Brian Semrau commited on
Commit
f1e2244
1 Parent(s): 516e93a

upload latest model, fix tokenizer

Browse files
README.md CHANGED
@@ -4,9 +4,13 @@ tags:
4
  - music
5
  ---
6
 
7
- Trained in 48hr on an RTX3090
 
8
 
9
- Trained using GiantMIDI-Piano dataset, augmented 30x
 
 
10
 
11
  To run, use the rwkv pip package: https://pypi.org/project/rwkv/
12
- To convert to MIDI, use https://github.com/briansemrau/MIDI-LLM-tokenizer
 
 
4
  - music
5
  ---
6
 
7
+ step650 model was trained in 48hr on an RTX3090,
8
+ step1100 model initially experimentally trained to add tiny attenion (RWKV-4b), but then finetuned again to remove it. Final finetuning was done with ctx 4096.
9
 
10
+ See training history: https://wandb.ai/brianflakes/rwkv-midi
11
+
12
+ Trained using GiantMIDI-Piano dataset, augmented 30x.
13
 
14
  To run, use the rwkv pip package: https://pypi.org/project/rwkv/
15
+
16
+ To convert to MIDI, use https://github.com/briansemrau/MIDI-LLM-tokenizer
rwkvmidipiano-12L_1024D_step1100_ctx4096.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff6664e7fb2edb9da16f13ffe1ddbcd657aa9eaaba0d8848bc242a5b442b4d3a
3
+ size 672771653
tokenizer-midipiano.json CHANGED
@@ -45,131 +45,131 @@
45
  "<pad>": 0,
46
  "<start>": 1,
47
  "<end>": 2,
48
- "t0": 3,
49
- "t1": 4,
50
- "t2": 5,
51
- "t3": 6,
52
- "t4": 7,
53
- "t5": 8,
54
- "t6": 9,
55
- "t7": 10,
56
- "t8": 11,
57
- "t9": 12,
58
- "t10": 13,
59
- "t11": 14,
60
- "t12": 15,
61
- "t13": 16,
62
- "t14": 17,
63
- "t15": 18,
64
- "t16": 19,
65
- "t17": 20,
66
- "t18": 21,
67
- "t19": 22,
68
- "t20": 23,
69
- "t21": 24,
70
- "t22": 25,
71
- "t23": 26,
72
- "t24": 27,
73
- "t25": 28,
74
- "t26": 29,
75
- "t27": 30,
76
- "t28": 31,
77
- "t29": 32,
78
- "t30": 33,
79
- "t31": 34,
80
- "t32": 35,
81
- "t33": 36,
82
- "t34": 37,
83
- "t35": 38,
84
- "t36": 39,
85
- "t37": 40,
86
- "t38": 41,
87
- "t39": 42,
88
- "t40": 43,
89
- "t41": 44,
90
- "t42": 45,
91
- "t43": 46,
92
- "t44": 47,
93
- "t45": 48,
94
- "t46": 49,
95
- "t47": 50,
96
- "t48": 51,
97
- "t49": 52,
98
- "t50": 53,
99
- "t51": 54,
100
- "t52": 55,
101
- "t53": 56,
102
- "t54": 57,
103
- "t55": 58,
104
- "t56": 59,
105
- "t57": 60,
106
- "t58": 61,
107
- "t59": 62,
108
- "t60": 63,
109
- "t61": 64,
110
- "t62": 65,
111
- "t63": 66,
112
- "t64": 67,
113
- "t65": 68,
114
- "t66": 69,
115
- "t67": 70,
116
- "t68": 71,
117
- "t69": 72,
118
- "t70": 73,
119
- "t71": 74,
120
- "t72": 75,
121
- "t73": 76,
122
- "t74": 77,
123
- "t75": 78,
124
- "t76": 79,
125
- "t77": 80,
126
- "t78": 81,
127
- "t79": 82,
128
- "t80": 83,
129
- "t81": 84,
130
- "t82": 85,
131
- "t83": 86,
132
- "t84": 87,
133
- "t85": 88,
134
- "t86": 89,
135
- "t87": 90,
136
- "t88": 91,
137
- "t89": 92,
138
- "t90": 93,
139
- "t91": 94,
140
- "t92": 95,
141
- "t93": 96,
142
- "t94": 97,
143
- "t95": 98,
144
- "t96": 99,
145
- "t97": 100,
146
- "t98": 101,
147
- "t99": 102,
148
- "t100": 103,
149
- "t101": 104,
150
- "t102": 105,
151
- "t103": 106,
152
- "t104": 107,
153
- "t105": 108,
154
- "t106": 109,
155
- "t107": 110,
156
- "t108": 111,
157
- "t109": 112,
158
- "t110": 113,
159
- "t111": 114,
160
- "t112": 115,
161
- "t113": 116,
162
- "t114": 117,
163
- "t115": 118,
164
- "t116": 119,
165
- "t117": 120,
166
- "t118": 121,
167
- "t119": 122,
168
- "t120": 123,
169
- "t121": 124,
170
- "t122": 125,
171
- "t123": 126,
172
- "t124": 127,
173
  "p:0:0": 128,
174
  "p:0:1": 129,
175
  "p:0:2": 130,
 
