta": 10000000,
|
| 77 |
+
"rope_type": "default"
|
| 78 |
+
},
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"transformers_version": "5.6.2",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 248320
|
| 83 |
+
}
|
Keen-qwen4b-20b/checkpoint-13449/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 248044,
|
| 4 |
+
"output_attentions": false,
|
| 5 |
+
"output_hidden_states": false,
|
| 6 |
+
"pad_token_id": 248044,
|
| 7 |
+
"transformers_version": "5.6.2",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
Keen-qwen4b-20b/checkpoint-13449/tokenizer_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"audio_bos_token": "<|audio_start|>",
|
| 4 |
+
"audio_eos_token": "<|audio_end|>",
|
| 5 |
+
"audio_token": "<|audio_pad|>",
|
| 6 |
+
"backend": "tokenizers",
|
| 7 |
+
"bos_token": null,
|
| 8 |
+
"clean_up_tokenization_spaces": false,
|
| 9 |
+
"eos_token": "<|endoftext|>",
|
| 10 |
+
"errors": "replace",
|
| 11 |
+
"extra_special_tokens": [
|
| 12 |
+
"<|im_start|>",
|
| 13 |
+
"<|im_end|>",
|
| 14 |
+
"<|object_ref_start|>",
|
| 15 |
+
"<|object_ref_end|>",
|
| 16 |
+
"<|box_start|>",
|
| 17 |
+
"<|box_end|>",
|
| 18 |
+
"<|quad_start|>",
|
| 19 |
+
"<|quad_end|>",
|
| 20 |
+
"<|vision_start|>",
|
| 21 |
+
"<|vision_end|>",
|
| 22 |
+
"<|vision_pad|>",
|
| 23 |
+
"<|image_pad|>",
|
| 24 |
+
"<|video_pad|>",
|
| 25 |
+
"<fact>",
|
| 26 |
+
"</fact>"
|
| 27 |
+
],
|
| 28 |
+
"image_token": "<|image_pad|>",
|
| 29 |
+
"model_max_length": 32768,
|
| 30 |
+
"model_specific_special_tokens": {
|
| 31 |
+
"audio_bos_token": "<|audio_start|>",
|
| 32 |
+
"audio_eos_token": "<|audio_end|>",
|
| 33 |
+
"audio_token": "<|audio_pad|>",
|
| 34 |
+
"image_token": "<|image_pad|>",
|
| 35 |
+
"video_token": "<|video_pad|>",
|
| 36 |
+
"vision_bos_token": "<|vision_start|>",
|
| 37 |
+
"vision_eos_token": "<|vision_end|>"
|
| 38 |
+
},
|
| 39 |
+
"pad_token": "<|endoftext|>",
|
| 40 |
+
"pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
| 41 |
+
"split_special_tokens": false,
|
| 42 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 43 |
+
"unk_token": null,
|
| 44 |
+
"video_token": "<|video_pad|>",
|
| 45 |
+
"vision_bos_token": "<|vision_start|>",
|
| 46 |
+
"vision_eos_token": "<|vision_end|>"
|
| 47 |
+
}
|
Keen-qwen4b-20b/checkpoint-15698/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Keen-qwen4b-20b: checkpoint-15698
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 15698
|
| 10 |
+
- Cumulative body tokens: 14,000,083,680
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Keen conditions passage reconstruction on `<fact>entity list</fact>`; the prefix participates in attention but is excluded from the training loss.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Keen-qwen4b-20b/checkpoint-15698"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("<fact>Paris, France</fact>\nParis", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Keen-qwen4b-20b/checkpoint-15698/config.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"attn_output_gate": true,
|
| 8 |
+
"bos_token_id": null,
|
| 9 |
+
"dtype": "bfloat16",
|
| 10 |
+
"eos_token_id": 248044,
|
| 11 |
+
"full_attention_interval": 4,
|
| 12 |
+
"head_dim": 256,
|
| 13 |
+
"hidden_act": "silu",
|
| 14 |
+
"hidden_size": 2560,
|
| 15 |
+
"initializer_range": 0.02,
|
| 16 |
+
"intermediate_size": 9216,
|
| 17 |
+
"layer_types": [
|
| 18 |
+
"linear_attention",
|
| 19 |
+
"linear_attention",
|
| 20 |
+
"linear_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"linear_attention",
|
| 23 |
+
"linear_attention",
|
| 24 |
+
"linear_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"linear_attention",
|
| 27 |
+
"linear_attention",
|
| 28 |
+
"linear_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"linear_attention",
|
| 31 |
+
"linear_attention",
|
| 32 |
+
"linear_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"linear_attention",
|
| 35 |
+
"linear_attention",
|
| 36 |
+
"linear_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"linear_attention",
|
| 39 |
+
"linear_attention",
|
| 40 |
+
"linear_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"linear_attention",
|
| 43 |
+
"linear_attention",
|
| 44 |
+
"linear_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"linear_attention",
|
| 47 |
+
"linear_attention",
|
| 48 |
+
"linear_attention",
|
| 49 |
+
"full_attention"
|
| 50 |
+
],
|
| 51 |
+
"linear_conv_kernel_dim": 4,
|
| 52 |
+
"linear_key_head_dim": 128,
|
| 53 |
+
"linear_num_key_heads": 16,
|
| 54 |
+
"linear_num_value_heads": 32,
|
| 55 |
+
"linear_value_head_dim": 128,
|
| 56 |
+
"mamba_ssm_dtype": "float32",
|
| 57 |
+
"max_position_embeddings": 262144,
|
| 58 |
+
"mlp_only_layers": [],
|
| 59 |
+
"model_type": "qwen3_5_text",
|
| 60 |
+
"mtp_num_hidden_layers": 1,
|
| 61 |
+
"mtp_use_dedicated_embeddings": false,
|
| 62 |
+
"num_attention_heads": 16,
|
| 63 |
+
"num_hidden_layers": 32,
|
| 64 |
+
"num_key_value_heads": 4,
|
| 65 |
+
"pad_token_id": 248044,
|
| 66 |
+
"partial_rotary_factor": 0.25,
|
| 67 |
+
"rms_norm_eps": 1e-06,
|
| 68 |
+
"rope_parameters": {
|
| 69 |
+
"mrope_interleaved": true,
|
| 70 |
+
"mrope_section": [
|
| 71 |
+
11,
|
| 72 |
+
11,
|
| 73 |
+
10
|
| 74 |
+
],
|
| 75 |
+
"partial_rotary_factor": 0.25,
|
| 76 |
+
"rope_theta": 10000000,
|
| 77 |
+
"rope_type": "default"
|
| 78 |
+
},
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"transformers_version": "5.6.2",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 248320
|
| 83 |
+
}
|
Keen-qwen4b-20b/checkpoint-15698/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 248044,
|
| 4 |
+
"output_attentions": false,
|
| 5 |
+
"output_hidden_states": false,
|
| 6 |
+
"pad_token_id": 248044,
|
| 7 |
+
"transformers_version": "5.6.2",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
Keen-qwen4b-20b/checkpoint-15698/tokenizer_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"audio_bos_token": "<|audio_start|>",
|
| 4 |
+
"audio_eos_token": "<|audio_end|>",
|
| 5 |
+
"audio_token": "<|audio_pad|>",
|
| 6 |
+
"backend": "tokenizers",
|
| 7 |
+
"bos_token": null,
|
| 8 |
+
"clean_up_tokenization_spaces": false,
|
| 9 |
+
"eos_token": "<|endoftext|>",
|
| 10 |
+
"errors": "replace",
|
| 11 |
+
"extra_special_tokens": [
|
| 12 |
+
"<|im_start|>",
|
| 13 |
+
"<|im_end|>",
|
| 14 |
+
"<|object_ref_start|>",
|
| 15 |
+
"<|object_ref_end|>",
|
| 16 |
+
"<|box_start|>",
|
| 17 |
+
"<|box_end|>",
|
| 18 |
+
"<|quad_start|>",
|
| 19 |
+
"<|quad_end|>",
|
| 20 |
+
"<|vision_start|>",
|
| 21 |
+
"<|vision_end|>",
|
| 22 |
+
"<|vision_pad|>",
|
| 23 |
+
"<|image_pad|>",
|
| 24 |
+
"<|video_pad|>",
|
| 25 |
+
"<fact>",
|
| 26 |
+
"</fact>"
|
| 27 |
+
],
|
| 28 |
+
"image_token": "<|image_pad|>",
|
| 29 |
+
"model_max_length": 32768,
|
| 30 |
+
"model_specific_special_tokens": {
|
| 31 |
+
"audio_bos_token": "<|audio_start|>",
|
| 32 |
+
"audio_eos_token": "<|audio_end|>",
|
| 33 |
+
"audio_token": "<|audio_pad|>",
|
| 34 |
+
"image_token": "<|image_pad|>",
|
| 35 |
+
"video_token": "<|video_pad|>",
|
| 36 |
+
"vision_bos_token": "<|vision_start|>",
|
| 37 |
+
"vision_eos_token": "<|vision_end|>"
|
| 38 |
+
},
|
| 39 |
+
"pad_token": "<|endoftext|>",
|
| 40 |
+
"pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
| 41 |
+
"split_special_tokens": false,
|
| 42 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 43 |
+
"unk_token": null,
|
| 44 |
+
"video_token": "<|video_pad|>",
|
| 45 |
+
"vision_bos_token": "<|vision_start|>",
|
| 46 |
+
"vision_eos_token": "<|vision_end|>"
|
| 47 |
+
}
|
Keen-qwen4b-20b/checkpoint-17953/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Keen-qwen4b-20b: checkpoint-17953
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 17953
|
| 10 |
+
- Cumulative body tokens: 16,000,387,372
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Keen conditions passage reconstruction on `<fact>entity list</fact>`; the prefix participates in attention but is excluded from the training loss.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Keen-qwen4b-20b/checkpoint-17953"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("<fact>Paris, France</fact>\nParis", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Keen-qwen4b-20b/checkpoint-17953/config.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"attn_output_gate": true,
|
| 8 |
+
"bos_token_id": null,
|
| 9 |
+
"dtype": "bfloat16",
|
| 10 |
+
"eos_token_id": 248044,
|
| 11 |
+
"full_attention_interval": 4,
|
| 12 |
+
"head_dim": 256,
|
| 13 |
+
"hidden_act": "silu",
|
| 14 |
+
"hidden_size": 2560,
|
| 15 |
+
"initializer_range": 0.02,
|
| 16 |
+
"intermediate_size": 9216,
|
| 17 |
+
"layer_types": [
|
| 18 |
+
"linear_attention",
|
| 19 |
+
"linear_attention",
|
| 20 |
+
"linear_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"linear_attention",
|
| 23 |
+
"linear_attention",
|
| 24 |
+
"linear_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"linear_attention",
|
| 27 |
+
"linear_attention",
|
| 28 |
+
"linear_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"linear_attention",
|
| 31 |
+
"linear_attention",
|
| 32 |
+
"linear_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"linear_attention",
|
| 35 |
+
"linear_attention",
|
| 36 |
+
"linear_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"linear_attention",
|
| 39 |
+
"linear_attention",
|
| 40 |
+
"linear_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"linear_attention",
|
| 43 |
+
"linear_attention",
|
| 44 |
+
"linear_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"linear_attention",
|
| 47 |
+
"linear_attention",
|
| 48 |
+
"linear_attention",
|
| 49 |
+
"full_attention"
|
| 50 |
+
],
|
| 51 |
+
"linear_conv_kernel_dim": 4,
|
| 52 |
+
"linear_key_head_dim": 128,
|
| 53 |
+
"linear_num_key_heads": 16,
|
| 54 |
+
"linear_num_value_heads": 32,
|
| 55 |
+
"linear_value_head_dim": 128,
|
| 56 |
+
"mamba_ssm_dtype": "float32",
|
| 57 |
+
"max_position_embeddings": 262144,
|
| 58 |
+
"mlp_only_layers": [],
|
| 59 |
+
"model_type": "qwen3_5_text",
|
| 60 |
+
"mtp_num_hidden_layers": 1,
|
| 61 |
+
"mtp_use_dedicated_embeddings": false,
|
| 62 |
+
"num_attention_heads": 16,
|
| 63 |
+
"num_hidden_layers": 32,
|
| 64 |
+
"num_key_value_heads": 4,
|
| 65 |
+
"pad_token_id": 248044,
|
| 66 |
+
"partial_rotary_factor": 0.25,
|
| 67 |
+
"rms_norm_eps": 1e-06,
|
| 68 |
+
"rope_parameters": {
|
| 69 |
+
"mrope_interleaved": true,
|
| 70 |
+
"mrope_section": [
|
| 71 |
+
11,
|
| 72 |
+
11,
|
| 73 |
+
10
|
| 74 |
+
],
|
| 75 |
+
"partial_rotary_factor": 0.25,
|
| 76 |
+
"rope_theta": 10000000,
|
| 77 |
+
"rope_type": "default"
|
| 78 |
+
},
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"transformers_version": "5.6.2",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 248320
|
| 83 |
+
}
|
Keen-qwen4b-20b/checkpoint-17953/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 248044,
|
| 4 |
+
"output_attentions": false,
|
| 5 |
+
"output_hidden_states": false,
|
| 6 |
+
"pad_token_id": 248044,
|
| 7 |
+
"transformers_version": "5.6.2",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
Keen-qwen4b-20b/checkpoint-17953/tokenizer_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"audio_bos_token": "<|audio_start|>",
|
| 4 |
+
"audio_eos_token": "<|audio_end|>",
|
| 5 |
+
