Upload MatformerForCausalLM
Browse files- config.json +2 -2
- generation_config.json +1 -1
config.json
CHANGED
|
@@ -58,7 +58,6 @@
|
|
| 58 |
"positional_encoding": "alibi",
|
| 59 |
"sliding_window_size": null
|
| 60 |
},
|
| 61 |
-
"dtype": "bfloat16",
|
| 62 |
"encoder": null,
|
| 63 |
"entropy": null,
|
| 64 |
"eos_token_id": 2,
|
|
@@ -79,8 +78,9 @@
|
|
| 79 |
"rms_norm_eps": 1e-06,
|
| 80 |
"rope_theta": 10000.0,
|
| 81 |
"sliding_type": null,
|
|
|
|
| 82 |
"training_objective": "autoregressive",
|
| 83 |
-
"transformers_version": "4.
|
| 84 |
"use_cache": false,
|
| 85 |
"vocab_size": 32768
|
| 86 |
}
|
|
|
|
| 58 |
"positional_encoding": "alibi",
|
| 59 |
"sliding_window_size": null
|
| 60 |
},
|
|
|
|
| 61 |
"encoder": null,
|
| 62 |
"entropy": null,
|
| 63 |
"eos_token_id": 2,
|
|
|
|
| 78 |
"rms_norm_eps": 1e-06,
|
| 79 |
"rope_theta": 10000.0,
|
| 80 |
"sliding_type": null,
|
| 81 |
+
"torch_dtype": "bfloat16",
|
| 82 |
"training_objective": "autoregressive",
|
| 83 |
+
"transformers_version": "4.53.0.dev0",
|
| 84 |
"use_cache": false,
|
| 85 |
"vocab_size": 32768
|
| 86 |
}
|
generation_config.json
CHANGED
|
@@ -3,6 +3,6 @@
|
|
| 3 |
"bos_token_id": 1,
|
| 4 |
"eos_token_id": 2,
|
| 5 |
"pad_token_id": 0,
|
| 6 |
-
"transformers_version": "4.
|
| 7 |
"use_cache": false
|
| 8 |
}
|
|
|
|
| 3 |
"bos_token_id": 1,
|
| 4 |
"eos_token_id": 2,
|
| 5 |
"pad_token_id": 0,
|
| 6 |
+
"transformers_version": "4.53.0.dev0",
|
| 7 |
"use_cache": false
|
| 8 |
}
|