latest.json: checkpoints/2584388k (step 78,160)
Browse files- README.md +4 -4
- latest.json +7 -7
- latest/config.json +2 -2
- latest/model.safetensors +1 -1
README.md
CHANGED
|
@@ -30,7 +30,7 @@ deliberately derived from it: Nirca began as a port of mini-AGI to MLX.
|
|
| 30 |
➡️ **Try it in your browser (WebGPU):** https://tg-techie-agents.github.io/nirca-mini-webgpu/mini/
|
| 31 |
|
| 32 |
This card is written by program from the checkpoint's own record, at step
|
| 33 |
-
78,
|
| 34 |
|
| 35 |
## Model details
|
| 36 |
|
|
@@ -115,9 +115,9 @@ Each step is listed because the run's records show it was done:
|
|
| 115 |
|
| 116 |
## Training procedure
|
| 117 |
|
| 118 |
-
- **Steps:** 78,
|
| 119 |
-
- **Tokens seen:** 2,
|
| 120 |
-
- **Quantization-aware training:** from step 14,655 of 78,
|
| 121 |
- **Objective:** next-byte prediction; on chats, loss on the assistant's turns only
|
| 122 |
- **Recurrence:** the number of passes is drawn per step during training, and
|
| 123 |
the halting head learns when to stop
|
|
|
|
| 30 |
➡️ **Try it in your browser (WebGPU):** https://tg-techie-agents.github.io/nirca-mini-webgpu/mini/
|
| 31 |
|
| 32 |
This card is written by program from the checkpoint's own record, at step
|
| 33 |
+
78,160 (2,584,388,242 tokens seen).
|
| 34 |
|
| 35 |
## Model details
|
| 36 |
|
|
|
|
| 115 |
|
| 116 |
## Training procedure
|
| 117 |
|
| 118 |
+
- **Steps:** 78,160
|
| 119 |
+
- **Tokens seen:** 2,584,388,242 (one token is one byte)
|
| 120 |
+
- **Quantization-aware training:** from step 14,655 of 78,160, so for 99% of its 2,584,388,242 tokens
|
| 121 |
- **Objective:** next-byte prediction; on chats, loss on the assistant's turns only
|
| 122 |
- **Recurrence:** the number of passes is drawn per step during training, and
|
| 123 |
the halting head learns when to stop
|
latest.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
{
|
| 2 |
"name": "nirca-mini",
|
| 3 |
-
"step":
|
| 4 |
"quant": "ternary",
|
| 5 |
"quant_scope": "all",
|
| 6 |
"n_experts": 32,
|
|
@@ -40,17 +40,17 @@
|
|
| 40 |
"pool_depth": 1,
|
| 41 |
"pool_max": 32
|
| 42 |
},
|
| 43 |
-
"tokens":
|
| 44 |
-
"checkpoint": "checkpoints/
|
| 45 |
"files": {
|
| 46 |
"config.json": {
|
| 47 |
-
"url": "https://huggingface.co/nirca/nirca-mini/resolve/main/checkpoints/
|
| 48 |
-
"sha256": "
|
| 49 |
"bytes": 873
|
| 50 |
},
|
| 51 |
"model.safetensors": {
|
| 52 |
-
"url": "https://huggingface.co/nirca/nirca-mini/resolve/main/checkpoints/
|
| 53 |
-
"sha256": "
|
| 54 |
"bytes": 32638048
|
| 55 |
}
|
| 56 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"name": "nirca-mini",
|
| 3 |
+
"step": 78160,
|
| 4 |
"quant": "ternary",
|
| 5 |
"quant_scope": "all",
|
| 6 |
"n_experts": 32,
|
|
|
|
| 40 |
"pool_depth": 1,
|
| 41 |
"pool_max": 32
|
| 42 |
},
|
| 43 |
+
"tokens": 2584388242,
|
| 44 |
+
"checkpoint": "checkpoints/2584388k/",
|
| 45 |
"files": {
|
| 46 |
"config.json": {
|
| 47 |
+
"url": "https://huggingface.co/nirca/nirca-mini/resolve/main/checkpoints/2584388k/config.json",
|
| 48 |
+
"sha256": "96f474d808b7e11eb1843a0c5b1e104366d15394fb272f8305b8f29ca12a1602",
|
| 49 |
"bytes": 873
|
| 50 |
},
|
| 51 |
"model.safetensors": {
|
| 52 |
+
"url": "https://huggingface.co/nirca/nirca-mini/resolve/main/checkpoints/2584388k/model.safetensors",
|
| 53 |
+
"sha256": "15ef10cc73e4286013656f9f36366cc960d7249a87cd4d4a0d4f6ffdbd9eeb28",
|
| 54 |
"bytes": 32638048
|
| 55 |
}
|
| 56 |
}
|
latest/config.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
{
|
| 2 |
"name": "nirca-mini",
|
| 3 |
-
"step":
|
| 4 |
-
"tokens":
|
| 5 |
"quant": "ternary",
|
| 6 |
"quant_scope": "all",
|
| 7 |
"n_experts": 32,
|
|
|
|
| 1 |
{
|
| 2 |
"name": "nirca-mini",
|
| 3 |
+
"step": 78160,
|
| 4 |
+
"tokens": 2584388242,
|
| 5 |
"quant": "ternary",
|
| 6 |
"quant_scope": "all",
|
| 7 |
"n_experts": 32,
|
latest/model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 32638048
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:15ef10cc73e4286013656f9f36366cc960d7249a87cd4d4a0d4f6ffdbd9eeb28
|
| 3 |
size 32638048
|