"audio_token": "<|audio_pad|>",
|
| 6 |
+
"backend": "tokenizers",
|
| 7 |
+
"bos_token": null,
|
| 8 |
+
"clean_up_tokenization_spaces": false,
|
| 9 |
+
"eos_token": "<|endoftext|>",
|
| 10 |
+
"errors": "replace",
|
| 11 |
+
"extra_special_tokens": [
|
| 12 |
+
"<|im_start|>",
|
| 13 |
+
"<|im_end|>",
|
| 14 |
+
"<|object_ref_start|>",
|
| 15 |
+
"<|object_ref_end|>",
|
| 16 |
+
"<|box_start|>",
|
| 17 |
+
"<|box_end|>",
|
| 18 |
+
"<|quad_start|>",
|
| 19 |
+
"<|quad_end|>",
|
| 20 |
+
"<|vision_start|>",
|
| 21 |
+
"<|vision_end|>",
|
| 22 |
+
"<|vision_pad|>",
|
| 23 |
+
"<|image_pad|>",
|
| 24 |
+
"<|video_pad|>",
|
| 25 |
+
"<fact>",
|
| 26 |
+
"</fact>"
|
| 27 |
+
],
|
| 28 |
+
"image_token": "<|image_pad|>",
|
| 29 |
+
"model_max_length": 32768,
|
| 30 |
+
"model_specific_special_tokens": {
|
| 31 |
+
"audio_bos_token": "<|audio_start|>",
|
| 32 |
+
"audio_eos_token": "<|audio_end|>",
|
| 33 |
+
"audio_token": "<|audio_pad|>",
|
| 34 |
+
"image_token": "<|image_pad|>",
|
| 35 |
+
"video_token": "<|video_pad|>",
|
| 36 |
+
"vision_bos_token": "<|vision_start|>",
|
| 37 |
+
"vision_eos_token": "<|vision_end|>"
|
| 38 |
+
},
|
| 39 |
+
"pad_token": "<|endoftext|>",
|
| 40 |
+
"pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
| 41 |
+
"split_special_tokens": false,
|
| 42 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 43 |
+
"unk_token": null,
|
| 44 |
+
"video_token": "<|video_pad|>",
|
| 45 |
+
"vision_bos_token": "<|vision_start|>",
|
| 46 |
+
"vision_eos_token": "<|vision_end|>"
|
| 47 |
+
}
|
Keen-qwen4b-20b/checkpoint-20218/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Keen-qwen4b-20b: checkpoint-20218
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 20218
|
| 10 |
+
- Cumulative body tokens: 18,000,030,679
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Keen conditions passage reconstruction on `<fact>entity list</fact>`; the prefix participates in attention but is excluded from the training loss.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Keen-qwen4b-20b/checkpoint-20218"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("<fact>Paris, France</fact>\nParis", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Keen-qwen4b-20b/checkpoint-20218/config.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"attn_output_gate": true,
|
| 8 |
+
"bos_token_id": null,
|
| 9 |
+
"dtype": "bfloat16",
|
| 10 |
+
"eos_token_id": 248044,
|
| 11 |
+
"full_attention_interval": 4,
|
| 12 |
+
"head_dim": 256,
|
| 13 |
+
"hidden_act": "silu",
|
| 14 |
+
"hidden_size": 2560,
|
| 15 |
+
"initializer_range": 0.02,
|
| 16 |
+
"intermediate_size": 9216,
|
| 17 |
+
"layer_types": [
|
| 18 |
+
"linear_attention",
|
| 19 |
+
"linear_attention",
|
| 20 |
+
"linear_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"linear_attention",
|
| 23 |
+
"linear_attention",
|
| 24 |
+
"linear_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"linear_attention",
|
| 27 |
+
"linear_attention",
|
| 28 |
+
"linear_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"linear_attention",
|
| 31 |
+
"linear_attention",
|
| 32 |
+
"linear_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"linear_attention",
|
| 35 |
+
"linear_attention",
|
| 36 |
+
"linear_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"linear_attention",
|
| 39 |
+
"linear_attention",
|
| 40 |
+
"linear_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"linear_attention",
|
| 43 |
+
"linear_attention",
|
| 44 |
+
"linear_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"linear_attention",
|
| 47 |
+
"linear_attention",
|
| 48 |
+
"linear_attention",
|
| 49 |
+
"full_attention"
|
| 50 |
+
],
|
| 51 |
+
"linear_conv_kernel_dim": 4,
|
| 52 |
+
"linear_key_head_dim": 128,
|
| 53 |
+
"linear_num_key_heads": 16,
|
| 54 |
+
"linear_num_value_heads": 32,
|
| 55 |
+
"linear_value_head_dim": 128,
|
| 56 |
+
"mamba_ssm_dtype": "float32",
|
| 57 |
+
"max_position_embeddings": 262144,
|
| 58 |
+
"mlp_only_layers": [],
|
| 59 |
+
"model_type": "qwen3_5_text",
|
| 60 |
+
"mtp_num_hidden_layers": 1,
|
| 61 |
+
"mtp_use_dedicated_embeddings": false,
|
| 62 |
+
"num_attention_heads": 16,
|
| 63 |
+
"num_hidden_layers": 32,
|
| 64 |
+
"num_key_value_heads": 4,
|
| 65 |
+
"pad_token_id": 248044,
|
| 66 |
+
"partial_rotary_factor": 0.25,
|
| 67 |
+
"rms_norm_eps": 1e-06,
|
| 68 |
+
"rope_parameters": {
|
| 69 |
+
"mrope_interleaved": true,
|
| 70 |
+
"mrope_section": [
|
| 71 |
+
11,
|
| 72 |
+
11,
|
| 73 |
+
10
|
| 74 |
+
],
|
| 75 |
+
"partial_rotary_factor": 0.25,
|
| 76 |
+
"rope_theta": 10000000,
|
| 77 |
+
"rope_type": "default"
|
| 78 |
+
},
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"transformers_version": "5.6.2",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 248320
|
| 83 |
+
}
|
Keen-qwen4b-20b/checkpoint-20218/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 248044,
|
| 4 |
+
"output_attentions": false,
|
| 5 |
+
"output_hidden_states": false,
|
| 6 |
+
"pad_token_id": 248044,
|
| 7 |
+
"transformers_version": "5.6.2",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
Keen-qwen4b-20b/checkpoint-20218/tokenizer_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"audio_bos_token": "<|audio_start|>",
|
| 4 |
+
"audio_eos_token": "<|audio_end|>",
|
| 5 |
+
"audio_token": "<|audio_pad|>",
|
| 6 |
+
"backend": "tokenizers",
|
| 7 |
+
"bos_token": null,
|
| 8 |
+
"clean_up_tokenization_spaces": false,
|
| 9 |
+
"eos_token": "<|endoftext|>",
|
| 10 |
+
"errors": "replace",
|
| 11 |
+
"extra_special_tokens": [
|
| 12 |
+
"<|im_start|>",
|
| 13 |
+
"<|im_end|>",
|
| 14 |
+
"<|object_ref_start|>",
|
| 15 |
+
"<|object_ref_end|>",
|
| 16 |
+
"<|box_start|>",
|
| 17 |
+
"<|box_end|>",
|
| 18 |
+
"<|quad_start|>",
|
| 19 |
+
"<|quad_end|>",
|
| 20 |
+
"<|vision_start|>",
|
| 21 |
+
"<|vision_end|>",
|
| 22 |
+
"<|vision_pad|>",
|
| 23 |
+
"<|image_pad|>",
|
| 24 |
+
"<|video_pad|>",
|
| 25 |
+
"<fact>",
|
| 26 |
+
"</fact>"
|
| 27 |
+
],
|
| 28 |
+
"image_token": "<|image_pad|>",
|
| 29 |
+
"model_max_length": 32768,
|
| 30 |
+
"model_specific_special_tokens": {
|
| 31 |
+
"audio_bos_token": "<|audio_start|>",
|
| 32 |
+
"audio_eos_token": "<|audio_end|>",
|
| 33 |
+
"audio_token": "<|audio_pad|>",
|
| 34 |
+
"image_token": "<|image_pad|>",
|
| 35 |
+
"video_token": "<|video_pad|>",
|
| 36 |
+
"vision_bos_token": "<|vision_start|>",
|
| 37 |
+
"vision_eos_token": "<|vision_end|>"
|
| 38 |
+
},
|
| 39 |
+
"pad_token": "<|endoftext|>",
|
| 40 |
+
"pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
| 41 |
+
"split_special_tokens": false,
|
| 42 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 43 |
+
"unk_token": null,
|
| 44 |
+
"video_token": "<|video_pad|>",
|
| 45 |
+
"vision_bos_token": "<|vision_start|>",
|
| 46 |
+
"vision_eos_token": "<|vision_end|>"
|
| 47 |
+
}
|
Keen-qwen4b-20b/checkpoint-22478/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Keen-qwen4b-20b: checkpoint-22478
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 22478
|
| 10 |
+
- Cumulative body tokens: 20,000,557,588
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Keen conditions passage reconstruction on `<fact>entity list</fact>`; the prefix participates in attention but is excluded from the training loss.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Keen-qwen4b-20b/checkpoint-22478"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("<fact>Paris, France</fact>\nParis", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Keen-qwen4b-20b/checkpoint-22478/config.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"attn_output_gate": true,
|
| 8 |
+
"bos_token_id": null,
|
| 9 |
+
"dtype": "bfloat16",
|
| 10 |
+
"eos_token_id": 248044,
|
| 11 |
+
"full_attention_interval": 4,
|
| 12 |
+
"head_dim": 256,
|
| 13 |
+
"hidden_act": "silu",
|
| 14 |
+
"hidden_size": 2560,
|
| 15 |
+
"initializer_range": 0.02,
|
| 16 |
+
"intermediate_size": 9216,
|
| 17 |
+
"layer_types": [
|
| 18 |
+
"linear_attention",
|
| 19 |
+
"linear_attention",
|
| 20 |
+
"linear_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"linear_attention",
|
| 23 |
+
"linear_attention",
|
| 24 |
+
"linear_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"linear_attention",
|
| 27 |
+
"linear_attention",
|
| 28 |
+
"linear_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"linear_attention",
|
| 31 |
+
"linear_attention",
|
| 32 |
+
"linear_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"linear_attention",
|
| 35 |
+
"linear_attention",
|
| 36 |
+
"linear_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"linear_attention",
|
| 39 |
+
"linear_attention",
|
| 40 |
+
"linear_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"linear_attention",
|
| 43 |
+
"linear_attention",
|
| 44 |
+
"linear_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"linear_attention",
|
| 47 |
+
"linear_attention",
|
| 48 |
+
"linear_attention",
|
| 49 |
+
"full_attention"
|
| 50 |
+
],
|
| 51 |
+
"linear_conv_kernel_dim": 4,
|
| 52 |
+
"linear_key_head_dim": 128,
|
| 53 |
+
"linear_num_key_heads": 16,
|
| 54 |
+
"linear_num_value_heads": 32,
|
| 55 |
+
"linear_value_head_dim": 128,
|
| 56 |
+
"mamba_ssm_dtype": "float32",
|
| 57 |
+
"max_position_embeddings": 262144,
|
| 58 |
+
"mlp_only_layers": [],
|
| 59 |
+
"model_type": "qwen3_5_text",
|
| 60 |
+
"mtp_num_hidden_layers": 1,
|
| 61 |
+
"mtp_use_dedicated_embeddings": false,
|
| 62 |
+
"num_attention_heads": 16,
|
| 63 |
+
"num_hidden_layers": 32,
|
| 64 |
+
"num_key_value_heads": 4,
|
| 65 |
+
"pad_token_id": 248044,
|
| 66 |
+
"partial_rotary_factor": 0.25,
|
| 67 |
+
"rms_norm_eps": 1e-06,
|
| 68 |
+
"rope_parameters": {
|
| 69 |
+
"mrope_interleaved": true,
|
| 70 |
+
"mrope_section": [
|
| 71 |
+
11,
|
| 72 |
+
11,
|
| 73 |
+
10
|
| 74 |
+
],
|
| 75 |
+
"partial_rotary_factor": 0.25,
|
| 76 |
+
"rope_theta": 10000000,
|
| 77 |
+
"rope_type": "default"
|
| 78 |
+
},
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"transformers_version": "5.6.2",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 248320
|
| 83 |
+
}
|
Keen-qwen4b-20b/checkpoint-22478/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 248044,
|
| 4 |
+
"output_attentions": false,
|
| 5 |
+
"output_hidden_states": false,
|
| 6 |
+
"pad_token_id": 248044,
|
| 7 |
+
"transformers_version": "5.6.2",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
Keen-qwen4b-20b/checkpoint-22478/tokenizer_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"audio_bos_token": "<|audio_start|>",
|
| 4 |
+
"audio_eos_token": "<|audio_end|>",
|
| 5 |
+
"audio_token": "<|audio_pad|>",
|
| 6 |
+
"backend": "tokenizers",
|
| 7 |
+
"bos_token": null,
|
| 8 |
+
"clean_up_tokenization_spaces": false,
|
| 9 |
+
"eos_token": "<|endoftext|>",
|
| 10 |
+
"errors": "replace",
|
| 11 |
+
"extra_special_tokens": [
|
| 12 |
+
"<|im_start|>",
|
| 13 |
+
"<|im_end|>",
|
| 14 |
+
"<|object_ref_start|>",
|
| 15 |
+
"<|object_ref_end|>",
|
| 16 |
+
"<|box_start|>",
|
| 17 |
+
"<|box_end|>",
|
| 18 |
+
"<|quad_start|>",
|
| 19 |
+
"<|quad_end|>",
|
| 20 |
+
"<|vision_start|>",
|
| 21 |
+
"<|vision_end|>",
|
| 22 |
+
"<|vision_pad|>",
|
| 23 |
+
"<|image_pad|>",
|
| 24 |
+
"<|video_pad|>",
|
| 25 |
+
"<fact>",
|
| 26 |
+
"</fact>"
|
| 27 |
+
],
|
| 28 |
+
"image_token": "<|image_pad|>",
|
| 29 |
+