45
  "<pad>": 0,
46
  "<start>": 1,
47
  "<end>": 2,
48
+ "t1": 3,
49
+ "t2": 4,
50
+ "t3": 5,
51
+ "t4": 6,
52
+ "t5": 7,
53
+ "t6": 8,
54
+ "t7": 9,
55
+ "t8": 10,
56
+ "t9": 11,
57
+ "t10": 12,
58
+ "t11": 13,
59
+ "t12": 14,
60
+ "t13": 15,
61
+ "t14": 16,
62
+ "t15": 17,
63
+ "t16": 18,
64
+ "t17": 19,
65
+ "t18": 20,
66
+ "t19": 21,
67
+ "t20": 22,
68
+ "t21": 23,
69
+ "t22": 24,
70
+ "t23": 25,
71
+ "t24": 26,
72
+ "t25": 27,
73
+ "t26": 28,
74
+ "t27": 29,
75
+ "t28": 30,
76
+ "t29": 31,
77
+ "t30": 32,
78
+ "t31": 33,
79
+ "t32": 34,
80
+ "t33": 35,
81
+ "t34": 36,
82
+ "t35": 37,
83
+ "t36": 38,
84
+ "t37": 39,
85
+ "t38": 40,
86
+ "t39": 41,
87
+ "t40": 42,
88
+ "t41": 43,
89
+ "t42": 44,
90
+ "t43": 45,
91
+ "t44": 46,
92
+ "t45": 47,
93
+ "t46": 48,
94
+ "t47": 49,
95
+ "t48": 50,
96
+ "t49": 51,
97
+ "t50": 52,
98
+ "t51": 53,
99
+ "t52": 54,
100
+ "t53": 55,
101
+ "t54": 56,
102
+ "t55": 57,
103
+ "t56": 58,
104
+ "t57": 59,
105
+ "t58": 60,
106
+ "t59": 61,
107
+ "t60": 62,
108
+ "t61": 63,
109
+ "t62": 64,
110
+ "t63": 65,
111
+ "t64": 66,
112
+ "t65": 67,
113
+ "t66": 68,
114
+ "t67": 69,
115
+ "t68": 70,
116
+ "t69": 71,
117
+ "t70": 72,
118
+ "t71": 73,
119
+ "t72": 74,
120
+ "t73": 75,
121
+ "t74": 76,
122
+ "t75": 77,
123
+ "t76": 78,
124
+ "t77": 79,
125
+ "t78": 80,
126
+ "t79": 81,
127
+ "t80": 82,
128
+ "t81": 83,
129
+ "t82": 84,
130
+ "t83": 85,
131
+ "t84": 86,
132
+ "t85": 87,
133
+ "t86": 88,
134
+ "t87": 89,
135
+ "t88": 90,
136
+ "t89": 91,
137
+ "t90": 92,
138
+ "t91": 93,
139
+ "t92": 94,
140
+ "t93": 95,
141
+ "t94": 96,
142
+ "t95": 97,
143
+ "t96": 98,
144
+ "t97": 99,
145
+ "t98": 100,
146
+ "t99": 101,
147
+ "t100": 102,
148
+ "t101": 103,
149
+ "t102": 104,
150
+ "t103": 105,
151
+ "t104": 106,
152
+ "t105": 107,
153
+ "t106": 108,
154
+ "t107": 109,
155
+ "t108": 110,
156
+ "t109": 111,
157
+ "t110": 112,
158
+ "t111": 113,
159
+ "t112": 114,
160
+ "t113": 115,
161
+ "t114": 116,
162
+ "t115": 117,
163
+ "t116": 118,
164
+ "t117": 119,
165
+ "t118": 120,
166
+ "t119": 121,
167
+ "t120": 122,
168
+ "t121": 123,
169
+ "t122": 124,
170
+ "t123": 125,
171
+ "t124": 126,
172
+ "t125": 127,
173
  "p:0:0": 128,
174
  "p:0:1": 129,
175
  "p:0:2": 130,