"model_max_length": 32768,
|
| 30 |
+
"model_specific_special_tokens": {
|
| 31 |
+
"audio_bos_token": "<|audio_start|>",
|
| 32 |
+
"audio_eos_token": "<|audio_end|>",
|
| 33 |
+
"audio_token": "<|audio_pad|>",
|
| 34 |
+
"image_token": "<|image_pad|>",
|
| 35 |
+
"video_token": "<|video_pad|>",
|
| 36 |
+
"vision_bos_token": "<|vision_start|>",
|
| 37 |
+
"vision_eos_token": "<|vision_end|>"
|
| 38 |
+
},
|
| 39 |
+
"pad_token": "<|endoftext|>",
|
| 40 |
+
"pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
| 41 |
+
"split_special_tokens": false,
|
| 42 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 43 |
+
"unk_token": null,
|
| 44 |
+
"video_token": "<|video_pad|>",
|
| 45 |
+
"vision_bos_token": "<|vision_start|>",
|
| 46 |
+
"vision_eos_token": "<|vision_end|>"
|
| 47 |
+
}
|
Keen-qwen4b-20b/checkpoint-2253/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Keen-qwen4b-20b: checkpoint-2253
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 2253
|
| 10 |
+
- Cumulative body tokens: 2,000,696,711
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Keen conditions passage reconstruction on `<fact>entity list</fact>`; the prefix participates in attention but is excluded from the training loss.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Keen-qwen4b-20b/checkpoint-2253"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("<fact>Paris, France</fact>\nParis", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Keen-qwen4b-20b/checkpoint-2253/config.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"attn_output_gate": true,
|
| 8 |
+
"bos_token_id": null,
|
| 9 |
+
"dtype": "bfloat16",
|
| 10 |
+
"eos_token_id": 248044,
|
| 11 |
+
"full_attention_interval": 4,
|
| 12 |
+
"head_dim": 256,
|
| 13 |
+
"hidden_act": "silu",
|
| 14 |
+
"hidden_size": 2560,
|
| 15 |
+
"initializer_range": 0.02,
|
| 16 |
+
"intermediate_size": 9216,
|
| 17 |
+
"layer_types": [
|
| 18 |
+
"linear_attention",
|
| 19 |
+
"linear_attention",
|
| 20 |
+
"linear_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"linear_attention",
|
| 23 |
+
"linear_attention",
|
| 24 |
+
"linear_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"linear_attention",
|
| 27 |
+
"linear_attention",
|
| 28 |
+
"linear_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"linear_attention",
|
| 31 |
+
"linear_attention",
|
| 32 |
+
"linear_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"linear_attention",
|
| 35 |
+
"linear_attention",
|
| 36 |
+
"linear_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"linear_attention",
|
| 39 |
+
"linear_attention",
|
| 40 |
+
"linear_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"linear_attention",
|
| 43 |
+
"linear_attention",
|
| 44 |
+
"linear_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"linear_attention",
|
| 47 |
+
"linear_attention",
|
| 48 |
+
"linear_attention",
|
| 49 |
+
"full_attention"
|
| 50 |
+
],
|
| 51 |
+
"linear_conv_kernel_dim": 4,
|
| 52 |
+
"linear_key_head_dim": 128,
|
| 53 |
+
"linear_num_key_heads": 16,
|
| 54 |
+
"linear_num_value_heads": 32,
|
| 55 |
+
"linear_value_head_dim": 128,
|
| 56 |
+
"mamba_ssm_dtype": "float32",
|
| 57 |
+
"max_position_embeddings": 262144,
|
| 58 |
+
"mlp_only_layers": [],
|
| 59 |
+
"model_type": "qwen3_5_text",
|
| 60 |
+
"mtp_num_hidden_layers": 1,
|
| 61 |
+
"mtp_use_dedicated_embeddings": false,
|
| 62 |
+
"num_attention_heads": 16,
|
| 63 |
+
"num_hidden_layers": 32,
|
| 64 |
+
"num_key_value_heads": 4,
|
| 65 |
+
"pad_token_id": 248044,
|
| 66 |
+
"partial_rotary_factor": 0.25,
|
| 67 |
+
"rms_norm_eps": 1e-06,
|
| 68 |
+
"rope_parameters": {
|
| 69 |
+
"mrope_interleaved": true,
|
| 70 |
+
"mrope_section": [
|
| 71 |
+
11,
|
| 72 |
+
11,
|
| 73 |
+
10
|
| 74 |
+
],
|
| 75 |
+
"partial_rotary_factor": 0.25,
|
| 76 |
+
"rope_theta": 10000000,
|
| 77 |
+
"rope_type": "default"
|
| 78 |
+
},
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"transformers_version": "5.6.2",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 248320
|
| 83 |
+
}
|
Keen-qwen4b-20b/checkpoint-2253/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 248044,
|
| 4 |
+
"output_attentions": false,
|
| 5 |
+
"output_hidden_states": false,
|
| 6 |
+
"pad_token_id": 248044,
|
| 7 |
+
"transformers_version": "5.6.2",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
Keen-qwen4b-20b/checkpoint-2253/tokenizer_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"audio_bos_token": "<|audio_start|>",
|
| 4 |
+
"audio_eos_token": "<|audio_end|>",
|
| 5 |
+
"audio_token": "<|audio_pad|>",
|
| 6 |
+
"backend": "tokenizers",
|
| 7 |
+
"bos_token": null,
|
| 8 |
+
"clean_up_tokenization_spaces": false,
|
| 9 |
+
"eos_token": "<|endoftext|>",
|
| 10 |
+
"errors": "replace",
|
| 11 |
+
"extra_special_tokens": [
|
| 12 |
+
"<|im_start|>",
|
| 13 |
+
"<|im_end|>",
|
| 14 |
+
"<|object_ref_start|>",
|
| 15 |
+
"<|object_ref_end|>",
|
| 16 |
+
"<|box_start|>",
|
| 17 |
+
"<|box_end|>",
|
| 18 |
+
"<|quad_start|>",
|
| 19 |
+
"<|quad_end|>",
|
| 20 |
+
"<|vision_start|>",
|
| 21 |
+
"<|vision_end|>",
|
| 22 |
+
"<|vision_pad|>",
|
| 23 |
+
"<|image_pad|>",
|
| 24 |
+
"<|video_pad|>",
|
| 25 |
+
"<fact>",
|
| 26 |
+
"</fact>"
|
| 27 |
+
],
|
| 28 |
+
"image_token": "<|image_pad|>",
|
| 29 |
+
"model_max_length": 32768,
|
| 30 |
+
"model_specific_special_tokens": {
|
| 31 |
+
"audio_bos_token": "<|audio_start|>",
|
| 32 |
+
"audio_eos_token": "<|audio_end|>",
|
| 33 |
+
"audio_token": "<|audio_pad|>",
|
| 34 |
+
"image_token": "<|image_pad|>",
|
| 35 |
+
"video_token": "<|video_pad|>",
|
| 36 |
+
"vision_bos_token": "<|vision_start|>",
|
| 37 |
+
"vision_eos_token": "<|vision_end|>"
|
| 38 |
+
},
|
| 39 |
+
"pad_token": "<|endoftext|>",
|
| 40 |
+
"pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
| 41 |
+
"split_special_tokens": false,
|
| 42 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 43 |
+
"unk_token": null,
|
| 44 |
+
"video_token": "<|video_pad|>",
|
| 45 |
+
"vision_bos_token": "<|vision_start|>",
|
| 46 |
+
"vision_eos_token": "<|vision_end|>"
|
| 47 |
+
}
|
Keen-qwen4b-20b/checkpoint-4453/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Keen-qwen4b-20b: checkpoint-4453
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 4453
|
| 10 |
+
- Cumulative body tokens: 4,000,757,038
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Keen conditions passage reconstruction on `<fact>entity list</fact>`; the prefix participates in attention but is excluded from the training loss.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Keen-qwen4b-20b/checkpoint-4453"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("<fact>Paris, France</fact>\nParis", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Keen-qwen4b-20b/checkpoint-4453/config.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"attn_output_gate": true,
|
| 8 |
+
"bos_token_id": null,
|
| 9 |
+
"dtype": "bfloat16",
|
| 10 |
+
"eos_token_id": 248044,
|
| 11 |
+
"full_attention_interval": 4,
|
| 12 |
+
"head_dim": 256,
|
| 13 |
+
"hidden_act": "silu",
|
| 14 |
+
"hidden_size": 2560,
|
| 15 |
+
"initializer_range": 0.02,
|
| 16 |
+
"intermediate_size": 9216,
|
| 17 |
+
"layer_types": [
|
| 18 |
+
"linear_attention",
|
| 19 |
+
"linear_attention",
|
| 20 |
+
"linear_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"linear_attention",
|
| 23 |
+
"linear_attention",
|
| 24 |
+
"linear_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"linear_attention",
|
| 27 |
+
"linear_attention",
|
| 28 |
+
"linear_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"linear_attention",
|
| 31 |
+
"linear_attention",
|
| 32 |
+
"linear_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"linear_attention",
|
| 35 |
+
"linear_attention",
|
| 36 |
+
"linear_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"linear_attention",
|
| 39 |
+
"linear_attention",
|
| 40 |
+
"linear_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"linear_attention",
|
| 43 |
+
"linear_attention",
|
| 44 |
+
"linear_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"linear_attention",
|
| 47 |
+
"linear_attention",
|
| 48 |
+
"linear_attention",
|
| 49 |
+
"full_attention"
|
| 50 |
+
],
|
| 51 |
+
"linear_conv_kernel_dim": 4,
|
| 52 |
+
"linear_key_head_dim": 128,
|
| 53 |
+
"linear_num_key_heads": 16,
|
| 54 |
+
"linear_num_value_heads": 32,
|
| 55 |
+
"linear_value_head_dim": 128,
|
| 56 |
+
"mamba_ssm_dtype": "float32",
|
| 57 |
+
"max_position_embeddings": 262144,
|
| 58 |
+
"mlp_only_layers": [],
|
| 59 |
+
"model_type": "qwen3_5_text",
|
| 60 |
+
"mtp_num_hidden_layers": 1,
|
| 61 |
+
"mtp_use_dedicated_embeddings": false,
|
| 62 |
+
"num_attention_heads": 16,
|
| 63 |
+
"num_hidden_layers": 32,
|
| 64 |
+
"num_key_value_heads": 4,
|
| 65 |
+
"pad_token_id": 248044,
|
| 66 |
+
"partial_rotary_factor": 0.25,
|
| 67 |
+
"rms_norm_eps": 1e-06,
|
| 68 |
+
"rope_parameters": {
|
| 69 |
+
"mrope_interleaved": true,
|
| 70 |
+
"mrope_section": [
|
| 71 |
+
11,
|
| 72 |
+
11,
|
| 73 |
+
10
|
| 74 |
+
],
|
| 75 |
+
"partial_rotary_factor": 0.25,
|
| 76 |
+
"rope_theta": 10000000,
|
| 77 |
+
"rope_type": "default"
|
| 78 |
+
},
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"transformers_version": "5.6.2",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 248320
|
| 83 |
+
}
|
Keen-qwen4b-20b/checkpoint-4453/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 248044,
|
| 4 |
+
"output_attentions": false,
|
| 5 |
+
"output_hidden_states": false,
|
| 6 |
+
"pad_token_id": 248044,
|
| 7 |
+
"transformers_version": "5.6.2",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
Keen-qwen4b-20b/checkpoint-4453/tokenizer_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"audio_bos_token": "<|audio_start|>",
|
| 4 |
+
"audio_eos_token": "<|audio_end|>",
|
| 5 |
+
"audio_token": "<|audio_pad|>",
|
| 6 |
+
"backend": "tokenizers",
|
| 7 |
+
"bos_token": null,
|
| 8 |
+
"clean_up_tokenization_spaces": false,
|
| 9 |
+
"eos_token": "<|endoftext|>",
|
| 10 |
+
"errors": "replace",
|
| 11 |
+
"extra_special_tokens": [
|
| 12 |
+
"<|im_start|>",
|
| 13 |
+
"<|im_end|>",
|
| 14 |
+
"<|object_ref_start|>",
|
| 15 |
+
"<|object_ref_end|>",
|
| 16 |
+
"<|box_start|>",
|
| 17 |
+
"<|box_end|>",
|
| 18 |
+
"<|quad_start|>",
|
| 19 |
+
"<|quad_end|>",
|
| 20 |
+
"<|vision_start|>",
|
| 21 |
+
"<|vision_end|>",
|
| 22 |
+
"<|vision_pad|>",
|
| 23 |
+
"<|image_pad|>",
|
| 24 |
+
"<|video_pad|>",
|
| 25 |
+
"<fact>",
|
| 26 |
+
"</fact>"
|
| 27 |
+
],
|
| 28 |
+
"image_token": "<|image_pad|>",
|
| 29 |
+
"model_max_length": 32768,
|
| 30 |
+
"model_specific_special_tokens": {
|
| 31 |
+
"audio_bos_token": "<|audio_start|>",
|
| 32 |
+
"audio_eos_token": "<|audio_end|>",
|
| 33 |
+
"audio_token": "<|audio_pad|>",
|
| 34 |
+
"image_token": "<|image_pad|>",
|
| 35 |
+
"video_token": "<|video_pad|>",
|
| 36 |
+
"vision_bos_token": "<|vision_start|>",
|
| 37 |
+
"vision_eos_token": "<|vision_end|>"
|
| 38 |
+
},
|
| 39 |
+
"pad_token": "<|endoftext|>",
|
| 40 |
+
"pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
| 41 |
+
"split_special_tokens": false,
|
| 42 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 43 |
+
"unk_token": null,
|
| 44 |
+
"video_token": "<|video_pad|>",
|
| 45 |
+
"vision_bos_token": "<|vision_start|>",
|
| 46 |
+
"vision_eos_token": "<|vision_end|>"
|
| 47 |
+
}
|
Keen-qwen4b-20b/checkpoint-6701/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Keen-qwen4b-20b: checkpoint-6701
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 6701
|
| 10 |
+
- Cumulative body tokens: 6,000,082,648
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Keen conditions passage reconstruction on `<fact>entity list</fact>`; the prefix participates in attention but is excluded from the training loss.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Keen-qwen4b-20b/checkpoint-6701"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("<fact>Paris, France</fact>\nParis", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Keen-qwen4b-20b/checkpoint-6701/config.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"attn_output_gate": true,
|
| 8 |
+
"bos_token_id": null,
|
| 9 |
+
"dtype": "bfloat16",
|
| 10 |
+
"eos_token_id": 248044,
|
| 11 |
+
"full_attention_interval": 4,
|
| 12 |
+
"head_dim": 256,
|
| 13 |
+
"hidden_act": "silu",
|
| 14 |
+
"hidden_size": 2560,
|
| 15 |
+
"initializer_range": 0.02,
|
| 16 |
+
"intermediate_size": 9216,
|
| 17 |
+
"layer_types": [
|
| 18 |
+
"linear_attention",
|
| 19 |
+
"linear_attention",
|
| 20 |
+
"linear_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"linear_attention",
|
| 23 |
+
"linear_attention",
|
| 24 |
+
"linear_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"linear_attention",
|
| 27 |
+
"linear_attention",
|
| 28 |
+
"linear_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"linear_attention",
|
| 31 |
+
"linear_attention",
|
| 32 |
+
"linear_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"linear_attention",
|
| 35 |
+
"linear_attention",
|
| 36 |
+
"linear_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"linear_attention",
|
| 39 |
+
"linear_attention",
|
| 40 |
+
"linear_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"linear_attention",
|
| 43 |
+
"linear_attention",
|
| 44 |
+
"linear_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"linear_attention",
|
| 47 |
+
"linear_attention",
|
| 48 |
+
"linear_attention",
|
| 49 |
+
"full_attention"
|
| 50 |
+
],
|
| 51 |
+
"linear_conv_kernel_dim": 4,
|
| 52 |
+
"linear_key_head_dim": 128,
|
| 53 |
+
"linear_num_key_heads": 16,
|
| 54 |
+
"linear_num_value_heads": 32,
|
| 55 |
+
"linear_value_head_dim": 128,
|
| 56 |
+
"mamba_ssm_dtype": "float32",
|
| 57 |
+
"max_position_embeddings": 262144,
|
| 58 |
+
"mlp_only_layers": [],
|
| 59 |
+
"model_type": "qwen3_5_text",
|
| 60 |
+
"mtp_num_hidden_layers": 1,
|
| 61 |
+
"mtp_use_dedicated_embeddings": false,
|
| 62 |
+
"num_attention_heads": 16,
|
| 63 |
+
"num_hidden_layers": 32,
|
| 64 |
+
"num_key_value_heads": 4,
|
| 65 |
+
"pad_token_id": 248044,
|
| 66 |
+
"partial_rotary_factor": 0.25,
|
| 67 |
+
"rms_norm_eps": 1e-06,
|
| 68 |
+
"rope_parameters": {
|
| 69 |
+
"mrope_interleaved": true,
|
| 70 |
+
"mrope_section": [
|
| 71 |
+
11,
|
| 72 |
+
11,
|
| 73 |
+
10
|
| 74 |
+
],
|
| 75 |
+
"partial_rotary_factor": 0.25,
|
| 76 |
+
"rope_theta": 10000000,
|
| 77 |
+
"rope_type": "default"
|
| 78 |
+
},
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"transformers_version": "5.6.2",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 248320
|
| 83 |
+
}
|
Keen-qwen4b-20b/checkpoint-6701/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 248044,
|
| 4 |
+
"output_attentions": false,
|
| 5 |
+
"output_hidden_states": false,
|
| 6 |
+
"pad_token_id": 248044,
|
| 7 |
+
"transformers_version": "5.6.2",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
Keen-qwen4b-20b/checkpoint-6701/tokenizer_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"audio_bos_token": "<|audio_start|>",
|
| 4 |
+
"audio_eos_token": "<|audio_end|>",
|
| 5 |
+
"audio_token": "<|audio_pad|>",
|
| 6 |
+
"backend": "tokenizers",
|
| 7 |
+
"bos_token": null,
|
| 8 |
+
"clean_up_tokenization_spaces": false,
|
| 9 |
+
"eos_token": "<|endoftext|>",
|
| 10 |
+
"errors": "replace",
|
| 11 |
+
"extra_special_tokens": [
|
| 12 |
+
"<|im_start|>",
|
| 13 |
+
"<|im_end|>",
|
| 14 |
+
"<|object_ref_start|>",
|
| 15 |
+
"<|object_ref_end|>",
|
| 16 |
+
"<|box_start|>",
|
| 17 |
+
"<|box_end|>",
|
| 18 |
+
"<|quad_start|>",
|
| 19 |
+
"<|quad_end|>",
|
| 20 |
+
"<|vision_start|>",
|
| 21 |
+
"<|vision_end|>",
|
| 22 |
+
"<|vision_pad|>",
|
| 23 |
+
"<|image_pad|>",
|
| 24 |
+
"<|video_pad|>",
|
| 25 |
+
"<fact>",
|
| 26 |
+
"</fact>"
|
| 27 |
+
],
|
| 28 |
+
"image_token": "<|image_pad|>",
|
| 29 |
+
"model_max_length": 32768,
|
| 30 |
+
"model_specific_special_tokens": {
|
| 31 |
+
"audio_bos_token": "<|audio_start|>",
|
| 32 |
+
"audio_eos_token": "<|audio_end|>",
|
| 33 |
+
"audio_token": "<|audio_pad|>",
|
| 34 |
+
"image_token": "<|image_pad|>",
|
| 35 |
+
"video_token": "<|video_pad|>",
|
| 36 |
+
"vision_bos_token": "<|vision_start|>",
|
| 37 |
+
"vision_eos_token": "<|vision_end|>"
|
| 38 |
+
},
|
| 39 |
+
"pad_token": "<|endoftext|>",
|
| 40 |
+
"pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
| 41 |
+
"split_special_tokens": false,
|
| 42 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 43 |
+
"unk_token": null,
|
| 44 |
+
"video_token": "<|video_pad|>",
|
| 45 |
+
"vision_bos_token": "<|vision_start|>",
|
| 46 |
+
"vision_eos_token": "<|vision_end|>"
|
| 47 |
+
}
|
Keen-qwen4b-20b/checkpoint-8956/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Keen-qwen4b-20b: checkpoint-8956
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 8956
|
| 10 |
+
- Cumulative body tokens: 8,000,600,622
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Keen conditions passage reconstruction on `<fact>entity list</fact>`; the prefix participates in attention but is excluded from the training loss.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Keen-qwen4b-20b/checkpoint-8956"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("<fact>Paris, France</fact>\nParis", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Keen-qwen4b-20b/checkpoint-8956/config.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"attn_output_gate": true,
|
| 8 |
+
"bos_token_id": null,
|
| 9 |
+
"dtype": "bfloat16",
|
| 10 |
+
"eos_token_id": 248044,
|
| 11 |
+
"full_attention_interval": 4,
|
| 12 |
+
"head_dim": 256,
|
| 13 |
+
"hidden_act": "silu",
|
| 14 |
+
"hidden_size": 2560,
|
| 15 |
+
"initializer_range": 0.02,
|
| 16 |
+
"intermediate_size": 9216,
|
| 17 |
+
"layer_types": [
|
| 18 |
+
"linear_attention",
|
| 19 |
+
"linear_attention",
|
| 20 |
+
"linear_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"linear_attention",
|
| 23 |
+
"linear_attention",
|
| 24 |
+
"linear_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"linear_attention",
|
| 27 |
+
"linear_attention",
|
| 28 |
+
"linear_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"linear_attention",
|
| 31 |
+
"linear_attention",
|
| 32 |
+
"linear_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"linear_attention",
|
| 35 |
+
"linear_attention",
|
| 36 |
+
"linear_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"linear_attention",
|
| 39 |
+
"linear_attention",
|
| 40 |
+
"linear_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"linear_attention",
|
| 43 |
+
"linear_attention",
|
| 44 |
+
"linear_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"linear_attention",
|
| 47 |
+
"linear_attention",
|
| 48 |
+
"linear_attention",
|
| 49 |
+
"full_attention"
|
| 50 |
+
],
|
| 51 |
+
"linear_conv_kernel_dim": 4,
|
| 52 |
+
"linear_key_head_dim": 128,
|
| 53 |
+
"linear_num_key_heads": 16,
|
| 54 |
+
"linear_num_value_heads": 32,
|
| 55 |
+
"linear_value_head_dim": 128,
|
| 56 |
+
"mamba_ssm_dtype": "float32",
|
| 57 |
+
"max_position_embeddings": 262144,
|
| 58 |
+
"mlp_only_layers": [],
|
| 59 |
+
"model_type": "qwen3_5_text",
|
| 60 |
+
"mtp_num_hidden_layers": 1,
|
| 61 |
+
"mtp_use_dedicated_embeddings": false,
|
| 62 |
+
"num_attention_heads": 16,
|
| 63 |
+
"num_hidden_layers": 32,
|
| 64 |
+
"num_key_value_heads": 4,
|
| 65 |
+
"pad_token_id": 248044,
|
| 66 |
+
"partial_rotary_factor": 0.25,
|
| 67 |
+
"rms_norm_eps": 1e-06,
|
| 68 |
+
"rope_parameters": {
|
| 69 |
+
"mrope_interleaved": true,
|
| 70 |
+
"mrope_section": [
|
| 71 |
+
11,
|
| 72 |
+
11,
|
| 73 |
+
10
|
| 74 |
+
],
|
| 75 |
+
"partial_rotary_factor": 0.25,
|
| 76 |
+
"rope_theta": 10000000,
|
| 77 |
+
"rope_type": "default"
|
| 78 |
+
},
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"transformers_version": "5.6.2",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 248320
|
| 83 |
+
}
|
Keen-qwen4b-20b/checkpoint-8956/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 248044,
|
| 4 |
+
"output_attentions": false,
|
| 5 |
+
"output_hidden_states": false,
|
| 6 |
+
"pad_token_id": 248044,
|
| 7 |
+
"transformers_version": "5.6.2",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
Keen-qwen4b-20b/checkpoint-8956/tokenizer_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"audio_bos_token": "<|audio_start|>",
|
| 4 |
+
"audio_eos_token": "<|audio_end|>",
|
| 5 |
+
"audio_token": "<|audio_pad|>",
|
| 6 |
+
"backend": "tokenizers",
|
| 7 |
+
"bos_token": null,
|
| 8 |
+
"clean_up_tokenization_spaces": false,
|
| 9 |
+
"eos_token": "<|endoftext|>",
|
| 10 |
+
"errors": "replace",
|
| 11 |
+
"extra_special_tokens": [
|
| 12 |
+
"<|im_start|>",
|
| 13 |
+
"<|im_end|>",
|
| 14 |
+
"<|object_ref_start|>",
|
| 15 |
+
"<|object_ref_end|>",
|
| 16 |
+
"<|box_start|>",
|
| 17 |
+
"<|box_end|>",
|
| 18 |
+
"<|quad_start|>",
|
| 19 |
+
"<|quad_end|>",
|
| 20 |
+
"<|vision_start|>",
|
| 21 |
+
"<|vision_end|>",
|
| 22 |
+
"<|vision_pad|>",
|
| 23 |
+
"<|image_pad|>",
|
| 24 |
+
"<|video_pad|>",
|
| 25 |
+
"<fact>",
|
| 26 |
+
"</fact>"
|
| 27 |
+
],
|
| 28 |
+
"image_token": "<|image_pad|>",
|
| 29 |
+
"model_max_length": 32768,
|
| 30 |
+
"model_specific_special_tokens": {
|
| 31 |
+
"audio_bos_token": "<|audio_start|>",
|
| 32 |
+
"audio_eos_token": "<|audio_end|>",
|
| 33 |
+
"audio_token": "<|audio_pad|>",
|
| 34 |
+
"image_token": "<|image_pad|>",
|
| 35 |
+
"video_token": "<|video_pad|>",
|
| 36 |
+
"vision_bos_token": "<|vision_start|>",
|
| 37 |
+
"vision_eos_token": "<|vision_end|>"
|
| 38 |
+
},
|
| 39 |
+
"pad_token": "<|endoftext|>",
|
| 40 |
+
"pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
| 41 |
+
"split_special_tokens": false,
|
| 42 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 43 |
+
"unk_token": null,
|
| 44 |
+
"video_token": "<|video_pad|>",
|
| 45 |
+
"vision_bos_token": "<|vision_start|>",
|
| 46 |
+
"vision_eos_token": "<|vision_end|>"
|
| 47 |
+
}
|
README.md
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
pipeline_tag: text-generation
|
| 4 |
+
language:
|
| 5 |
+
- en
|
| 6 |
+
tags:
|
| 7 |
+
- qwen3_5
|
| 8 |
+
- pretraining
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# AnonymousModel-mht
|
| 12 |
+
|
| 13 |
+
Inference-only release of 20 budget-boundary checkpoints from two matched,
|
| 14 |
+
randomly initialized Qwen3.5-4B text-model pretraining runs on approximately
|
| 15 |
+
20B body tokens. Each family contains ten approximately 2B-spaced checkpoints
|
| 16 |
+
from 2B through 20B. The additional final checkpoints at approximately 20.156B
|
| 17 |
+
tokens are intentionally excluded from this release.
|
| 18 |
+
|
| 19 |
+
- [Keen-qwen4b-20b](Keen-qwen4b-20b/README.md): factual-list-conditioned pretraining.
|
| 20 |
+
- [Vanilla-qwen4b-20b](Vanilla-qwen4b-20b/README.md): ordinary text pretraining.
|
| 21 |
+
|
| 22 |
+
The training context was 4,096 tokens. These are base models, not QA-SFT or
|
| 23 |
+
instruction-tuned models. Each checkpoint directory is independently loadable
|
| 24 |
+
with standard Transformers 5.6.2 and contains BF16 safetensors weights,
|
| 25 |
+
model/generation configuration and a tokenizer. Select the checkpoint using
|
| 26 |
+
`subfolder`; this repository root is a collection and has no root model weights.
|
| 27 |
+
Optimizer states, scheduler states and training logs are not distributed.
|
| 28 |
+
|
| 29 |
+
Checkpoint loading examples and exact token counts appear in the family and
|
| 30 |
+
checkpoint READMEs. `checkpoints.json` indexes the release. `manifest.json`
|
| 31 |
+
records artifact hashes. Configuration enables inference caching; the released
|
| 32 |
+
weights and tokenizer vocabulary are unchanged from their training checkpoints.
|
Vanilla-qwen4b-20b/README.md
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Vanilla-qwen4b-20b
|
| 2 |
+
|
| 3 |
+
Ten budget-boundary checkpoints, approximately every 2B body tokens from 2B through 20B. The additional final checkpoint at approximately 20.156B tokens is intentionally excluded.
|
| 4 |
+
|
| 5 |
+
| Checkpoint | Step | Body tokens | Status |
|
| 6 |
+
|---|---:|---:|---|
|
| 7 |
+
| [checkpoint-1910](checkpoint-1910) | 1,910 | 2,000,045,638 | Intermediate |
|
| 8 |
+
| [checkpoint-3818](checkpoint-3818) | 3,818 | 4,000,211,289 | Intermediate |
|
| 9 |
+
| [checkpoint-5726](checkpoint-5726) | 5,726 | 6,000,159,072 | Intermediate |
|
| 10 |
+
| [checkpoint-7635](checkpoint-7635) | 7,635 | 8,000,913,248 | Intermediate |
|
| 11 |
+
| [checkpoint-9543](checkpoint-9543) | 9,543 | 10,000,619,298 | Intermediate |
|
| 12 |
+
| [checkpoint-11452](checkpoint-11452) | 11,452 | 12,000,591,962 | Intermediate |
|
| 13 |
+
| [checkpoint-13362](checkpoint-13362) | 13,362 | 14,000,166,403 | Intermediate |
|
| 14 |
+
| [checkpoint-15273](checkpoint-15273) | 15,273 | 16,000,320,768 | Intermediate |
|
| 15 |
+
| [checkpoint-17188](checkpoint-17188) | 17,188 | 18,000,257,314 | Intermediate |
|
| 16 |
+
| [checkpoint-19106](checkpoint-19106) | 19,106 | 20,000,220,345 | Intermediate |
|
Vanilla-qwen4b-20b/checkpoint-1910/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Vanilla-qwen4b-20b: checkpoint-1910
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 1910
|
| 10 |
+
- Cumulative body tokens: 2,000,045,638
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Vanilla uses ordinary passage continuation without a facts prefix.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Vanilla-qwen4b-20b/checkpoint-1910"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("Paris", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Vanilla-qwen4b-20b/checkpoint-9543/README.md
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Vanilla-qwen4b-20b: checkpoint-9543
|
| 2 |
+
|
| 3 |
+
Text-only Qwen3.5-4B architecture, trained from random initialization.
|
| 4 |
+
This is a base pretraining checkpoint, not an instruction-tuned chat model.
|
| 5 |
+
Training used 4,096-token packs with document-isolated attention.
|
| 6 |
+
The larger positional limit in the architecture configuration does not imply
|
| 7 |
+
validated long-context capability beyond the training context.
|
| 8 |
+
|
| 9 |
+
- Optimizer step: 9543
|
| 10 |
+
- Cumulative body tokens: 10,000,619,298
|
| 11 |
+
- Parameters: 4,205,751,296
|
| 12 |
+
- Precision: BF16
|
| 13 |
+
- Special document-end token: `<|endoftext|>`
|
| 14 |
+
|
| 15 |
+
Vanilla uses ordinary passage continuation without a facts prefix.
|
| 16 |
+
|
| 17 |
+
Use Transformers 5.6.2 (the training/evaluation version) with its native
|
| 18 |
+
`Qwen3_5ForCausalLM` implementation. Install PyTorch and Accelerate for GPU loading.
|
| 19 |
+
No remote custom model code is required. Use plain text prompts rather than a
|
| 20 |
+
chat template. Inference caching is enabled in the released configuration;
|
| 21 |
+
weights and tokenizer vocabulary are unchanged.
|
| 22 |
+
|
| 23 |
+
```python
|
| 24 |
+
import torch
|
| 25 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 26 |
+
|
| 27 |
+
repo = "meihant/AnonymousModel-mht"
|
| 28 |
+
subfolder = "Vanilla-qwen4b-20b/checkpoint-9543"
|
| 29 |
+
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
|
| 30 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 31 |
+
repo, subfolder=subfolder, dtype=torch.bfloat16,
|
| 32 |
+
device_map="auto", attn_implementation="eager",
|
| 33 |
+
).eval()
|
| 34 |
+
inputs = tokenizer("Paris", return_tensors="pt").to(model.device)
|
| 35 |
+
with torch.inference_mode():
|
| 36 |
+
output = model.generate(**inputs, max_new_tokens=64, do_sample=False,
|
| 37 |
+
use_cache=True, pad_token_id=tokenizer.eos_token_id)
|
| 38 |
+
print(tokenizer.decode(output[0], skip_special_tokens=False))
|
| 39 |
+
```
|
| 40 |
+
|
Vanilla-qwen4b-20b/checkpoint-9543/config.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3_5ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"attn_output_gate": true,
|
| 8 |
+
"bos_token_id": null,
|
| 9 |
+
"dtype": "bfloat16",
|
| 10 |
+
"eos_token_id": 248044,
|
| 11 |
+
"full_attention_interval": 4,
|
| 12 |
+
"head_dim": 256,
|
| 13 |
+
"hidden_act": "silu",
|
| 14 |
+
"hidden_size": 2560,
|
| 15 |
+
"initializer_range": 0.02,
|
| 16 |
+
"intermediate_size": 9216,
|
| 17 |
+
"layer_types": [
|
| 18 |
+
"linear_attention",
|
| 19 |
+
"linear_attention",
|
| 20 |
+
"linear_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"linear_attention",
|
| 23 |
+
"linear_attention",
|
| 24 |
+
"linear_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"linear_attention",
|
| 27 |
+
"linear_attention",
|
| 28 |
+
"linear_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"linear_attention",
|
| 31 |
+
"linear_attention",
|
| 32 |
+
"linear_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"linear_attention",
|
| 35 |
+
"linear_attention",
|
| 36 |
+
"linear_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"linear_attention",
|
| 39 |
+
"linear_attention",
|
| 40 |
+
"linear_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"linear_attention",
|
| 43 |
+
"linear_attention",
|
| 44 |
+
"linear_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"linear_attention",
|
| 47 |
+
"linear_attention",
|
| 48 |
+
"linear_attention",
|
| 49 |
+
"full_attention"
|
| 50 |
+
],
|
| 51 |
+
"linear_conv_kernel_dim": 4,
|
| 52 |
+
"linear_key_head_dim": 128,
|
| 53 |
+
"linear_num_key_heads": 16,
|
| 54 |
+
"linear_num_value_heads": 32,
|
| 55 |
+
"linear_value_head_dim": 128,
|
| 56 |
+
"mamba_ssm_dtype": "float32",
|
| 57 |
+
"max_position_embeddings": 262144,
|
| 58 |
+
"mlp_only_layers": [],
|
| 59 |
+
"model_type": "qwen3_5_text",
|
| 60 |
+
"mtp_num_hidden_layers": 1,
|
| 61 |
+
"mtp_use_dedicated_embeddings": false,
|
| 62 |
+
"num_attention_heads": 16,
|
| 63 |
+
"num_hidden_layers": 32,
|
| 64 |
+
"num_key_value_heads": 4,
|
| 65 |
+
"pad_token_id": 248044,
|
| 66 |
+
"partial_rotary_factor": 0.25,
|
| 67 |
+
"rms_norm_eps": 1e-06,
|
| 68 |
+
"rope_parameters": {
|
| 69 |
+
"mrope_interleaved": true,
|
| 70 |
+
"mrope_section": [
|
| 71 |
+
11,
|
| 72 |
+
11,
|
| 73 |
+
10
|
| 74 |
+
],
|
| 75 |
+
"partial_rotary_factor": 0.25,
|
| 76 |
+
"rope_theta": 10000000,
|
| 77 |
+
"rope_type": "default"
|
| 78 |
+
},
|
| 79 |
+
"tie_word_embeddings": true,
|
| 80 |
+
"transformers_version": "5.6.2",
|
| 81 |
+
"use_cache": true,
|
| 82 |
+
"vocab_size": 248320
|
| 83 |
+
}
|
Vanilla-qwen4b-20b/checkpoint-9543/generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 248044,
|
| 4 |
+
"output_attentions": false,
|
| 5 |
+
"output_hidden_states": false,
|
| 6 |
+
"pad_token_id": 248044,
|
| 7 |
+
"transformers_version": "5.6.2",
|
| 8 |
+
"use_cache": true
|
| 9 |
+
}
|
Vanilla-qwen4b-20b/checkpoint-9543/tokenizer_config.json
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"audio_bos_token": "<|audio_start|>",
|
| 4 |
+
"audio_eos_token": "<|audio_end|>",
|
| 5 |
+
"audio_token": "<|audio_pad|>",
|
| 6 |
+
"backend": "tokenizers",
|
| 7 |
+
"bos_token": null,
|
| 8 |
+
"clean_up_tokenization_spaces": false,
|
| 9 |
+
"eos_token": "<|endoftext|>",
|
| 10 |
+
"errors": "replace",
|
| 11 |
+
"extra_special_tokens": [
|
| 12 |
+
"<|im_start|>",
|
| 13 |
+
"<|im_end|>",
|
| 14 |
+
"<|object_ref_start|>",
|
| 15 |
+
"<|object_ref_end|>",
|
| 16 |
+
"<|box_start|>",
|
| 17 |
+
"<|box_end|>",
|
| 18 |
+
"<|quad_start|>",
|
| 19 |
+
"<|quad_end|>",
|
| 20 |
+
"<|vision_start|>",
|
| 21 |
+
"<|vision_end|>",
|
| 22 |
+
"<|vision_pad|>",
|
| 23 |
+
"<|image_pad|>",
|
| 24 |
+
"<|video_pad|>",
|
| 25 |
+
"<fact>",
|
| 26 |
+
"</fact>"
|
| 27 |
+
],
|
| 28 |
+
"image_token": "<|image_pad|>",
|
| 29 |
+
"model_max_length": 32768,
|
| 30 |
+
"model_specific_special_tokens": {
|
| 31 |
+
"audio_bos_token": "<|audio_start|>",
|
| 32 |
+
"audio_eos_token": "<|audio_end|>",
|
| 33 |
+
"audio_token": "<|audio_pad|>",
|
| 34 |
+
"image_token": "<|image_pad|>",
|
| 35 |
+
"video_token": "<|video_pad|>",
|
| 36 |
+
"vision_bos_token": "<|vision_start|>",
|
| 37 |
+
"vision_eos_token": "<|vision_end|>"
|
| 38 |
+
},
|
| 39 |
+
"pad_token": "<|endoftext|>",
|
| 40 |
+
"pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
|
| 41 |
+
"split_special_tokens": false,
|
| 42 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 43 |
+
"unk_token": null,
|
| 44 |
+
"video_token": "<|video_pad|>",
|
| 45 |
+
"vision_bos_token": "<|vision_start|>",
|
| 46 |
+
"vision_eos_token": "<|vision_end|>"
|
| 47 |
+
}
|
checkpoints.json
ADDED
|
@@ -0,0 +1,202 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"subfolder": "Keen-qwen4b-20b/checkpoint-2253",
|
| 4 |
+
"step": 2253,
|
| 5 |
+
"body_tokens": 2000696711,
|
| 6 |
+
"final": false,
|
| 7 |
+
"layout": {
|
| 8 |
+
"tensors": 426,
|
| 9 |
+
"serialized_parameters": 4205751296
|
| 10 |
+
}
|
| 11 |
+
},
|
| 12 |
+
{
|
| 13 |
+
"subfolder": "Keen-qwen4b-20b/checkpoint-4453",
|
| 14 |
+
"step": 4453,
|
| 15 |
+
"body_tokens": 4000757038,
|
| 16 |
+
"final": false,
|
| 17 |
+
"layout": {
|
| 18 |
+
"tensors": 426,
|
| 19 |
+
"serialized_parameters": 4205751296
|
| 20 |
+
}
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"subfolder": "Keen-qwen4b-20b/checkpoint-6701",
|
| 24 |
+
"step": 6701,
|
| 25 |
+
"body_tokens": 6000082648,
|
| 26 |
+
"final": false,
|
| 27 |
+
"layout": {
|
| 28 |
+
"tensors": 426,
|
| 29 |
+
"serialized_parameters": 4205751296
|
| 30 |
+
}
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"subfolder": "Keen-qwen4b-20b/checkpoint-8956",
|
| 34 |
+
"step": 8956,
|
| 35 |
+
"body_tokens": 8000600622,
|
| 36 |
+
"final": false,
|
| 37 |
+
"layout": {
|
| 38 |
+
"tensors": 426,
|
| 39 |
+
"serialized_parameters": 4205751296
|
| 40 |
+
}
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"subfolder": "Keen-qwen4b-20b/checkpoint-11202",
|
| 44 |
+
"step": 11202,
|
| 45 |
+
"body_tokens": 10000294657,
|
| 46 |
+
"final": false,
|
| 47 |
+
"layout": {
|
| 48 |
+
"tensors": 426,
|
| 49 |
+
"serialized_parameters": 4205751296
|
| 50 |
+
}
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"subfolder": "Keen-qwen4b-20b/checkpoint-13449",
|
| 54 |
+
"step": 13449,
|
| 55 |
+
"body_tokens": 12000429703,
|
| 56 |
+
"final": false,
|
| 57 |
+
"layout": {
|
| 58 |
+
"tensors": 426,
|
| 59 |
+
"serialized_parameters": 4205751296
|
| 60 |
+
}
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"subfolder": "Keen-qwen4b-20b/checkpoint-15698",
|
| 64 |
+
"step": 15698,
|
| 65 |
+
"body_tokens": 14000083680,
|
| 66 |
+
"final": false,
|
| 67 |
+
"layout": {
|
| 68 |
+
"tensors": 426,
|
| 69 |
+
"serialized_parameters": 4205751296
|
| 70 |
+
}
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"subfolder": "Keen-qwen4b-20b/checkpoint-17953",
|
| 74 |
+
"step": 17953,
|
| 75 |
+
"body_tokens": 16000387372,
|
| 76 |
+
"final": false,
|
| 77 |
+
"layout": {
|
| 78 |
+
"tensors": 426,
|
| 79 |
+
"serialized_parameters": 4205751296
|
| 80 |
+
}
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"subfolder": "Keen-qwen4b-20b/checkpoint-20218",
|
| 84 |
+
"step": 20218,
|
| 85 |
+
"body_tokens": 18000030679,
|
| 86 |
+
"final": false,
|
| 87 |
+
"layout": {
|
| 88 |
+
"tensors": 426,
|
| 89 |
+
"serialized_parameters": 4205751296
|
| 90 |
+
}
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"subfolder": "Keen-qwen4b-20b/checkpoint-22478",
|
| 94 |
+
"step": 22478,
|
| 95 |
+
"body_tokens": 20000557588,
|
| 96 |
+
"final": false,
|
| 97 |
+
"layout": {
|
| 98 |
+
"tensors": 426,
|
| 99 |
+
"serialized_parameters": 4205751296
|
| 100 |
+
}
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"subfolder": "Vanilla-qwen4b-20b/checkpoint-1910",
|
| 104 |
+
"step": 1910,
|
| 105 |
+
"body_tokens": 2000045638,
|
| 106 |
+
"final": false,
|
| 107 |
+
"layout": {
|
| 108 |
+
"tensors": 426,
|
| 109 |
+
"serialized_parameters": 4205751296
|
| 110 |
+
}
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"subfolder": "Vanilla-qwen4b-20b/checkpoint-3818",
|
| 114 |
+
"step": 3818,
|
| 115 |
+
"body_tokens": 4000211289,
|
| 116 |
+
"final": false,
|
| 117 |
+
"layout": {
|
| 118 |
+
"tensors": 426,
|
| 119 |
+
"serialized_parameters": 4205751296
|
| 120 |
+
}
|
| 121 |
+
},
|
| 122 |
+
{
|
| 123 |
+
"subfolder": "Vanilla-qwen4b-20b/checkpoint-5726",
|
| 124 |
+
"step": 5726,
|
| 125 |
+
"body_tokens": 6000159072,
|
| 126 |
+
"final": false,
|
| 127 |
+
"layout": {
|
| 128 |
+
"tensors": 426,
|
| 129 |
+
"serialized_parameters": 4205751296
|
| 130 |
+
}
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"subfolder": "Vanilla-qwen4b-20b/checkpoint-7635",
|
| 134 |
+
"step": 7635,
|
| 135 |
+
"body_tokens": 8000913248,
|
| 136 |
+
"final": false,
|
| 137 |
+
"layout": {
|
| 138 |
+
"tensors": 426,
|
| 139 |
+
"serialized_parameters": 4205751296
|
| 140 |
+
}
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"subfolder": "Vanilla-qwen4b-20b/checkpoint-9543",
|
| 144 |
+
"step": 9543,
|
| 145 |
+
"body_tokens": 10000619298,
|
| 146 |
+
"final": false,
|
| 147 |
+
"layout": {
|
| 148 |
+
"tensors": 426,
|
| 149 |
+
"serialized_parameters": 4205751296
|
| 150 |
+
}
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"subfolder": "Vanilla-qwen4b-20b/checkpoint-11452",
|
| 154 |
+
"step": 11452,
|
| 155 |
+
"body_tokens": 12000591962,
|
| 156 |
+
"final": false,
|
| 157 |
+
"layout": {
|
| 158 |
+
"tensors": 426,
|
| 159 |
+
"serialized_parameters": 4205751296
|
| 160 |
+
}
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"subfolder": "Vanilla-qwen4b-20b/checkpoint-13362",
|
| 164 |
+
"step": 13362,
|
| 165 |
+
"body_tokens": 14000166403,
|
| 166 |
+
"final": false,
|
| 167 |
+
"layout": {
|
| 168 |
+
"tensors": 426,
|
| 169 |
+
"serialized_parameters": 4205751296
|
| 170 |
+
}
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"subfolder": "Vanilla-qwen4b-20b/checkpoint-15273",
|
| 174 |
+
"step": 15273,
|
| 175 |
+
"body_tokens": 16000320768,
|
| 176 |
+
"final": false,
|
| 177 |
+
"layout": {
|
| 178 |
+
"tensors": 426,
|
| 179 |
+
"serialized_parameters": 4205751296
|
| 180 |
+
}
|
| 181 |
+
},
|
| 182 |
+
{
|
| 183 |
+
"subfolder": "Vanilla-qwen4b-20b/checkpoint-17188",
|
| 184 |
+
"step": 17188,
|
| 185 |
+
"body_tokens": 18000257314,
|
| 186 |
+
"final": false,
|
| 187 |
+
"layout": {
|
| 188 |
+
"tensors": 426,
|
| 189 |
+
"serialized_parameters": 4205751296
|
| 190 |
+
}
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"subfolder": "Vanilla-qwen4b-20b/checkpoint-19106",
|
| 194 |
+
"step": 19106,
|
| 195 |
+
"body_tokens": 20000220345,
|
| 196 |
+
"final": false,
|
| 197 |
+
"layout": {
|
| 198 |
+
"tensors": 426,
|
| 199 |
+
"serialized_parameters": 4205751296
|
| 200 |
+
}
|
| 201 |
+
}
|
| 202 |
+
]
|
manifest.json
ADDED
|
@@ -0,0 +1,626 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"repo": "meihant/AnonymousModel-mht",
|
| 3 |
+
"checkpoint_count": 20,
|
| 4 |
+
"files": [
|
| 5 |
+
{
|
| 6 |
+
"path": "Keen-qwen4b-20b/checkpoint-2253/README.md",
|
| 7 |
+
"bytes": 1795,
|
| 8 |
+
"sha256": "7114cdabbf634f4503318ed7d94c629a782caa9d06362748d9c4b053d731bd14"
|
| 9 |
+
},
|
| 10 |
+
{
|
| 11 |
+
"path": "Keen-qwen4b-20b/checkpoint-2253/config.json",
|
| 12 |
+
"bytes": 1980,
|
| 13 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 14 |
+
},
|
| 15 |
+
{
|
| 16 |
+
"path": "Keen-qwen4b-20b/checkpoint-2253/generation_config.json",
|
| 17 |
+
"bytes": 204,
|
| 18 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 19 |
+
},
|
| 20 |
+
{
|
| 21 |
+
"path": "Keen-qwen4b-20b/checkpoint-2253/model.safetensors",
|
| 22 |
+
"bytes": 8411554440,
|
| 23 |
+
"sha256": "f7816a1cae0b08b1e0901c378b916df2769b9b7fa7ebfac63e98854c33b656a4"
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"path": "Keen-qwen4b-20b/checkpoint-2253/tokenizer.json",
|
| 27 |
+
"bytes": 19989692,
|
| 28 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"path": "Keen-qwen4b-20b/checkpoint-2253/tokenizer_config.json",
|
| 32 |
+
"bytes": 1422,
|
| 33 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 34 |
+
},
|
| 35 |
+
{
|
| 36 |
+
"path": "Keen-qwen4b-20b/checkpoint-4453/README.md",
|
| 37 |
+
"bytes": 1795,
|
| 38 |
+
"sha256": "6c8cb0371d6f20969ad10759cf3d67d67f5afb95a8edf5ffb1c29dddd268929e"
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"path": "Keen-qwen4b-20b/checkpoint-4453/config.json",
|
| 42 |
+
"bytes": 1980,
|
| 43 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 44 |
+
},
|
| 45 |
+
{
|
| 46 |
+
"path": "Keen-qwen4b-20b/checkpoint-4453/generation_config.json",
|
| 47 |
+
"bytes": 204,
|
| 48 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"path": "Keen-qwen4b-20b/checkpoint-4453/model.safetensors",
|
| 52 |
+
"bytes": 8411554440,
|
| 53 |
+
"sha256": "9bcfa8a9a42d9b1362d339b13e364fe04ae52c9460f58a715901f3cac64c1a44"
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"path": "Keen-qwen4b-20b/checkpoint-4453/tokenizer.json",
|
| 57 |
+
"bytes": 19989692,
|
| 58 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"path": "Keen-qwen4b-20b/checkpoint-4453/tokenizer_config.json",
|
| 62 |
+
"bytes": 1422,
|
| 63 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"path": "Keen-qwen4b-20b/checkpoint-6701/README.md",
|
| 67 |
+
"bytes": 1795,
|
| 68 |
+
"sha256": "203f225133c82bb316fddbd327a72923c3d783cc5640c97ae9d5b6dabbb02321"
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"path": "Keen-qwen4b-20b/checkpoint-6701/config.json",
|
| 72 |
+
"bytes": 1980,
|
| 73 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"path": "Keen-qwen4b-20b/checkpoint-6701/generation_config.json",
|
| 77 |
+
"bytes": 204,
|
| 78 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 79 |
+
},
|
| 80 |
+
{
|
| 81 |
+
"path": "Keen-qwen4b-20b/checkpoint-6701/model.safetensors",
|
| 82 |
+
"bytes": 8411554440,
|
| 83 |
+
"sha256": "94e659545b1ba9eecb5c9a2f03206e893bd544f9b4388e591b4316cabf7f1120"
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"path": "Keen-qwen4b-20b/checkpoint-6701/tokenizer.json",
|
| 87 |
+
"bytes": 19989692,
|
| 88 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"path": "Keen-qwen4b-20b/checkpoint-6701/tokenizer_config.json",
|
| 92 |
+
"bytes": 1422,
|
| 93 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"path": "Keen-qwen4b-20b/checkpoint-8956/README.md",
|
| 97 |
+
"bytes": 1795,
|
| 98 |
+
"sha256": "a23693eb784917f5e3437d7f29dedadd7dc38e7f35029c8eff52def5c91083c2"
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"path": "Keen-qwen4b-20b/checkpoint-8956/config.json",
|
| 102 |
+
"bytes": 1980,
|
| 103 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"path": "Keen-qwen4b-20b/checkpoint-8956/generation_config.json",
|
| 107 |
+
"bytes": 204,
|
| 108 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"path": "Keen-qwen4b-20b/checkpoint-8956/model.safetensors",
|
| 112 |
+
"bytes": 8411554440,
|
| 113 |
+
"sha256": "8d54e7eb42dffdfc89613913ba3a6c7bfcf0e9eb66ad193c978608eb652efc4a"
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"path": "Keen-qwen4b-20b/checkpoint-8956/tokenizer.json",
|
| 117 |
+
"bytes": 19989692,
|
| 118 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 119 |
+
},
|
| 120 |
+
{
|
| 121 |
+
"path": "Keen-qwen4b-20b/checkpoint-8956/tokenizer_config.json",
|
| 122 |
+
"bytes": 1422,
|
| 123 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"path": "Keen-qwen4b-20b/checkpoint-11202/README.md",
|
| 127 |
+
"bytes": 1799,
|
| 128 |
+
"sha256": "87177522c0ac02e3ddf88e6d405b64baacfaab0094fb6c7f7831a2b4040351d5"
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"path": "Keen-qwen4b-20b/checkpoint-11202/config.json",
|
| 132 |
+
"bytes": 1980,
|
| 133 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 134 |
+
},
|
| 135 |
+
{
|
| 136 |
+
"path": "Keen-qwen4b-20b/checkpoint-11202/generation_config.json",
|
| 137 |
+
"bytes": 204,
|
| 138 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 139 |
+
},
|
| 140 |
+
{
|
| 141 |
+
"path": "Keen-qwen4b-20b/checkpoint-11202/model.safetensors",
|
| 142 |
+
"bytes": 8411554440,
|
| 143 |
+
"sha256": "9e5c178a11a7f4d1f2c4486a1e21a47acfba0998e2484122c6dbe2e9f61f9101"
|
| 144 |
+
},
|
| 145 |
+
{
|
| 146 |
+
"path": "Keen-qwen4b-20b/checkpoint-11202/tokenizer.json",
|
| 147 |
+
"bytes": 19989692,
|
| 148 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 149 |
+
},
|
| 150 |
+
{
|
| 151 |
+
"path": "Keen-qwen4b-20b/checkpoint-11202/tokenizer_config.json",
|
| 152 |
+
"bytes": 1422,
|
| 153 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"path": "Keen-qwen4b-20b/checkpoint-13449/README.md",
|
| 157 |
+
"bytes": 1799,
|
| 158 |
+
"sha256": "4cddbfbdfa777f37aa7b2bbac8c48efd699b71c87fd7d8c9f2090b9ea8c39716"
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"path": "Keen-qwen4b-20b/checkpoint-13449/config.json",
|
| 162 |
+
"bytes": 1980,
|
| 163 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"path": "Keen-qwen4b-20b/checkpoint-13449/generation_config.json",
|
| 167 |
+
"bytes": 204,
|
| 168 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 169 |
+
},
|
| 170 |
+
{
|
| 171 |
+
"path": "Keen-qwen4b-20b/checkpoint-13449/model.safetensors",
|
| 172 |
+
"bytes": 8411554440,
|
| 173 |
+
"sha256": "5fe83ee3b190b8772bef64da62b031b09516d3666f23deaa2fb32b10416050a6"
|
| 174 |
+
},
|
| 175 |
+
{
|
| 176 |
+
"path": "Keen-qwen4b-20b/checkpoint-13449/tokenizer.json",
|
| 177 |
+
"bytes": 19989692,
|
| 178 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"path": "Keen-qwen4b-20b/checkpoint-13449/tokenizer_config.json",
|
| 182 |
+
"bytes": 1422,
|
| 183 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 184 |
+
},
|
| 185 |
+
{
|
| 186 |
+
"path": "Keen-qwen4b-20b/checkpoint-15698/README.md",
|
| 187 |
+
"bytes": 1799,
|
| 188 |
+
"sha256": "26ecd7389adb3c8f1c7be24063315bfcc8350a76a031b9b1ded461004120db40"
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"path": "Keen-qwen4b-20b/checkpoint-15698/config.json",
|
| 192 |
+
"bytes": 1980,
|
| 193 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 194 |
+
},
|
| 195 |
+
{
|
| 196 |
+
"path": "Keen-qwen4b-20b/checkpoint-15698/generation_config.json",
|
| 197 |
+
"bytes": 204,
|
| 198 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"path": "Keen-qwen4b-20b/checkpoint-15698/model.safetensors",
|
| 202 |
+
"bytes": 8411554440,
|
| 203 |
+
"sha256": "e19c76755aa3b2ce6c0cdf0d170f8d35df339a20e07fb25c9d82cb1d7d22786c"
|
| 204 |
+
},
|
| 205 |
+
{
|
| 206 |
+
"path": "Keen-qwen4b-20b/checkpoint-15698/tokenizer.json",
|
| 207 |
+
"bytes": 19989692,
|
| 208 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 209 |
+
},
|
| 210 |
+
{
|
| 211 |
+
"path": "Keen-qwen4b-20b/checkpoint-15698/tokenizer_config.json",
|
| 212 |
+
"bytes": 1422,
|
| 213 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 214 |
+
},
|
| 215 |
+
{
|
| 216 |
+
"path": "Keen-qwen4b-20b/checkpoint-17953/README.md",
|
| 217 |
+
"bytes": 1799,
|
| 218 |
+
"sha256": "6dbada8dc77e92db73827c7b2e5b048be9cc4d1873562c84e3b7c6aba7e1a311"
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"path": "Keen-qwen4b-20b/checkpoint-17953/config.json",
|
| 222 |
+
"bytes": 1980,
|
| 223 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"path": "Keen-qwen4b-20b/checkpoint-17953/generation_config.json",
|
| 227 |
+
"bytes": 204,
|
| 228 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"path": "Keen-qwen4b-20b/checkpoint-17953/model.safetensors",
|
| 232 |
+
"bytes": 8411554440,
|
| 233 |
+
"sha256": "701a90729ccc31bada5fc4dff6e90a8ac7b90afbbb85f8f80816d693e65dc296"
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"path": "Keen-qwen4b-20b/checkpoint-17953/tokenizer.json",
|
| 237 |
+
"bytes": 19989692,
|
| 238 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 239 |
+
},
|
| 240 |
+
{
|
| 241 |
+
"path": "Keen-qwen4b-20b/checkpoint-17953/tokenizer_config.json",
|
| 242 |
+
"bytes": 1422,
|
| 243 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"path": "Keen-qwen4b-20b/checkpoint-20218/README.md",
|
| 247 |
+
"bytes": 1799,
|
| 248 |
+
"sha256": "63381a07180f5cd5cf041a9a8375d744efb21433d4295f06b0efe781583704df"
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"path": "Keen-qwen4b-20b/checkpoint-20218/config.json",
|
| 252 |
+
"bytes": 1980,
|
| 253 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 254 |
+
},
|
| 255 |
+
{
|
| 256 |
+
"path": "Keen-qwen4b-20b/checkpoint-20218/generation_config.json",
|
| 257 |
+
"bytes": 204,
|
| 258 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 259 |
+
},
|
| 260 |
+
{
|
| 261 |
+
"path": "Keen-qwen4b-20b/checkpoint-20218/model.safetensors",
|
| 262 |
+
"bytes": 8411554440,
|
| 263 |
+
"sha256": "afdc1e1f396a76fe3cabfac49d643dda0e1019743e5350f8a4e67cd16e668866"
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"path": "Keen-qwen4b-20b/checkpoint-20218/tokenizer.json",
|
| 267 |
+
"bytes": 19989692,
|
| 268 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"path": "Keen-qwen4b-20b/checkpoint-20218/tokenizer_config.json",
|
| 272 |
+
"bytes": 1422,
|
| 273 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 274 |
+
},
|
| 275 |
+
{
|
| 276 |
+
"path": "Keen-qwen4b-20b/checkpoint-22478/README.md",
|
| 277 |
+
"bytes": 1799,
|
| 278 |
+
"sha256": "501e75a95dd70b1bba8cf470dcf80bfc17458a4f648fd43d8bb68d68910cb966"
|
| 279 |
+
},
|
| 280 |
+
{
|
| 281 |
+
"path": "Keen-qwen4b-20b/checkpoint-22478/config.json",
|
| 282 |
+
"bytes": 1980,
|
| 283 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 284 |
+
},
|
| 285 |
+
{
|
| 286 |
+
"path": "Keen-qwen4b-20b/checkpoint-22478/generation_config.json",
|
| 287 |
+
"bytes": 204,
|
| 288 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 289 |
+
},
|
| 290 |
+
{
|
| 291 |
+
"path": "Keen-qwen4b-20b/checkpoint-22478/model.safetensors",
|
| 292 |
+
"bytes": 8411554440,
|
| 293 |
+
"sha256": "b869467d5db824e6df4a995bfaca0ae74a7b8e43899593b877b774bcc3861811"
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"path": "Keen-qwen4b-20b/checkpoint-22478/tokenizer.json",
|
| 297 |
+
"bytes": 19989692,
|
| 298 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"path": "Keen-qwen4b-20b/checkpoint-22478/tokenizer_config.json",
|
| 302 |
+
"bytes": 1422,
|
| 303 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 304 |
+
},
|
| 305 |
+
{
|
| 306 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-1910/README.md",
|
| 307 |
+
"bytes": 1693,
|
| 308 |
+
"sha256": "011a083a5362e976bd5256d29e456d2ed322ddae0283bfcc32808cde5bf8429b"
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-1910/config.json",
|
| 312 |
+
"bytes": 1980,
|
| 313 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 314 |
+
},
|
| 315 |
+
{
|
| 316 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-1910/generation_config.json",
|
| 317 |
+
"bytes": 204,
|
| 318 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-1910/model.safetensors",
|
| 322 |
+
"bytes": 8411554440,
|
| 323 |
+
"sha256": "ebbecc5de3ed774dbfce3387c20cba60244884b3690bf9aa59c26a11769e5557"
|
| 324 |
+
},
|
| 325 |
+
{
|
| 326 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-1910/tokenizer.json",
|
| 327 |
+
"bytes": 19989692,
|
| 328 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-1910/tokenizer_config.json",
|
| 332 |
+
"bytes": 1422,
|
| 333 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 334 |
+
},
|
| 335 |
+
{
|
| 336 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-3818/README.md",
|
| 337 |
+
"bytes": 1693,
|
| 338 |
+
"sha256": "ddd2df66b19338dc39a1bf3d9881fd9b53ef170401377397ec262ad254b07006"
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-3818/config.json",
|
| 342 |
+
"bytes": 1980,
|
| 343 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 344 |
+
},
|
| 345 |
+
{
|
| 346 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-3818/generation_config.json",
|
| 347 |
+
"bytes": 204,
|
| 348 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 349 |
+
},
|
| 350 |
+
{
|
| 351 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-3818/model.safetensors",
|
| 352 |
+
"bytes": 8411554440,
|
| 353 |
+
"sha256": "4f00f81c6fc90de35737f7e36e1ac56e4cc381b3dba06da29b60a0474f91fa21"
|
| 354 |
+
},
|
| 355 |
+
{
|
| 356 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-3818/tokenizer.json",
|
| 357 |
+
"bytes": 19989692,
|
| 358 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 359 |
+
},
|
| 360 |
+
{
|
| 361 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-3818/tokenizer_config.json",
|
| 362 |
+
"bytes": 1422,
|
| 363 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-5726/README.md",
|
| 367 |
+
"bytes": 1693,
|
| 368 |
+
"sha256": "88dfa4be8fe7a95a4f3898838c84a0c35aa14fac5f81b53e676b136bf3c7f5df"
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-5726/config.json",
|
| 372 |
+
"bytes": 1980,
|
| 373 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-5726/generation_config.json",
|
| 377 |
+
"bytes": 204,
|
| 378 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 379 |
+
},
|
| 380 |
+
{
|
| 381 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-5726/model.safetensors",
|
| 382 |
+
"bytes": 8411554440,
|
| 383 |
+
"sha256": "67da1a0d5464184db38d0612ed09cee7e38c962bd0fc15063caf52ee11b55997"
|
| 384 |
+
},
|
| 385 |
+
{
|
| 386 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-5726/tokenizer.json",
|
| 387 |
+
"bytes": 19989692,
|
| 388 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-5726/tokenizer_config.json",
|
| 392 |
+
"bytes": 1422,
|
| 393 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 394 |
+
},
|
| 395 |
+
{
|
| 396 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-7635/README.md",
|
| 397 |
+
"bytes": 1693,
|
| 398 |
+
"sha256": "8f8507357a56e749dca3e1e791add779b44d35ebb6dafdc034ba0eb7c47c23a1"
|
| 399 |
+
},
|
| 400 |
+
{
|
| 401 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-7635/config.json",
|
| 402 |
+
"bytes": 1980,
|
| 403 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-7635/generation_config.json",
|
| 407 |
+
"bytes": 204,
|
| 408 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 409 |
+
},
|
| 410 |
+
{
|
| 411 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-7635/model.safetensors",
|
| 412 |
+
"bytes": 8411554440,
|
| 413 |
+
"sha256": "b718e6cf6d61bfa6e864e85d6bee37079a3f4f030c420311885ca8b4297a1c03"
|
| 414 |
+
},
|
| 415 |
+
{
|
| 416 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-7635/tokenizer.json",
|
| 417 |
+
"bytes": 19989692,
|
| 418 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 419 |
+
},
|
| 420 |
+
{
|
| 421 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-7635/tokenizer_config.json",
|
| 422 |
+
"bytes": 1422,
|
| 423 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 424 |
+
},
|
| 425 |
+
{
|
| 426 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-9543/README.md",
|
| 427 |
+
"bytes": 1694,
|
| 428 |
+
"sha256": "ae63402e545913d61faea0073e1ed777b545b809cd4481ac0bca5a02d82225e2"
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-9543/config.json",
|
| 432 |
+
"bytes": 1980,
|
| 433 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 434 |
+
},
|
| 435 |
+
{
|
| 436 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-9543/generation_config.json",
|
| 437 |
+
"bytes": 204,
|
| 438 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 439 |
+
},
|
| 440 |
+
{
|
| 441 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-9543/model.safetensors",
|
| 442 |
+
"bytes": 8411554440,
|
| 443 |
+
"sha256": "7bb6c5406dcb41f68006815bf25e67a42320b08c78d822b7ac4bd9c29fd3a13a"
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-9543/tokenizer.json",
|
| 447 |
+
"bytes": 19989692,
|
| 448 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 449 |
+
},
|
| 450 |
+
{
|
| 451 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-9543/tokenizer_config.json",
|
| 452 |
+
"bytes": 1422,
|
| 453 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 454 |
+
},
|
| 455 |
+
{
|
| 456 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-11452/README.md",
|
| 457 |
+
"bytes": 1697,
|
| 458 |
+
"sha256": "96b6a47607525e2053bc0eeb4274c8b370e010519072fc9c8ecd01eadda805c4"
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-11452/config.json",
|
| 462 |
+
"bytes": 1980,
|
| 463 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 464 |
+
},
|
| 465 |
+
{
|
| 466 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-11452/generation_config.json",
|
| 467 |
+
"bytes": 204,
|
| 468 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 469 |
+
},
|
| 470 |
+
{
|
| 471 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-11452/model.safetensors",
|
| 472 |
+
"bytes": 8411554440,
|
| 473 |
+
"sha256": "a7009776caa28cd43561b57da0d4e9f3538296df42ffa2807f86a729172c20df"
|
| 474 |
+
},
|
| 475 |
+
{
|
| 476 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-11452/tokenizer.json",
|
| 477 |
+
"bytes": 19989692,
|
| 478 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 479 |
+
},
|
| 480 |
+
{
|
| 481 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-11452/tokenizer_config.json",
|
| 482 |
+
"bytes": 1422,
|
| 483 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 484 |
+
},
|
| 485 |
+
{
|
| 486 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-13362/README.md",
|
| 487 |
+
"bytes": 1697,
|
| 488 |
+
"sha256": "ac547bb2f3a971bcba7e805436d2454d734cf1d5c0e8dcd8a414cc7ce5891b44"
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-13362/config.json",
|
| 492 |
+
"bytes": 1980,
|
| 493 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 494 |
+
},
|
| 495 |
+
{
|
| 496 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-13362/generation_config.json",
|
| 497 |
+
"bytes": 204,
|
| 498 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 499 |
+
},
|
| 500 |
+
{
|
| 501 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-13362/model.safetensors",
|
| 502 |
+
"bytes": 8411554440,
|
| 503 |
+
"sha256": "aa377d8aebe815164abafc2dd8a85ccc3e44fadc7552e4cb20014167ca09ee74"
|
| 504 |
+
},
|
| 505 |
+
{
|
| 506 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-13362/tokenizer.json",
|
| 507 |
+
"bytes": 19989692,
|
| 508 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-13362/tokenizer_config.json",
|
| 512 |
+
"bytes": 1422,
|
| 513 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 514 |
+
},
|
| 515 |
+
{
|
| 516 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-15273/README.md",
|
| 517 |
+
"bytes": 1697,
|
| 518 |
+
"sha256": "3b875f6a12b86f42691c7650699702b8f3125f87d4009d7552be28973741eef7"
|
| 519 |
+
},
|
| 520 |
+
{
|
| 521 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-15273/config.json",
|
| 522 |
+
"bytes": 1980,
|
| 523 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 524 |
+
},
|
| 525 |
+
{
|
| 526 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-15273/generation_config.json",
|
| 527 |
+
"bytes": 204,
|
| 528 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 529 |
+
},
|
| 530 |
+
{
|
| 531 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-15273/model.safetensors",
|
| 532 |
+
"bytes": 8411554440,
|
| 533 |
+
"sha256": "9233f915f89cd02a49cbd8690c5de7625f483dcc697cc2a451137522613e464a"
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-15273/tokenizer.json",
|
| 537 |
+
"bytes": 19989692,
|
| 538 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 539 |
+
},
|
| 540 |
+
{
|
| 541 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-15273/tokenizer_config.json",
|
| 542 |
+
"bytes": 1422,
|
| 543 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 544 |
+
},
|
| 545 |
+
{
|
| 546 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-17188/README.md",
|
| 547 |
+
"bytes": 1697,
|
| 548 |
+
"sha256": "6c5264ec80652a50f225b979cf4a6686caca332055081b49ab1b20211e78796d"
|
| 549 |
+
},
|
| 550 |
+
{
|
| 551 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-17188/config.json",
|
| 552 |
+
"bytes": 1980,
|
| 553 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-17188/generation_config.json",
|
| 557 |
+
"bytes": 204,
|
| 558 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 559 |
+
},
|
| 560 |
+
{
|
| 561 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-17188/model.safetensors",
|
| 562 |
+
"bytes": 8411554440,
|
| 563 |
+
"sha256": "67d469d450411055d454f462e379f69c74b39a771ef14ffc569589a61978117d"
|
| 564 |
+
},
|
| 565 |
+
{
|
| 566 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-17188/tokenizer.json",
|
| 567 |
+
"bytes": 19989692,
|
| 568 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 569 |
+
},
|
| 570 |
+
{
|
| 571 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-17188/tokenizer_config.json",
|
| 572 |
+
"bytes": 1422,
|
| 573 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 574 |
+
},
|
| 575 |
+
{
|
| 576 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-19106/README.md",
|
| 577 |
+
"bytes": 1697,
|
| 578 |
+
"sha256": "02d4488e974a25842481c18679dbbf013105cfc62b5f7ec6e1d24dbe99130fa6"
|
| 579 |
+
},
|
| 580 |
+
{
|
| 581 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-19106/config.json",
|
| 582 |
+
"bytes": 1980,
|
| 583 |
+
"sha256": "eac5a6e0ba2e18c79768899a66a50997393c6ca2e50df8763d2c6787fb077c27"
|
| 584 |
+
},
|
| 585 |
+
{
|
| 586 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-19106/generation_config.json",
|
| 587 |
+
"bytes": 204,
|
| 588 |
+
"sha256": "aaba716c1c3b8d85ed91268153d53cf1caeef458d56c4aad008d7b51810ec9ea"
|
| 589 |
+
},
|
| 590 |
+
{
|
| 591 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-19106/model.safetensors",
|
| 592 |
+
"bytes": 8411554440,
|
| 593 |
+
"sha256": "8ae4543af3abe122675c0252d7bc0d1f8fcfcac8292e5064a9d824a2a75eb8e3"
|
| 594 |
+
},
|
| 595 |
+
{
|
| 596 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-19106/tokenizer.json",
|
| 597 |
+
"bytes": 19989692,
|
| 598 |
+
"sha256": "4f8a9ce416e1039464b3e8c00e1e284f037b63350c952ce2d5f641690e3a9a1f"
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"path": "Vanilla-qwen4b-20b/checkpoint-19106/tokenizer_config.json",
|
| 602 |
+
"bytes": 1422,
|
| 603 |
+
"sha256": "488d44148dce4f2d03f01b1dad1ed33f25a17eb3cfb099535be47b33c5b9a877"
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"path": "README.md",
|
| 607 |
+
"bytes": 1433,
|
| 608 |
+
"sha256": "873c031bb2f94a8de8ab4b7103ab10f29a25da9291a67d82dfe90ed5606325f3"
|
| 609 |
+
},
|
| 610 |
+
{
|
| 611 |
+
"path": "checkpoints.json",
|
| 612 |
+
"bytes": 4387,
|
| 613 |
+
"sha256": "e776a321c70c7e44cd1f6bfa61c3fb13822f8a45797194d501baacb135beca55"
|
| 614 |
+
},
|
| 615 |
+
{
|
| 616 |
+
"path": "Keen-qwen4b-20b/README.md",
|
| 617 |
+
"bytes": 1069,
|
| 618 |
+
"sha256": "aa9f4a05570d7fbcdbbc35c85009f637b25bf498a3813c0c238e5b97e9e67017"
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"path": "Vanilla-qwen4b-20b/README.md",
|
| 622 |
+
"bytes": 1069,
|
| 623 |
+
"sha256": "d86c96bfd5aff631e6c0c757a83f4275f8b742c45cc31dea5840755fbbfce70c"
|
| 624 |
+
}
|
| 625 |
+
]
|
| 626 |
+
}
|